Epidemiology

Epidemiology_For_Public_Health_Practice_Friis_Epidemiology_for_Public_Health_Practice_5th_Edition.pdf

Home >Environmental science homework help >Epidemiology

Epidemiology for Public Health Practice Robert H. Friis, PhD Professor, Emeritus, and Chair Emeritus Health Science Department California State University Long Beach, California

Thomas A. Sellers, PhD, MPH Director Moffitt Cancer Center & Research Institute Tampa, Florida

FIFTH EDITION

51589_FM00_Printer.indd 1 15/02/13 3:11 AM

World Headquarters Jones & Bartlett Learning 5 Wall Street Burlington, MA 01803 978-443-5000 [email protected] www.jblearning.com

Jones & Bartlett Learning books and products are available through most bookstores and online book- sellers. To contact Jones & Bartlett Learning directly, call 800-832-0034, fax 978-443-8000, or visit our website, www.jblearning.com.

Substantial discounts on bulk quantities of Jones & Bartlett Learning publications are available to corporations, professional associations, and other qualified organizations. For details and specific discount information, contact the special sales department at Jones & Bartlett Learning via the above contact information or send an email to [email protected].

All rights reserved. No part of the material protected by this copyright may be reproduced or utilized in any form, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without written permission from the copyright owner.

Epidemiology for Public Health Practice, Fifth Edition is an independent publication and has not been authorized, sponsored, or otherwise approved by the owners of the trademarks or service marks referenced in this product.

Some images in this book feature models. These models do not necessarily endorse, represent, or participate in the activities represented in the images.

The screenshots in this product are for educational and instructive purposes only. All trademarks displayed are the trademarks of the parties noted therein. Such use of trademarks is not an endorsement by said parties of Jones & Bartlett Learning, its products, or its services, nor should such use be deemed an endorsement by Jones & Bartlett Learning of said third party’s products or services.

This publication is designed to provide accurate and authoritative information in regard to the Subject Matter covered. It is sold with the understanding that the publisher is not engaged in rendering legal, accounting, or other professional service. If legal advice or other expert assistance is required, the service of a competent professional person should be sought.

Production Credits Publisher: Michael Brown Managing Editor: Maro Gartside Editorial Assistant: Kayla Dos Santos Editorial Assistant: Chloe Falivene Production Assistant: Alyssa Lawrence Senior Marketing Manager: Sophie Fleck Teague

Manufacturing and Inventory Control Supervisor: Amy Bacus Composition: diacriTech Cover Design: Kristin E. Parker Cover Image: © W7/ ShutterStock, Inc. Printing and Binding: Edwards Brothers Malloy Cover Printing: Edwards Brothers Malloy

Printed in the United States of America 17 16 15 14 13 10 9 8 7 6 5 4 3 2 1

To order this product, use ISBN: 978-1-4496-6549-4

Library of Congress Cataloging-in-Publication Data Friis, Robert H. Epidemiology for public health practice / Robert H. Friis and Thomas Sellers.—5th ed. p. ; cm. Includes bibliographical references and index. ISBN 978-1-4496-5158-9 (pbk.) I. Sellers, Thomas A. II. Title. [DNLM: 1. Epidemiology. 2. Epidemiologic Methods. 3. Public Health. WA 105] 614.4—dc23

2012039130

6048

51589_FM00_Printer.indd 2 15/02/13 3:11 AM

New to This Edition ................................................... ix

Introduction ............................................................... xiii

Preface ........................................................................ xvii

Acknowledgments ...................................................... xix

About the Authors ............................................................ xxiii

Chapter 1 History and Scope of Epidemiology ........................... 1 Introduction ................................................................... 2 Epidemiology Defined .................................................... 8 Foundations of Epidemiology ......................................... 15 Historical Antecedents of Epidemiology ......................... 23 Recent Applications of Epidemiology ............................. 41 Conclusion ..................................................................... 48 Study Questions and Exercises ........................................ 49 References ....................................................................... 51

Chapter 2 Practical Applications of Epidemiology ..................... 55 Introduction ................................................................... 56 Applications for the Assessment of the Health Status of Populations and Delivery of Health Services .............. 59 Applications Relevant to Disease Etiology ....................... 83 Conclusion ..................................................................... 101 Study Questions and Exercises ........................................ 101 References ....................................................................... 104

Contents

iii

51589_FM00_Printer.indd 3 15/02/13 3:11 AM

Chapter 3 Measures of Morbidity and Mortality Used in Epidemiology ......................................................... 107

Introduction ................................................................... 108 Definitions of Count, Ratio, Proportion, and Rate ......... 108 Risk Versus Rate; Cumulative Incidence ......................... 121 Interrelationship Between Prevalence and Incidence ....... 124 Applications of Incidence Data ....................................... 126 Crude Rates .................................................................... 126 Specific Rates and Proportional Mortality Ratio ............. 138 Adjusted Rates ................................................................ 144 Conclusion ..................................................................... 151 Study Questions and Exercises ........................................ 152 References ....................................................................... 155

Chapter 4 Descriptive Epidemiology: Person, Place, Time ......... 157 Introduction ................................................................... 158 Characteristics of Persons ................................................ 163 Characteristics of Place ................................................... 203 Characteristics of Time ................................................... 217 Conclusion ..................................................................... 223 Study Questions and Exercises ........................................ 223 References ....................................................................... 225 Appendix 4—Project: Descriptive Epidemiology of a Selected Health Problem ......................................... 233

Chapter 5 Sources of Data for Use in Epidemiology ................... 235 Introduction ................................................................... 236 Criteria for the Quality and Utility of Epidemiologic Data ................................................... 239 Online Sources of Epidemiologic Data ........................... 241 Confidentiality, Sharing of Data, and Record Linkage .... 244 Statistics Derived from the Vital Registration System ..... 247 Reportable Disease Statistics ........................................... 254 Screening Surveys ........................................................... 259 Disease Registries ............................................................ 260 Morbidity Surveys of the General Population ................. 262 Insurance Data................................................................ 267 Clinical Data Sources ...................................................... 267 Absenteeism Data ........................................................... 271 School Health Programs ................................................. 272

iv C o n t e n t s

51589_FM00_Printer.indd 4 15/02/13 3:11 AM

Morbidity in the Armed Forces: Data on Active Personnel and Veterans .............................................. 272 Other Sources: Census Data ........................................... 273 Conclusion ..................................................................... 274 Study Questions and Exercises ........................................ 274 References ....................................................................... 276

Chapter 6 Study Designs: Ecologic, Cross-Sectional, Case-Control .......................................................... 279

Introduction ................................................................... 280 Observational Versus Experimental Approaches in Epidemiology ............................................................. 281 Overview of Study Designs Used in Epidemiology ......... 282 Ecologic Studies .............................................................. 287 Cross-Sectional Studies ................................................... 294 Case-Control Studies ...................................................... 303 Conclusion ..................................................................... 317 Study Questions and Exercises ........................................ 317 References ....................................................................... 319

Chapter 7 Study Designs: Cohort Studies ................................... 323 Introduction ................................................................... 324 Cohort Studies Defined .................................................. 325 Sampling and Cohort Formation Options ...................... 335 Temporal Differences in Cohort Designs ........................ 341 Practical Considerations .................................................. 344 Measures of Effect: Their Interpretation and Examples ... 347 Summary of Cohort Studies ............................................ 358 Conclusion ..................................................................... 359 Study Questions and Exercises ........................................ 362 References ....................................................................... 363

Chapter 8 Experimental Study Designs ...................................... 367 Introduction ................................................................... 368 Hierarchy of Study Designs ............................................ 371 Intervention Studies ........................................................ 373 Clinical Trials ................................................................. 374 Community Trials .......................................................... 392 Conclusion ..................................................................... 404 Study Questions and Exercises ........................................ 405 References ....................................................................... 406

C o n t e n t s v

51589_FM00_Printer.indd 5 15/02/13 3:11 AM

Chapter 9 Measures of Effect ...................................................... 409 Introduction ................................................................... 410 Absolute Effects .............................................................. 410 Relative Effects ............................................................... 414 Statistical Measures of Effect ........................................... 420 Evaluating Epidemiologic Associations ........................... 423 Models of Causal Relationships ...................................... 425 Conclusion ..................................................................... 430 Study Questions and Exercises ........................................ 431 References ....................................................................... 432 Appendix 9—Cohort Study Data for Coffee Use and Anxiety ................................................................ 433

Chapter 10 Data Interpretation Issues .......................................... 435 Introduction ................................................................... 436 Validity of Study Designs ............................................... 437 Sources of Error in Epidemiologic Research .................... 440 Techniques to Reduce Bias ............................................. 449 Methods to Control Confounding .................................. 450 Bias in Analysis and Publication ...................................... 454 Conclusion ..................................................................... 456 Study Questions and Exercises ........................................ 456 References ....................................................................... 458

Chapter 11 Screening for Disease in the Community ................... 461 Introduction ................................................................... 462 Screening for Disease ...................................................... 464 Appropriate Situations for Screening Tests and Programs .................................................................... 468 Characteristics of a Good Screening Test ........................ 471 Evaluation of Screening Tests ......................................... 471 Sources of Unreliability and Invalidity ............................ 476 Measures of the Validity of Screening Tests .................... 476 Effects of Prevalence of Disease on Screening Test Results........................................................................ 479 Relationship Between Sensitivity and Specificity ............. 482 Evaluation of Screening Programs ................................... 483 Issues in the Classification of Morbidity and Mortality ... 485 Conclusion ..................................................................... 486 Study Questions and Exercises ........................................ 487 References ....................................................................... 488 Appendix 11—Data for Problem 6 ................................. 490

vi C o n t e n t s

51589_FM00_Printer.indd 6 15/02/13 3:11 AM

Chapter 12 Epidemiology of Infectious Diseases .......................... 491 Introduction ................................................................... 492 Agents of Infectious Disease ............................................ 493 Characteristics of Infectious Disease Agents .................... 496 Host ............................................................................... 497 The Environment ........................................................... 499 Means of Transmission: Directly or Indirectly from Reservoir .................................................................... 500 Measures of Disease Outbreaks ....................................... 506 Procedures Used in the Investigation of Infectious Disease Outbreaks ...................................................... 511 Epidemiologically Significant Infectious Diseases in the Community ............................................................... 513 Conclusion ..................................................................... 539 Study Questions and Exercises ........................................ 539 References ....................................................................... 542 Appendix 12—Data from a Foodborne Illness Outbreak in a College Cafeteria ................................. 545

Chapter 13 Epidemiologic Aspects of Work in the Environment .......................................................... 547

Introduction ................................................................... 548 Health Effects Associated with Environmental Hazards .. 550 Study Designs Used in Environmental Epidemiology ..... 550 Toxicologic Concepts Related to Environmental Epidemiology ............................................................. 555 Types of Agents .............................................................. 557 Environmental Hazards Found in the Work Setting ....... 571 Noteworthy Community Environmental Health Hazards ...................................................................... 575 Conclusion ..................................................................... 588 Study Questions and Exercises ........................................ 591 References ....................................................................... 592

Chapter 14 Molecular and Genetic Epidemiology ........................ 599 Introduction ................................................................... 600 Definitions and Distinctions: Molecular Versus Genetic Epidemiology ............................................................. 605 Epidemiologic Evidence for Genetic Factors ................... 609 Causes of Familial Aggregation ....................................... 610 Shared Family Environment and Familial Aggregation ... 612 Gene Mapping: Segregation and Linkage Analysis .......... 616

C o n t e n t s vii

51589_FM00_Printer.indd 7 15/02/13 3:11 AM

Genome-Wide Association Studies (GWAS) .................. 626 Linkage Disequilibrium Revisited: Haplotypes ............... 628 Application of Genes in Epidemiologic Designs .............. 631 Genetics and Public Health ............................................ 638 Conclusion ..................................................................... 642 Study Questions and Exercises ........................................ 642 References ....................................................................... 643

Chapter 15 Social, Behavioral, and Psychosocial Epidemiology ... 649 Introduction ................................................................... 650 Research Designs Used in Psychosocial, Behavioral, and Social Epidemiology ................................................... 655 The Social Context of Health ......................................... 657 Independent Variables .................................................... 660 Moderating Variables ...................................................... 669 Dependent (Outcome) Variables: Physical and Mental Health ........................................................................ 684 Conclusion ..................................................................... 691 Study Questions and Exercises ........................................ 692 References ....................................................................... 694

Chapter 16 Epidemiology as a Profession ..................................... 701 Introduction ................................................................... 702 Specializations within Epidemiology ............................... 703 Career Roles for Epidemiologists .................................... 705 Epidemiology Associations and Journals ......................... 708 Competencies Required of Epidemiologists .................... 711 Resources for Education and Employment ..................... 712 Professional Ethics in Epidemiology ............................... 714 Conclusion ..................................................................... 719 Study Questions and Exercises ........................................ 720 References ....................................................................... 721

Appendix A—Guide to the Critical Appraisal of an Epidemiologic/Public Health Research Article ...... 723

Appendix B—Answers to Selected Study Questions ... 727

Glossary ...................................................................... 737

Index ............................................................................. 759

viii C o n t e n t s

51589_FM00_Printer.indd 8 15/02/13 3:11 AM

New to This Edition

Chapter 1: History and Scope of Epidemiology

●● New and updated images ●● Updated chart: three presentations of epidemiologic data ●● Updated chart: pneumonia and influenza mortality ●● New chart on the interdisciplinary nature of epidemiology ●● Glossary of terms used in the yearly bill of mortality for 1632 ●● Expanded information on cholera and John Snow

Chapter 2: Pract ical Applicat ions of Epidemiology

●● Updated information on leading causes of death from 1900 to 2009 ●● Expanded discussion of population dynamics and predictions about the future ●● More information provided on the health of the community and health

disparities, including the GINI index

Chapter 3: Measures of Morbidity and Mortal i ty Used in Epidemiology

●● Expanded coverage of epidemiologic measures (e.g., sex ratios) ●● More information on prevalence given with figure to show interrelationships

between prevalence and incidence ●● Further clarification of perinatal mortality provided

51589_FM00_Printer.indd 9 15/02/13 3:11 AM

Chapter 4: Descript ive Epidemiology: Person, Place, Time

●● Updated coverage of morbidity and mortality data by descriptive epidemiologic variables provided throughout the chapter

●● New examples of case studies and case series ●● New information on age effects associated with morbidity and mortality ●● Many new charts added to this chapter ●● Updates from the 2010 Census, with current definitions of race/ ethnicity

Chapter 5: Sources of Data for Use in Epidemiology

●● Updated information on data sources including notifiable diseases ●● Further clarification of criteria for the quality of epidemiologic data ●● Rationale strengthened for the need for high-quality epidemiologic data

Chapter 6: Study Designs: Ecologic, Cross- Sectional , Case-Control

●● Clarification regarding design and applications of case-control studies ●● More information on matching in case-control studies ●● Clearer definitions of terms provided ●● Further discussion of comparisons between cross-sectional and case-

control studies

Chapter 7: Study Designs: Cohort Studies

●● Introduction updated ●● Additional clarification of terminology used in cohort studies ●● Exhibit on life table methods updated to the most recent information

Chapter 8: Experimental Study Designs

●● Expanded coverage of intervention studies ●● Several new images, including an image of a scurvy victim ●● Discussion of phase 4 clinical trials ●● New table and a glossary of terms used in clinical trials ●● Applications of epidemiology to vaccines and prevention: HPV vaccine

x n e w t o t h i s e d i t i o n

51589_FM00_Printer.indd 10 15/02/13 3:11 AM

Chapter 9: Measures of Effect

●● Introduction revised ●● STROBE guidelines and quality of epidemiologic studies ●● Meta-analysis and systematic reviews

Chapter 10: Data Interpretat ion Issues

●● More information on Simpson’s Paradox, including a new figure ●● Information bias and screening mammography

Chapter 11: Screening for Disease in the Community

●● New figure showing participants in a mammogram and a blood pressure screening test

●● New figure showing participation rates in screening for colorectal cancer, breast cancer, and cervical cancer

●● Updated discussion on controversies in screening ●● Difficulties with false positive screening test results

Chapter 12: Epidemiology of Infect ious Diseases

●● Many updated charts showing data on disease incidence and prevalence (e.g., measles, malaria, hepatitis, valley fever, Lyme disease)

●● Information on the cholera epidemic in Haiti ●● Revised exhibit on viral hepatitis

Chapter 13: Epidemiologic Aspects of Work in the Environment

●● New information on methodologic topics (e.g., exposure assessments, clustering, and confounding)

●● Updated data on blood lead levels and mercury advisories ●● New topics include global warming, the BP oil spill, and the Japanese

tsunami and its effects on the Fukushima nuclear reactor ●● Many new images to capture students’ interest in this topic

n e w t o t h i s e d i t i o n xi

51589_FM00_Printer.indd 11 15/02/13 3:11 AM

Chapter 14: Molecular and Genetic Epidemiology

●● New diagram of Mendelian inheritance ●● Additional discussion of the population genetics concept of linkage

disequilibrium ●● Expanded discussion of the concept of haplotypes ●● A thorough update of this chapter with the latest developments in the field

Chapter 15: Social , Behavioral , and Psychosocial Epidemiology

●● Many new illustrations added to this chapter ●● The concept of community-based participatory research added ●● New information on the social context of health (e.g., poverty, the Glasgow

effect) ●● Healthy People 2020 overarching goals included ●● Update on depression

Chapter 16: Epidemiology as a Profession

●● Updated to show current professional resources and issues

Other

●● Exciting new figures, tables, and exhibits provided throughout ●● Additional exercises and study questions

xii n e w t o t h i s e d i t i o n

51589_FM00_Printer.indd 12 15/02/13 3:11 AM

Introduction

Epidemiology is an important, exciting, and rewarding field for the public health practitioner! Almost daily, one hears dramatic media reports about flare-ups of diseases, either previously known or seemingly new conditions. These accounts demonstrate how epidemiologists help to uncover the causes of human illnesses in the population and thereby underscore the importance of epidemiology to society. Deadly outbreaks of communicable diseases, the ongoing threat of resur- gent epidemics, and the possible intentional spread of pathogenic microorgan- isms through acts of bioterrorism present challenges to the field. By assisting the reader in understanding why and how diseases occur and how they may be pre- vented, epidemiology is a valuable pursuit. In this text you will learn that many epidemiologic investigations into the causes of mysterious outbreaks are similar to detective work.

One of the challenges for the authors has been to distill with sufficient breadth and depth all of the fascinating components of this discipline. As the Fifth Edition is being finalized, new and resurgent health conditions challenge public health practitioners; some current examples are resurgent whooping cough, outbreaks of foodborne diseases, hantavirus infections (which normally are infrequent) in a national park, fungal meningitis associated with epidural steroid injections, and a West Nile virus epidemic. Thus, the ongoing flow of accounts of disease outbreaks (noted in the First Edition) has not been staunched and, in fact, is con- tinuing unabated during the second decade of the 21st century.

Since the publication of the earlier editions of this book, the wealth of epidemiologic research findings has continued to proliferate and win the atten- tion of the popular media and professional journals. For example, some of these recent discoveries relate to continuing advances in genetics and molecu- lar biology, recognition of emerging infections, and the growing use of the Internet. As a result, the Second Edition introduced several enhancements: a new chapter on molecular and genetic epidemiology, a new chapter on experimental

xiii

51589_FM00_Printer.indd 13 15/02/13 3:11 AM

epidemiology, material on epidemiology Internet sites, and updated charts and tables throughout the text.

The Third Edition incorporated a new chapter on cohort designs, a glossary, and an expanded coverage of ecologic and case-control study designs. The Third Edition also included new material on the role of epidemiology in policy making, epidemiology and geographic information systems, and the definition of race used in Census 2000. A new Appendix A provided an extended guide to critiqu- ing published research studies in public health and epidemiology. Several new tables summarized unadjusted measures of morbidity and mortality, contrasted different types of observational study designs, and compared observational versus intervention study designs.

The Fourth Edition presented new information on infectious disease threats associated with E. coli foodborne illness and avian influenza as well as expanded coverage of the historical background of epidemiology. Chapter 3, “Measures of Morbidity and Mortality Used in Epidemiology,” was updated to reflect the use of the 2000 standard population in age standardization. A new Chapter 16, titled “Epidemiology as a Profession,” covered methods for accessing the profes- sion and employment opportunities in the field.

The Fifth Edition provides an extensive update of information from the previ- ous editions. Examples are coverage of the 2009 H1N1 influenza epidemic, the 2010 U.S. Census, and numerous additional and updated figures, charts, and photographs throughout the book. Trends in morbidity have been updated to reflect the most recently available information. New information is presented throughout the text: for example, in Chapter 12 (infectious diseases), Chapter 13 (environmental health), and Chapter 14 (molecular and genetic epidemiol- ogy). Definitions used in the text have been aligned with the 2008 Dictionary of Epidemiology, a standard reference in the field.

We intend the audience for the textbook to be beginning public health mas- ter’s degree students, undergraduate and graduate health education and social ecology students, undergraduate medical students, nursing students, residents in primary care medicine, and applicants who are preparing for medical board examinations. These students are similar to those with whom both authors have worked over the years. Students from the social and behavioral sciences also have found epidemiology to be a useful tool in medical sociology and behavioral med- icine. We have included study questions and exercises at the end of each chapter; this material would be helpful to review for board examinations. Appendix B contains an expanded answer set to selected problems.

xiv i n t r o d u C t i o n

51589_FM00_Printer.indd 14 15/02/13 3:11 AM

Each chapter begins with a list of learning objectives and an outline to help focus the reader’s attention to key points. Some of the major issues and examples are highlighted in text boxes and tables. Chapter 1, which defines epidemiol- ogy and provides a historical background for the discipline, is complemented by Chapter 2, which provides examples of practical applications of epidemiology as well as a discussion of causal inference. Although examples of epidemiologic statistical techniques are interspersed throughout the book, Chapter 3 focuses on the “nuts and bolts” of measures of morbidity and mortality. Chapters 4 through 11 deal with the important topics of descriptive epidemiology: data sources, study designs, measures of effect, data interpretation, and screening. Chapters 12 through 15 focus on four content areas in epidemiology: infec- tious diseases, occupational and environmental health, molecular and genetic epidemiology, and psychosocial epidemiology. Finally, Chapter 16 covers pro- fessional issues in epidemiology. This text provides a thorough grounding in the key areas of methodology, causality, and the complex issues that surround chronic and infectious disease investigations. The authors assume that the reader will have had some familiarity with introductory biostatistics, although the text is intelligible to those who do not have such familiarity. A companion website for students is available for the text. This website provides extensive resources for students, including the student study guide that was included with the last edition. We recommend that students and instructors navigate through the site during class time. For example, the flashcards available may be used as part of an in-class activity to drill students for the class examinations. Dr. Friis uses in-class Internet navigation in order to show students how to locate resources for the project shown in the Appendix at the end of Chapter 4. Completion of the proj- ect can be one of the major assignments in an epidemiology class. In addition to completing a written version of the assignment, students may enjoy delivering a brief PowerPoint presentation of their research to the entire class. Students’ motivation and success in an epidemiology course are enhanced by reviewing the various activities provided.

i n t r o d u C t i o n xv

51589_FM00_Printer.indd 15 15/02/13 3:11 AM

51589_FM00_Printer.indd 16 15/02/13 3:11 AM

Preface

My interest in epidemiology began during the 1960s when, as an undergraduate student at the University of California at Berkeley and a graduate student at Columbia University, I observed the student revolts and activism that occurred during that era. Student unrest was, I believed, a phenomenon that occurred in large groups and could be explained by a theoretical framework, perhaps one that would include such concepts as alienation or anomie. I became interested in studying the distribution of these psychological states in student populations. Unknowingly, I had embarked upon epidemiologic research. I find epidemiol- ogy to be a field that has great personal appeal because it is capable of impacting the health of large groups of people through improvements in social conditions and environmental modifications.

My formal training in epidemiology began at the Institute for Social Research of the University of Michigan, where I spent 2 years as a postdoctoral fellow. My first professional position in epidemiology was as an assistant professor in the Division of Epidemiology at the School of Public Health, Columbia University. As a fledgling professor, I found epidemiology to be a fascinating discipline, and began to develop this textbook from my early teaching experiences. I concluded that there was a need for a textbook that would be oriented toward the begin- ning practitioner in the field, would provide coverage of a wide range of topics, and would emphasize the social and behavioral foundations of epidemiology as well as the medical model. This textbook has evolved from my early teaching experience at Columbia as well as later teaching and research positions at Albert Einstein College of Medicine, Brooklyn College, the University of California at Irvine, and the California State University system. Practical experience in epide- miology, as an epidemiologist in a local health department in Orange County, California, is also reflected in the book.

—Robert H. Friis

xvii

51589_FM00_Printer.indd 17 15/02/13 3:11 AM

Like many others now reading this book, I had absolutely no idea what epidemiology was before I took my first required class in it at Tulane University School of Public Health and Tropical Medicine. What I discovered was a method to combine my training in nutrition and interest in health with an aptitude for math and analytical reasoning. This led to a change in majors and ultimately a PhD in epidemiology.

My first faculty appointment was at the University of Minnesota School of Public Health. Before I knew it, I was assigned to teach the introduction to epi- demiology course during the winter quarter. This was the time of year when only nonmajors enrolled. I quickly learned, as had my predecessors, that my teaching and learning style was quite different from those of my students. Moreover, most of the textbooks available at that time were geared toward epidemiology majors. For 9 years, I studied learning styles (and even co-developed and co-taught a graduate course on teaching) and experimented to find new ways to present the fundamentals of epidemiology in a nontechnical, nontheoretical, intuitive man- ner. This text reflects these learning experiences.

—Thomas A. Sellers

xviii P r e f a C e

51589_FM00_Printer.indd 18 15/02/13 3:11 AM

Acknowledgments

First, I express my gratitude to my teachers and colleagues at the settings where I have worked during the past 4 decades. Their insights and suggestions have helped me clarify my thinking about epidemiology. Among these individu- als are the late Dr. Sidney Cobb and the late Dr. John R. P. French, Jr., who were my postdoctoral supervisors at the University of Michigan’s Institute for Social Research. Dr. Mervyn Susser offered me my first professional employ- ment in epidemiology at the School of Public Health, Columbia University. He and Dr. Zena Stein helped me to greatly increase my fund of knowledge about research and teaching in the field. The late Professor Anna Gelman pro- vided me with many practical ideas regarding how to teach epidemiology. Dr. Stephen A. Richardson also contributed to my knowledge about epidemiologic research. Finally, Dr. Jeremiah Tilles, former Associate Dean, California College of Medicine, University of California at Irvine, helped to increase my insights regarding the epidemiology of infectious diseases.

I also thank students in my epidemiology classes who contributed their suggestions and read early drafts of the first edition. The comments of anony- mous reviewers were particularly helpful in revising the manuscript. Jonathan Horowitz, former instructor in Health Science at California State University, Long Beach, spent a great deal of time reviewing several chapters of a very early version of the text, and I acknowledge his contributions. Sherry Stock, a former student in medical sociology at Long Beach, typed the first draft and provided much additional valuable assistance in securing bibliographic research materials. Dr. Yee-Lean Lee, Professor, Infectious Disease Division in the Department of Medicine at the University of California at Irvine, reviewed and commented on the chapter dealing with the epidemiology of infectious diseases. Also, Dr. Harold Hunter, Professor Emeritus of Health Care Administration, California State University, Long Beach, reviewed several chapters of the manuscript.

xix

51589_FM00_Printer.indd 19 15/02/13 3:11 AM

Finally, my wife, Carol Friis, typed the final version of the manuscript and made helpful comments. Without her support and assistance, completion of the text would not have been possible.

For the second edition of the text, I again thank my epidemiology students, who continued to provide much useful feedback. Graduate students Janelle Yamashita, Cindy Bayliss, and Jocelin Sabado were extremely helpful in con- ducting literature searches and preparing the text. Sharon Jean assisted with typ- ing the manuscript.

With respect to the third edition, I would like to thank students at my home university and at other universities who provided many worthwhile sugges- tions for enhancement of the text. I am also grateful for the informal feedback I received from faculty members (across the United States and in several for- eign countries) who adopted this text in their courses. Former California State University graduate student Ibtisam Khoury, now a lecturer in the Health Science Department, conducted background research, provided ideas for clari- fication of complex concepts, and helped to develop several new tables. Faculty members Dr. Javier Lopez-Zetina and Dr. Dennis Fisher, housed at the same university, reviewed several of the chapters. Critiques from anonymous reviewers also were instrumental in development of the third edition. Once again, I am deeply indebted to my wife, Carol Friis, who assisted with editing and typing the manuscript. Without her keen eye, writing this book would have been a much more difficult task.

Regarding the fourth edition, I once again acknowledge my students’ sugges- tions for continued improvement of this book. Although many students are wor- thy of recognition, I would especially like to thank graduate student Lesley Shen. Claire Garrido-Ortega, a former student and now a lecturer in the Department of Health Science, contributed her ideas to the new edition. I have received many suggestions from the readers of the previous edition of this text; I would like to thank them also—particularly Dr. Lee Caplan at Morehouse University. Once more, I recognize the support of my wife, Carol Friis, who helped with preparation of the text.

The fifth edition benefited from the input of students and faculty members in the Department of Health Science. Particularly noteworthy were the sugges- tions provided by faculty member Dr. Javier Lopez-Zetina and former graduate students (and now faculty members) Ibtisam Khoury, Che Wanke, and Claire Garrido-Ortega. Jaina Pallasigui, MPH graduate, helped with background research for this revision. Roxanne Garza reviewed the manuscript.

—R.H.F.

xx a C k n o w l e d g m e n t s

51589_FM00_Printer.indd 20 15/02/13 3:11 AM

I have been most fortunate to receive training and guidance from a significant number of individuals. First and foremost, I thank Dr. Dorothy Clemmer, who taught me my first course in epidemiology at Tulane University School of Public Health and Tropical Medicine. Her enthusiasm and support helped me to “see the light.” The early years of my education included mentorship with Dr. Gerald Berenson and Dr. Robert C. Elston. Both have been extremely influential in my practical and theoretical understanding of this discipline. Dr. J. Michael Sprafka was a great supporter and colleague for those first precarious episodes of teach- ing. I owe many thanks to the numerous bright and challenging public health students at the University of Minnesota for their support, encouragement, and patience while I experimented with methods of presentation to find out what worked best for “nonmajors.” Finally, I acknowledge my father, Gene R. Sellers, who has published many fine textbooks and gave me the courage to attempt this project; my loving wife, Barbara, for her understanding and enduring belief in me; and my two sons, Jamison Thomas and Ryan Austin, who are my inspira- tion and loves of my life.

For the second edition, I acknowledge the encouragement of the students and colleagues who had used the first edition of this text. I also thank our publisher and their staff for their professionalism. Finally, I acknowledge the drive and creativity of Bob Friis, whose energies made this book a reality and a success.

For the fourth edition, I would like to particularly thank my wonderful friends and colleagues at the Moffitt Cancer Center (especially Yifan Huang, Cathy Phelan, Jong Park, and Anna Giuliano) and the Mayo Cancer Center (espe- cially Ellen Goode, Jim Cerhan, Celine Vachon, and Shane Pankratz) for their brilliance and dedication. I’ve learned that the application of the epidemiologic method can be fun if you work with the right team. I have certainly benefited from being around such a wonderful cast of bright and stimulating people. This has translated into exciting research projects, new knowledge, and practical insights added to this edition. Moreover, they share my hope and dream for an end to cancer and the terrible impact of this disease.

For the fifth edition, I want to add a posthumous note of love and apprecia- tion to my mother for always believing in me and for encouraging my pursuit of an academic career dedicated to cancer research. That she lost her life to the disease has reconfirmed my determination to make an impact through applica- tion of the epidemiologic method.

—T.A.S.

a C k n o w l e d g m e n t s xxi

51589_FM00_Printer.indd 21 15/02/13 3:11 AM

51589_FM00_Printer.indd 22 15/02/13 3:11 AM

About the Authors

Robert H. Friis, PhD, is a Professor Emeritus of Health Science and Chair Emeritus of the Department of Health Science at California State University, Long Beach, and former Director of the CSULB-VAMC, Long Beach, Joint Studies Institute. He is also a former Clinical Professor of Community and Environmental Medicine at the University of California at Irvine. Previously, he was an Associate Clinical Professor in the Department of Medicine, Department of Neurology, and School of Social Ecology, University of California at Irvine. His entire professional career has been devoted to the field of epidemiology. He has conducted research and taught epidemiology and related subjects for more than 4 decades at universities in New York City and Southern California. In addition to previous employment in a local health department as an epidemi- ologist, he has conducted research and has published and presented numerous papers related to mental health, chronic disease, disability, minority health, and psychosocial epidemiology. His textbook, Essentials of Environmental Health, Second Edition, is also published by Jones & Bartlett Learning. Dr. Friis has been principal investigator or co-investigator on grants and contracts from University of California’s Tobacco-Related Disease Research Program, from the National Institutes of Health, and from other agencies for research on geriatric health, depression in Hispanic populations, nursing home infections, and environmental health issues. His research interests have led him to conduct research in Mexico City and European countries. He has been a visiting professor at the Center for Nutrition and Toxicology, Karolinska Institute, Stockholm, Sweden; the Max Planck Institute, Munich, Germany; and Dresden Technical University, also in Germany. He reviews articles for scientific journals and is a member of the editorial board of Public Health. Dr. Friis is a member of the Society for Epidemiologic Research, the American Public Health Association (epidemiology section), is a past president of the Southern California Public Health Association,

xxiii

51589_FM00_Printer.indd 23 15/02/13 3:11 AM

and is a fellow of the Royal Academy of Public Health. Among his awards are a postdoctoral fellowship (for study at the Institute for Social Research, University of Michigan), and the Achievement Award for Scholarly and Creative Activity from California State University, Long Beach. His biography is listed in Who’s Who in America.

Thomas A. Sellers, PhD, MPH, is Director of the Moffitt Cancer Center & Research Institute and Executive Vice President of the H. Lee Moffitt Cancer Center and Research Institute. Prior to this position in sunny, warm Tampa, Florida, he was Professor of Epidemiology in the Department of Health Sciences Research at the Mayo Clinic and the Deputy Director of the Mayo Clinic Cancer Center. He began his career at the University of Minnesota School of Public Health, where he taught the Introduction to Epidemiology course to nonmajors for 9 years. His primary research interests include understanding the etiology of common adult cancers, particularly breast and ovarian cancer. He has published more than 300 peer-reviewed scientific articles, reviews, and book chapters, and now serves as a Deputy Editor of Cancer Epidemiology, Biomarkers, and Prevention and as Associate Editor of the American Journal of Epidemiology. Dr. Sellers is a long-standing member of the American Association for Cancer Research and the American Society for Preventive Oncology, and is a founding member of the International Genetic Epidemiology Society. Dr. Sellers has been an invited member of Advisory Committees to the National Cancer Institute, has provided invited lectures worldwide, and has served on numerous grant review panels.

xxiv a b o u t t h e a u t h o r s

51589_FM00_Printer.indd 24 15/02/13 3:11 AM

1Chapte r

History and Scope of Epidemiology

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● define the term epidemiology ●● def ine the components of epidemiology (determinants,

distribution, morbidity, and mortality) ●● name and describe characteristics of the epidemiologic approach ●● discuss the importance of Hippocrates’ hypothesis and how it dif-

fered from the common beliefs of the time ●● discuss Graunt’s contributions to biostatistics and how they

affected modern epidemiology ●● explain what is meant by the term natural experiments, and give at

least one example

CHAPTER OUTLINE

I. Introduction II. Epidemiology Defined

III. Foundations of Epidemiology IV. Historical Antecedents of Epidemiology V. Recent Applications of Epidemiology

VI. Conclusion VII. Study Questions and Exercises

51589_CH01_Printer.indd 1 09/02/13 12:00 PM

2 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Introduction

Controversies and speculations regarding the findings of epidemiologic research are frequent topics of media reports; these findings sometimes arouse public hysteria. Examples of the questions raised by media reports include: “Is it more dangerous to vaccinate an entire population against smallpox (with result- ing complications from the vaccine) or to risk infection with the disease itself through a terrorist attack?” “Is Ebola virus a danger to the general public?” “Should I give up eating fatty foods?” “Is it safe to drink coffee or alcoholic beverages?” “Will chemicals in the environment cause cancer?” “Should one purchase bottled water instead of consuming tap water from public drinking supplies?” “Will medications for chronic diseases (long-standing illnesses that are difficult to eradicate) such as diabetes cause harmful side effects?” “Will the foods that I purchase in the supermarket make me sick?” “When can we expect the next global pandemic influenza and what shall be the response?”

Consider the 2009–2010 episode of influenza first identified in the United States1 and eventually called 2009 H1N1 influenza. Ultimately the 2009 H1N1 outbreak threatened to become an alarming pandemic that public health officials feared could mimic the famous 1918 “killer flu.” In April 2009, 2 cases of 2009 H1N1 came to the attention of the Centers for Disease Control and Prevention (CDC), which investigates outbreaks of infectious diseases such as influenza. Thereafter, the number of cases expanded rapidly in the United States and then worldwide. When the epidemic eventually subsided during summer 2010, an estimated 60 million cases had occurred in the United States. According to the CDC, people in the age range of 18–64 years were most heavily affected by the virus; less affected were those 65 years of age and older. Exhibit 1–1 provides an account of the pandemic.

the 2009 h1N1 pandemic

During spring 2009, a 10-year-old California child was diagnosed with an unusual variety of influenza. Soon afterwards a case of the same flu strain was identified in an 8-year-old who lived approxi- mately 130 miles from the first patient. This was an alarming event e

x h

ib it

1 –1

continues

51589_CH01_Printer.indd 2 09/02/13 12:00 PM

i n t r o d u C t i o n 3

Exhibit 1–1 continued

in several respects. The type of influenza virus was usually found among swine. However, the newly identified virus appeared to have been transmit- ted among humans. Secondly, the appearance of these two unusual cases raised public health officials’ suspicions that a deadly flu pandemic similar to the 1918 pandemic might be under way.

Scientists named the new virus 2009 H1N1. The agent was “. . . a unique combination of influenza virus genes never previously identified in either animals or people.”1 The genes of the new virus were closely related to North American swine-lineage H1N1 influenza viruses. Before this outbreak, human-to-human spread of swine-origin influenza viruses was highly unusual. During the previous three years (from December 2005 to January 2009), only 12 U.S. cases of swine influenza had been reported. The vast majority (n = 11) had indicated some contact with pigs. One of the unusual features of infections with the 2009 H1N1 virus were reports of a high prevalence of obesity among influenza-affected patients in inten- sive care units.

Following the identification of the initial cases in California, swine flu spread across the United States and jumped international borders. In response to a potential widespread epidemic, some schools and pub- lic health officials implemented pandemic preparedness plans, which included school closures and social distancing. In June, the World Health Organization (WHO) declared that a global pandemic was under way. Here is a brief chronology of the events that transpired during the

pandemic.

●● April 15, 2009—first case of pandemic influenza (2009 H1N1) identi-

fied in a 10-year-old California patient. ●● April 17—eight-year-old child living 130 miles away from first case devel-

ops influenza. ●● April 21—Centers for Disease Control and Prevention (CDC) began

work on a vaccine against the virus. ●● April 22—three new cases are identified in San Diego County and Impe-

rial County. ●● April 23—two new cases identified in Texas.

continues

51589_CH01_Printer.indd 3 09/02/13 12:00 PM

4 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Exhibit 1–1 continued

●● April 23—seven samples from Mexico were positive for 2009 H1N1. ●● April 25—WHO declares a “Public Health Emergency of International

Concern.” ●● April 25—cases diagnosed in New York City, Kansas, and Ohio. ●● April 29—WHO raises the influenza pandemic alert from phase 4 to

phase 5. ●● May 6—CDC recommends prioritized testing and antiviral treatment

for people at high risk of complications from flu. ●● June 11—WHO raises the worldwide pandemic alert level to phase 6

and declares the global pandemic is under way. ●● June 11—more than 70 countries have reported cases of pandemic

influenza. ●● June through July—the number of countries reporting influenza has

nearly doubled; all 50 states in the U.S. have reported cases. ●● Summer and fall—extraordinary influenza-like illness activity reported

in the U.S. ●● September 30—initial supplies of 2009 H1N1 vaccine distributed on a

limited basis. ●● December—vaccine made available to all who wanted it. ●● Summer 2010—flu activity reaches normal summer time levels in the U.S.

According to the CDC approximately 60 million people became infected with 2009 H1N1 between April 2009 and March 13, 2010. The estimated range of the number of cases was between 43 million and 88 million. The process of estimating the number of flu cases is imprecise because many patients who become ill do not seek medical care, and those who do are not tested for the virus. Figure 1–1 reports CDC estimates of 2009 H1N1 cases in the US by age group. n

Source: Data from Centers for Disease Control and Prevention. The 2009 H1N1pandemic: summary highlights, April 2009—April 2010. Available at: http://www.cdc.gov/h1n1flu/ cdcresponse.htm. Accessed July 19, 2012.

continues

51589_CH01_Printer.indd 4 09/02/13 12:00 PM

i n t r o d u C t i o n 5

Another example of a disease that elicited public hysteria was the outbreak of Escherichia coli (E. coli) infections during late summer and fall 2006. The outbreak affected multiple states in the United States and captured media head- lines for several months. Known as E. coli O157:H7, this bacterial agent can be ingested in contaminated food. The agent is an enteric pathogen, which can pro- duce bloody diarrhea, and in some instances, the hemolytic-uremic syndrome (HUS), a type of kidney failure. Severe cases of E. coli O157:H7 can be fatal.

The 2006 outbreak was a mysterious event that gradually unfolded over time. The outbreak sickened 199 persons across United States and caused 3 deaths (as of October 6, 2006, when the outbreak appeared to have subsided). Figure 1–2 shows the affected states. The 2006 outbreak caused 102 (51%) of the ill persons to be hospi- talized; in all, 31 patients (16%) were afflicted with HUS. The majority of cases (141, 71%) were female. A total of 22 children 5 years of age and younger were affected.2

FiGURE 1–1 CDC estimates of 2009 H1N1 cases in the United States by age group. Source: Reproduced from Centers for Disease Control and Prevention. The CDC Estimates of 2009 H1N1 Influenza Cases, Hospitalizations and Deaths in the United States, April 2009–March 13, 2010. Available at: http://www.cdc.gov/h1n1flu/estimates/April_March_13. htm. Accessed August 23, 2012.

60,000,000

50,000,000

40,000,000

30,000,000

20 09

H 1N

1 C

as es

20,000,000

10,000,000

0– 17

Y rs

April – Oct 17, 2009

April – Nov 14, 2009

April 2009 – Dec 12, 2010

April 2009 – Jan 16, 2010

April – Feb 13, 2010

April – March 13, 2010

Age Group Data by Date Range

18 –6

4 Y

≥6 5

Y rs

≥6 5

Y rs

0– 17

Y rs

18 –6

4 Y

≥6 5

Y rs

0– 17

Y rs

18 –6

4 Y

≥6 5

Y rs

0– 17

Y rs

18 –6

4 Y

≥6 5

Y rs

0– 17

Y rs

18 –6

4 Y

≥6 5

Y rs

0– 17

Y rs

18 –6

4 Y

Exhibit 1–1 continued

51589_CH01_Printer.indd 5 09/02/13 12:00 PM

6 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Tracking down the mysterious origins of the outbreak required extensive detective work. The outbreak was linked to prepackaged spinach as the most likely vehicle. Investigators traced the spinach back to its source, Natural Selec- tion Foods near Salinas, California. The producer announced a recall of spinach on September 15, 2006.3 The FDA and State of California conducted a trace- back investigation, which implicated four ranches in Monterey and San Benito Counties. Cattle feces from one of the four ranches contained a strain of E. coli O157:H7 that matched the strain that had contaminated the spinach and also matched the strain found in the 199 cases.4 The mechanism for contamination of the spinach with E. coli bacteria was never established definitively.

Noteworthy is the fact that subsequent to this major outbreak, E. coli O157:H7 continues to threaten the food supply of the United States, not only from spinach but also from other foods.5 During November and December 2006, Taco Bell restaurants in the northeastern United States experienced a major outbreak that caused at least 71 persons to fall ill. Contamination of Topp’s brand frozen ground beef patties and Totino’s or Jeno’s brand frozen pizzas with E. coli O157:H7 is believed to have sickened more than 60 residents

FiGURE 1–2 Distribution of Escherichia coli serotype O157:H7 cases across the United States, September 2006. Source: Reproduced from Centers for Disease Control and Prevention. Ongoing multistate outbreak of Escherichia coli serotype O157:H7 infections associated with consumption of fresh spinach— United States, September 2006. MMWR. 2006;55:1045–1046.

OR ID

MN WI

IN OH

VA KY

1–4 5–9 10–14 15 or higher

51589_CH01_Printer.indd 6 09/02/13 12:00 PM

i n t r o d u C t i o n 7

of the eastern half of the United States during summer and early fall 2007. In 2008 and 2009, E. coli outbreaks were associated with ground beef and prepack- aged cookie dough. Ground beef, cheese, romaine lettuce, bologna, and hazel- nuts caused outbreaks during 2010 and 2011. A major outbreak of E. coli O104 occurred in Germany in 2011; 6 travelers from the United States were made ill, with one of the six dying. During summer 2012, a multistate outbreak caused by E. coli O145 sickened 18 persons and caused 9 deaths.

In summary, the 2009 H1N1 flu pandemic (Exhibit 1–1) and the E. coli spinach-associated outbreak illustrate that epidemiologic research methods are a powerful tool for studying the health of populations. In many instances, epidemiology resembles detective work, because the causes of disease occurrence are often unknown. Both examples raise several issues that are typical of many epidemiologic research studies:

●● When there is a linkage or association between a factor (i.e., contaminants in food and water; animal reservoirs for disease agents) and a health out- come, does this observation mean that the factor is a cause of disease?

●● If there is an association, how does the occurrence of disease vary according to the demographic characteristics and geographic locations of the affected persons?

●● Based on the observation of such an association, what practical steps should individuals and public health departments take? What should the individual consumer do?

●● Do the findings from an epidemiologic study merit panic or a measured response?

●● How applicable are the findings to settings other than the one in which the research was conducted? What are the policy implications of the findings?

In this chapter we answer the foregoing questions. We discuss the stages that are necessary to unravel mysteries about diseases, such as those due to environ- mental exposures or those for which the cause is entirely unknown.

Epidemiology is a discipline that describes, quantifies, postulates causal mechanisms for diseases in populations, and develops methods for the control of diseases. Using the results of epidemiologic studies, public health practitioners are aided in their quest to control health problems such as foodborne disease out- breaks and influenza pandemics. The investigation into the spinach-associated E. coli outbreak illustrates some of the classic methods of epidemiology; first, describing all of the cases, enumerating them, and then following up with addi- tional studies. Extensive detective work was involved in identifying the cause of the outbreak. The hypothesized causal mechanism that was ultimately linked to

51589_CH01_Printer.indd 7 09/02/13 12:00 PM

8 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

contaminated spinach was the bacterium E. coli. All of the features described in the investigation are hallmarks of the epidemiologic approach. In this example, the means by which E. coli contaminated the spinach remains an unresolved issue.

The 2009 H1N1 pandemic demonstrated the use of epidemiologic data to identify the source of the initial outbreaks, describe pandemic spread, and mount a public health response to control a pandemic. Officials created public awareness of the need to be vaccinated against the virus and to prevent spread of the virus by covering up one’s mouth when coughing and washing one’s hands frequently.

Epidemiology Defined

The word epidemiology derives from epidemic, a term that provides an immediate clue to its subject matter. Epidemiology originates from the Greek words epi (upon) + demos (people) + logy (study of). Although some conceptions of epidemiology are quite narrow, we suggest a broadened scope and propose the following definition:

Epidemiology is concerned with the occurrence, distribution, and determinants of “health-related states or events”6 (e.g., health and diseases, morbidity, injuries, disabil- ity, and mortality in populations). Epidemiologic studies are applied to the control of health problems in populations. The key aspects of this definition are determinants, distribution, population, and health phenomena (e.g., morbidity and mortality).

Determinants Determinants are factors or events that are capable of bringing about a change in health. Some examples are specific biologic agents (e.g., bacteria) that are associated with infectious diseases or chemical agents that may act as carcino- gens. Other potential determinants for changes in health may include less spe- cific factors, such as stress or adverse lifestyle patterns (lack of exercise or a diet

Case 1: Intentional Dissemination of Bacteria That Cause Anthrax

After the United States experienced its worst terrorist attack on September 11, 2001, reports appeared in the media about cases of anthrax in Florida beginning in early October. In the United States, anthrax usually affects herbivores (livestock and some wild animals);

continues

51589_CH01_Printer.indd 8 09/02/13 12:00 PM

e p i d e m i o l o g y d e f i n e d 9

human cases are unusual. Anthrax is an acute bacterial disease caused by exposure to Bacillus anthracis. Cutaneous anthrax affects the skin, producing lesions that develop into a black scab. Untreated cutaneous anthrax has a case-fatality rate of 5–20%. The much more severe inhala- tional form, which affects the lungs and later becomes disseminated by the bloodstream, has a high case fatality rate.7 Observations of an alert infectious disease specialist along with the support of laboratory staff led to the suspicion that anthrax had been deliberately sent through the postal system.8 The CDC, in collaboration with officials at the state and local levels, identified a total of 21 anthrax cases (16 confirmed and 5 suspected) as of October 31, 2001. The majority of the cases occurred among employees located in four areas: Florida, New York City, New Jersey, and the District of Columbia.9–12 Figure 1–3 portrays the dis- tribution of the 21 cases in 4 geographic areas of the United States. n

As of October 31, 2001, 21 cases were reported

in four states and one isolated case in

Connecticut (not linked to any exposure source)

Florida (2 cases) Workers at America

Media, Inc First case was 63-year-old

worker who dies from inhalational anthrax.

Second case identified in co-worker with positive

nasal sample. Environmental sample from workplace tests positive for anthrax.

New York City (7 cases) 1 inhalational (confirmed), 6 cutaneous (3 confirmed,

3 suspected) cases in 4 media companies

One of six cases involved suspected mail room

contact with letter that contained anthrax.

Case seven: inhalational case (patient worked in

hospital stockroom)

New Jersey (7 cases at 2 postal facilities) 5 cases confirmed 2 cases suspected

5 cases at a New Jersey postal facility

No contaminated letters identified, but contaminated

mail suspected. 2 cases at a second New Jersey postal center (mail sorter and another worker)

District of Columbia (5 cases, all inhalational)

4 cases in DC postal facility 1 case in U.S. State

Department mail facility (This facility receives

mail from the DC facility that had 4 cases.)

FiGURE 1–3 Occurrence of anthrax cases during the 2001 terrorist incident according to the investigation by the Centers for Disease Control and Prevention.

CasE 1 continued

high in saturated fats). The following four vignettes illustrate the concern of epidemiology with disease determinants. For example, consider the steps taken to track down the source of the bacteria that caused anthrax and were sent through the mail; contemplate the position of an epidemiologist once again. Imagine a possible scenario for describing, quantifying, and identifying the determinants for each of the vignettes.

51589_CH01_Printer.indd 9 09/02/13 12:00 PM

10 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Case 2: Outbreak of Fear

When a 36-year-old lab technician known as Kinfumu checked into the general hospital in Kikwit, Zaire, complaining of diarrhea and a fever, anyone could have mistaken his illness for the dysentery that was plagu- ing the city. Nurses, doctors, and nuns did what they could to help the young man. They soon saw that his disease wasn’t just dysentery. Blood began oozing from every orifice in his body. Within 4 days he was dead. By then the illness had all but liquefied his internal organs.

That was just the beginning. The day Kinfumu died, a nurse and a nun who had cared for him fell ill. The nun was evacuated to another town 70 miles to the west where she died—but not until the contagion had spread to at least three of her fellow nuns. Two subsequently died. In Kikwit, the disease raged through the ranks of the hospital’s staff. Inhabitants of the city began fleeing to neighboring villages. Some of the fugitives carried the deadly illness with them. Terrified health officials in Kikwit sent an urgent message to the World Health Organization. The Geneva-based group summoned expert help from around the globe: a team of experienced virus hunters composed of tropical-medicine specialists, microbiologists, and other researchers. They grabbed their lab equipment and their bubble suits and clambered aboard transport planes headed for Kikwit.13 n

Case 3: Fear on Seventh Avenue

On normal workdays, the streets of New York City’s garment district are lively canyons bustling with honking trucks, scurrying buyers, and sweat- ing rack boys pushing carts loaded with suits, coats, and dresses. But during September 1978 a tense new atmosphere was evident. Sanitation trucks cruised the side streets off Seventh Avenue flushing pools of stag- nant water from the gutters and spraying out disinfectant. Teams of health officers drained water towers on building roofs. Air condition- ers fell silent for inspection, and several chilling signs appeared on 35th Street: “The New York City Department of Health has been advised of

continues

51589_CH01_Printer.indd 10 09/02/13 12:00 PM

e p i d e m i o l o g y d e f i n e d 11

Case 4: Red Spots on Airline Flight Attendants

From January 1 to March 10, 1980, Eastern Airlines received 190 reports of episodes of red spots appearing on the skin of flight attendants (FAs) during various flights. Complaints of symptoms accompanying the spots were rare, but some FAs expressed concern that the spots were caused by bleeding through the skin and might indicate a serious health hazard. On March 12, investigators from the CDC traveled to Miami to assist in the investigation. No evidence of damage to underlying skin was noted on these examinations, nor was any noted by consultant dermatologists who examined affected FAs after the spots had disappeared. Chemical tests on clinical specimens for the presence of blood were negative. Airline personnel had investigated the ventilation systems, cleaning materials and procedures, and other environmental factors on affected aircraft. Airflow patterns and cabin temperatures, pressures, and rela- tive humidity were found to be normal. Cleaning materials and routines had been changed, but cases continued to occur. Written reports by FAs of 132 cases occurring in January and February showed that 91 different FAs had been affected, 68 once and 23 several times. Of these cases, 119 (90%) had occurred on a single type of aircraft. Of the 119 cases from implicated aircraft, 96% occurred on north- or southbound flights between the New York City and Miami metropolitan areas, flights that are partially over water. Only rarely was a case reported from the same airplane when flying transcontinental or other east-west routes.15 n

possible cases of Legionnaires’ disease in this building.” By the weekend, there were 6 cases of the mysterious disease, 73 more suspected, and 2 deaths. In the New York City outbreak, three brothers were the first vic- tims. Carlisle, Gilbert, and Joseph Leggette developed the fever, muscle aches, and chest congestion that make the disease resemble pneumo- nia. Joseph and Gilbert recovered; Carlisle did not. “He just got sick and about a week later he was dead,” said John Leggette, a fourth brother who warily returned to his own job in the garment district the next week. “I’m scared,” he said. “But what can you do?”14 n

CasE 3 continued

51589_CH01_Printer.indd 11 09/02/13 12:00 PM

12 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Health departments, the CDC in Atlanta, and epidemiologic researchers frequently confront a problem that has no clear determinants or etiologic basis. The methods and findings of epidemiologic studies may direct one to, or suggest, particular causal mechanisms underlying health-related events or conditions, such as the four examples cited in the vignettes: anthrax, the suspected outbreak of Ebola virus, Legionnaires’ disease, and red spots on airline flight attendants. Read the solution to Case 4 to clear up the mystery of Case 4.

Solution to Case 4: Red Spots

The investigation then concentrated on defining the clinical picture more clearly. An Eastern Airlines (EAL) physician, a consultant dermatologist, and a physician from the National Institute for Occupational Safety and Health (NIOSH) rode on implicated flights on March 14 and examined three new cases considered by the EAL physician and other flight atten- dants (FAs) to be typical cases. Although the spots observed consisted of red liquid, they did not resemble blood. To identify potential environmen- tal sources of red-colored material, investigators observed the standard activities of FAs on board implicated flights. At the beginning of each flight FAs routinely demonstrated the use of life vests, required in emer- gency landings over water. Because the vests used for demonstration were not actually functional, they were marked in bright red ink with the words “Demo Only.” When the vests were demonstrated, the red ink areas came into close contact with the face, neck, and hands of the demonstrator. Noting that on some vests the red ink rubbed or flaked off easily, inves- tigators used red material from the vests to elicit the typical clinical pic- ture on themselves. On preliminary chemical analyses, material in clinical specimens of red spots obtained from cases was found to match red-ink specimens from demonstration vests. On March 15 and 16, EAL removed all demonstration model life vests from all its aircraft and instructed FAs to use the standard, functional, passenger-model vests for demonstra- tion purposes. The airline . . . continue[d] to request reports of cases to verify the effectiveness of this action. Although all demonstration vests were obtained from the same manufacturer, the vests removed from spe- cific aircraft were noted to vary somewhat in the color of fabric and in the color and texture of red ink, suggesting that many different production lots may have been in use simultaneously on any given aircraft.15 n

51589_CH01_Printer.indd 12 09/02/13 12:00 PM

e p i d e m i o l o g y d e f i n e d 13

Distribution Frequency of disease occurrence and mortality rates vary from one population group to another in the United States. For example, in 2006 death rates from coronary heart disease (CHD) and stroke were higher among African-Americans (blacks) than among American Indians/Alaskan natives, Asian/Pacific islanders, or whites.16 In comparison with other racial/ethnic groups, Hispanics have lower mortality rates for CHD than non-Hispanics.16,17 Such variations in disease fre- quency illustrate how disease may have different distributions depending upon the underlying characteristics of the populations being studied. Population sub- groups that have higher occurrence of adverse health outcomes are defined as having health disparities, which need to be targeted for appropriate interventions.

Population Epidemiology examines disease occurrence among population groups rather than among individuals. Lilienfeld18 noted that this focus is a widely accepted feature of epidemiology. For this reason, epidemiology is often referred to as “population medicine.” As a result, the epidemiologic and clinical descriptions of a disease are quite different. Sometimes, when a new disease is first recognized, clinical descriptions of the condition are the first data available. These initial clinical descriptions can lead to subsequent epidemiologic investigations.

Note the different descriptions of toxic shock syndrome (TSS), a condition that showed sharp increases during 1980 in comparison with the immediately previous years. TSS is a severe illness that in the 1980 outbreak was found to be associated with vaginal tampon use. The clinical description of TSS would include specific signs and symptoms, such as high fever, headache, malaise, and other more dramatic symp- toms, such as vomiting and profuse watery diarrhea. The epidemiologic description would indicate which age groups would be most likely to be affected, time trends, geographic trends, and other variables that affect the distribution of TSS.

A second example is myocardial infarction (MI; heart attack). A clinical description of MI would list specific signs and symptoms, such as chest pain, heart rate, nausea, and other individual characteristics of the patient. The epi- demiologic description of the same condition would indicate which age groups would be most likely to be affected, seasonal trends in heart attack rates, geo- graphic variations in frequency, and other characteristics of persons associated with the frequency of heart attack in populations.

Referring again to the vignettes, one may note that the problem that plagued Kinfumu in Case 2 was recognized as a particularly acute problem for epidemiol- ogy when similar complaints from other patients were discovered and the disease

51589_CH01_Printer.indd 13 09/02/13 12:00 PM

14 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

began to spread. If more than one person complains about a health problem, the health provider may develop the suspicion that some widespread exposure rather than something unique to an individual is occurring. The clinical observation might suggest further epidemiologic investigation of the problem.

Health Phenomena As indicated in the definition, epidemiology is used to investigate many differ- ent kinds of health outcomes. These range from infectious diseases to chronic diseases and various states of health, such as disability, injury, limitation of activ- ity, and mortality.19 Other health outcomes have included individuals’ positive functioning and active life expectancy as well as adverse health-related events, including mental disorders, suicide, substance abuse, and injury. Epidemiol- ogy’s concern with positive states of health is illustrated by research into active life expectancy among geriatric populations. This research seeks to determine the factors associated with optimal mental and physical functioning as well as enhanced quality of life and ultimately aims to limit disability in later life.

Morbidity and Mortality Two other terms central to epidemiology are morbidity and mortality. The for- mer, morbidity, designates illness, whereas the latter, mortality, refers to death. Note that most measures of morbidity and mortality are defined for specific types of morbidity or causes of death.

Aims and Levels The preceding sections hinted at the complete scope of epidemiology. As the basic method of public health, epidemiology is concerned with efforts to describe, explain, predict, and control. The term levels denotes the hierarchy of tasks that epidemiologic studies seek to accomplish (e.g., description of the occurrence of diseases is a less-demanding task and therefore ranks lower on the hierarchy of levels than explaining the causes of a disease and predicting and controlling them). More information will be provided later in the chapter.

●● To describe the health status of populations means to enumerate the cases of disease, to obtain relative frequencies of the disease within subgroups, and to discover important trends in the occurrence of disease.

●● To explain the etiology of disease means to discover causal factors as well as to determine modes of transmission.

●● To predict the occurrence of disease is to estimate the actual number of cases that will develop as well as to identify the distribution within populations.

51589_CH01_Printer.indd 14 09/02/13 12:00 PM

f o u n d a t i o n s o f e p i d e m i o l o g y 15

Such information is crucial to planning interventions and allocation of healthcare resources.

●● To control the distribution of disease, the epidemiologic approach is used to prevent the occurrence of new cases of disease, to eradicate existing cases, and to prolong the lives of those with the disease.

The implication of these aims is that epidemiology has two different goals: one related to the distribution of health outcomes and the second to controlling diseases. The first goal is to achieve an improved understanding of the natural history of disease and the factors that influence its distribution. With the knowl- edge that is obtained from such efforts, one can then proceed to accomplish the second goal, which is control of disease via carefully designed interventions.

Foundations of Epidemiology

Epidemiology Is Interdisciplinary Refer to Figure 1–4, which characterizes the interdisciplinary foundations of epidemiology. As an interdisciplinary field, epidemiology draws from biostatistics and the social and behavioral sciences as well as from the medically related fields such as toxicology, pathology, virology, genetics, microbiology, and clinical medi- cine. Terris20 pointed out that epidemiology is an extraordinarily rich and complex science that derives techniques and methodologies from many disciplines. He wrote that epidemiology “must draw upon and synthesize knowledge from the biological sciences of man and of his parasites, from the numerous sciences of the physical environment, and from the sciences concerned with human society.”20(p 203)

FiGURE 1–4 The interdisciplinary foundations of epidemiology.

Microbiology Virology

BiostatisticsToxicology

Epidemiology is Interdisciplinary

Social and Behavioral Sciences, Demography

Clinical Medicine/ Pathology

51589_CH01_Printer.indd 15 09/02/13 12:00 PM

16 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Here are some illustrations of the contributions of other disciplines to epidemiology. Microbiology, the science of microorganisms, yields information about specific disease agents, including their morphology and modes of transmission. Related fields are bacteriology and virology. The previously discussed investigations of anthrax, Legionnaires’ disease, and TSS utilized microbiologic techniques to identify possible infectious agents. Another example is epidemiologic studies of foodborne illnesses (e.g., E. coli); these studies apply microbiologic procedures to reveal the commonalities of bacteria involved in an outbreak in order to define whether it was caused by a common source.

Clinical medicine is involved in the diagnosis of the patient’s state of health, particularly when defining whether the patient has a specific disease or condition. A pathologist’s expertise may help differentiate between normal and diseased tis- sue. From our previous examples, clinical medicine diagnosed the individuals’ symptoms or signs of ill health. Astute physicians and nurses may suggest epide- miologic research on the basis of clinical observations.

Toxicology, the science of poisons, is concerned with the presence and health effects of chemical agents, particularly those found in the environment and the workplace. A crucial issue for the United States is the fate of hazardous chemicals once they have performed their function. In the past, numerous toxic chemicals (e.g., pesticides) were deposited in an unsafe manner into waste sites that were later designated as hazardous. Toxicologic knowledge helps determine the pres- ence of noxious chemical agents in hazardous waste sites and whether any health effects observed are consistent with the known effects of exposure to toxic agents. When responses to exogenous agents vary from person to person, geneticists may become part of the research team via the disciplines of molecular and genetic epidemiology. Frequently, toxicologists and epidemiologists collaborate in envi- ronmental and occupational investigations.

Social and behavioral sciences elucidate the role of race, social class, edu- cation, cultural group membership, and behavioral practices in health-related phenomena. Social and behavioral science disciplines, that is, sociology and psychology, are devoted respectively to the development of social theory and the study of behavior. The special concern of social epidemiologic approaches is the study of social conditions and disease processes.21 Furthermore, the social sciences provide a great deal of the methodology on sampling; measurement; questionnaire development, design, and delivery; and group comparisons. Increasingly, community interventions have drawn upon the fund of knowledge from the social sciences. Demography is the study of data related to the structure of human populations.

51589_CH01_Printer.indd 16 09/02/13 12:00 PM

f o u n d a t i o n s o f e p i d e m i o l o g y 17

Finally, the field of biostatistics is critical to the evaluation of epidemiologic data, especially when one is trying to separate chance from meaningful obser- vations. Epidemiology profits from the interdisciplinary approach because the causality of a particular disease in a population may involve the interaction of multiple factors. The contributions of many disciplines help unravel the factors associated with a particular disease.

Methods and Procedures The empirical dimensions of epidemiologic studies require quantification of rele- vant factors. Quantification refers to the translation of qualitative impressions into numbers. Qualitative sources of information about disease may be, for example, a physician’s observations derived through medical practice about the types of people among whom a disease seems to be common. Epidemiologists enumerate cases of disease to objectify subjective impressions; the standard epidemiologic measures often require counting the number of cases of disease and examining their distri- bution according to demographic variables, such as age, sex, race, and other vari- ables as well as exposure category and clinical features. The following quotation

As of March 26, [2003] CDC has received 51 reports of suspected SARS cases from 21 states . . . identified using the CDC updated interim case definition . . . The first suspected case was identified on March 15, in a man aged 53 years who traveled to Singapore and became ill on March 10. Four clusters of suspected cases have been identified, three of which involved a traveler who had visited Southeast Asia (including Guangdong province, Hong Kong, or Vietnam) and a single family contact. One of these clusters involved suspected cases in patients L and M . . . who had stayed together at hotel M during March 1–6, when other hotel guests were symptomatic. Patient L became sick on March 13 after returning to the United States. His wife, patient M, became ill several days after the onset of her husband’s symptoms, suggesting secondary transmission.

The Language of Quantification: Severe Acute Respiratory Syndrome (SARS) in the United States

continues

51589_CH01_Printer.indd 17 09/02/13 12:00 PM

18 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

illustrates a summary of the characteristics of 51 suspected cases of severe acute respiratory syndrome (SARS) that were reported to the CDC as of early 2003.

Sometimes epidemiologists present quantified information as tables, maps, charts, and graphs. Both charts and graphs are pictorial illustrations of the fre- quency of disease. (Refer to the later section on John Snow for an example of a map.) Quantification facilitates the epidemiologic investigation of the sources of variation of a disease by the characteristics of time, place, and person: When did the case occur? Where was it located? Who was affected?

Key methods for the graphic presentation of data are the use of pie charts, bar graphs, and line graphs. Figure 1–5 shows an example of each type: a pie chart (A, admission diagnoses of discharged hospice care patients); a bar graph (B, diabetes prevalence among adults); and a line graph (C, obesity among children). Epidemiologists use these types of graphs to describe characteristics of data, such as subgroup differences and time trends.23

Use of Special Vocabulary Epidemiology employs a unique vocabulary of terms to describe the frequency of occurrence of disease. Examples from this vocabulary are the words epidemic and pandemic.

Dorland’s Illustrated Medical Dictionary defines the word epidemic as “attack- ing many people at the same time, widely diffused and rapidly spreading.” More precisely, an epidemic refers to an excessive occurrence of a disease: “Most current definitions [of epidemic] stress the concept of excessive prevalence as its basic implication in both lay and professional usage.”24(p 2) The following pas- sage illustrates this notion by defining an epidemic as:

The occurrence, in a defined community or region, of cases of an illness (or an outbreak) with a frequency clearly in excess of normal expectancy. The number of cases indicating presence of an epidemic varies according to the infectious agent, size and type of population exposed, previous experience or lack of exposure to the disease, and time and place of occurrence; epidemicity is thus relative to usual frequency of the disease . . .25(p 705)

Three patients in the United States with suspected SARS (patients I, L, and M) reported staying at hotel M when other persons staying in the hotel were symptomatic. The fourth cluster began with a suspected case in a person who traveled in Guangdong province and Hong Kong. Two [healthcare workers] subsequently became ill at the U.S. hospital where this patient was admitted.22(p 244) n

51589_CH01_Printer.indd 18 09/02/13 12:00 PM

f o u n d a t i o n s o f e p i d e m i o l o g y 19

FiGURE 1–5 Examples of three different presentations of epidemiologic data. (A) Pie chart. Primary admission diagnoses of discharged hospice care patients: United States, 2007 (B) Bar graph. Diabetes prevalence among adults 20 years of age and over, by age: United States, 1988–1994 and 2005–2008. (C) Line graph. Obesity among children, by age: United States, 1988–1994 through 2007–2008. Source: Adapted and Reprinted from National Center for Health Statistics. Health, United States, 2010: with Special Feature on Death and Dying. Hyattsville, MD, 2011.

*Chronic lower respiratory disease.

All other 26%

Stroke 5%

CLRD* 5%

Heart disease

11%

Cancer 43%

Alzheimer’s and other dementia

11%

20–44 years 3 1988–1994

2005–2008 4

20 30 Percent

100

45–64 years

65 years and over

6–11 years

12–19 years

2–5 years

P er

ce n

1988– 1994

1999– 2000 Year

2001– 2002

2003– 2004

2005– 2006

2007– 2008

51589_CH01_Printer.indd 19 09/02/13 12:00 PM

20 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

The “usual frequency” means the disease’s typical occurrence at the same time, within the same population, and in the same geographic area. Further, when a communicable disease has disappeared and a single case reappears, that event represents an epidemic. Also, the occurrence of two cases of a new disease (“first invasion”) linked in time and place may be considered to be an epidemic, as this happening suggests disease transmission.

Communicable disease—An illness caused by an infectious agent that can be transmitted from one person to another. Infectious disease—A synonym for a communicable disease. Outbreak—A localized disease epidemic (e.g., in a town or healthcare facility). n

Explanation of Key Terms Used in the Definition of “Epidemic”

In current thinking, an epidemic is not confined to infectious diseases. Take, for example, the Love Canal incident that generated spirited public debate and media attention during the late 1970s. Love Canal was a toxic waste disposal site located in Niagara Falls, New York. It was the destina- tion for burial of thousands of chemical-filled drums deposited by the Hooker Chemicals & Plastics Corporation. Eventually, the waste disposal site was covered and converted into a housing tract. Subsequently, residents of the area reported several different types of health effects, including miscarriages, birth defects, and impaired cognitive functioning. The Love Canal site was the focus of extensive health effects studies and epidemiologic research. The threat posed by Love Canal and other hazardous waste sites led to the creation of the Superfund in 1980. Its purpose was to promote the cleanup of hazard- ous wastes.

By referring to the case studies reported in this text, you have seen additional examples—red spots among airline FAs and TSS—that illustrate two instances in which epidemiologic methodology was employed to study noninfectious condi- tions. TSS and red spots among airline FAs both represented apparent epidemics because the usual or expected rate was nil. Epidemiologic methods also are used to investigate occupationally associated illness (e.g., brown lung disease among

51589_CH01_Printer.indd 20 09/02/13 12:00 PM

f o u n d a t i o n s o f e p i d e m i o l o g y 21

textile workers and asbestosis among shipyard workers), environmental health hazards (e.g., toxic chemicals and air pollution), and conditions associated with lifestyle (e.g., unintentional injuries, ischemic heart disease, and certain forms of cancer).

Related to the term epidemic is the term pandemic, which refers to an epidemic on a worldwide scale; during a pandemic, large numbers of persons may be affected and a disease may cross international borders. Examples are flu pandemics, such as the pandemic of 1918 and more recent flu pandemics that occur periodically. The term endemic is used to characterize a disease that is habitually present in a particular geographical region. To illustrate, malaria is endemic to some tropical areas of Asia, and cholera is endemic to less developed countries where sanitation is lacking. Previously, during the 19th century, cholera was endemic to Western countries, such as England and the United States. However, cholera is no longer endemic to these two countries because of the introduction of sanitation and other public health measures.

Methods for Ascertainment of Epidemic Frequency of Disease The CDC and vital statistics departments of state and local governments gather surveillance data on a continuing basis to determine whether an epidemic is taking place. The word surveillance denotes the systematic collection of data per- taining to the occurrence of specific diseases, the analysis and interpretation of these data, and the dissemination of consolidated and processed information to contributors to the surveillance program and other interested persons. Com- mon surveillance activities include monitoring foodborne disease outbreaks, collecting information on communicable and infectious diseases, and tracking influenza.

As noted previously, an epidemic refers to the occurrence of disease in excess of normal expectancy. In order to ascertain epidemic trends, one must have data about the usual occurrence of a disease. Providing such information is the function of surveillance. For example, suppose a health practitioner states that 500 CHD deaths were reported in an upstate New York community during a particular year and that an epidemic is taking place. This information by itself would be insufficient to justify the assertion that an epidemic of CHD deaths has occurred. The usual frequency of CHD deaths would need to be determined via ongoing surveillance programs in the same community at some prior time. In addition, the size, age, and sex distribution of the population would need to be

51589_CH01_Printer.indd 21 09/02/13 12:00 PM

22 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

known. With this information at hand, one could determine whether or not an epidemic of CHD deaths has occurred.

A second example of determining epidemic frequency is shown in Figure 1–6 for influenza and pneumonia deaths. The figure displays weekly pneumonia and influenza deaths in the United States from winter 2007 to spring 2012. The chart demonstrates that influenza (flu) has an underlying sea- sonal baseline, reflected in cyclic seasonal increases and declines in mortality. In the United States and other countries in the Northern hemisphere, flu occurs most frequently during the winter months (i.e., from October through April).26 Therefore, the flu season spans the latter part of one calendar year and the early part of the following year (e.g., the 2011–2012 flu season). In the figure, the lower line denotes the usual number of total deaths to be expected from pneumonia and influenza during each week of the year. An upper paral- lel line indicates the frequency of disease at the epidemic threshold, that is, the minimum number of deaths that would support the conclusion that an epidemic was under way. The epidemic threshold is based on statistical projec- tions. Figure 1–6 demonstrates that the combined pneumonia and influenza deaths peaked substantially above the epidemic threshold during early 2008, late 2009, and early 2011.

FiGURE 1–6 Percentage of all deaths attributable to pneumonia and influenza (P&I), by surveillance week and year—122 Cities Mortality Reporting System, United States, 2007–May 19, 2012. Source: Reproduced Centers for Disease Control and Prevention. Update Influenza Activity—United States, 2011–12 Season and Composition of the 2012–13 Influenza Vaccine. MMWR. 2012;61:418.

Epidemic threshold

% o

f al

l d ea

th s

at tr

ib u

te d

t o

P &

Seasonal baseline

105040 4

20 30

20082007

40 50 10 20 30

2009

40 50 10 20 30

2010

40 50 10 20 30

2011

10 20

2012

40 50

Surveillance week and year

51589_CH01_Printer.indd 22 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 23

Historical Antecedents of Epidemiology

Epidemiology is often thought of as a relatively new discipline. However, this viewpoint is not entirely correct. The history of epidemiology began with the classical period of the Greeks and Romans and included major developments that occurred during later eras: the medieval period, the Renaissance, the late 1800s and early 1900s, and more recently the mid- to late 20th century, when the pace of epidemiologic activities exploded.

It may be said that epidemiology began with the Greeks, who in their concern for the ancient epidemics and deadly toll of diseases, attributed disease causality to environmental factors. Early causal explanations for epidemics included the wrath of the gods, the breakdown of religious beliefs and morality, the influence of weather, and “bad air.” During the medieval period, the Black Death caused by plague killed more than 25% of the European population. Another terrible scourge was smallpox: Edward Jenner’s work led to the development of an effec- tive vaccination against smallpox. During the late Renaissance, pioneering bio- statisticians quantified morbidity and mortality trends.

When the 19th century arrived, deadly cholera epidemics impacted Europe and the United States The disease is thought to have been spread along trade routes from India to Asia, the Middle East, and Russia. Cholera is a life- threatening condition caused by a bacterium; victims retch from severe (but painless) vomiting and diarrhea and eventually die from dehydration and elec- trolyte disturbances. An example that memorializes the assault of cholera on Europe is the Cholera Fountain (Cholera Brunnen) in Dresden, Germany. Resi- dents constructed the fountain in the mid-1800s to express their gratitude for having escaped a cholera epidemic that threatened the city. (See Figure 1–7.) Often cited as a major historical development is John Snow’s investigations of London cholera outbreaks, reported in Snow on Cholera.27

A contemporary of Snow, William Farr, promoted innovative uses of vital sta- tistics data. During the 19th century, early microbiologists formalized the germ theory of disease, which attributed diseases to specific organisms. At the begin- ning of the 20th century, a flu pandemic killed more than 50 million people worldwide. Each of these historical developments that contributed to the genesis of epidemiology is discussed in turn below.

Environment as a Factor in Disease Causation The following account by Thucydides records, in detail, the ravages produced by a deadly disease, “Thucydides’ plague”28; such graphic descriptions of major

51589_CH01_Printer.indd 23 09/02/13 12:00 PM

24 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

FiGURE 1–7 The Cholera Fountain in Dresden, Germany.

epidemics in history indicate this early author’s concern with the causality of these remarkable phenomena:

Others, who were in perfect health, were taken suddenly, without any apparent cause, with violent heats in their heads, and with redness and inflammations in their eyes. Their tongues and throats within became immediately bloody; their breath in great disorder and offensive. A sneezing and a hoarseness ensued; and, in a short time, the pain descended into the breast, attended with a violent cough. When it was once settled about the mouth of the stomach, a retching, and vomit- ing of bilious stuff, in as great a variety as ever was known among physicians, suc- ceeded, but not without the greatest anxiety imaginable. Many were seized with a hiccup, that brought up nothing, but occasioned a violent convulsion, which in some went off presently, but in others continued much longer. The body out- wardly was neither very hot to the touch, nor pale, but reddish, livid, and flow- ered (as it were) all over with little pimply eruptions, and ulcers; but inwardly the heat was so exceedingly great, that they could not endure the slightest covering, or the finest linen, or any thing short of absolute nakedness. It was also an infinite pleasure to them to plunge into cold water; and many of those who were not well attended did so, running to the wells, to quench their insatiable thirst: not that it signified whether they drank much or little; a great uneasiness and restless- ness attending them, together with a continual watching. While the distemper was

51589_CH01_Printer.indd 24 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 25

advancing to the height, the body did not fall away, but resisted the vehemence of it beyond expectation; so that many of them died the ninth and the seventh day of the inward burning, some strength yet remaining; or, if they held out longer, many of them afterwards died of weakness; the distemper descending into the belly, and there producing violent ulcerations, and fluxes of the simple or unmixed kind.28

Hippocrates, in On Airs, Waters, and Places,29 gave birth in about 400 bc to the idea that disease might be associated with the physical environment; his thinking represented a movement away from supernatural explanations of dis- ease causation to a rational account of the origin of humankind’s illnesses. Note in the following passage his reference to climate and physical environment:

Whoever wishes to investigate medicine properly should proceed thus: in the first place to consider the seasons of the year, and what effects each of them produces (for they are not at all alike, but differ much from themselves in regard to their changes). Then the winds, the hot and the cold, especially such as are common to all countries, and then such as are peculiar to each locality. We must also consider the qualities of the waters, for as they differ from one another in taste and weight, so also do they differ much in their qualities. In the same manner, when one comes into a city to which he is a stranger, he ought to consider its situation, how it lies as to the winds and the rising of the sun; for its influence is not the same whether it lies to the north or the south, to the rising or to the setting sun. These things one ought to consider most attentively, and concerning the waters which the inhabit- ants use, whether they be marshy and soft, or hard, and running from elevated and rocky situations, and then if saltish and unfit for cooking; and the ground, whether it be naked and deficient in water, or wooded and well watered, and whether it lies in a hollow, confined situation, or is elevated and cold; and the mode in which the inhabitants live, and what are their pursuits, whether they are fond of drinking and eating to excess, and given to indolence, or are fond of exercise and labor, and not given to excess in eating and drinking.29(pp 156–157)

The Black Death Occurring between 1346 and 1352, the Black Death is a dramatic example of a pandemic of great historical significance to epidemiology.30 The Black Death is noteworthy because of the scope of human mortality that it produced as well as for its impact upon medieval civilization. Estimates suggest that the Black Death claimed about one-quarter to one-third of the population of Europe. Northern Africa and the near Middle East also were affected severely; at the inception of the outbreak, the population of this region including Europe numbered about 100 million people; 20–30 million people are believed to have died in Europe.

Historians attribute the Black Death to bubonic plague, which is the most common of the three forms of plague.30,31 The bacterium Yersinia pestis pro- duces swelling of the lymph nodes in the groin and other sites of the body.

51589_CH01_Printer.indd 25 09/02/13 12:00 PM

26 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

These painful swellings, called buboes, are followed in several days by high fever and the appearance of black splotches on the skin. The reservoir for Y. pestis is vari- ous types of rodents, including rats. Plague can be transmitted when fleas that feed on rodents bite a human host. At the time of the Black Death, no method for treat- ment of plague existed. Most victims died within a few days after the occurrence of buboes. Currently, plague is treatable with antibiotics. In addition, improvement in sanitary conditions has led to the decline in plague cases; 2,118 cases were reported worldwide in 2003.31 From 1,000 to 2,000 plague cases are reported annually to the World Health Organization (according to data available in 2012).32

Use of Mortality Counts In 1662, John Graunt published Natural and Political Observations Mentioned in a Following Index, and Made Upon the Bills of Mortality.33 This work recorded descriptive characteristics of birth and death data, including seasonal variations, infant mortality, and excess male over female differences in mortality. Graunt’s work made a fundamental contribution by discovering regularities in medical and social phenomena. He is said to be the first to employ quantitative meth- ods in describing population vital statistics by organizing mortality data in a mortality table and has been referred to as the Columbus of statistics. Graunt’s procedures allowed the discovery of trends in births and deaths due to specific causes. Although his conclusions were sometimes erroneous, his development of statistical methods was highly important.34

Concerning sex differences in death rates, Graunt wrote:

Of the difference between the numbers of Males and Females. The next Observation is, That there be more Males than Females . . . There have been Buried from the year 1628, to the year 1662, exclusive, 209436 Males, and but 190474 Females: but it will be objected, That in London it may be indeed so, though otherwise elsewhere; because London is the great Stage and Shop of business, wherein the Masculine Sex bears the greatest part. But we Answer, That there have been also Christened within the same time 139782 Males, and but 130866 Females, and that the Country- Accounts are consonant enough to those of London upon this matter.33(p 44)

Figure 1–8 shows the 10 leading causes of mortality from the Yearly Mortality Bill for 1632. A legend at bottom of the figure defines the archaic terms used in Graunt’s time.

Edward Jenner and Smallpox Vaccination The term vaccination derives from the Latin word for cow (vacca), the source of the cowpox virus that was used to create a vaccine against smallpox. A precursor

51589_CH01_Printer.indd 26 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 27

of smallpox vaccination was variolation, which referred to an early Asian method of conferring immunity to smallpox by introducing dried scabs from smallpox patients into the noses of potential victims who wished to be protected from this disease.35 Variolation often produced a milder case of disease with a much lower fatality rate than that caused by community-acquired smallpox. The method gained popularity in Europe during the early 1700s, when the procedure was modified by injecting infectious material under the skin; variolation was first tested among abandoned children and prisoners. When it was declared safe, members of the English royal family were inoculated.

Edward Jenner (Figure 1–9) is credited with the development of the smallpox vaccination, a lower-risk method for conferring immunity against smallpox than

FiGURE 1–8 Yearly Mortality Bill for 1632: The 10 leading causes of mortality in Graunt’s Time. Source: Data from Graunt J. Natural and Political Observations, Mentioned in a Following Index, and Made upon the Bills of Mortality, 2nd ed. London: Tho. Roycroft; 1662: p. 8.

Chrisom

Consumption

Fever

Collick, Stone, and Strangury

Flux and Small Pox

Bloody Flux, Scowring, and Flux

Dropsie and Swelling

Glossary of Terms Used in Chart

Bloody flux

Chrisom

Consumption

Dropsie

Flox

Flux

Liver grown

Scowring

Small pox

Stone

Strangury

Dysentery

Death of a child within one month of baptism

Tuberculosis

Dropsy—edema

Hemorrhagic smallpox

Excessive flow or discharge from the boby

Having an enlarged liver

Scouring—purging of the bowels; probably referring to diarrhea

Smallpox

Calculus, e.g., gallstone

Slow and painful discharge of urine

Convulsion

Childbed

Liver grown

Number of deaths

0 500 1000 1500 2000 2500

51589_CH01_Printer.indd 27 09/02/13 12:00 PM

28 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

variolation.36 He was fascinated by folk wisdom, which suggested that dairy- maids who had contracted cowpox seemed to be immune to smallpox. Infec- tion with the cowpox virus produced a much less severe form of disease than smallpox. Jenner conducted an experiment in which he used scabs from the cow- pox lesions on the arm of a dairymaid, Sarah Nelmes (Figure 1–10), to create a smallpox vaccine. He then used the material to vaccinate an 8-year-old boy, James Phipps. Following the vaccination, Phipps appeared to develop immu- nity to the smallpox virus to which he was reexposed several times subsequently.

FiGURE 1–9 Edward Jenner vaccinating a child. Source: Images from the History of Medicine, National Library of Medicine.

51589_CH01_Printer.indd 28 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 29

Later, Jenner vaccinated his own son and several other children, obtaining similar positive findings, which were published in 1798. (In 1978 smallpox was finally eliminated worldwide. Since 1972, routine vaccination of the nonmilitary popu- lation of the United States has been discontinued.)37

Use of Natural Experiments A natural experiment refers to “[n]aturally occurring circumstances in which sub- sets of the population have different levels of exposure to a supposed causal factor in a situation resembling an actual experiment, where human subjects would be randomly allocated to groups. The presence of persons in a particular group is typically nonrandom;”6 the following section is an account of John Snow’s natu- ral experiment.

FiGURE 1–10 Arm of Sarah Nelmes with lesions of cowpox. Source: Reproduced from the National Library of Medicine. Smallpox: A great and terrible scourge: Vaccination. Available at: http://www.nlm.nih.gov/exhibition/smallpox/ sp_vaccination.html. Accessed July 19, 2012.

51589_CH01_Printer.indd 29 09/02/13 12:00 PM

30 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

During the 19th century, water from the highly polluted Thames River was London’s primary source of drinking water. Figure 1–11 expresses concerns about the cleanliness of Thames River water during this time period. In this context John Snow conducted a famous natural experiment.

Snow investigated a cholera epidemic that occurred during the mid-19th century in Broad Street, Golden Square, London. Snow’s work, a classic study that linked the cholera epidemic to contaminated water supplies, is noteworthy because it utilized many of the features of epidemiologic inquiry: a spot map of cases and tabulation of fatal attacks and deaths. Through the application of his keen powers of observation and inference, he developed the hypothesis that contaminated water might be associated with outbreaks of cholera. He made several observations that others had not previously made. One observation was that cholera was associated with water from one of two water supplies that served the Golden Square district of London.38 Broad Street was served by two sepa- rate water companies, the Lambeth Company and the Southwark and Vauxhall Company. Lilienfeld and Lilienfeld39 wrote:

In London, several water companies were responsible for supplying water to differ- ent parts of the city. In 1849, Snow noted that the cholera rates were particularly

FiGURE 1–11 George Cruickshank, 1792–1878, artist. Salus Populi Suprema Lex Source of the South Warwick Water Works. Source: Images from the History of Medicine, National Library of Medicine.

51589_CH01_Printer.indd 30 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 31

high in those areas of London that were supplied by the Lambeth Company and the Southwark and Vauxhall Company, both of whom obtained their water from the Thames River at a point heavily polluted with sewage.39(p 36)

Snow’s account of the outbreak of 1849 is found in Exhibit 1–2. Between 1849 and 1854 the Lambeth Company had its source of water relo-

cated to a less contaminated part of the Thames. In 1854, another epidemic of

Snow on Cholera

The most terrible outbreak of cholera which ever occurred in this kingdom, is probably that which took place in Broad Street, Golden Square, and the adjoining streets, a few weeks ago. Within two hun- dred and fifty yards of the spot where Cambridge Street joins Broad Street, there were upwards of five hundred fatal attacks of cholera in ten days. The mortality in this limited area probably equals any

that was ever caused in this country, even by the plague; and it was much more sudden, as the greater number of cases terminated in a few hours. The mortality would undoubtedly have been much greater had it not been for the flight of the population. Persons in furnished lodgings left first, then other lodgers went away, leaving their furniture to be sent for when they could meet with a place to put it in. Many houses were closed alto- gether, owing to the death of the proprietors; and, in a great number of instances, the tradesmen who remained had sent away their families: so that in less than six days from the commencement of the outbreak, the most afflicted streets were deserted by more than three-quarters of their inhabitants.

There were a few cases of cholera in the neighbourhood of Broad Street, Golden Square, in the latter part of August; and the so-called outbreak, which commenced in the night between the 31st August and the 1st September, was, as in all similar instances, only a violent increase of the malady. As soon as I became acquainted with the situation and extent of this irruption of cholera, I suspected some contamination of the water of the much-frequented street-pump in Broad Street, near the end of Cambridge Street; but on examining the water, on the evening of the 3rd September, I found so little impurity in it of an organic nature, that I hesi- tated to come to a conclusion. Further inquiry, however, showed me that

e x

h ib

it 1

–2

continues

51589_CH01_Printer.indd 31 09/02/13 12:00 PM

32 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

there was no other circumstance or agent common to the circumscribed locality in which this sudden increase of cholera occurred, and not extend- ing beyond it, except the water of the above mentioned pump. I found, moreover, that the water varied, during the next two days, in the amount of organic impurity, visible to the naked eye, on close inspection, in the form of small white, flocculent particles; and I concluded that, at the com- mencement of the outbreak, it might possibly have been still more impure.

The deaths which occurred during this fatal outbreak of cholera are indicated in the accompanying map (Figure 1–12), as far as I could ascer- tain them . . . The dotted line on the map surrounds the sub-districts of Golden Square, St. James’s, and Berwick Street, St. James’s, together with the adjoining portion of the sub-district of St. Anne, Soho, extending from Wardour Street to Dean Street, and a small part of the sub-district of St. James’s Square enclosed by Marylebone Street, Titchfield Street, Great Windmill Street, and Brewer Street. All the deaths from cholera which were registered in the six weeks from 19th August to 30th September within this locality, as well as those of persons removed into Middlesex Hospital, are shown in the map by a black line in the situation of the house in which it occurred, or in which the fatal attack was contracted . . . The pump in Broad Street is indicated on the map, as well as all the surrounding pumps to which the public had access at the time. It requires to be stated that the water of the pump in Marlborough Street, at the end of Carnaby Street, was so impure that many people avoided using it. And I found that the persons who died near this pump in the beginning of September, had water from the Broad Street pump. With regard to the pump in Rupert Street, it will be noticed that some streets which are near to it on the map, are in fact a good way removed, on account of the circuitous road to it. These circumstances being taken into account, it will be observed that the deaths either very much diminished, or ceased altogether at every point where it becomes decidedly nearer to send to another pump than to the one in Broad Street. It may also be noticed that the deaths are most numer- ous near to the pump where the water could be more readily obtained . . . The greatest number of attacks in any one day occurred on the 1st of September, immediately after the outbreak commenced. The following

Exhibit 1–2 continued

continues

51589_CH01_Printer.indd 32 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 33

Exhibit 1–2 continued

FiGURE 1–12 Cholera deaths in the neighborhood of Broad Street, August 19 to September 30, 1849. Source: Reproduced from John Snow’s dot map of the Broad Street and Golden Square area of London, in Snow on Cholera by John Snow, Commonwealth Fund: New York, 1936.

continues

51589_CH01_Printer.indd 33 09/02/13 12:00 PM

34 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Exhibit 1–2 continued

FiGURE 1–13 The 1849 cholera outbreak in Golden Square district, London. Fatal attacks and deaths, August 31– September 8. Source: Data from Table I, Snow J. Snow on Cholera, p. 49, Harvard University Press, © 1965.

Aug. 31 Sept. 1 Sept. 2 Sept. 3 Sept. 4 Sept. 5 Sept. 6 Sept. 7 Sept. 8

160

140

120

100

N u

m b

Date

Legend

Fatal Attacks

Deaths

day the attacks fell from one hundred and forty-three to one hundred and sixteen, and the day afterwards to fifty-four . . . The fresh attacks contin- ued to become less numerous every day. On September the 8th—the day when the handle of the pump was removed—there were twelve attacks; on the 9th, eleven; on the 10th, five; on the 11th, five; on the 12th, only one; and after this time, there were never more than four attacks on one day. During the decline of the epidemic the deaths were more numerous than the attacks, owing to the decrease of many persons who had lingered for

several days in consecutive fever (Figure 1–13). n

Source: Reprinted from Snow J. Snow on Cholera. Cambridge, MA: Harvard University Press: 1965:38–51.

cholera occurred. This epidemic was in an area that consisted of two-thirds of London’s resident population south of the Thames and was being served by both companies. In this area, the two companies had their water mains laid out in an interpenetrating manner, so that houses on the same street were receiving their water from different sources.39

51589_CH01_Printer.indd 34 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 35

This was a naturally occurring situation, a “natural experiment,” if you will, because in 1849 all residents received contaminated water from the two water companies. After 1849, the Lambeth Company used less contaminated water by relocating its water supply. Snow demonstrated that a disproportionate number of residents who contracted cholera in the 1854 outbreak used water from one water company, which used polluted water, in comparison with the other com- pany, which used relatively unpolluted water.

Snow’s methodology maintains contemporary relevance. His methods utilized logical organization of observations, a natural experiment, and a quantitative approach.39 All these methods are hallmarks of present-day epi- demiologic inquiry. Note that it is possible to visit the site of the pump that figured so prominently in Snow’s investigation of cholera; a London public house on the original site of the pump has been named in Snow’s honor. A replica of the pump is located nearby. Refer to Exhibit 1–3 for pictures of the site and the pump with a reproduction of the text on the base of the replica.

Another study, occurring during the mid-19th century, also used nascent epidemiologic methods. Ignaz Semmelweis,40 in his position as a clinical assis- tant in obstetrics and gynecology at a Vienna hospital, observed that women in the maternity wards were dying at high rates from puerperal fever. In 1840, when the medical education system changed, he found a much higher mortality rate among the women on the teaching wards for medical students

a Visit to the broad Street pump and the Sir John Snow public house, Located at 39 broadwick Street, London, england W1F9QJ

Figure 1–14 shows John Snow, Figure 1–15 displays a replica of the Broad Street pump. Broad Street has been renamed Broadwick Street. Figure 1–16 shows a plaque titled “The Soho Chloera

Epidemic” at the base of the pump. Figure 1–17 presents a picture of the John Snow Pub.n

e x

h ib

it 1

–3

continues

51589_CH01_Printer.indd 35 09/02/13 12:00 PM

36 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Exhibit 1–3 continued

FiGURE 1–16 Plaque commemorating the Soho cholera epidemic, 1854.

FiGURE 1–15 Replica of Broad Street pump near its approximate original location.

continues

FiGURE 1–14 Photograph of John Snow. Source: © National Library of Medicine.

51589_CH01_Printer.indd 36 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 37

Exhibit 1–3 continued

FiGURE 1–17 The John Snow Pub named in honor of the British Anesthesiologist.

and physicians than on the teaching wards for midwives. He postulated that medical students and physicians had contaminated their hands during autopsies. As a result, they transmitted infections while attending women in the maternity wards.41 When the practice of hand washing with chlorinated solutions was introduced, the death rate for puerperal fever in the wards for medical students and physicians dropped to a rate equal to that in the wards for midwives.

William Farr A contemporary of John Snow, William Farr assumed the post of “Compiler of Abstracts” at the General Register Office (located in England) in 1839 and held this position for 40 years. Among Farr’s contributions to public health and epidemiology was the development of a more sophisticated system for codify- ing medical conditions than was previously in use. Farr’s classification scheme, which departed from a narrow medical view, provided the foundation for the International Classification of Diseases in use today. Also noteworthy is the fact

51589_CH01_Printer.indd 37 09/02/13 12:00 PM

38 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

that Farr used data such as census reports to study occupational mortality in England. In addition, he explored the possible linkage between mortality rates and population density, showing that both the average number of deaths and births per 1,000 living persons increased with population density (defined as number of persons per square mile). Because of the excess of births over deaths in all except the most crowded areas, the population tended to increase in the less crowded areas. With respect to deaths in high mortality districts, such as Liverpool, which had a mortality rate more than 22 per 1,000 greater than that experienced in healthier districts, he attributed mortality to factors such as “. . . impurities of water, pernicious dirts, floating dusts, zynotic contagions, [and] crowdings in lodgings . . .”42(p 90) The healthier districts had “. . . a salu- brious soil, and supply the inhabitants with water generally free from organic impurities.”42(pp 90–91)

Identification of Specific Agents of Disease In the late 1800s, Robert Koch verified that a human disease was caused by a specific living organism. His epoch-making study, Die Aetiologie der Tuberkulose, was published in 1882. This breakthrough made possible greater refinement of the classification of disease by specific causal organisms.43 Previously, the group- ing together of diseases according to grosser classifications had hampered their epidemiologic study.

King44 noted that Koch’s postulates are usually formatted as follows:

1. The microorganism must be observed in every case of the disease. 2. It must be isolated and grown in pure culture. 3. The pure culture must, when inoculated into a susceptible animal,

reproduce the disease. 4. The microorganism must be observed in, and recovered from, the

experimentally diseased animal.44

King noted, “What Koch accomplished, in brief, was to demonstrate for the first time in any human disease a strict relation between a micro-organism and a disease.”44(p 351) This specification of the causal disease organism provided a defi- nite criterion for the identification of a disease, rather than the vague standards Koch’s predecessors and contemporaries had employed.

Increasing awareness of the role of microbial agents in the causation of human illness—the germ theory of disease—eventually reached the public health com- munity. One method to limit the spread of infectious disease was through the use of cartoons published in the popular media. Figure 1–18 suggested that skirts

51589_CH01_Printer.indd 38 09/02/13 12:00 PM

h i s t o r i C a l a n t e C e d e n t s o f e p i d e m i o l o g y 39

FiGURE 1–18 Samuel D. Erhart, Communicable diseases spread by household and street dust. Source: Images from the History of Medicine, National Library of Medicine.

that trail on the ground (in fashion around the turn of the 20th century) could bring deadly germs into the household.45

51589_CH01_Printer.indd 39 09/02/13 12:00 PM

40 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

The 1918 Influenza Pandemic So great was its impact, this outbreak has been referred to as “the Mother of All Pandemics.”46 Also known as the Spanish Flu, the pandemic that occurred dur- ing the period of 1918–1919 killed from 50 to 100 million persons worldwide. Estimates suggest that one-third of the world’s population of 1.5 billion at the time was infected and developed clinically observable illness. This very severe form of influenza had case-fatality rates of approximately 2.5% compared with the 0.1% or lower rates observed in other influenza pandemics. Differentiating this form of influenza from other outbreaks was its impact on healthy young adults; persons aged 20–40 accounted for nearly half of the mortality toll in this pandemic, whereas influenza deaths normally are more frequent among the very young and the very old.47,48 The pandemic spread in three distinct waves during a one-year period throughout Europe, Asia, and North America; the first wave began in spring 1918, with two subsequent waves occurring during the fall and winter of 1918–1919. In the United States, the flu’s impact was so great that healthcare facilities were taxed to the limit. As a result of large numbers of deaths, the bodies of victims accumulated in morgues awaiting burial, which was delayed because of a shortage of coffins and morticians.

A repeat of the 1918 pandemic is within the realm of possibility, as suggested by the 2009 H1N1 influenza pandemic. This event raised questions about how modern society would cope with a global outbreak of influenza or other highly communicable disease. Will healthcare facilities have adequate “surge” capacity to deal with a sudden and large increase in the number of patients? Will it be necessary to enforce “social distancing” to reduce the spread of epidemic dis- eases? How will essential services be maintained? These are examples of issues for which the public health community will need to be prepared.

Other Significant Historical Developments Alexander Fleming, Alexander Langmuir, Wade Hampton Frost, and Joseph Goldberger made several other historically significant contributions. Scottish researcher Fleming is credited with discovering the antimicrobial properties of the mold Pencillium notatum in 1928. This discovery led to development of the antibiotic penicillin, which became available toward the end of World War II. Langmuir, regarded as the father of infectious disease epidemiology, in 1949 established the epidemiology section of the federal agency presently called the Centers for Disease Control and Prevention. This section later came to be

51589_CH01_Printer.indd 40 09/02/13 12:00 PM

r e C e n t a p p l i C a t i o n s o f e p i d e m i o l o g y 41

known as the Epidemic Intelligence Service (EIS), which celebrated its 60th anniversary in 2011. Frost, who held the first professorship of epidemiology in the United States beginning in 1930 at Johns Hopkins University, advocated the use of quantitative methods (e.g., a procedure known as cohort analysis) to illuminate public health problems, although his concept of epidemiology tended to be restricted narrowly to the study of infectious diseases. Finally, Goldberger’s discovery of the cure for pellagra, a nutritional deficiency disease characterized by the so-called three Ds (dermatitis, diarrhea, and dementia), led to reductions in the occurrence of the disease, which had gained attention in the early 1900s.

Recent Applicat ions of Epidemiology

Epidemiologic activity has exploded during the past several decades.49 For example, the ongoing Framingham Heart Study, begun in 1948, is one of the pioneering research investigations of risk factors for coronary heart dis- ease. Refer to the classic article by Kannell and Abbott for a description of the study.50 Another development, occurring after World War II, was research on the association between smoking and lung cancer.51 An example is the histori- cally significant work of Doll and Peto,52 based on a fascinating study of British physicians.

The computer and powerful statistical software have aided the proliferation of epidemiologic research studies. Popular interest in epidemiologic findings is also intense. Almost every day now, one encounters media reports of epide- miologic research into such diverse health concerns as acquired immune defi- ciency syndrome, chemical spills, breast cancer screening, and the health effects of secondhand cigarette smoke. Table 1–1 reports triumphs in epidemiology; these are examples in which epidemiologists have identified risk factors for can- cer, heart disease, infectious diseases, and many other conditions. One triumph in Table 1–1 is how epidemiology helped to uncover the association between the human papillomavirus and cervical cancer. On June 8, 2006, the FDA announced the licensing of the first vaccine (Gardisil®) to prevent cervical cancer caused by four types of human papillomavirus and approved its use in females aged 9–26 years. Returning to Table 1–1, the reader should note that although many of the terms used in the table have not yet been discussed in this book, later sections of the text will cover some of them. Additional examples of applica- tions of epidemiology are provided in the following sections.

51589_CH01_Printer.indd 41 09/02/13 12:00 PM

42 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

Table 1–1 Triumphs in Epidemiology

Risk Factor Categories Disease Risk Factors Direction

Alcohol Esophageal cancer alcohol (interaction with smoking)

Viruses Liver cancer hepatitis B virus IR Burkitt lymphoma Epstein Barr virus IR Kaposi sarcoma Herpes simplex virus type B IR Cervical cancer “something transmitted

sexually” (human papilloma virus)

Nasopharyngeal carcinoma Epstein Barr virus IR Yellow fever “something transmitted by

mosquitos” (Flavivirus) IR

New variant (nv) Creutzfeldt-Jacob disease

prions (interaction with genotype)

Bacteria Cholera “something in water” (Vibrio cholera)

Peptic ulcer Helicobactor pylori IR Puerperal fever “something on doctor’s

hands” (group B Streptococcus)

Nutrition Pellagra “something in bread” (niacin) P Neural tube defects folic acid, folate P Oral clefts folic acid P

Occupation Lung cancer asbestos (interaction with smoking)

Bladder cancer aniline dye IR Mesothelioma asbestos IR Angiosarcoma vinyl chloride IR Infertility (male) DBCP IR Nasal cancer nickel smelting IR Lung cancer “something in uranium mines”

(interaction with smoking) IR

Environment Dental caries fluoride [deficiency] P Cancer arsenic IR

Drugs/ Devices Myocardial infarction aspirin P Micoagthnia iso-retinene during pregnancy IR Pelvic inflammatory disease Dalkon Shield IUD IR Septic abortion Dalkon Shield IUD IR

continues

Infectious Diseases in the Community Infectious disease epidemiology, one of the most familiar types of epidemiol- ogy, investigates the occurrence of epidemics of infectious and communicable diseases. Examples are studying diseases caused by bacteria, viruses, and micro- biologic agents; tracking down the cause of foodborne illness; and investigating

51589_CH01_Printer.indd 42 09/02/13 12:00 PM

r e C e n t a p p l i C a t i o n s o f e p i d e m i o l o g y 43

Risk Factor Categories Disease Risk Factors Direction

Hormones Clear cell adenocarcinoma of the vagina

diethylstilbestrol prenatally IR

Venous thromboembolism combined estrogen/progestin (oral contraceptives)

Venous thromboembolism postmenopausal estrogen IR Ovarian cancer oral contraceptives P Endometrial cancer combined estrogen/progestin P Endometrial cancer oral contraceptives:

postmenopausal estrogen IR

Iron deficiency anemia oral contraceptives P Benign breast disease oral contraceptives P Myocardial infarction oral contraceptives

(interaction with smoking) IR

Ischemic stroke oral contraceptives (interaction with hypertension; modified by dose)

Genetics Breast cancer “something genetic” (BRCA1, BRCA2 mutations)

Ovarian cancer “something genetic” (BRCA2 mutations)

Colon cancer “something genetic” (APC1 mutations)

Miscellaneous Toxic shock syndrome super absorbent tampons IR SIDS prone sleep position IR Reyes syndrome aspirin (interaction with infec-

tion) IR

Smoking Lung cancer smoking IR Coronary disease smoking IR Hemorrhagic stroke smoking IR Ischemic stroke smoking IR Abdominal aortic aneurism smoking IR Peripheral vascular disease smoking IR Parkinson’s disease smoking P Ulcerative colitis smoking P Laryngeal cancer smoking IR Intrauterine growth retardation smoking during pregnancy IR Toxemia/pre-eclampsia smoking during pregnancy P

Table 1–1 continued

Abbreviations: IR, increased risk; P, protective (see Chapters 3, 6, and 7).

Source: Compiled by Diane Petitti. Adapted with permission from The Epidemiology Monitor. October 2001; 6.

new diseases such as SARS, pandemic influenza 2009 H1N1 (Exhibit 1–1), and avian influenza (Exhibit 1–4). An illustration is the use of epidemiologic meth- ods to attempt to eradicate, when possible, polio, measles, smallpox, and other communicable diseases. Another example is outbreaks of infectious diseases in

51589_CH01_Printer.indd 43 09/02/13 12:00 PM

44 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

hospitals (nosocomial infections). The role of the Epidemic Intelligence Service

in investigating disease outbreaks is defined as follows:

The Centers for Disease Control and Prevention (CDC) in Atlanta, Georgia tracks disease outbreaks that occur in the United States and throughout the world. As one facet of this process, the CDC supports a training program for personnel who

highly pathogenic avian influenza (hpai) avian influenza (h5N1)

Investigations into an outbreak of highly pathogenic avian influenza (HPAI) demonstrate the role of epidemiology in containing out- breaks of infectious diseases that threaten the health of the popula- tion. The arrival of avian influenza (caused by the H5N1 virus) that began in the late 1990s is an example of the occurrence of an infec-

tious disease with potential to impact a specific community as well as the entire world. This highly fatal condition worried public health authorities who were concerned that avian influenza could create a worldwide pan- demic, mirroring the 1918 pandemic and lesser influenza epidemics that occurred later in the 20th century. The emergence of a pandemic might be the consequence of mutation of the virus into a version that could be com- municated rapidly on a person-to-person basis.

Beginning in 1997, avian influenza appeared in Hong Kong, with an initial 18 human cases, of which 6 were fatal.54 These human cases coin- cided with outbreaks among poultry on farms and in markets that sold live poultry. Authorities destroyed the entire chicken population in Hong Kong; subsequently, no additional human cases linked to the source in Hong Kong were reported. Two additional human cases were reported in Hong Kong in 2003 and were associated with travel to mainland China.

The epidemic did not end in Hong Kong: Additional cases began appearing in Southeast Asia during late 2003. Virus outbreaks involving animals and humans were limited primarily to Vietnam and some other areas of Southeast Asia (e.g., Thailand). One case of probable person- to-person spread of H5N1 virus is believed to have occurred in Thailand. Then, in 2005, the virus manifested itself in central Asia, spreading to Europe, Africa, and the Middle East. From December 1, 2003 to April 30, 2006, nine countries reported a total of 205 laboratory-verified cases to

e x

h ib

it 1

–4

continues

51589_CH01_Printer.indd 44 09/02/13 12:00 PM

r e C e n t a p p l i C a t i o n s o f e p i d e m i o l o g y 45

Exhibit 1–4 continued

the World Health Organization, with 113 of these illnesses being fatal. At about the same time, infection with the virus was reported among flocks of domestic and wild birds in 50 countries.

Officials were concerned that migrating flocks of wild birds, which cover vast geographical areas, could spread H5N1 to domestic poultry in many parts of the world (Figure 1–19).55 Humans who come into contact with these domestic birds would be at risk of contracting the highly pathogenic virus. As of 2012, the following conclusions have been reached about HPAI (H5N1):

●● Since November 2003 more than 600 human cases (with about a 60% case fatality rate) have been reported worldwide from 15 countries. Nations with the greatest number of cases are Indonesia, Vietnam, and Egypt.

●● The virus can cause severe infections (e.g., severe respiratory illness and death) in humans.

●● Human contact with infected poultry has been associated with most cases.

●● The virus does not show evidence of efficient person-to-person transmission. n

FiGURE 1–19 Pathogenic avian influenza (H5N1) can appear in wild and domestic avian flocks.

51589_CH01_Printer.indd 45 09/02/13 12:00 PM

46 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

respond to requests for assistance in investigating diseases and offer other forms of epidemiologic expertise. Known as the Epidemic Intelligence Service (EIS), this two-year program provides educational opportunities in applied epidemiology. Since 1951, more than 3,000 EIS officers have applied their training to tackle complex health problems. Selected EIS candidates are physicians, nurses, and indi- viduals who have had public health training. Examples of EIS activities include investigating outbreaks of foodborne illnesses such as salmonellosis and listeriosis, potential transmission of hepatitis, and occurrence of Legionnaires’ disease. A clas- sic example of EIS detective work is the investigation of a 1987 cholera outbreak in an inland village in Guinea-Bissau, Africa. The EIS linked this episode, which killed 11 people, to the body of a dockworker smuggled from the coast to an inland village for burial. More than half of the participants at a funeral feast for the deceased later developed cholera. Traditional practices such as washing bodies of the dead and preparation of funeral feasts (in an unsanitary environment) might have contributed to the cholera outbreak in the village.53

Health and the Environment Toxic chemicals used in industry, air pollutants, contaminants in drinking water, unsafe homes and vehicles, and other environmental factors agents may affect human health. Both occupational and environmental epidemiology address the occurrence and distribution of adverse health outcomes such as dust-associated conditions, occupational dermatoses, and diseases linked to harmful physical energy (e.g., ionizing radiation from X-ray machines and other sources). Many of the diseases studied by environmental epidemiologists have agents and mani- festations similar to those in occupational epidemiology, for example, the role of pesticides in causing environmentally associated illness. Injury control epide- miology studies risk factors associated with unintentional injuries (e.g., motor vehicle crashes, bicycle injuries, falls, and occupational injuries). Findings may suggest preventive measures including environmental modifications, safer design of vehicles, and safety laws to prevent injuries. Reproductive and perinatal epi- demiology investigates environmental and occupational exposures and birth outcomes. Related topics are sudden infant death syndrome, epidemiology of neonatal brain hemorrhage, early pregnancy, and methodological issues in drug epidemiology.

Chronic Disease, Lifestyle, and Health Promotion An example of this category is the role of lifestyle (e.g., exercise, diet, smoking, and alcohol consumption) in physical health outcomes such as obesity, coronary heart disease, arthritis, diabetes, and cancer. Hypothesized risk factors studied include

51589_CH01_Printer.indd 46 09/02/13 12:00 PM

r e C e n t a p p l i C a t i o n s o f e p i d e m i o l o g y 47

antecedent variables within the person’s physical and psychosocial environment that may be associated with health and disease. To illustrate, epidemiologic research has explored the relationship between obesity and the tendency of the built environment to dissuade people from walking. Also, poor dietary choices, smoking, substance abuse, and excessive alcohol consumption are linked to many chronic illnesses. Regarding the psychosocial environment, cultural prac- tices affect behaviors that are linked to health and disease. Epidemiologic studies are central to the identification of the causes and methods for addressing health disparities in society.

Psychological and Social Factors in Health Stress, social support, and socioeconomic status affect the occurrence and out- comes of mental and physical health. Research has examined the relationship between the psychological and dimensions and illnesses such as arthritis, some gastrointestinal conditions, and essential hypertension. A related topic involves epidemiologic studies of personality factors and disease, exemplified by the type A personality (coronary prone) and its potential link to heart disease. Psychiatric epidemiology is concerned with the distribution and determinants of mental dis- orders. Examples are the definition and measurement of mental disorders, social factors related to them, and urban and rural differences in their frequency. Major research programs conducted in the community have investigated the epidemiol- ogy of depressive symptomatology.

Also studied as psychosocial determinants are factors that affect the distri- bution of disabilities (e.g., impaired cognition in children, genetic syndromes, autism). Social, cultural, and demographic factors (socioeconomic status, gender, employment, marital status, and race) are demonstrated correlates of mental and physical health status. An important aspect of this branch of epidemiology is the role of such determinants in health disparities.

Molecular and Genetic Epidemiology Numerous advances in molecular and genetic epidemiology have taken place during the genomics age. The field of molecular epidemiology applies the tech- niques of molecular biology to epidemiologic studies. An illustration is using genetic and molecular markers (e.g., deoxyribonucleic acid (DNA) typing) to examine behavioral outcomes and host susceptibility to disease. Genetic epi- demiology studies the distribution of genetically associated diseases among the

51589_CH01_Printer.indd 47 09/02/13 12:00 PM

48 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

population. For example, research has demonstrated inherited susceptibility to severe breast and ovarian cancer as well as to alcohol use disorders (AUDs). Refer to the National Cancer Institute website for more information on cancer genetics.56 With respect to alcohol use disorders, researchers have examined the contribution of specific genes to the increased mortality found in individuals with AUDs.57

A landmark of the genomics age was the completion of the Human Genome Project. Two excellent overview articles discuss how epidemiology interacts with genomics,58 and how the genomics revolution has transformed epide- miology.59 Khoury et al. write that, “[e]pidemiology is essential to fulfill the promise of genomics for clinical and public health practice. . . . Genomics can enhance potential for epidemiology to contribute to multidisciplinary scientific research.”58(p 936)

Conclusion

Epidemiology is concerned with the occurrence, distribution, and determinants of health-related states or events (e.g., health and diseases, morbidity, injuries, disability, and mortality in populations). Epidemiologic studies are applied to the control of health problems in populations. As a result, sometimes the discipline is called population medicine. Several examples demonstrated that the etiologic bases of disease and health conditions in the population are often unknown. Epidemiology is used as a tool to suggest factors associated with occurrence of disease and introduce methods to stop the spread of infectious and communi- cable disease.

Three aspects characterize the epidemiologic approach. The first is quanti- fication, which is counting of cases of disease and construction of tables that show variation of disease by time, place, and person. The second is use of special vocabulary, for example, epidemic and epidemic frequency of disease. The third is interdisciplinary composition, which draws from microbiology, biostatistics, social and behavioral sciences, and clinical medicine.

The historical antecedents of epidemiology began with Hippocrates, who implicated the environment as a factor in disease causation. Second, Graunt, one of the biostatistics pioneers, compiled vital statistics in the mid-1600s. Third, Snow used natural experiments to track a cholera outbreak in Golden Square, London. Finally, Koch’s postulates advanced the theory of specific disease agents. At present, epidemiology is relevant to many kinds of health problems found in the community.

51589_CH01_Printer.indd 48 09/02/13 12:00 PM

s t u d y Q u e s t i o n s a n d e x e r C i s e s 49

Study Questions and Exercises

1. Using your own words, give a definition of epidemiology. Before you read Chapter 1, what were your impressions regarding the scope of epi- demiology? Based on the material presented in this chapter, what topics are covered by epidemiology? That is, to what extent does epidemiol- ogy focus exclusively upon the study of infectious diseases or upon other types of diseases and conditions?

2. How would the clinical and epidemiologic descriptions of a disease dif- fer, and how would they be similar?

3. To what extent does epidemiology rely on medical disciplines for its con- tent, and to what extent does it draw upon other disciplines? Explain the statement that epidemiology is interdisciplinary.

4. Describe the significance for epidemiology of the following historical developments: a. associating the environment with disease causality b. use of vital statistics c. use of natural experiments d. identification of specific agents of disease

5. Explain what is meant by the following components of the definition of epidemiology: a. determinants b. distribution c. morbidity and mortality

6. The following questions pertain to the term epidemic. a. What is meant by an epidemic? Give a definition in your own words. b. Describe a scenario in which only one or two cases of disease may

represent an epidemic. c. What is the purpose of surveillance? d. Give an example of a disease that has cyclic patterns. e. What is the epidemic threshold for a disease? In what sense is it pos-

sible to conceive of the epidemic threshold as a statistical concept? 7. Epidemiologic research and findings often receive dramatic media cover-

age. Find an article in a media source (e.g., The New York Times) on a topic related to epidemiology. In a one-page essay, summarize the find- ings and discuss how the article illustrates the approach of epidemiol- ogy to the study of diseases (health conditions) in populations. You may search online for an appropriate article.

51589_CH01_Printer.indd 49 09/02/13 12:00 PM

50 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

8. During the next week, read and review health-related articles available on the Internet or in your local or national newspaper. Try to find the following terms used in newspaper articles; keep a record of them and describe how they are used: a. epidemiology b. epidemiologist c. infectious disease d. chronic disease e. clinical trial f. increased risk of mortality associated with a new medication

9. What is the definition of a natural experiment? Identify any recent exam- ples of natural experiments. To what extent might changes in legislation to limit smoking in public places or to increase the speed limit on high- ways be considered natural experiments?

10. Review Exhibit 1–2, Snow on Cholera. What do you believe was the purpose of each of the following observations by Snow? a. “small white, flocculent particles” in the water from the Broad

Street pump b. the location of cholera deaths as shown in Figure 1–12 c. people who died avoided the pump in Marlborough Street and instead

had the water from the Broad Street pump d. “the greatest number of attacks in any one day occurred on the 1st of

September, . . .” e. “On September 8th—the day when the handle of the pump was

removed . . .” To what extent do you think removing the pump han- dle was effective in stopping the disease outbreak?

11. How does quantification support the accomplishment of the four aims of epidemiology?

12. How did Koch’s postulates contribute to the advancement of epidemiol- ogy? To what extent is identification of specific agent factors a prerequi- site for tracking down the causes of disease outbreaks?

13. What are the characteristics that distinguish pandemic disease from epidemic disease? Name some examples of notorious pandemics that occurred in history. Why did the “Spanish Flu” of 1918 qualify as a pan- demic? In giving your answer, be sure to distinguish among the terms epidemic, pandemic, and endemic.

14. Identify some infectious diseases that could reach pandemic occurrence during the 21st century. What conditions do you believe exist at present

51589_CH01_Printer.indd 50 09/02/13 12:00 PM

r e f e r e n C e s 51

that could incite the occurrence of pandemics? Why have public health officials been concerned about the emergence of new diseases such as “bird flu”? Speculate about what might happen to organized society and the healthcare system should an outbreak of pandemic influenza occur.

15. The Black Death that occurred during the Middle Ages eradicated a large proportion of the world population at that time. Estimate how likely it would be for a similar epidemic of plague to develop during the current decade.

16. In developed countries, many safeguards exist for the prevention of foodborne illness. Discuss how it would be possible for a foodborne ill- ness outbreak such as the one caused by E. coli to erupt in a developed country.

References 1. Centers for Disease Control and Prevention. The 2009 H1N1 pandemic: Summary

highlights, April 2009–April 2010. http://www.cdc.gov/h1n1flu/cdcresponse.htm. Accessed July 19, 2012.

2. Centers for Disease Control and Prevention. Update on multi-state outbreak of E. coli O157:H7 infections from fresh spinach, October 6, 2006. http://www.cdc .gov/foodborne/ecolispinach/100606.htm. Accessed July 19, 2012.

3. Centers for Disease Control and Prevention. Ongoing multistate outbreak of Escherichia coli serotype O157:H7 infections associated with consumption of fresh spinach—United States, September 2006. MMWR. 2006;55:1045–1046.

4. U.S. Food and Drug Administration. FDA news: FDA statement on foodborne E. coli O157:H7 outbreak in spinach. http://www.fda.gov/NewsEvents/Newsroom/ PressAnnouncements/2006/ucm108767.htm. Accessed July 19, 2012.

5. Centers for Disease Control and Prevention. Reports of selected E. coli outbreak investigations. http://www.cdc.gov/ecoli/outbreaks.html. Accessed July 19, 2012.

6. Porta M. A Dictionary of Epidemiology, 5th ed. New York: Oxford University Press, 2008.

7. Karin M. Anthrax invades and evades the immune system to cause widespread infection. Environmental Health News, Highlights in Environmental Health Sciences Research, 2002 Highlights. Division of Extramural Research and Training, National Institute of Environmental Health Sciences. 2002:1–2. http://www.niehs.nih.gov/ research/supported/sep/2002/anthrax/. Accessed July 19, 2012.

8. Hughes JM, Gerberding JL. Anthrax bioterrorism: Lessons learned and future direc- tions. Emerg Infect Dis. 2002;8:1013–1014.

9. Centers for Disease Control and Prevention. Update: Investigation of bioterrorism- related anthrax and interim guidelines for clinical evaluation of persons with possible anthrax. MMWR. 2001;50:941–948.

51589_CH01_Printer.indd 51 09/02/13 12:00 PM

52 C h a p t e r 1 h i s t o r y a n d s C o p e o f e p i d e m i o l o g y

10. Maillard J-M, Fischer M, McKee KT Jr, et al. First case of bioterrorism-related inhalational anthrax, Florida, 2001: North Carolina investigation. Emerg Infect Dis. 2002;8(10):1035–1038.

11. Centers for Disease Control and Prevention. Ongoing investigation of anthrax— Florida, October 2001. MMWR. 2001;50:877.

12. Centers for Disease Control and Prevention. Update: Investigation of bioterrorism- related anthrax—Connecticut, 2001. MMWR. 2001;50:1077–1079.

13. Mathews T, Lee ED. Outbreak of fear. Newsweek. 1995; May 22:48, 50. 14. Fear on Seventh Ave. Newsweek. 1978; September 18:30. 15. Centers for Disease Control and Prevention. Red spots on airline flight attendants.

MMWR. 1980;29:141. 16. Keenan NL, Shaw KM. Coronary heart disease and stroke deaths—United States,

2006. MMWR; 60:62–66. 17. Friis RH, Nanjundappa G, Prendergast T, et al. Hispanic coronary heart disease mor-

tality and risk in Orange County, California. Public Health Rep. 1981;96:418–422. 18. Lilienfeld DE. Definitions of epidemiology. Am J Epidemiol. 1978;107:87–90. 19. Mausner JS, Kramer S. Epidemiology: An Introductory Text, 2nd ed. Philadelphia:

Saunders; 1985. 20. Terris M. The epidemiologic tradition. Public Health Rep. 1979;94:203–209. 21. Syme SL. Behavioral factors associated with the etiology of physical disease: A social

epidemiological approach. Am J Public Health. 1974;64:1043–1045. 22. Centers for Disease Control and Prevention. Update: Outbreak of severe acute respi-

ratory syndrome—worldwide, 2003. MMWR. 2003;52:244. 23. Fingerhut LA, Warner M. Injury Chartbook. Health, United States, 1996–97.

Hyattsville, MD: National Center for Health Statistics; 1997. 24. MacMahon B, Pugh TF. Epidemiology Principles and Methods. Boston: Little,

Brown; 1970. 25. Heymann DL, ed. Control of Communicable Diseases Manual, 19th ed. Washington,

DC: American Public Health Association; 2008. 26. Centers for Disease Control and Prevention. The flu season. http://www.cdc.gov/flu/

about/season/flu-season.htm. Accessed July 14, 2012. 27. Snow J. Snow on Cholera. Cambridge, MA: Harvard University Press; 1965. 28. Hippocrates, The Writings of Hippocrates and Galen. Epitomised from the Original

Latin Translations, by John Redman Coxe. Philadelphia: Lindsay and Blakiston; 1846. 29. Hippocrates. On Airs, Waters, and Places. In: Adams F, ed. The Genuine Works of

Hippocrates. New York: Wood; 1886. 30. McEvedy C. The bubonic plague. Scientific American. 1988;258(Feb):118–123. 31. National Institutes of Health, National Institute of Allergy and Infectious Diseases.

Plague. http://www.niaid.nih.gov/topics/plague/Pages/default.aspx. Accessed July 19, 2012.

32. Centers for Disease Control and Prevention. Maps and statistics–plague. Plague worldwide. http://www.cdc.gov/plague/maps/index.html. Accessed July 19, 2012.

33. Graunt J. Natural and Political Observations, Mentioned in a Following Index, and Made Upon the Bills of Mortality, 2nd ed. London: Tho. Roycroft; 1662.

34. Kargon R. John Graunt, Francis Bacon, and the Royal Society. The reception of statistics. J Hist Med Allied Sci. 1963;October:337–348.

35. National Library of Medicine. Smallpox: A Great and Terrible Scourge: Variolation. http://www.nlm.nih.gov/exhibition/smallpox/sp_variolation.html. Accessed July 19, 2012.

51589_CH01_Printer.indd 52 09/02/13 12:00 PM

r e f e r e n C e s 53

36. National Library of Medicine. Smallpox: A Great and Terrible Scourge: Vaccination. http://www.nlm.nih.gov/exhibition/smallpox/sp_vaccination.html. Accessed July 19, 2012.

37. Centers for Disease Control and Prevention. Smallpox. http://www.bt.cdc.gov/train- ing/smallpoxvaccine/reactions/smallpox.html. Accessed July 19, 2012.

38. Enterline PE. Epidemiology: “Nothing more than common sense?” Occup Health Saf. 1979;January/February:45–47.

39. Lilienfeld AM, Lilienfeld DE. Foundations of Epidemiology, 2nd ed. New York: Oxford University Press; 1980.

40. Semmelweis IP, Murphy FB, trans. The etiology, the concept and prophylaxis of childbed fever (1861). In: Med Classics. 1941;5:350–773.

41. Iffy L, Kaminetzky HA, Maidman JE, et al. Control of perinatal infection by tradi- tional preventive measures. Obstet Gynecol. 1979;54:403–411.

42. Whitehead M. William Farr’s legacy to the study of inequalities in health. Bull World Health Organ. 2000;78:86–96.

43. Susser M. Causal Thinking in the Health Sciences. New York: Oxford University Press; 1973.

44. King LS. Dr Koch’s postulates. J Hist Med. Autumn 1952:350–361. 45. Hansen B. The image and advocacy of public health in American caricature and

cartoons from 1860 to 1900. Am J Public Health. 1997;87:1798–1807. 46. Taubenberger JK, Morens DM. 1918 influenza: The mother of all pandemics. Emerg

Infect Dis. 2006;12:15–22. 47. Flu.gov. Pandemic flu history. http://www.flu.gov/pandemic/history/index.html.

Accessed July 19, 2012. 48. Billings M. The influenza pandemic of 1918. http://www.stanford.edu/group/virus/

uda/. Accessed July 19, 2012. 49. Rothman KJ. Modern Epidemiology. Boston: Little, Brown; 1986. 50. Kannell WB, Abbott RD. Incidence and prognosis of unrecognized myocardial

infarction: An update on the Framingham study. N Engl J Med. 1984;311:1144–1147. 51. Monson RR. Occupational Epidemiology. Boca Raton, FL: CRC Press; 1990. 52. Doll R, Peto R. Mortality in relation to smoking: 20 years’ observation on male

British doctors. Br Med J. 1976;2:1525–1536. 53. Jaret P. The disease detectives. National Geogr Mag. 1991;January:116–140. 54. World Health Organization. Epidemiology of WHO-confirmed human cases of avian

influenza A (H5N1) infection. Weekly Epidemiological Record. 2006;81:249–260. 55. Centers for Disease Control and Prevention. Highly pathogenic avian influenza A

(H5N1) in people. http://www.cdc.gov/flu/avianflu/h5n1-people.htm. Accessed July 19, 2012.

56. National Cancer Institute and the National Institutes of Health. Cancer genetics. http://www.cancer.gov/cancertopics/genetics. Accessed July 20, 2012.

57. Berggren U, Fahlke C, Berglund KJ, et al. Dopamine D2 receptor genotype is associ- ated with increased mortality at a 10-year follow-up of alcohol-dependent individu- als. Alcohol Alcohol. 2010;(45):1–5.

58. Khoury MJ., Millikan R, Little J, Gwinn M. An emergence of epidemiology in the genomics age. Int J Epidemiol. 2004;33:936–944.

59. Palmer LJ. The new epidemiology: Putting the pieces together in complex disease aetiology. Int J Epidemiol. 2004;33:925–928.

51589_CH01_Printer.indd 53 09/02/13 12:00 PM

51589_CH01_Printer.indd 54 09/02/13 12:00 PM

2Chapte r

Practical Applications of Epidemiology

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

● discuss uses and applications of epidemiology ● define the influence of population dynamics on community health ● state how epidemiology may be used for operations research ● discuss the clinical applications of epidemiology ● cite causal mechanisms from the epidemiologic perspective

CHAPTER OUTLINE

I. Introduction II. Applications for the Assessment of the Health Status of

Populations and Delivery of Health Services III. Applications Relevant to Disease Etiology IV. Conclusion V. Study Questions and Exercises

51589_CH02_Printer.indd 55 09/02/13 12:08 PM

56 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

Introduction

This chapter provides a broad overview of the range of applications of the epidemiologic approach. As the basic method of public health, epidemiology touches many aspects of the health sciences. The late Jerry Morris, professor of community health at the London School of Hygiene and Tropical Medicine, articulated seven uses for epidemiology.1 (Refer to Figure 2–1.) These uses include one group related to health status and health services and another set related to disease etiology. The first part of this chapter covers applications in health status and health services. For example, by describing the occurrence of disease in the community, epidemiology helps public health practitioners and administrators plan for allocation of resources. Once needed services are imple- mented, the epidemiologic approach can help evaluate their function and utility. (See Exhibit 2–1 for a statement of seven uses of epidemiology.)

The second part of the chapter focuses on applications of epidemiology that are relevant to disease etiology. The causes of many diseases remain unknown; epidemiologists in research universities and federal and private agencies continue to search for clues as to the nature of disease. Knowledge that is acquired through such research may be helpful in efforts to prevent the occurrence

Seven Uses of Epidemiology

Disease Etiology Health Status and Health Services

Study history of the health

of populations

Diagnose the health

of the community

Examine the working of

health services

Estimate individual risks and chances

Identify syndromes

Complete the clinical

picture

Search for causes

51589_CH02_Printer.indd 56 09/02/13 12:08 PM

i n t r o d u C t i o n 57

Seven Uses of epidemiology

The epidemiological method is the only way of asking some questions in medicine, one way of asking others, and no way at all to ask many. Several uses of epidemiology have been described:

1. To study the history of the health of populations, and of the rise and fall of diseases and changes in their character. Useful projections into the future may be possible.

2. To diagnose the health of the community and the condition of the people, to measure the true dimensions and distribution of ill- health in terms of incidence, prevalence, disability, and mortal- ity; to set health problems in perspective and define their relative importance; to identify groups needing special attention. Ways of life change, and with them the community’s health; new mea- surements for monitoring them must therefore constantly be sought.

3. To study the working of health services with a view to their improve- ment. Operational research translates knowledge of (chang- ing) community health and expectations in terms of needs for services and measure [sic] how these are met. The success of services delivered in reaching stated norms, and the effects on community health—and its needs—have to be appraised, in rela- tion to resources. Such knowledge may be applied in action research pioneering better services, and in drawing up plans for the future. Timely information on health and health services is itself a key service requiring much study and experiment. Today, information is required at many levels, from the local district to the international.

4. To estimate from the group experience what are the individual risks on average of disease, accident and defect, and the chances of avoiding them.

5. To identify syndromes by describing the distribution and associa- tion of clinical phenomena in the population.

6. To complete the clinical picture of chronic diseases and describe their natural history: by including in due proportion all kinds of

e x

h ib

it 2

–1

continues

51589_CH02_Printer.indd 57 09/02/13 12:08 PM

58 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

of disease. Results of these epidemiologic studies are often quite newsworthy and sometimes controversial. More and more frequently, medical journals such as the New England Journal of Medicine (NEJM) are publishing reports of epide- miologic studies.2 Among the key reasons for the proliferation of these studies are, first, that they concentrate on associations between diseases and possible life- style factors, such as a habit, type of behavior, or some element of the diet, that presumably can be changed. Consequently, “The reports are . . . often of great interest to the popular media and the public, as well as to physicians interested in preventive medicine.”2(p 823) A second reason is that the major diseases that are predominant in American society are “chronic, degenerative diseases that prob- ably have several contributing causes, some of which have to do with lifestyle, operating over long periods.”2(p 823) An NEJM editorial pointed out:

It is usually very difficult to investigate such risk factors through experimental (or interventional) studies. In some cases it is impractical and in some it is unethical.

patients, wherever they present, together with the undemanding and the symptomless cases who do not present and whose needs may be as great; by following the course of remission and relapse, adjustment and disability in defined populations. Follow-up of cohorts is necessary to detect early subclinical and perhaps reversible disease and to discover precursor abnormali- ties during the pathogenesis, which may offer opportunities for prevention.

7. To search for causes of health and disease by computing the expe- rience of groups defined by their composition, inheritance and experience, their behaviour [sic] and environments. To confirm par- ticular causes of the chronic diseases and the patterns of multiple causes, describing their mode of operation singly and together, and to assess their importance in terms of the relative risks of those exposed. Postulated causes will often be tested in naturally occurring experiments of opportunity and sometimes by planned experiments. n

ExhIbIt 2–1 continued

51589_CH02_Printer.indd 58 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 59

For example, researchers cannot expose half of a group of children to lead for 10 years to compare their IQs 20 years later with those of the unexposed children. We must therefore rely on epidemiologic (or observational) studies.2(p 823)

Because of the increasingly important function that epidemiology performs in clinical decision-making, this chapter also touches on some of the valuable considerations of this application. Finally, a few words of caution are presented on limitations of epidemiology in determining the cause of disease. Coverage of the general concept of causality will permit a fuller understanding of these issues. The term causality refers to the relationship between cause and effect.

Applicat ions for the Assessment of the Health Status of Populat ions and Delivery of Health Services

As Morris noted, principal uses of epidemiology under this category include the history of the health of populations, diagnosis of the health of the community, and the working of health services.1

Historical Use of Epidemiology: Study of Past and Future Trends in Health and Illness An example of the historical use of epidemiology is the study of changes in disease frequency over time. (These changes are known as secular trends.) Illnesses and causes of mortality that afflict humanity, with certain exceptions, have shown dramatic changes in industrialized nations from the beginning of modern medicine to the present day. In general, chronic condi- tions have replaced acute infectious diseases as the major causes of morbidity and mortality in contemporary industrialized societies. Mortality data shed light on the overall health status of populations, suggest long-term trends in health, and help to identify subgroups of the population that are at greater risk of mortality than other subgroups.

Figure 2–2 identifies the top 10 causes of death for two contrasting years: 1900 and 2009, a period of more than one century. The data show that influenza and pneumonia dropped from the top position in 1900 to eight in 2009. In 2009 diseases of the heart were the leading cause of death, followed in second place by cancer. The overall crude death rate from all causes declined greatly during this period of about one century—from 1719.1 to 793.7 per 100,000 population.

51589_CH02_Printer.indd 59 09/02/13 12:08 PM

60 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

Since the early 1960s, the leading causes of death over decades of time have shown marked changes (Figure 2–3). For example, death rates for heart dis- ease, cancer, and stroke have shown long-term declining trends. Increases have been reported for Alzheimer’s disease, kidney disease, and hypertension.

FIGURE 2–2 The ten leading causes of mortality, 1900 and 2009, rank, cause, and crude death rate per 100,000 (not age-adjusted). Data for 1900 exclude infant mortality. Sources: Data from U.S. Bureau of the Census, Statistical Abstract of the United States: 1957, p. 69 ; United States Public Health Service, Vital Statistics Rates in the United States 1900–1940, Washington, DC: United States Government Printing Office, 1947; and from Kochanek KD, Xu JQ, Murphy SL, at al. Deaths: Preliminary Data for 2009, National Vital Statistics Reports. Vol 59, No 4, p. 5. Hyattsville, MD: National Center for Health Statistics, 2011.

Influenza and pneumonia, 202.2

Tuberculosis (all forms), 194.4

Diarrhea and Enteritis, 139.9

Diseases of the Heart, 137.4

Cerebrovascular Diseases, 106.9

Nephritis (Kidney Disease), 81.0

Accidents, 72.3

Cancer, 64.0 Senility, 50.2

Diphtheria, 40.3

Diabetes, 22.3

Cancer, 185.2

Cerebrovascular Diseases, 43.5

Chronic Lower Repiratory

Diseases, 44.7

Diseases of the Heart, 195.0

Flu-Pneu., 17.5

Kidney Dis., 15.7

Suicide, 11.9

Mortality in 1900

Mortality in 2009

Alzheimer's Disease, 25.7

Accidents, 38.2

51589_CH02_Printer.indd 60 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 61

In determining the reasons for these trends, one must take into account certain conditions that may affect the reliability of observed changes. According to MacMahon and Pugh, these are “variation in diagnosis, reporting, case fatality, or some other circumstance other than a true change of incidence.”3(p 159) Specific examples follow:

● Lack of comparability over time due to altered diagnostic criteria. The diagnostic criteria used in a later time period reflect new knowledge about disease; some categories of disease used in earlier eras may be omitted altogether. The diagnostic criteria may be more precise at a later time; for instance, considerable information has been obtained over three quarters of a century about chronic diseases. In some cases, when changes in diagnostic procedures are due to known alterations in diagnostic coding systems, the changes will be abrupt and readily identifiable.

● Aging of the general population. As the population ages due to the reduced impact of infectious diseases, improved medical care, and a decline in the death rate, there may be greater uncertainty about the precise cause of

FIGURE 2–3 Age-adjusted death rates for selected leading causes of death: United States, from 1958 to 2008. Source: Reproduced from Miniño AM, Murphy SL, Xu JQ, Kochanek KD. Deaths: Final Data for 2008. National Vital Statistics Reports; Vol. 59, No. 10. Hyattsville, MD: National Center for Health Statistics. 2011.

1,000.0 ICD-7 ICD-8

Diseases of heart

Malignant neoplasms

Cerebrovascular diseases

Hypertension

Parkinson’s disease

Alzheimer’s disease

Accidents (unintentional injuries)

ICD-9 ICD-10

100.0

10.0

R at

e p

er 1

00 ,0

00 U

.S . s

ta n

d ar

d p

o p

u la

ti o

1.0

19581960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2008 0.1

Nephritis, nephrotic syndrome, and nephrosis

Notes: ICD is the International Classification of Diseases. Circled numbers indicate ranking of conditions as leading causes of death in 2008. Age-adjusted death rates per 100,000 U.S. standard population; see “Technical Notes.”

Year

51589_CH02_Printer.indd 61 09/02/13 12:08 PM

62 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

death. Also, there may be inaccurate assignment of the underlying cause of death when older individuals are affected by chronic disease because multiple organ systems may fail simultaneously.

● Changes in the fatal course of the condition. Such changes would be reflected over the long run in decreases in the number of people with dis- ease who actually die of it.

Despite the factors that reduce the reliability of observed changes in morbid- ity and mortality, Figure 2–4 identifies four trends in disorders: disappearing, residual, persisting, and new epidemic disorders.4 Changes in the occurrence and patterns of morbidity and mortality are the results of a range of factors including improvements in medical care (e.g., development of new immunizations and medicines), alterations in environmental conditions (e.g., increased levels of pol- lution in the presence of toxic chemicals in our food), and appearance of new or more virulent forms of microbial disease agents. The four trends are defined as follows:

● Disappearing disorders are those disorders that were formerly common sources of morbidity and mortality in developed countries but that at pres- ent have nearly disappeared in their epidemic form. Under this category are smallpox (currently eradicated), poliomyelitis, and other diseases such as

FIGURE 2–4 Four trends in disorders.

Sexually transm. infections Tobacco use

Infant mortality

Lung cancer

HIV/AIDS

Obesity

Smallpox (eradicated)

Polio

Measles

Cancer (some forms)

Mental disorders Cerebrovascular diseases

Residual

New EpidemicPersisting

Disappearing

51589_CH02_Printer.indd 62 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 63

measles that have been brought under control by means of immunizations, improvement in sanitary conditions, and the use of antibiotics and other medications.

● Residual disorders are diseases for which the key contributing factors are largely known but specific methods of control have not been effectively implemented. Sexually transmitted diseases, perinatal and infant mortality among the economically disadvantaged, and health problems associated with use of tobacco and alcohol are examples.

● Persisting disorders are diseases that remain common because an effective method of prevention or cure evades discovery. Some forms of cancer and mental disorders are representative of this category.

● New epidemic disorders are diseases that are increasing markedly in fre- quency in comparison with previous time periods. The reader may surmise that examples of these are lung cancer and, most recently, acquired immune deficiency syndrome (AIDS). The emergence of new epidemics of diseases may be a result of the increased life expectancy of the population, new envi- ronmental exposures, or changes in lifestyle, diet, and other practices asso- ciated with contemporary life. Increases in the levels of obesity and type 2 diabetes in many parts of the world, notably in developed countries and also in developing areas, are examples of this category of disorders.

Predictions About the Future The study of population dynamics in relation to sources of morbidity and mortality reveals much about possible future trends in a population’s health. A population pyramid represents the age and sex composition of the popula- tion of an area or country at a point in time.5 By examining the distribution of a population by age and sex, one may view the impacts of mortality from acute and chronic conditions as well as the quality of medical care available to a population.

Figure 2–5 shows the age and sex distribution of the population of developed and developing countries for three time periods: 1950, 1990, and 2030. The left and right sides of each chart compare males and females, respectively. The x-axis (bottom of each chart) gives the number of the population in millions. The y-axis (left side of each chart) presents ages grouped into 5-year intervals. The following trends in the age and sex distributions are evident:

● Developing countries. In 1950 and 1990, less developed countries had a triangular population distribution. A triangular distribution is associ- ated with high death rates from infections, high birth rates, and other

51589_CH02_Printer.indd 63 09/02/13 12:08 PM

64 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

FIGURE 2–5 Population age distribution for developing and developed countries, by age group and sex–worldwide, 1950, 1990, and 2030. Source: Adapted and reprinted from Centers for Disease Control and Prevention, MMWR 2003;52(6):103. The United Nations and the U.S. Bureau of the Census are the authors of the original material.

Male

Developed countries

Developing countries

Female

Male Female

1950

1990

Projections for 2030

75–79

A g

e g

ro u

≥80

60–64 55–59 50–54 45–49 40–44 35–39 30–34 25–29 20–24 15–19 10–14

5–9 0–4

300 200 200100

Population (millions)

1000400 300 400

65–69 70–74

75–79

A g

e g

ro u

≥80

60–64 55–59 50–54 45–49 40–44 35–39 30–34 25–29 20–24 15–19 10–14

5–9 0–4

300 200 200100

Population (millions)

1000400 300 400

65–69 70–74

75–79

A g

e g

ro u

≥80

60–64 55–59 50–54 45–49 40–44 35–39 30–34 25–29 20–24 15–19 10–14

5–9 0–4

300 200 200100

Population (millions)

1000400 300 400

65–69 70–74

Male Female

51589_CH02_Printer.indd 64 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 65

conditions that take a heavy toll during the childhood years. These deaths result from a constellation of factors associated with poverty and depri- vation: poor nutrition, lack of potable water, and unavailability of basic immunizations, antibiotics, and sewage treatment. Consequently, fewer children survive into old age, causing smaller numbers of the population in the older groups. By 2030, improvements in health in developing coun- tries are likely to result in greater survival of younger persons, causing a projected change in the shape of the population distribution.

● Developed countries (industrialized societies). These countries manifest a rectangular population distribution. This rectangular shape was consis- tent for 1950 and 1990 and, with some exceptions, is projected also for 2030. Characteristically, infections take a smaller toll than in developing countries, causing a greater proportion of children to survive into old age; approximately equal numbers of individuals are present in each age group except among the very oldest age groups, with larger numbers of older women than men who survive. Because of reduced mortality due to infec- tious diseases and improved medical care in comparison with less devel- oped regions, residents of developed countries enjoy greater life expectancy. With continuing advances in medical care, the population of developed countries will grow increasingly older. The U.S. Bureau of the Census esti- mates that about one-fifth of the U.S. population in 2030 will be 65 years of age and older. There will be a need for health services that affect aging and all of its associated dimensions. One illustration is increasing the avail- ability of programs for the major chronic diseases, both with respect to preventive care in the early years and direct care in the older years.

Population Dynamics and Epidemiology Population dynamics denote changes in the demographic structure of popu- lations associated with such factors as births and deaths and immigration and emigration. This section presents definitions of two types of populations, fixed populations and dynamic populations, and illustrates how populations grow and wane. Noteworthy related concepts are the demographic transition and the epi- demiologic transition.

Terminology: fixed populations and dynamic populations A population may be either fixed or dynamic. A fixed population is one distinguished by a specific happening and consequently adds no new members; therefore, the population decreases in size as a result of deaths only.

51589_CH02_Printer.indd 65 09/02/13 12:08 PM

66 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

Examples of a fixed population are survivors of the 9–11 terrorist attack in New York, residents of New Orleans during Hurricane Katrina, and persons who have had a medical procedure such as hip replacement. A dynamic population is one that adds new members through immigration and births or loses members through emigration and deaths.7 An example of a dynamic population is the population of a county, city, or state in the United States.

Influences on population size Three major factors affect the sizes of populations: births, deaths, and migration.5 The latter term includes immigration and emigration—permanent movement into and out of a country, respectively. Figure 2–6 demonstrates how the three variables affect the net size of a population.

FIGURE 2–6 How births, deaths, and migration affect the net size of a population.

Emigration

Immigration

Births

Immigration Births

Births

Emigration Deaths

Stable Population (Equilibrium)

Deaths

Emigration Deaths

Population Size

Increasing Population

Decreasing Population

51589_CH02_Printer.indd 66 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 67

● Population in equilibrium or a steady state—the three factors do not contribute to net increases or decreases in the number of persons, meaning that the number of members exiting for various reasons equals the number entering.

● Population increasing in size—the net effect caused by the number of persons immigrating plus the number of births exceeds the number of per- sons emigrating plus the number of deaths.

● Population decreasing in size—the net effect caused by the number of persons emigrating plus the number of deaths exceeds the number of per- sons immigrating plus the number of births.

As the population pyramid portends, population characteristics are related to health patterns found in the community. The term demographic transition refers to the historical shift from high birth and death rates found in agrarian societ- ies to much lower birth and death rates found in developed countries.5 A decline in the death rate has been attributed in part to improvement in general hygienic and social conditions. Industrialization and urbanization contribute to a decline in the birth rate. The term epidemiologic transition is used to describe a shift in the pattern of morbidity and mortality from causes related primarily to infectious and communicable diseases to causes associated with chronic, degenerative dis- eases. The epidemiologic transition accompanies the demographic transition. The demographic transition, however, is not without its own set of consequences: Both industrialization and urbanization have led to environmental contamination, con- centration of social and health problems in the urban core areas of the United States, and out-migration of inner city residents to the suburbs.

Health of the Community One of the important applications in epidemiology is to provide methodolo- gies used to describe the overall health of a particular community. The result- ing description may then provide a key to the types of problems that require attention and also accentuate the need for specific health services. A complete epidemiologic description would include indices of health as well as indicators of the psychosocial milieu of the community. A representative list of variables that might be covered in a description of the health of the community is given in Exhibit 2–2.

Demographic and social variables Age and sex distribution: Referring to Exhibit 2–2, note that the first set of variables shown are demographic and social variables. Consider the example of

51589_CH02_Printer.indd 67 09/02/13 12:08 PM

68 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

Descriptive Variables for the health of the Community

Demographic and social variables:

1. Age and sex distribution 2. Socioeconomic status 3. Family structure, including marital status and number of single-

parent families 4. Racial, ethnic, and religious composition

Variables related to community infrastructure:

1. Availability of social and health services including hospitals and emergency rooms

2. Quality of housing stock including presence of lead-based paint and asbestos

3. Social stability (residential mobility) 4. Community policing 5. Employment opportunities

Health-related outcome variables:

1. Homicide and suicide rates 2. Infant mortality rate 3. Mortality from selected conditions (cause specific) 4. Scope of chronic and infectious diseases 5. Alcoholism and substance abuse rates 6. Teenage pregnancy rates 7. Occurrence of sexually transmitted diseases 8. Birth rate

Environmental variables:

1. Air pollution from stationary and mobile sources 2. Access to parks/recreational facilities 3. Availability of clean water 4. Availability of markets that supply healthful groceries 5. Number of liquor stores and fast-food outlets 6. Nutritional quality of foods and beverages vended to school-

children 7. Soil levels of radon n

e x

h ib

it 2

–2

51589_CH02_Printer.indd 68 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 69

the relationship between the age and sex composition of the population and typical health problems. In a community that consists primarily of senior citizens (as in a retirement community), health problems related to aging would tend to predominate. Chronic diseases (e.g., cancer, heart disease, and stroke) increase in prevalence among the elderly. Because of the longer life expectancy of women, an older population would tend to have a majority of elderly women, who might have unique health needs such as screening and interventions for osteoporosis, risk of falling, and other conditions associated with aging.

In contrast, a younger community would also have a distinctive morbidity and mortality profile. If there are many young children and teenagers, health officials might be particularly concerned with providing immunizations against vaccine-preventable infectious diseases. Another topic would be the prevention of sexually transmitted diseases (e.g., HIV/AIDS) and health education pro- grams for avoidance of substance use and smoking. Finally, attention would need to be directed to the control of unintentional injuries and deaths, which are the leading cause of mortality among younger persons, particularly young males.

Socioeconomic status (SES): SES, which comprises income level, educa- tional attainment, and type of occupation, is a major determinant of the com- munity’s health. Often, persons who have inadequate income and employment opportunities lack health insurance and access to health care. By definition, an aspect of low SES is low education levels. Individuals who have low education levels in comparison with more highly educated persons may be less aware of dietary and exercise practices that promote good health. Service employment in comparison with professional occupations usually does not does not carry a full range of health benefits.

Racial, ethnic, and religious composition: The racial and ethnic composi- tion of the community is related to its health profile. Some health outcomes are more common in one racial or ethnic group than in another, for example, sickle cell anemia among African Americans or diabetes mellitus among Latinos. Tay-Sachs disease tends to be more common among persons of Eastern European Jewish extraction than among other groups.

A community may demonstrate characteristic health patterns associated with members of a religious denomination if that group has settled in the commu- nity. Adherents of some religious denominations may adopt lifestyle and dietary practices that may affect the community health profile. For example, members of some religious groups may abstain from alcohol consumption and smoking or avoid certain foods that are high in saturated fats or increase cancer risks. Consequently, such communities would be expected to have lower frequencies of

51589_CH02_Printer.indd 69 09/02/13 12:08 PM

70 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

adverse health outcomes related to alcohol consumption, tobacco use, and diet. Thus, the health of the community may be determined to some extent by racial, ethnic, and religious factors.

Variables related to community infrastructure Availability of health and social services: The socioeconomic characteristics of the community relate in part to the availability of health and social services and ability to pay for healthcare services. Wealthy communities, because of greater tax resources, have the capacity to provide a greater range of social and health- related services, which may be more up-to-date and conveniently located than in less affluent areas. Low-income residents may utilize, as their primary source of medical care, public health services, which may be overcrowded and inaccessible by public transportation. Often, when state and federal funding are curtailed, wealthy communities have the means to back fill lost revenue with local funding resources, whereas poorer locales do not have this option.

Quality of housing stock: Safe and clean housing is essential to the health of the community. The presence of toxic lead, dangerous asbestos, and vermin in older housing detract from the quality of housing stock and contribute to adverse health outcomes. The U.S. Census Bureau operates the American Hous- ing Survey, which provides statistical information on the quality of housing in the United States.6 Figure 2–7 presents data for 2007 and 2009. In both years, slightly more than 5% of housing units were classified as inadequate and 23% as unhealthy, meaning that housing had rodent infestations, absence of smoke alarms, leaks, and peeling paint.

Social stability: Some of the newer communities, such as those in the Sunbelt of the southern United States, have highly mobile residents. The constant shift- ing of residents contributes to a sense of social instability, alienation, and lack of social connectedness. In turn, social pathology and adverse mental health prob- lems may result. Less affluent urban communities of some parts of the United States have high unemployment levels that encourage out-migration of younger residents who are seeking better economic prospects, leaving behind a majority of older and indigent individuals.

Community policing programs reinforce social stability by reducing violent crime. Communities that form partnerships with the police force (e.g., through neighborhood watch programs) often are more successful at policing the com- munity and maintaining lower crime rates than in communities where such part- nerships do not exist.

51589_CH02_Printer.indd 70 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 71

Health-related outcome variables: Measures of health outcomes shown in the exhibit are a barometer of community health status and suggest needed social and health-related services.

● Infant mortality rate: An elevated infant mortality rate may reflect inade- quate prenatal care, inadequate maternal diet, or a deficit of relevant social and health services.

● Suicide rates: Depression, social isolation, and alienation within the com- munity may contribute to increased suicide rates and also elevated rates of alcoholism and substance abuse.

● Chronic and infectious diseases: Often, chronic conditions (e.g., obesity and type 2 diabetes) reflect poor dietary choices and the existence of “food deserts” in the community. A resurgence of preventable infectious diseases, such as measles and tuberculosis, may reflect the failure of immunization and community infectious disease surveillance programs.

FIGURE 2–7 The quality of housing in the United States. Source: Data from Raymond J, Wheeler W, Brown MJ. Inadequate and Unhealthy Housing, 2007 and 2009. MMWR. 2011;60:22, 23, 25, 26.

Unhealthy Housing

Inadequate Housing

No Smoke Alarm

Peeling Paint

Leaks

Rodents

∗Percentage

0.0 5.0 10.0 15.0 20.0 25.0 30.0 35.0 40.0 45.0 50.0

43.1

37.3 33.9

31.9

11.3 14.7

8.3 6.6

2.8 3.3

2.4 1.8

4.9 6.0

9.3 7.1

6.5 4.3

5.6 4.2

21.9 27.1

19.2∗

27.1

West South Midwest Northeast

51589_CH02_Printer.indd 71 09/02/13 12:08 PM

72 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

● Teenage pregnancy rates/sexually transmitted diseases: Increases in the occurrence of pregnancies, births, and sexually transmitted diseases among teenagers within specific communities suggest the need for appropriate education and counseling services targeted to this age group.

● Homicide rates: High firearm death rates and homicide rates are indica- tors of the adverse conditions within the community. Figure 2–8 portrays motor vehicle, homicide, and firearm death rates for the South Atlantic states (plus Washington, D.C.) in the United States during 2003. Accord- ing to data reported for 2003, Washington, D.C., led all the other areas in mortality caused by assault and firearms, with age-adjusted death rates of 31.5 and 26.9, respectively, per 100,000 population.

FIGURE 2–8 Motor vehicle, assault, and firearm injury death rates (age adjusted), South Atlantic States, United States, 2003. Source: Data from Hoyert DL, Heron MP, Murphy SL, Kung H. Deaths: Final data for 2003, National Vital Statistics Reports. Vol. 46, No 13, p. 5. Hyattsville, MD: National Center for Health Statistics, 2006.

South Atlantic States

West Virginia

Virginia

South Carolina

North Carolina

Maryland

Georgia

Florida

District of Columbia

Delaware

0 20 40 60 80

Deaths per 100,000 population (age adjusted)

Motor vehicle accidents

Assault (homicide)

Injury by firearms

51589_CH02_Printer.indd 72 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 73

Environmental variables Numerous adverse environmental factors are implicated in the health of the community. Members of some economically disadvantaged communities have high levels of exposure to air pollution that emanate from diesel trucks and other vehicles on freeways that traverse the community. Other sources of air pollution include nearby industrial and power plants as well as port facilities where ships are off-loaded. Access to playgrounds and public parks may be limited as may be access to nutritious and healthful foods, particularly meals supplied to school- children. In some communities, the dominant food source may be snacks from liquor stores, the fare sold by fast-food outlets, and sugar-laden beverages sold in vending machines. Some low socioeconomic status communities are over- crowded and more likely to have associated unsanitary conditions, which can be linked to ill-health and transmission of infectious diseases.

Health disparities Using epidemiology to describe the health of the community relates to Goal 2, “Eliminate Health Disparities,” of Healthy People 2010. Goal 2 strives “. . . to eliminate health disparities among segments of the population, including dif- ferences that occur by gender, race or ethnicity, education or income, disabil- ity, geographic location, or sexual orientation.”8(p 11) A later document, Healthy People 2020, continues to express this goal. One of the four overarching goals of Healthy People 2020 is to “[a]chieve health equity, eliminate disparities, and improve the health of all groups.”9

Health disparities have been defined as, “. . . differences in health outcomes that are closely linked with social, economic, and environmental disadvantage.”10(p 1) Six areas are the focus of the U.S. Department of Health and Human Services: infant mortality, cancer screening and management, cardiovascular disease, diabetes, human immunodeficiency virus infection/AIDS, and immunizations. In a 2011 report, the CDC noted that “increasingly, the research, policy, and public health practice lit- erature report substantial disparities in life expectancy, morbidity, risk factors, and quality of life, as well does persistence of these disparities among segments of the population . . .” 11(p 3) As the U.S. population ages and becomes more ethnically and socioeconomically diverse, health disparities are likely to increase in the future.

For example, consider infant mortality, which as noted previously is an indi- cator of the health of the community. While the infant mortality rate in the United States has trended downward, it is 27th (based on 2006 data) in compari- son with other developed nations. Within the United States, African-American infants have approximately 2.45 times the mortality rate of white infants

51589_CH02_Printer.indd 73 09/02/13 12:08 PM

74 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

(14.01 per 1,000 versus 6.85 per 1,000 in 2003).12 When epidemiology is used to study the health of the community, this discipline can identify geographic areas that have elevated rates of infant mortality (as well as other adverse health conditions) and assist in identifying risk factors for these elevated rates.

Income inequality is one of the factors associated with health disparities. A common measure of income inequality is known as the Gini index, which is a number that ranges from 0 to 1. The closer the index is to one, the greater is the level of inequality. For example, a value of zero indicates total equality and a value of one total inequality. Income inequality is highest among advanced developed economies; in 2007 the Gini index for the United States was 0.46.11 In order to portray the effects of income inequality, statisticians report associations between the Gini index and health outcomes such as inequality in the number of healthy days. Figure 2–9 shows the state-specific index of inequality in the number of healthy days and the average number of healthy days in the United States for 2007. The lowest levels of health inequality and highest mean number of healthy days occurred in Utah, Connecticut, and North Dakota, the three states that had the lowest Gini scores. At the bottom of the list were Tennessee, Kentucky, and West Virginia, which had the three highest Gini scores and, consequently, the highest health inequality and lowest average number of healthy days.

Policy evaluation Regarding the health of the community, epidemiology is not only a descriptive tool but also plays a role in policy evaluation. As Ibrahim has pointed out, “Health planning and policy formulation in the ideal sense should apply to total commu- nities and employ a centralized process, which facilitates an overview of the whole rather than selected health problems.”13(p 4) Samet and Lee wrote: “The findings of epidemiologic research figure prominently in nearly all aspects of developing policies to safeguard the public’s health. Epidemiologic evidence receives consid- eration at the national and even global levels, while also directly and indirectly influencing individual decisions concerning lifestyle, work, and family.”14(p S1)

Legislators and government officials are charged with the responsibility of enact- ing laws, enforcing them, and creating policies, many of which have substantial impacts on public health. Numerous examples that have occurred in distant and recent history come to mind, including fluoridation of water, helmet protection for motorcycle riders, mandatory seat belt use in motor vehicles, and requiring auto- mobile manufacturers to install air bags in vehicles. Other examples of laws that impact health are shown in Table 2–1. The remainder of this section will advocate for an increasing role of epidemiologists in informing the policy-making process.

51589_CH02_Printer.indd 74 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 75

F IG

U R

E 2

–9

St at

e- sp

ec if

ic G

in i

in de

x of

i n

eq u

al it

y in

n u

m be

r of

h ea

lt h

y da

ys a

n d

av er

ag e

n u

m be

r of

h ea

lt h

da ys

–U n

it ed

S ta

te s,

2 00

7. S

ou rc

e: R

ep ro

du ce

d fr

om C

en te

rs f

or D

is ea

se a

n d

C on

tr ol

a n

d P

re ve

n ti

on .

M M

W R

. 20

11 ;6

0 (s

u pp

le m

en t)

:7 .

N u

m b

er o

f h

ea lt

h y

d ay

15 10

U ta

h C

on ne

ct ic

ut N

or th

D ak

ot a

H aw

ai i

D .C

. N

eb ra

sk a

C ol

or ad

o lll

in oi

s V

er m

on t

M in

ne so

ta Id

ah o

C al

ifo rn

W is

co ns

in Io

w a

M ar

yl an

d V

irg in

ia N

ew Y

or k

N ew

H am

ps hi

re M

as sa

ch us

et ts

S ou

th D

ak ot

a W

as hi

ng to

n M

on ta

na K

an sa

s O

re go

State

D el

aw ar

e A

la sk

a P

en ns

yl va

ni a

M ai

ne G

eo rg

ia M

ic hi

ga n

In di

an a

O hi

o N

ew je

rs ey

M is

so ur

i R

ho de

Is la

nd N

ev ad

a A

riz on

a A

rk an

sa s

N ew

M ex

ic o

F lo

rid a

S ou

th C

ar ol

in a

M is

si ss

ip pi

Te xa

s O

kl ah

om a

N or

th C

ar ol

in a

Lo ui

si an

a A

la ba

m a

Te nn

es se

e K

en tu

ck y

W es

t V irg

in ia

0 0.

05 0.

15 0.

25 0.

35 0.

1 0.

G in

i i n

d ex

G in

i i nd

A ve

ra ge

h ea

lth y

da ys

0. 3

W yo

m in

20 25

51589_CH02_Printer.indd 75 09/02/13 12:08 PM

76 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

The question arises as to whether enacted policies merely satisfy public whim, appease well-meaning interest groups, or in fact do have an established scien- tific rationale and documented efficacy. The term evidence-based decisions, as applied to public health policies, implies the enactment of laws that have empiri- cal support for their need as well as for their effectiveness.

Support for the involvement of epidemiologists in public health policy- making has been advocated strongly.15 Epidemiologists have an important role to play in the development of evidence-based decisions because of their expertise in studying about risks associated with certain exposures and their familiarity with findings based on human subjects.16 A clear illustration of epidemiologists’ involvement in risk assessment arises in the determination of health effects associated with vary- ing levels of exposure to potentially toxic agents in environmental health studies.

Further, in their traditional activities, epidemiologists participate in policy-making related to education, research, and publication of manuscripts.15 Expertise in these areas can be applied readily to other policy arenas. Matanoski pointed out that “. . . epidemiologists can predict future risks based on current trends and knowledge of changing risk factors in the population. Planning for future needs and setting goals to meet these needs will require population-based thinking, for which epidemiologists are well trained.”16(p 541)

The extremely complex issue of public health policy development encompasses five phases known as the policy cycle. These phases include examination of population health, assessment of potential interventions, alternative policy choices, policy imple- mentation, and policy evaluation17 (see Figure 2–10 for a diagram of factors that influence policy decision-making). As you can see from Figure 2–10, an

Table 2–1 Examples of Laws and Ordinances That Affect Public Health

Tobacco control policies Smoke-free bar, restaurant, and worksite laws in the United States (as of 2012, 29 U.S. states ban smoking in restaurants and bars and 23 states ban smoking in restaurants and bars and nonhospitality worksites) Prohibition of smoking in commercial aircraft Prohibition of smoking in airports (United States, Germany, England, Spain, and other countries) Prohibition of smoking in shopping malls Prohibition of smoking in outdoor areas (e.g., public parks and beaches, outdoor stadiums) Prohibition of smoking in automobiles when children are present Drug treatment systems for nonprescription drugs, such as cocaine and heroin Needle distribution programs for prevention of needle sharing among intravenous drug users Laws to regulate amount of particulate matter emitted from automobiles Ban on plastic bags in some communities (or local ordinances that provide for recycling of plastic bags) Removal of high fat and high sugar content foods from vending machines in schools Printing nutritional information on restaurant menus Prohibition of drivers’ use of cells phones unless the devices are hands-free; texting generally not permitted

51589_CH02_Printer.indd 76 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 77

epidemiologist might be able to provide input into a number of phases of decision-making, for example, those phases that pertain to scientific fact (human health), interpretation of science, cost/benefit analysis, and risk assessment. Refer to the case study (Exhibit 2–3) in which we describe an applied epidemiologic study (conducted by Robert Friis and Julia Lee) to evaluate responses to the Smoke free Bars Law in California.

FIGURE 2–10 Factors influencing policy decision-making. Source: Reproduced from Matanoski GM. Conflicts between two cultures: implications for epidemiologic researchers in communicating with policy-makers. Am J Epidemiol. 2001: 154(Suppl 12):S37. Reprinted by permission of Oxford University Press.

Decisions

Actions

Interpretation of Science Cost/Benefit Analysis Effectiveness of Options Uncertainties

Scientific Fact Human Health Ecology

Risk Assessment Management Options

Economics

Public

Industry

Environmental Groups

Political Bodies

Laws and

Case Study: Using epidemiologic Methods to Conduct a policy evaluation of the Smokefree bars Law

This research project investigated a community’s response to the California Smokefree Bars (SFB) Law, a change in tobacco con- trol policy that was implemented as Assembly Bill (AB) 3037 on January 1, 1998. The SFB Law removed the exemption for bars, tav-

erns, and lounges that had been included in AB 13, the 1995 Workplace Safety Law. AB13/3037 banned smoking in all bars throughout the state (with some exemptions for bars with no employees). For our epidemio- logic research, the SFB Law was viewed as a natural experiment, with its scope and timing under the control of the California State Legislature.

e x

h ib

it 2

–3

continues

51589_CH02_Printer.indd 77 09/02/13 12:08 PM

78 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

ExhIbIt 2–3 continued

Tobacco control policy in the form of laws and local ordinances is occur- ring with increasing frequency as part of the antitobacco efforts to reduce the deleterious first- and secondhand health effects of cigarette smoke. Evi- dence suggests that secondhand smoke has harmful health consequences from which customers and workers in alcohol-serving establishments need protection. These adverse effects include cancer, emphysema and other lung disorders, and heart disease. Policies to reduce exposure to second- hand smoke need to be investigated to understand their potential to effect health-related changes in population groups and to suggest recommenda- tions regarding their efficacy.

Our policy analysis of the response to the SFB Law was conducted within Long Beach, which is the fifth-largest city (population, 460,000) in the state of California and the second largest in Los Angeles County, the county in which Long Beach is located. Noteworthy is the fact that Long Beach has a distinguished record of local tobacco control. In September 1994, Long Beach was one of 22 cities in the state recognized for protecting the health of its residents through strong tobacco control policies. The Long Beach Smok- ing Ordinance, enacted in 1991, prohibited smoking in all enclosed work- places and public places. In 1993, the Long Beach City Council strengthened the ordinance by prohibiting smoking in all restaurants and restaurant/bar combinations. Additionally, Long Beach is one of the few cities in the state with its own health department, a key factor for the positive community response to both local and statewide tobacco control. Over the years, a very active Tobacco Education Program within the city’s health department has worked closely with the city to educate the citizens regarding antitobacco concerns and also to implement various tobacco control policies.

In order to determine the response to the California SFB Law, we directed our efforts to gathering data from five different perspectives: bar personnel, residents, economic data from the restaurant business, compliance at the bars, and print media. The study was conducted over a 4-year period (July 1998–June 2002). Trained interviewers were sent to a sample of alcohol- serving establishments, such as restaurant bars and stand-alone bars.

Observations of compliance at Long Beach bars showed a continuing decrease in the proportion of bars with inside ashtrays; no restaurant bars

continues

51589_CH02_Printer.indd 78 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 79

Working of Health Services: Operations Research and Program Evaluation The term operations research (operational research) is defined as “[t]he systematic study, by observations and experiment, of the working of a sys- tem (e.g., health services), with a view to improvement.”18 Epidemiol- ogy applied to operations research refers to the study of the placement of health services in a community and the optimum utilization of such services. “The usual epidemiologic approaches—descriptive, analytic, and experi- mental—are all used in health services research and, in addition, methods of evaluation have been expanded through their application to problems in health services.”19(p 140) A major contribution of epidemiology to operations research is the development of research designs, analytic techniques, and

in the sample had ashtrays during fall 2000 or spring 2001. Inside smoking increased only for stand-alone bars during fall 2000, and then decreased in spring 2001. No inside smoking was observed in any restaurant bars in either fall 2000 or spring 2001. Outside smoking continued to increase during the third year. The extent of the smell of smoke was significantly higher in stand-alone bars than in restaurant-bars, whether measured dur- ing daytime or early evening hours during fall 2000 and spring 2001. Based upon the odor of smoke, we concluded that compliance with the law was higher within restaurant-bars than stand-alone bars, although smoking continued in some restaurant-bars.

In year 1 (with a follow-up in year 3), a telephone survey of a cross- sectional sample of Long Beach residents was conducted with over 1,500 respondents. A key result was that approval for the SFB Law increased from 66% in year 1 to 73% in year 3. Other results demonstrated that 68% approved a ban on smoking on a nearby, wooden ocean pier; 75% approved of a cigarette tax to fund early childhood development programs; and 83% approved of smokefree zones in parks frequented by children. In conclusion, this case study demonstrated how epidemiologic methods (e.g., cross-sectional surveys and other analyses of population-based data) could be used in public health policy evaluation. n

Supported by Grant 7RT-0185, University of California Tobacco-Related Disease Research Program.

ExhIbIt 2–3 continued

51589_CH02_Printer.indd 79 09/02/13 12:08 PM

80 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

measurement procedures. Operations research strives to answer the following kinds of questions, among others:

● What health services are not being supplied by an agency in the community? ● Is a particular health service unnecessarily duplicated in the community? ● What segments of the community are the primary utilizers of a service, and

which segments are being underserved? ● What is the most efficient organizational and staff power configuration? ● What characteristics of the community, providers, and patients affect

service delivery and outcome? ● What procedures could be used to assess, match, and refer patients to

service facilities?

The perspective of operations research reveals the extent to which health ser- vices are harmonized. Coordination and integration of services helps to optimize use of available funds and services. Uncoordinated programs result in wasted resources, fragmentation, low efficiency, duplication, service gaps, lack of ser- vice continuity, and delays in securing services. Usually a single agency or pro- gram is unable to provide a full spectrum of needed health services, especially to individuals who are afflicted with severe health problems such as multiple sclerosis or mental disorders. One agency may specialize in diagnosis, evaluation, and treatment of the client’s physical problems, whereas another may emphasize mental health issues. Because the mental and physical dimensions of the person are intertwined, the holistic medical concept argues that there should be greater coordination among various healthcare agencies that specialize in a particular component of health services. Operations research facilitates such coordination.

During the 1970s, Robert Friis directed a project to improve the coordination of health services to severely developmentally disabled children in the Bronx, New York. Some of the goals of the project were to identify unmet needs for ser- vices, to identify overlapping services, and to assist the referral of clients from one agency to another. In brief, for every severely developmentally disabled youngster in the Bronx (individuals with an IQ lower than 50) who was under the age of 21, the following representative items of information were collected:

● the facility from which medical treatment or follow-up was received ● drugs or medications that the person received ● diagnostic tests received in the past ● enrollment in educational, recreational, and other specified programs ● specific conditions and disabilities presented

51589_CH02_Printer.indd 80 09/02/13 12:08 PM

a p p l i C a t i o n s f o r t h e a s s e s s m e n t o f t h e h e a l t h s t a t u s 81

The project aimed to quantify the characteristics of service utilization. Examples were clients’ diagnoses, the number of separate agencies that each client visited, and the types of medical and other services. By linking these types of information, the project would inform health about the numbers and kinds of services needed in the community and make projections for funding of health services. Although this description is a simplification of the goals of the project, it illustrates how the epidemiologic approach may be utilized for operations research purposes.

Two additional examples of the application of epidemiologic methods to oper- ations research are quantification of methods of payment of healthcare services and description of the residents of residential care facilities. The National Ambu- latory Medical Care Survey (NAMCS) is a “survey of the private office-based, non-Federal physicians practicing in the United States.”20(p 1) Figure 2–11 (part A) presents NAMCS data for the percent distribution of office visits by primary expected source of payment (e.g., Medicare, Medicaid, private medical insur- ance, and self-pay) according to patient’s age. Among persons aged 18–64 years, more than 60% were funded by private medical insurance; the majority of per- sons aged 65 years and older were funded by Medicare.

Figure 2–11 (part B) shows the age and sex of residents of residential care facilities. The majority of residents were non-Hispanic whites, females, and per- sons aged 85 years and older. Quantitative information such as the characteristics of residential care patients and the method of payment for medical care contrib- utes to improvement of access to health care in the United States.

The foregoing examples illustrate the role of epidemiology in evaluation of healthcare utilization and needs assessment. A related application is program evaluation. Specifically, how well does a health program meet certain stated goals? To illustrate, if the goal of a national health insurance program is to pro- vide equal access to health services, an evaluation of the program should include utilization by socioeconomic status variables. The program would be on target if the analysis revealed little discrepancy in service utilization by social class. Epidemiologic methods may be employed to answer this question by providing the following methodologic input:

● methods for selecting target populations to be included in the evaluation ● design of instruments for data collection ● delimitation of types of health-related data to collect ● methods for assessment of healthcare needs

Evaluation of a clinic program or other health service can make use of epide- miologic tools. An example of an issue to include in the evaluation is the extent

51589_CH02_Printer.indd 81 09/02/13 12:08 PM

82 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

Percent distribution

1SCHIP is State Children’s Health Insurance Program. 2Includes no charge or charity.

All ages

P at

ie nt

’s a

ge in

y ea

Under 18

18-64

65 and over

60 80 10040

Private insurance

Medicare Medicaid or SCHIP1

Self pay2

Other Unknown/ blank

FIGURE 2–11 (A) Percent distribution of office visits by primary expected source of payment according to patient’s age: United States, 2003; (B) Selected characteristics of residential care residents: United States, 2010. Sources: (A) Data from Hing E, Cherry DK, Woodwell DA. National Ambulatory Medical Care Survey: 2003 Summary. Advance data from Vital and Health Statistics: No. 385. Hyattsville, Maryland: National Center for Health Statistics, 2005. (B) Data from Caffrey C, Sengupta M, Park-Lee E, et al. Residents Living in Residential Care Facilities: United States, 2010. NCHS Data Brief. 2012;No. 91:1.

Female

P er

ce nt

Non-Hispanic white All residents

Under 65

65–74

75–84

85 and over

Age (years):

5470

91 100

51589_CH02_Printer.indd 82 09/02/13 12:08 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 83

to which a program reaches minority individuals or socially and economically disadvantaged persons, the aged, or other targeted groups. An evaluation also might address the issue of changes or improvements in the overall health status of a target population. Other epidemiologic evaluations have studied patient sat- isfaction with medical care.

Demographic and socioeconomic indices may be employed in the epidemio- logic evaluation of utilization of surgical operations. An example is the use of surgical interventions for benign prostatic hyperplasia (BHP), a common clinical diagnosis that is costly for the healthcare system.21 BHP causes weak urine flow, urgency, and other urinary tract symptoms among some older men. Data from the Southern Community Cohort Study (n = 21,949 men) demonstrated that the diagnosis of BHP was twice as common among Caucasian men as among African- American men (4.1% versus 9.9%, respectively). However, surgery for persons who reported a BHP diagnosis was more prevalent among African- American men than among Caucasian men (12.9% versus 9.1%, respectively). Among study participants who had lower income, diagnosis with BHP was less common than among individuals with higher income. In summary, the researchers found that race and socioeconomic status were independently linked with BHP.

Another example of this use of epidemiology is in the evaluation of minority populations’ access to health insurance coverage. In an analysis of epidemiologic data from the Hispanic Health and Nutrition Examination Survey, 1982–1984, researchers examined the percentages of health insurance coverage among three major subpopulations (Mexican, Puerto Rican, and Cuban) of adult Latinas. The findings demonstrated that Mexican-origin women had the lowest level of any health insurance coverage, about 64% in comparison to the two other groups (about 74% and 81%, respectively). The disparity was even more pronounced for older women, particularly those between 50 and 64 years of age.22 A more recent study based on data from the Third National Health and Nutrition Exam- ination Survey found that lack of health insurance at the time of interview was associated with 40% greater risk of death during follow-up.23

Applicat ions Relevant to Disease Etiology

The second group of applications encompasses uses of epidemiology that are connected with disease etiology (e.g., determining the causes of infectious and chronic diseases such as tuberculosis and cancer as well as preventing them).

51589_CH02_Printer.indd 83 09/02/13 12:08 PM

84 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

Under this general area, Morris1 noted the search for causes, individual risks, and specific clinical concerns. (See Figure 2–1.)

Causality in Epidemiologic Research As an observational science, epidemiology is frequently subject to criticism. The prestigious journal Science ran a special news report entitled, “Epidemiol- ogy Faces Its Limits.”24 The subtitle read: “The Search for Subtle Links between Diet, Life Style, or Environmental Factors and Disease Is an Unending Source of Fear—but Often Yields Little Certainty.” A portion of the report follows:

The news about health risks comes thick and fast these days, and it seems almost constitutionally contradictory. In January of last year [1994], for instance, a Swedish study found a significant association between residential radon expo- sure and lung cancer. A Canadian study did not. Three months later, it was pes- ticide residues. The Journal of the National Cancer Institute published a study in April reporting—contrary to previous, less powerful studies—that the presence of DDT metabolites in the bloodstream seemed to have no effect on the risk of breast cancer. In October, it was abortions and breast cancer. Maybe yes. Maybe no. In January of this year it was electromagnetic fields (EMF) from power lines . . .

These are not isolated examples of the conflicting nature of epidemiologic studies; they are just a few to hit the newspapers. Over the years, such studies have come up with a mind-numbing array of potential disease-causing agents, from hair dyes (lymphomas, myelomas, and leukemia), to coffee (pancreatic cancer and heart disease), to oral contraceptives and other hormone treatments (virtually every disorder known to women). The pendulum swings back and forth, subjecting the public to an “epidemic of anxiety,” as Lewis Thomas wrote many years ago. Indeed, the New England Journal of Medicine published an edi- torial by editors Marcia Angell and Jerome Kassirer asking the pithy question, “What Should the Public Believe?” “Health-conscious Americans,” they wrote, “increasingly find themselves beset by contradictory advice. No sooner do they learn the results of one research study than they hear of one with the opposite message.”24

Part of the reason for the skepticism about epidemiologic research is the inability of the discipline to “prove” anything. The contributions of Koch are considered by some as a basis for this skepticism. His postulates, first developed by Henle, adapted in 1877, and further elaborated in 1882, also are referred to as the Henle–Koch postulates. They were instrumental in efforts to prove (or disprove) the causative involvement of a microorganism in the pathogenesis of

51589_CH02_Printer.indd 84 09/02/13 12:08 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 85

an infectious disease. The postulates specified that the agent must be present in every case of the disease, must be isolated and grown in pure culture, must reproduce the disease when reintroduced into a healthy susceptible animal, and must be recovered and grown again in a pure culture. In addition, the agent should occur in no other disease: the one agent–one disease criterion. This clas- sical Henle–Koch concept of causality, sometimes referred to as pure determin- ism, becomes problematic when one attempts to apply it to the chronic diseases prevalent in modern eras. Let us examine separately three of the four criteria that form part of Koch’s concept of causality:

1. Agent present in every case of the disease. How well would this criterion apply to cardiovascular disease (CVD)? Decades of research have estab- lished that individuals who develop CVD tend to be overweight, physi- cally inactive, cigarette smokers, and have high blood pressure and high total cholesterol. If we were to apply Koch’s postulates strictly, then every case of CVD would have all these characteristics. Clearly not true.

2. One agent–one disease. How would this criterion hold up against cig- arette smoking? We just pointed out that smokers are more likely to develop CVD than nonsmokers. Is CVD the only disease associated with smoking? No. In fact, smoking is associated with lung cancer, pancreatic cancer, oral cancer, nasopharyngeal cancer, cervical cancer, emphysema, chronic obstructive pulmonary disease, and stroke, to name just a few. Therefore, the one agent–one disease criterion is not particularly helpful, especially for diseases of noninfectious origin.

3. Exposure of healthy subjects to suspected agents. The ethical conduct of research on humans forbids exposure of subjects to risks that exceed potential benefits. Would it be reasonable to suspect the smoking–lung cancer association even if such an experiment was never conducted? As pointed out in the introduction to this chapter, there are simply some exposures that cannot be evaluated in the context of controlled experi- mental studies. Epidemiology must be relied upon to provide such information.

In addition to the three issues just discussed that are direct tests of Koch’s postulates, there are others that must be considered. It is relatively straight- forward to categorize individuals with respect to the presence or absence of an exposure when the exposure is an infectious agent; one is either exposed or not exposed. However, even this simplification ignores the complicating issue

51589_CH02_Printer.indd 85 09/02/13 12:08 PM

86 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

of biologically effective dose. What about something such as blood pressure? Individuals with “elevated” blood pressure are more likely to develop a stroke than individuals with “low” blood pressure. Where does one draw the line between elevated and normal (or low)? At what level should an individual be considered obese?

A more subtle concept to consider is the fact that, for diseases of unknown etiology, we are dealing with imperfect knowledge. For example, although we may know that smokers are 20 times more likely to develop lung cancer than nonsmokers, why is it that not all smokers develop the disease? There must be other factors (e.g., diet, alcohol intake, and host susceptibility) that are part of the total picture of causality. When not all the contributing factors are known, it is problematic indeed to know truly and accurately the complete cause of a given disease. The issue of causality and epidemiology has been the focus of debate for decades. Some of the early writings are still fascinating and relevant today. For example, refer to Causal Thinking in the Health Sciences by Mervyn Susser.25 The work Eras in Epidemiology by Susser and coauthor Zena Stein presents informa- tion on the historical evolution of epidemiologic ideas.26 This book illustrates how epidemiology relies on and contributes to carefully formulated concepts of cause whether derived experimentally or observationally in the laboratory or general environment, both physical and social.

To summarize, the doctrine of multiple causality (instead of single causal agents) is now accepted widely; current research indicates that a framework of multiple causes for chronic diseases such as heart disease, cancer, and diabetes mellitus is appropriate. Noted epidemiologist the late John Cassel was an artic- ulate proponent of multifactorial causality for contemporary diseases. In the fourth Wade Hampton Frost Lecture, Cassel noted that early theories stated “disease occurred as a result of new exposure to a pathogenic agent.” The single agent causal model was extended to “the well-known triad of host, agent and environment in epidemiologic thinking.”27(pp 107–108) The formulation was satisfactory to explain diseases of importance during the late 19th and early 20th centuries, when agents of overwhelming pathogenicity and virulence produced conditions such as typhoid and smallpox. Cassel suggested that the triad of agent, host, and environment is no longer satisfactory because, “In a modern society the majority of citizens are protected from these overwhelming agents and most of the agents associated with current diseases are ubiquitous in our environment . . . [There may be] categories or classes of environmen- tal factors that are capable of changing human resistance in important ways.” One group of factors, Cassel argued, was the social environment (“presence of

51589_CH02_Printer.indd 86 09/02/13 12:08 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 87

other members of the same species”), which might be capable of profoundly influencing host susceptibility to environmental disease agents, whether they are microbiologic or physiochemical.27

Risk Factors Defined Because of the uncertainty of “causal” factors in epidemiologic research, it is customary to refer to an exposure that is associated with a disease as a risk factor. There are three requisite criteria for risk factors:

1. The frequency of the disease varies by category or value of the factor. Consider cigarette smoking and lung cancer. Light smokers are more likely to develop lung cancer than nonsmokers, and heavy smokers are more likely still to develop the disease.

2. The risk factor must precede the onset of disease. This criterion, known as temporality, applies to the smoking–lung cancer example. We now know that smoking causes lung cancer. Nevertheless, hypothetically speaking, if individuals with lung cancer began to smoke after the onset of disease, smoking would not be a likely cause of their condition. The issue of the temporal relationship between exposure and disease is par- ticularly relevant to chronic diseases such as cancer. Epidemiologists may not be able to determine when exposure occurred in relationship to onset of the disease.

3. The observed association must not be due to any source of error. In illustration, researchers could introduce methodological errors at any of several points during an epidemiologic investigation. These errors might occur in the selection of study groups, measurement of exposure and dis- ease, and data analysis.

Modern Concepts of Causality The 1964 Surgeon General’s Report Causal inferences derived from epidemiologic research (especially in the realm of noninfectious diseases) gained increasing popularity as a topic of formal discus- sion as a result of findings (in the early 1950s) regarding the association between smoking and lung cancer.28 The publication of Smoking and Health, Report of the Advisory Committee to the Surgeon General of the Public Health Service listed five criteria for the judgment of the causal significance of an association29 and, based on these criteria, concluded that smoking was a cause of lung cancer among men. (Exhibit 2–4 provides a description of the report.) These criteria were addressed subsequently in other writings by Susser,30 Rothman,7 and Hill.31

51589_CH02_Printer.indd 87 09/02/13 12:08 PM

88 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

Case Study: Does Smoking Cause Lung Cancer?

The first Surgeon General’s report on smoking and health was published in 1964. This report generated global reaction by stating that cigarette smoking is a cause of lung cancer in men and is linked to other disabling or fatal diseases. Five criteria were identified as necessary for the establishment of a causal relationship between

smoking and lung cancer. The report’s authors concluded that, to judge the causal significance of the association between cigarette smoking and lung cancer, several of these criteria would have to be taken into account in combination and no single criterion would, in itself, be “pathogno- monic” (pathognomonic means characteristic or diagnostic). The criteria of judgment were strength of association, time sequence, consistency of relationship upon repetition, specificity of association, and coherence of explanation.

1. Strength of association: The report stated that the relative risk ratio is the most direct measure of the strength of association between smoking and lung cancer; several retrospective and pro- spective studies completed up to the time of the report dem- onstrated high relative risks for lung cancer among smokers and nonsmokers. Thus, it was concluded that the criterion of strength of association was supported.

2. Time sequence: The report argued that early exposure to tobacco smoke and late manifestation seems to meet the criterion of time sequence, at least superficially.

3. Consistency upon repetition: With regard to the causal relation- ship between smoking and health, the report asserted that this criterion was strongly confirmed for the relationship between smoking and lung cancer. Numerous retrospective and pro- spective studies demonstrated highly significant associations between smoking and lung cancer; it is unlikely that these find- ings would be obtained unless the associations were causal or else due to unknown factors.

e x

h ib

it 2

–4

continues

51589_CH02_Printer.indd 88 09/02/13 12:08 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 89

Sir Austin Bradford Hill In 1965 Sir Austin Bradford Hill, Professor Emeritus of Medical Statistics at the University of London, published one of the seminal articles that elaborated on the five criteria for causality in epidemiologic research.31 The article, which was his President’s Address to the Section of Occupational Medicine of the Royal Society of Medicine, lists nine aspects of an empirical association to consider when one is trying to decide whether the association is consistent with cause and effect. Refer to Table 2–2. These were not intended to be interpreted as criteria of causality, but nonetheless they have been presented as such in several textbooks. The following is a quotation from his article:

I have no wish, nor the skill, to embark upon a philosophical discussion of the meaning of “causation.” The “cause” of illness may be immediate and direct, it

4. Specificity: The hypothesis that smoking causes lung cancer has been attacked because of the lack of specificity of the relation- ship; smoking has been linked to a wide range of conditions, including cardiovascular disease (CVD), low birth weight, and bladder cancer. The report claimed, however, that rarely in the biologic realm does an agent always predict the occurrence of a disease; in addition, accumulating evidence about chronic dis- eases suggests that a given disease may have multiple causes.

5. Coherence of explanation: The report contended that the associa- tion between cigarette smoking and lung cancer was supported for this criterion. Evidence noted included the rise in lung cancer mor- tality with increases in per capita consumption of cigarettes and increases in lung cancer mortality as a function of age cohort pat- terns of smoking among men and women; the sex differential in mortality was consistent with sex differences in tobacco use. Gen- eral smoking rates were higher among men than among women; the report noted that young women were increasing their rates of smoking, however. n

Source: Data from U.S. Department of Health, Education and Welfare, Public Health Service. Smoking and Health, Report of the Advisory Committee to the Surgeon General of the Public Health Service. Public Health Service publication 1103. Washington, DC: Government Printing Office; 1964.

ExhIbIt 2–3 continued

51589_CH02_Printer.indd 89 09/02/13 12:08 PM

90 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

may be remote and indirect, underlying the observed association. But with the aims of occupational, and almost synonymously preventive, medicine in mind, the decisive question is whether the frequency of the undesirable event B will be influenced by a change in the environmental feature A. 31(p 295)

Further elaborating on his statement, Hill asserted that in some instances much research would be required to determine the existence of a causal associa- tion. In other cases, a smaller body of information would be adequate. Thus, making causal inferences depends upon the circumstances of an association. Hill’s landmark article identified nine issues that are relevant to causality and epidemiologic research. (Refer to Table 2–2.)

1. Strength of association. One example cited by Hill was the observation of Percival Pott that chimney sweeps in comparison to other workers had an enormous increase in scrotal cancer; the mortality was more than 200 times that of workers not exposed to tar and mineral oils. A strong association is less likely to be the result of errors.

2. Consistency upon repetition. This term refers to whether the association between agent and putative health effects has been observed by differ- ent persons in different places, circumstances, and times. The Surgeon General’s report of 1964 cited a total of 36 different studies that found an association between smoking and lung cancer.29 Hill felt that consistency was especially important when the exposure was rare.

3. Specificity. With respect to occupational exposures, Hill noted that if “the association is limited to specific workers and to particular sites and types of disease and there is no association between the work and other modes of dying, then clearly that is a strong argument in favor of causation.”31(p 297) He later went on to acknowledge that the criterion of specificity should be used as evidence in favor of causality; however, if evidence of a specific association cannot be obtained, this fact is not necessarily a refutation of a causal association.

Table 2–2 Aspects of an Association That Suggest Causality

1. Strength 2. Consistency 3. Specificity 4. Temporality 5. Biological gradient 6. Plausibility 7. Coherence 8. Experiment 9. Analogy

Source: Data from Hill AB. The environment and disease: association or causation? Proceedings of the Royal Society of Medicine. 1965; 58:295–300.

51589_CH02_Printer.indd 90 09/02/13 12:08 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 91

4. Time sequence. In Hill’s words, “Which is the cart and which is the horse?” For example, if one is trying to identify the role of diet in the pathogenesis of colon cancer, one has to be careful to sort out dietary pref- erences that lead to colon cancer versus dietary changes that result from early stages of the disease. There is some evidence that low intakes of cal- cium are associated with increased risk of colon cancer. If early stages of disease create problems with digestion of milk products (which are good sources of calcium), individuals may lower their intake of milk (and cal- cium) as a consequence of the disease. The shorter the duration between exposure to an agent and development of the disease (i.e., the latency period), the more certain one is regarding the hypothesized cause of the disease. For this reason, many of the acute infectious diseases or chemi- cal poisonings are relatively easy to pinpoint as to cause. Diseases having longer latency periods (many forms of cancer, for example) are more dif- ficult to relate to a causal agent; it is said that the onset of chronic diseases is insidious and that one is ignorant of the precise induction periods for chronic diseases. Many different causal factors could intervene during the latency period. This is why a great deal of detective work was needed to link early exposure to asbestos in shipyards to subsequent development of mesothelioma, a form of cancer of the lining of the abdominal cavity.

5. Biologic gradient. Evidence of a dose–response curve is another impor- tant criterion. Hill notes, “the fact that the death rate from lung cancer increases linearly with the number of cigarettes smoked daily adds a great deal to the simpler evidence that cigarette smokers have a higher death rate than non-smokers.”31(p 298) MacMahon and Pugh state, “the exis- tence of a dose-response relationship—that is, an increase in disease risk with increase in the amount of exposure—supports the view that an asso- ciation is a causal one.”3(p 235) Figure 2–12 illustrates a dose–response relationship between number of cigarettes smoked per day and lung can- cer mortality among male British physicians.

6. Plausibility. If an association is biologically plausible, it is credible on the basis of existing biomedical knowledge.18 The weakness of this line of evidence is that it is necessarily dependent upon the biologic knowledge of the day.

7. Coherence of explanation. The association must not seriously conflict with what is already known about the natural history and biology of the disease. Data from laboratory experiments on animals may be most help- ful. For example, the ability of tobacco extracts to cause skin cancer in

51589_CH02_Printer.indd 91 09/02/13 12:08 PM

92 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

mice is coherent with the theory that consumption of tobacco products in humans causes lung cancer.

8. Experiment. In some instances there may be “natural experiments” that shed important light on a topic. The observation that communities with naturally fluoridated water had fewer dental caries among their citizens than communities without fluoridated water is one example.

9. Analogy. The examples Hill cites are thalidomide and rubella. Thalidomide, administered in the early 1960s as an antinausea drug for use during pregnancy, was associated subsequently with severe birth defects. Rubella (German measles), if contracted during pregnancy, has been linked to birth defects, stillbirths, and miscarriages. Given that such associations have already been demonstrated, “we would surely be ready to accept slighter but similar evidence with another drug or another viral disease in pregnancy.”31(p 299)

Although it is not critical that all these lines of evidence be substantiated to uphold a causal association, the more that are supported, the more the case of causality is strengthened. More important, careful consideration of these

FIGURE 2–12 Dose–response relationships between smoking and lung cancer mortality among British physicians. Source: Data from R Doll, R Peto. Mortality in Relation to Smoking: 20 Year’s Observation on Male British Doctors. British Medical Journal, vol 2 (6051), pp 1525–1536, BMJ Publishing Group, © 1976.

25 +15–241–14Nonsmoker 0

Number of cigarettes

M or

ta lit

y ra

tio s 20

51589_CH02_Printer.indd 92 09/02/13 12:08 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 93

concepts is helpful in trying to decide at what point one needs to take action. One of Hill’s concluding remarks was particularly apropos: “All scientific work is incomplete, whether it be observational or experimental. All scientific work is liable to be upset or modified by advancing knowledge. That does not confer upon us a freedom to ignore the knowledge we already have, or to postpone the action that it appears to demand at a given time.”31(p 300) Evans,32 in a compel- ling discussion of causality, drew an analogy between ascertainment of causal- ity and establishment of guilt in a criminal trial. Evans’ detailed arguments are found in Exhibit 2–5.

Frequently, the processes of causal inference and statistical inference overlap yet represent different principles. According to Susser, “Formal statistical tests are framed to give mathematical answers to structured questions leading to judg- ments, whereas in any field practitioners must give answers to unstructured ques- tions leading from judgment to decision and implementation.”30(p 1)

Study of Risks to Individuals In many instances, epidemiologic research on disease etiology involves collec- tion of data on a number of individual members of different study groups or study populations. Epidemiologists use two main types of observational studies for research on disease etiology: case-control and cohort studies. A case-control design compares a group of individuals who have a disease of interest (the cases) with a group who does not have the disease (the controls). The two groups are compared with respect to a variety of hypothesized exposures (e.g., diet, exercise habits, or use of sunscreens). Differences in exposure that are observed between the two groups may suggest why one group has the disease and the other does not. Another research method is the cohort study. In this approach, a study group free from disease is assembled and measured with respect to a variety of exposures that are hypothesized to increase (or decrease) the chance of getting the disease. One then follows the group over time for the development of disease, comparing the frequency with which disease develops in the group exposed to the factor and the group not exposed to the factor. Either type of study may demonstrate that a disease or other outcome is more likely to occur in those with a particular exposure.

The issue of whether the results of an epidemiologic study influence clinical decision-making is in part determined by the criteria of causality covered in the previous section. How large is the effect? How consistent is the finding with previous research? Is there biologic plausibility? All these issues are important,

51589_CH02_Printer.indd 93 09/02/13 12:08 PM

94 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

rules of evidence: Criminality and Causality

In criminal law, the presence of the criminal at the scene of the crime would be equivalent to the presence of the agent in a lesion of the disease. Premeditation would be similar to the requirement that the causal expo- sure should precede the onset of the disease. The presence of accessories at the scene of the crime might be compared to the presence of cofactors and/or multiple causes for human diseases. The severity of the crime or the consequence of death might be loosely equivalent to susceptibility and the host responses, which determine the severity of the illness. The motivation involved in a crime should make sense in terms of reward to the criminal, just as the role of the causal agent should make biologic sense. The absence of other suspects and their elimination in a criminal trial would be similar to that of the exclusion of other putative causes in human illness. Finally, need that the proof of guilt must be established beyond a reasonable doubt would be true for both criminal justice and for disease causation. n

Source: Adapted from Evans AS. Causation and disease: A chronological journey, American Journal of Epidemiology, 108(4):254–255; with permission of the Johns Hopkins University, School of Hygiene and Public Health. © 1978.

e x

h ib

it 2

–5

Mayhem or murder and criminal law

1. Criminal present at scene of crime.

2. Premeditation.

3. Accessories involved in the crime.

4. Severity or death related to state of victim.

5. Motivation: The crime must make sense in terms of gain to the criminal.

6. No other suspect could have committed.

7. The proof of the guilt must be established beyond a reasonable doubt.

Morbidity, mortality, and causality

1. Agent present in lesion of the disease.

2. Causal events precede onset of disease.

3. Cofactors and/or multiple causalities involved.

4. Susceptibility and host response determine severity.

5. The role of the agent in the disease must make biologic and common sense.

6. No other agent could have caused the disease under the circumstances given.

7. The proof of causation must be established beyond reason- able doubt or role of chance.

51589_CH02_Printer.indd 94 09/02/13 12:08 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 95

but a major issue for the clinician is the relevance to each particular patient. Epidemiologic studies employ groups of individuals; the studies provide evi- dence that groups with particular exposures or lifestyle characteristics are more or less likely to develop disease than groups of individuals without the expo- sures. Extrapolation to the individual from findings based on observations of groups should be made with caution. The observation that cigarette smokers are 20 times more likely to develop lung cancer than nonsmokers does not neces- sarily entitle someone to tell a smoker, “You are 20 times more likely to get lung cancer than a nonsmoker.” The problem is that there are a number of other factors that may be important contributors to the cause of lung cancer. A more accurate statement would be, “Collectively, groups of individuals who smoke are 20 times more likely to develop lung cancer than nonsmokers.” The difference is subtle, yet important.

Another issue for the clinician is the size of the risk; an example is the slight risk of mortality from CVD associated with a high serum cholesterol level. If the risk is small, a person may reasonably not wish to change his or her lifestyle.33 The 1990 editorial in the New England Journal of Medicine is particularly illus- trative.2 Suppose that the 10-year risk of death is 1.7% in middle-aged men with cholesterol levels below 200 mg/dL but 4.9% if the cholesterol level is above 240 mg/dL.34 This difference in risk of approximately 3.0% may not be suf- ficient to induce an otherwise healthy man to try to lower his cholesterol level. Conversely, even if the risk factor is strong, it may still be unimportant to indi- vidual patients if the disease is rare.

Thus, the extrapolation of epidemiologic research to individuals is compli- cated. Another aspect of risk concerns public health implications. A risk fac- tor that may be relatively unimportant for individuals may be important indeed when the effect is multiplied over the population as a whole, especially if the disease is common.

Another example of this application of epidemiology is predicting the indi- vidual’s prognosis and likelihood of survival if afflicted by a serious medical con- dition. Clinicians can use such information to aid the patient in decision-making about whether to undergo invasive surgical procedures or debilitating treatments for cancer. Information about prognosis helps demonstrate the efficacy of medi- cal interventions (e.g., coronary bypass surgery) by showing whether the prac- tice yields an increase in long-term survival for the population. Some additional illustrations of the use of epidemiology to study risks to the individual are mak- ing predictions of mortality from cancer and other serious chronic illnesses and developing assessments of morbidity and mortality from infectious diseases.

51589_CH02_Printer.indd 95 09/02/13 12:08 PM

96 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

Epidemiologic research indicates that there is an important contribution to mortality from common infectious diseases, such as influenza and colds. Some- times mortality results from complications that can occur in high risk groups such as neonates, elderly persons, and immunocompromised individuals. With- out population-based data, mortality from these “minor” diseases might not be obvious. In 2008, influenza was responsible for 0.6 deaths per 100,000 individu- als in the United States.12

Epidemiologic data may be used to predict cancer prognosis and mortality. Both vary by site of the tumor, type, and a number of social variables, such as socioeconomic status, race, and sex. Figure 2–13 presents the 5-year relative survival rate for selected forms of cancer by race from 2002 to 2008. Differences in survival are evident by both cancer type and race. Among African- Americans in comparison with whites, the 5-year survival rates for all cancer sites were 59.9% and 68.9%, respectively. Survival rates for cancer of the pancreas and lung (6.0% and 16.9%, respectively) were lower than the rates for prostate can- cer (99.9%) and breast cancer (90.2%) 35

Another illustration of the study of risks to the individual involves prognosis of survival from coronary bypass surgery. The Veterans Administration Cooperative Study36 traced the survival of 596 patients treated by medication or by surgery

FIGURE 2–13 Five-year relative and period survival (%) from invasive cancer by race and sex in the United States, 2002–2008. Source: Data from Howlander N, Noone AM, Krapcho, et al. SEER Cancer Statistics Review, 1975–2009 (Vintage 2009 Populations), National Cancer Institute. Bethesda, MD, updated April 30, 2012.

67.7

All Cancer Sites 0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

80.0

90.0

100.0

Cervix Lung

P er

ce nt

ag e

Pancreas Breast Prostate

68.9

59.9

68.870.2

61.1

16.9 17.3 13.5

6.0 6.2 4.8

90.2 91.7

78.0

99.9 99.997.7

All Whites Blacks

51589_CH02_Printer.indd 96 09/02/13 12:08 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 97

for chronic stable angina in a large-scale prospective, randomized study. Findings indicated no differences in survival at 21 and 36 months between surgery patients and medically treated patients. Thus, the factors of mortality from surgery itself and expense of the operation need to be weighed against increases in life expec- tancy and improvement in the quality of life due to improved arterial circulation. This is an epidemiologic question that may be raised about risks associated with other types of surgical procedures as well.

Enlargement of the Clinical Picture of Disease When a new disease first gains the attention of health authorities, usually the most dramatic cases are the ones observed initially. One may conclude incor- rectly that the new disease is an extremely acute or fatal condition; later epi- demiologic studies may reveal that the most common form of the new disease is a mild, subclinical illness that occurs widely in the population. To develop a full clinical picture of the disease, thorough studies are necessary to find out about the subacute cases; an adequate study may require a survey of a complete population.

One example of this use of epidemiology was the investigation of the 1976 Legionnaires’ disease outbreak, which at first seemed to be a highly virulent and new condition. The outbreak of a mysterious illness that ravaged participants at the American Legion’s July 1976 convention in Philadelphia riveted public attention. Concerned officials appealed to local and federal epidemiologists to investigate the outbreak. Disease detectives ascertained that Legionnaires’ disease was associated with a previously unidentified bacterium, Legionella pneumophila. Although the Philadelphia outbreak suggested initially that Legionnaires’ disease was highly fatal, subsequent research found a much lower case fatality rate; about 15% of the people who developed the disease died from it. The previously unrec- ognized disease had probably occurred sporadically in other areas of the country before 1976.

Prevention of Disease One of the potential applications of research on disease etiology is to identify where, in the disease’s natural history, effective intervention might be imple- mented. The natural history of disease refers to the course of disease from its beginning to its final clinical end points. Figure 2–14 illustrates the natural his- tory of any disease in humans. As the figure demonstrates, the natural history signifies the progression of disease over time.

51589_CH02_Printer.indd 97 09/02/13 12:08 PM

98 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

The period of prepathogenesis occurs before the precursors of disease (e.g., the bacterium that causes Legionnaires’ disease) have interacted with the host (the person who gets the disease). The period of pathogenesis occurs after the precur- sors have interacted with the host, an event that is marked by initial appearance of disease (the presymptomatic stage) and is characterized by tissue and physi- ologic changes. Later stages of the natural history include development of active signs and symptoms, and eventually recovery, disability, or death (all examples of clinical end points).

According to the model that Leavell and Clark36 advanced, three strategies for disease prevention—primary, secondary, and tertiary—coincide with the periods of prepathogenesis and pathogenesis. Figure 2–15 demonstrates these three lev- els of prevention, which are described in more detail in the following sections.

Primary Prevention Primary prevention occurs during the period of prepathogenesis. As shown in Figure 2–15, primary prevention includes health promotion and specific protec- tion against diseases. The former is analogous to a type of prevention known as primordial prevention. The term primordial prevention denotes “. . . conditions, actions and measures that minimize hazards to health and that hence inhibit the emergence and establishment of processes and factors (environmental,

FIGURE 2–14 Prepathogenesis and pathogenesis periods of natural history. Source: Modified with permission from Leavell HR, Clark EG. Preventive Medicine for the Doctor in His Community: An Epidemiologic Approach, 3rd ed. New York: McGraw-Hill Book Company; 1965, p. 18.

Prepathogenesis Period

Before man is diseased The course of the disease in man

Period of pathogenesis

NATURAL HISTORY OF ANY DISEASE PROCESS IN MAN

Interaction of:

ENVIRONMENTAL factors that produce disease STIMULUS

Interaction of HOST

and STIMULUS

Early pathogenesis

Discernible early disease

Convalescence Advanced disease

CLINICAL HORIZON

Disease AGENT

Human HOST

Death

Chronic state

Disability

Recovery

51589_CH02_Printer.indd 98 09/02/13 12:09 PM

a p p l i C a t i o n s r e l e v a n t t o d i s e a s e e t i o l o g y 99

economic, social, behavioral, cultural) known to increase the risk of disease.”18 Primordial prevention is concerned with minimizing health hazards in general, whereas primary prevention seeks to lower the occurrence of disease. Primordial prevention is achieved in part through health promotion, which includes health education programs in general, marriage counseling, sex education, and provi- sion of adequate housing.

Examples of primary prevention that involve specific protection against disease-causing hazards are wearing protective devices to prevent occupational injuries, utilization of specific dietary supplements to prevent nutritional defi- ciency diseases, immunizations against specific infectious diseases, and educa- tion about the hazards of starting smoking. Interventions to reduce the number

FIGURE 2–15 Levels of application of preventive measures in the natural history of disease. Source: Modified with permission from Leavell HR, Clark EG. Preventive Medicine for the Doctor in His Community: An Epidemiologic Approach, 3rd ed. New York: McGraw-Hill Book Company; 1965, p. 18.

T H E N A T U R A L H I S T O R Y O F A N Y D I S E A S E I N M A N

P r e p a t h o g e n e s i s P e r i o d P e r i o d o f P a t h o g e n e s i s

L E V E L S o f A P P L I C A T I O N o f P R E V E N T I V E M E A S U R E S

HEALTH PROMOTION

SPECIFIC PROTECTION EARLY DIAGNOSIS and PROMPT TREATMENT

DISABILITY LIMITATION

REHABILITATION Health educat ion

Use of speci f ic immunizat ions Case-f inding measures,

indiv idual and mass

Adequate t reatment to arrest the disease process and to prevent fur ther compl icat ions and sequelae

Provis ion of hospi ta l and community faci l i t ies for retraining and educat ion for maximum use of remaining capaci t ies

Educat ion of the publ ic and industr y to ut i l ize the rehabi l i tated

As ful l employment as possible

Select ive placement

Work therapy in hospi ta ls

Use of shel tered colony

Provis ion of faci l i t ies to l imi t d isabi l i ty and to prevent death

Screening surveys

Select ive examinat ions

To cure and prevent disease processes

To prevent the spread of communicable diseases

To prevent compl icat ions and sequelae

To shor ten per iod of disabi l i ty

Object ives:

Attent ion to personal hygiene

Use of environmental sani tat ion

Protect ion against occupat ional hazards

Protect ion f rom accidents

Use of speci f ic nutr ients

Protect ion f rom carcinogens

Avoidance of a l lergens

Good standard of nutr i t ion adjusted to developmental phases of l i fe

Attent ion to person- al i ty development

Provis ion of adequate housing, recreat ion, and agreeable working condi t ions

Marr iage counsel ing and sex educat ion

Genet ics

Per iodic select ive examinat ions

Ter t iary Prevent ionSecondary Prevent ionPr imary Prevent ion

Interrelat ions of Agent, Host, and Environmental factors

Product ion of STIMULUS Ear ly

pathogenesis Discernible

ear ly lesions Advanced disease

Convalescence

React ion of the HOST to the STIMULUS

51589_CH02_Printer.indd 99 09/02/13 12:09 PM

100 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

of alcohol-related traffic accidents similarly may focus on education, media campaigns, and warning labels on alcohol-containing beverages.

Primary prevention may be either active or passive. Active prevention neces- sitates behavior change on the part of the subject. Wearing protective devices and obtaining vaccinations require involvement of the individual to receive the benefit. Passive interventions, on the other hand, do not require any behavior change. Fluoridation of public water supplies and vitamin fortification of milk and bread products achieve their desired effects without any voluntary effort of the recipients.

Secondary Prevention Secondary prevention, which takes place during the pathogenesis phase of the natural history of disease, encompasses early diagnosis and prompt treatment as well as disability limitation. One example of secondary prevention is early diagnosis and prompt treatment linked to cancer screening programs, which are efforts to detect cancer in its early stages (when it is treated more successfully) among apparently healthy individuals. One should note that in the instance of a positive screening result confirmed by a diagnostic workup, cancer is already present; however, detection of the tumor before the onset of clinical symptoms reduces the likelihood of progression to death. Most cancer screening programs are forms of secondary prevention. However, screening for colorectal cancer can be considered also as primary prevention: Because most colorectal cancers arise through a precancerous lesion (adenomatous polyp), screening that detects and removes polyps can prevent cancer, rather than merely detect cancer early.

Later in the natural history of disease (when discernible lesions or advanced disease have appeared), there occurs a type of secondary prevention called dis- ability limitation, which is designed to limit and shorten the period of disabil- ity and prevent death from a disease. Another goal of disability limitation is to prevent the side effects and complications that may be associated with a disease.

Tertiary Prevention Tertiary prevention takes place during late pathogenesis (advanced disease and convalescence stages). Thus, disease already has occurred and has been treated clinically, but rehabilitation is needed to restore the patient to an optimal func- tional level. Examples include physical therapy for stroke victims, halfway houses for persons recovering from alcohol abuse, sheltered homes for the developmen- tally disabled, and fitness programs for heart attack patients. This category of

51589_CH02_Printer.indd 100 09/02/13 12:09 PM

s t u d y Q u e s t i o n s a n d e x e r C i s e s 101

prevention seeks to achieve maximum use of the capacities of persons who have disabilities and help them regain full employment.

Conclusion

This chapter identified seven uses of epidemiology. The historical use of epidemiology traced changes in rates of disease from early in this century to the present. Dramatic changes in morbidity and mortality rates were noted. Pre- dictions of future trends in health status incorporate population dynamics or shifts in the demographic composition of populations. Operations research and program evaluation are examples of using epidemiologic methods to improve healthcare services. Public health practitioners and researchers employ epidemio- logic methods for describing the health of the community, identifying causes of disease, and studying risks to individuals. One of the most important epidemio- logic applications is the study of the causality of disease; a detailed account of causality was provided. The chapter concluded with a review of primary, second- ary, and tertiary prevention of diseases.

Study Questions and Exercises

1. Define in your own words the following terms: a. secular changes b. operations research c. risk factor d. the natural history of disease e. demographic transition f. epidemiologic transition g. disorders: disappearing, residual, persisting, epidemic h. population pyramid

2. Name three approaches for prevention (primary, secondary, and tertiary) of each of the following health problems/conditions: a. motor vehicle accidents b. obesity c. hepatitis A d. hepatitis B and C e. foodborne illness on cruise ships f. mortality due to gang violence

51589_CH02_Printer.indd 101 09/02/13 12:09 PM

102 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

3. Apply the seven uses of epidemiology (as formulated by Morris1) to a public health issue (e.g., reduction of health disparities). For example, use of number two, “Diagnose the health of the community,” might involve identification of groups in the community that are at high risk for sexually transmitted diseases. Similarly, the remaining six uses could be applied to other aspects of health disparities. Are the uses of epidemi- ology defined in the chapter distinct or overlapping? Can you think of other uses of epidemiology not identified in the chapter? Do all of the uses belong exclusively to the domain of epidemiology?

4. Describe a role for epidemiology in the field of policy evaluation. Con- sider how the field of epidemiology might inform policy evaluation of laws that regulate tobacco consumption in public places.

5. How are the rules of evidence for criminality similar or different from the rules of evidence for disease causality? (Refer to Exhibit 2–5 to help with your answer.)

6. Clinicians and epidemiologists differ in their assessment of the impor- tance of risks. State how the clinical and epidemiologic approaches differ. Give an example by using a disease or condition that is important for society.

7. Describe how it is possible for an infectious disease, when it first comes to the attention of public health authorities, to be considered an extremely acute or fatal condition, and then later is found to be mild or benign in its most common form. Give an example of such a disease.

8. This chapter stated how epidemiology may be applied to the study of the causality of disease. Suggest other examples of how epidemiology might be applied to study the causality of disease.

9. The following questions refer to Table 2A–1. a. Calculate the percentage decline in the death rate for all causes. What

generalizations can be made about changes in disease rates that have occurred between 1900 and the present?

b. Contrast the changes in death rates due to cancer, heart disease, and cerebrovascular diseases. What additional information would be use- ful to specify better the changes in these conditions?

c. Note the decline in mortality for the four communicable diseases (1, 2, 3, and 10) since 1900. With the exception of pneumonia and influenza, these are no longer among the 10 leading causes of death. Can you speculate regarding how much of each is due to environmen- tal improvements and how much to specific preventive and curative practices?

51589_CH02_Printer.indd 102 09/02/13 12:09 PM

s t u d y Q u e s t i o n s a n d e x e r C i s e s 103

d. Among the 10 leading causes of death in 2009 were chronic lower respiratory diseases (44.7 per 100,000—rank 3), diabetes (22.3 per 100,000—rank 7), Alzheimer’s disease (25.7 per 100,000—rank 6), and suicide (11.9 per 100,000—rank 10). (Note: Data are not shown in Table 2A–1.) In 1900, these were not among the 10 leading causes of death. How do you account for these changes?

10. The following questions refer to Figure 2–3. a. List and describe the trends in death rates by the five leading causes

of death. b. Describe the trend for hypertension and Parkinson’s disease. Can you

suggest an explanation for the trends in hypertension and Parkinson’s disease deaths?

c. Does the curve for accidental deaths correspond to our expectations from various publicity reports?

d. What is the trend for Alzheimer’s disease? Can you offer an explanation?

Table 2A–1 Leading Causes of Death and Rates for Those Causes in 1900 and 2009, United States

Rate per 100,000 Population

Rank 1900 Cause of Death* 1900 20091

All causes 1,719.1 793.7 1 Influenza and pneumonia, except

pneumonia of newborn 202.2 17.5

2 Tuberculosis, all forms 194.4 NA† 3 Diarrhea and enteritis 139.9 NA† 4 Disease of heart 137.4 195.0 5 Cerebrovascular diseases 106.9 41.9 6 Chronic nephritis 81.0 15.9 7 Accidents and adverse effects2 72.3 38.2 8 Malignant neoplasms 64.0 185.2 9 Senility 50.2 NA† 10 Diphtheria 40.3 NA†

* Some categories may not be strictly comparable because of change in classification. † NA: These are no longer listed among the top 10 causes of death. 1 Crude death rate 2 Accidents (unintentional injuries)

Sources: U.S. Bureau of the Census. Statistical Abstract of the United States: 1957. Washington, D.C: U.S. Bureau of the Census; 1957: 69; U.S. Public Health Service. Vital Statistics Rates in the United States, 1900–1940. Washington, D. C.: U.S. Government Printing Office: 1947; Kochanek KD, Xu JQ, Murphy SL, et al. Deaths: Preliminary data for 2009. National vital statistics reports; 59(4):5. Hyattsville, MD: National Center for Health Statistics: 2011.

51589_CH02_Printer.indd 103 09/02/13 12:09 PM

104 C h a p t e r 2 p r a C t i C a l a p p l i C at i o n s o f e p i d e m i o l o g y

References

1. Morris JN. Uses of Epidemiology, 3rd ed. Edinburgh, UK: Churchill Livingstone; 1975.

2. Angell M. The interpretation of epidemiologic studies. Editorial. N Engl J Med. 1990;323:823–825.

3. MacMahon B, Pugh TF. Epidemiology Principles and Methods. Boston, MA: Little, Brown; 1970.

4. Susser MW, Watson W, Hopper K. Sociology in Medicine, 3rd ed. New York, NY: Oxford University Press; 1985.

5. Mausner JS, Kramer S. Epidemiology: An Introductory Text, 2nd ed. Philadelphia: Saunders; 1985.

6. Raymond J, Wheeler W, Brown MJ. Inadequate and unhealthy housing, 2007 and 2009. MMWR. 2011;60 (supplement):21–27.

7. Rothman KJ. Modern Epidemiology. Boston, MA: Little, Brown; 1986. 8. U.S. Department of Health and Human Services. Healthy People 2010: Understanding

and Improving Health, 2nd ed. Washington, DC: U.S. Government Printing Office, November 2000.

9. US Department of Health and Human Services. Office of Disease Prevention and Health Promotion. ODPHP Publication No. B0132 November 2010. www .healthypeople.gov.

10. US Department of Health and Human Services. HHS Action Plan to Reduce Racial and Ethnic Health Disparities: A Nation Free of Disparities in Health and Health Care. Washington, D.C.: US Department of Health and Human Services, April 2011.

11. Truman BI, Smith CK, Roy K, et al. Rationale for regular reporting on health dispari- ties and inequalities–United States. MMWR. 2011;60(supplement):3–10.

12. Miniño AM, Murphy SL, Xu JQ, Kochanek KD. Deaths: Final data for 2008. National Vital Statistics Reports. 2011;59(10). Hyattsville, MD: National Center for Health Statistics. 2011.

13. Ibrahim MA. Epidemiology and Health Policy. Gaithersburg, MD: Aspen Publishers; 1985.

14. Samet JM, Lee NL. Bridging the gap: Perspectives on translating epidemiologic evi- dence into policy. Am J Epidemiol. 2001;154(12 Suppl):S1–S3.

15. Weed DL, Mink PJ. Roles and responsibilities of epidemiologists. Ann Epidemiol. 2002;12:67–72.

16. Matanoski GM. Conflicts between two cultures: Implications for epide- miologic researchers in communicating with policy-makers. Am J Epidemiol. 2001;154(12 Suppl):S36–S42.

17. Spasoff RA. Epidemiologic Methods for Health Policy. New York, NY: Oxford University Press; 1999.

18. Porta M. A Dictionary of Epidemiology, 5th ed. New York: Oxford University Press; 2008.

19. Hulka BS. Epidemiological applications to health services research. J Community Health. 1978;4:140–149.

20. Woodwell DA. Office Visits to General Surgeons 1989–90, National Ambulatory Medical Care Survey. Advance Data from Vital and Health Statistics. Report 228. Hyattsville, MD: National Center for Health Statistics; December 22, 1993.

51589_CH02_Printer.indd 104 09/02/13 12:09 PM

r e f e r e n C e s 105

21. Fowke JH, Murff, HJ, Senorello LB, at all. Race and socioeconomic status are independently associated with benign prostatic hyperplasia (BPH). J Urol. 2008;180(5):2091–2096.

22. de la Torre A, Friis R, Hunter HR, Garcia L. The health insurance of U.S. Latino women: A profile from the 1982–1984 HHANES. Am J Public Health. 1996;86: 533–537.

23. Wilper AP, Woolhandler S, Lasser KE, et al. Health insurance and mortality in US adults. Am J Public Health. 2009;99:2289–95.

24. Epidemiology faces its limits. Science. 1995;269:164–169. 25. Susser M. Causal Thinking in the Health Sciences. New York, NY: Oxford University

Press; 1973. 26. Susser M and Stein Z. Eras in Epidemiology. New York, NY: Oxford University Press;

2009. 27. Cassel J. The contribution of the social environment to host resistance. Am J

Epidemiol. 1976;104:107–123. 28. Winkelstein W Jr. Invited commentary on “Judgment and causal inference: criteria in

epidemiologic studies.” Am J Epidemiol. 1995;141:699–700. 29. U.S. Department of Health, Education and Welfare, Public Health Service. Smoking

and Health, Report of the Advisory Committee to the Surgeon General of the Public Health Service. Public Health Service publication 1103. Washington, DC: Government Printing Office; 1964.

30. Susser M. Judgment and causal inference: Criteria in epidemiologic studies. Am J Epidemiol. 1977;105:1–15.

31. Hill AB. The environment and disease: Association or causation? Proc R Soc Med. 1965;58:295–300.

32. Evans AS. Causation and disease: A chronological journey. Am J Epidemiol. 1978;108:254–255.

33. Brett AS. Treating hypercholesterolemia: How should practicing physicians interpret the published data for patients? N Engl J Med. 1989;321:676–680.

34. Pekkanen J, Linn S, Heiss G, et al. Ten-year mortality from cardiovascular disease in relation to cholesterol level among men with and without preexisting cardiovascular disease. N Engl J Med. 1990;322:1700–1707.

35. Howlader N, Noone AM, Krapcho M, et al. (eds). SEER Cancer Statistics Review, 1975–2009 (Vintage 2009 Populations), National Cancer Institute. Bethesda, MD.

36. Murphy M, Hultgren HN, Detre K, et al. Treatment of chronic stable angina: A preliminary report of survival data of the randomized Veterans Administration Cooperative Study. N Engl J Med. 1977;297:621–627.

37. Leavell HR, Clark EG. Preventive Medicine for the Doctor in His Community: An Epidemiologic Approach, 3rd ed. New York: McGraw-Hill Book Company; 1965.

51589_CH02_Printer.indd 105 09/02/13 12:09 PM

51589_CH02_Printer.indd 106 09/02/13 12:09 PM

107

3Chapte r

Measures of Morbidity and

Mortality Used in Epidemiology

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● define and distinguish among ratios, proportions, and rates ●● explain the term population at risk ●● identify and calculate commonly used rates for morbidity, mortality,

and natality ●● state the meanings and applications of incidence rates and prevalence ●● discuss limitations of crude rates and alternative measures for

crude rates ●● apply direct and indirect methods to adjust rates ●● explain when either direct or indirect rate adjustment should be used

CHAPTER OUTLINE

I. Introduction II. Definitions of Count, Ratio, Proportion, and Rate

III. Risk Versus Rate; Cumulative Incidence IV. Interrelationship Between Prevalence and Incidence V. Applications of Incidence Data

VI. Crude Rates VII. Specific Rates and Proportional Mortality Ratio

VIII. Adjusted Rates IX. Conclusion X. Study Questions and Exercises

51589_CH03_Printer.indd 107 09/02/13 12:19 PM

108 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Introduction

The 2009 H1N1 influenza pandemic illustrated how a potentially deadly virus could spread rapidly from the United States to other countries worldwide. At one point in the growing epidemic, public health officials pondered whether the 2009 pandemic was a repeat of the 1918 “killer” flu. When a flu outbreak occurs, what quantitative measures inform public health professionals that an epidemic caused by a killer virus is occurring? How fast is the virus spreading? How many deaths is the potentially virulent and lethal agent causing? In order to answer questions such as these, the work of the epidemiologist involves enu- merating cases of diseases and health-related phenomena as well as describing the occurrence and patterns of disease in the population. Epidemiology exam- ines risk factors associated with adverse health outcomes and identifies potential causal associations between exposures and diseases.

This chapter explains disease occurrence measures used commonly in public health practice for quantifying health outcomes. The foundation of studies designed to identify etiology, monitor trends, and evaluate public health interventions rests on the bedrock of our ability to measure the occurrence of morbidity and mortality care- fully and accurately. This chapter defines four categories of epidemiologic measures (counts, ratios, proportions, and rates), differentiates between the concepts of risk and rate, discusses relationships among measures, and illustrates their applications.

Definit ions of Count, Ratio, Proport ion, and Rate

The four types of epidemiologic measures covered in this section are counts, ratios, proportions, and rates. Refer to Figure 3–1 for an overview of the mea- sures discussed in this chapter. The figure identifies the measures and indicates their hierarchy and interrelationships.

Count The simplest and most frequently performed quantitative measure in epidemi- ology is a count. As the term implies, a count refers merely to the number of cases of a disease or other health phenomenon being studied. Several examples of counts are the number of:

●● cases of influenza reported in Westchester County, New York, during January of a particular year

●● traffic fatalities in the borough of Manhattan during a 24-hour period

51589_CH03_Printer.indd 108 09/02/13 12:19 PM

d e f i n i t i o n s o f C o u n t , r a t i o , p r o p o r t i o n , a n d r a t e 109

●● participants screened positive in a hypertension screening program organized by an industrial plant in northern California

●● college dorm residents who had mono ●● stomach cancer patients who were foreign born

Ratio A ratio is defined as “[t]he value obtained by dividing one quantity by another. RATE, PROPORTION, and PERCENTAGE are types of ratios.”1 A ratio therefore consists of a numerator and a denominator. The most general form of a ratio does not necessarily have any specified relationship between the numerator and denominator. A ratio may be expressed as follows: ratio = X/Y. An example of a ratio is the sex ratio, which is shown in three variations:

1. Simple sex ratio: Of 1,000 motorcycle fatalities, 950 victims are men and 50 are women. The sex ratio for motorcycle fatalities is:

Figure 3–1 Overview of epidemiologic measures.

Epidemiologic Measures

RatioCount

Proportion

Prevalence

Point Prevalence Period Prevalence

Incidence Rate (Attack Rate)

Rate (Crude, Specific, Adjusted)

Number of male cases

Number of female cases

950

50 19:1 male to female= =

2. Demographic sex ratio: This ratio refers to the number of males per 100 females. In the United States (2010), the sex ratio for the entire pop- ulation was 96.7, indicating more females than males.

Sex ratio Number of males

Number of females 100

151,781,326

156,964,212 100 96.7= × = × =

51589_CH03_Printer.indd 109 09/02/13 12:19 PM

110 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

3. Sex ratio at birth: the sex ratio at birth is defined as: (the number of male births divided by the number of female births) multiplied by 1,000.

Sex ratio at birth Number of male births

Number of female births 1,000= ×

Figure 3–2 shows that between 1940 and 2002, the sex ratio at birth exceeded 1.0 and made significant transitions in 1942, 1959, and 1971. Nevertheless, the sex ratio trended downward since 1940.

Proportion A proportion is a type of ratio in which the numerator is part of the denominator; proportions may be expressed as percentages. Let us consider how a proportion can be helpful in describing health issues by reexamining a count. For a count to be descriptive of a group, it usually should be seen relative to the size of the group. Suppose there were 10 college dorm residents who had hepatitis. How large a problem did these 10 cases represent? To answer this question, one would need to know whether the dormitory housed 20 students or 500 students. If there were only 20 students, then 50% (or 0.50) were ill. Conversely, if there were 500 students in the dormitory, then only 2% (or 0.02) were ill. Clearly,

Figure 3–2 Sex ratio at birth, 1940–2002. Source: Reproduced from TJ Mathews, BE Hamilton.Trend Analysis of the Sex Ratio at Birth in the United States. National Vital Statistics Reports, Vol. 55, No. 20, p. 1. Hyattsville, MD: National Center for Health Statistics; 2005.

Join point sex ratio trend

Year

NOTES: Sex ratio at birth is the number of male births divided by the number of female births multiplied by 1,000.

S ex

r at

io a

t b ir

1940 0

1,048

1,052

1,056

1,060

1,044

1950 1960 1970 1980 1990 2002

Observed sex ratio trend

51589_CH03_Printer.indd 110 09/02/13 12:19 PM

d e f i n i t i o n s o f C o u n t , r a t i o , p r o p o r t i o n , a n d r a t e 111

these two scenarios paint a completely different picture of the magnitude of the problem. In this situation, expressing the count as a proportion is indeed helpful.

Table 3–1 illustrates the calculation of the proportion of African-American male deaths among African-American and white boys aged 5–14 years.

In most situations, it will be informative to have some idea about the size of the denominator. Although the construction of a proportion is straightfor- ward, one of the central concerns of epidemiology is to find and enumerate appropriate denominators to describe and compare groups in a meaningful and useful way.

The previous discussion may leave the reader with the impression that counts, in and of themselves, are of little value in epidemiology; this is not true, however. In fact, case reports of patients with particularly unusual presentations or com- binations of symptoms often spur epidemiologic investigations. In addition, for some diseases even a single case is sufficient to be of public health importance. For example, if a case of smallpox or Ebola virus were reported, the size of the denominator would be irrelevant. That is, in these instances a single case, regard- less of the size of the population at risk, would stimulate an investigation.

Rate A rate also is a type of ratio; however, a rate differs from a proportion because the denominator involves a measure of time. The numerator consists of the fre- quency of a disease over a specified period of time, and the denominator is a unit size of population (Exhibit 3–1). It is critical to remember that to calculate a rate, two periods of time are involved: the beginning of the period and the end of the period.

Medical publications may use the terms ratio, proportion, and rate without strict adherence to the mathematical definitions for these terms. Hence, one must be alert to how a measure is defined and calculated.2 In the formula shown in Exhibit 3–1, the denominator also is termed the reference population and by

Table 3–1 Calculation of the Proportion of African-American Male Deaths Among African-American and White Boys Aged 5 to 14 Years

A B Total (A + B)

Number of deaths among African-American boys

Number of deaths among white boys

Total

1,150 3,810 4,960

Proportion = A/(A + B) × 100 = (1,150/4,960) × 100 = 23.2%

51589_CH03_Printer.indd 111 09/02/13 12:19 PM

112 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

definition is the population from which cases of a disease have been taken. For example, in calculating the annual death rate (crude mortality rate) in the United States, one would count all the deaths that occurred in the country during a certain year and assign this value to the numerator. The value for the denomina- tor would be the size of the population of the country during a particular year. The best estimate of the population would probably be the population around the midpoint of the year, if such information could be obtained. Referring to Exhibit 3–1, one calculates the U.S. crude mortality rate as 803.6 per 100,000 persons for 2007.

Rates improve one’s ability to make comparisons, although they also have lim- itations. Rates of mortality or morbidity for a specific disease (see the section on cause-specific mortality rates later in this chapter) reduce that standard of com- parison to a common denominator, the unit size of population. To illustrate, the U.S. crude death rate for diseases of the heart in 2003 was 235.6 per 100,000.

rate Calculation

Rate: A ratio that consists of a numerator and a denominator and in which time forms part of the denominator.

Epidemiologic rates contain the following elements:

●● disease frequency ●● unit size of population ●● time period during which an event occurs

Example:

Crude death rate Number of deaths in a given year

Reference population (during midpoint of the year)

100,000= ×

(Either rate per 1,000 or 100,000 is used as the multiplier)

Calculation problem (crude death rate in the United States):

Number of deaths in the United States during 2007 = 2,423,712 Population of the United States as of July 1, 2007 = 301,621,157

Crude death rate 2,423,712

301,621,157 100,000= × = 8003 6. per 100,000

e x

h ib

it 3

–1

51589_CH03_Printer.indd 112 09/02/13 12:19 PM

d e f i n i t i o n s o f C o u n t , r a t i o , p r o p o r t i o n , a n d r a t e 113

One also might calculate the heart disease death rate for geographic subdivisions of the country (also expressed as frequency per 100,000 individuals). These rates could then be compared with one another and with the rate for the United States to judge whether the rates found in each geographic area are higher or lower. For example, the crude death rates for diseases of the heart in New York and Texas were 288.0 and 188.9 per 100,000, respectively. It would appear that the death rate is higher in New York than in Texas based on the crude death rates. This may be a specious conclusion, however, because there may be important differences in population composition (e.g., age differences between popula- tions) that would affect mortality experience. Later in this chapter, the procedure to adjust for age differences or other factors is discussed.

Rates can be expressed in any form that is convenient (e.g., per 1,000, per 100,000, or per 1,000,000). Many of the rates that are published and routinely used as indicators of public health are expressed in particular conventions. For example, cancer rates are typically expressed per 100,000 population, and infant mortality is expressed per 1,000 live births. One of the determinants of the size of the denominator is whether the numerator is large enough to permit the rate to be expressed as an integer or an integer plus a trailing decimal (e.g., 4 or 4.2). For example, it would be preferable to describe the occurrence of disease as 4 per 100,000 (or 4.2 per 100,000) rather than 0.04 per 1,000 (or 0.042 per 1,000), even though both are perfectly correct. Throughout this chapter, the multiplier for a given morbidity or mortality statistic is provided.

Exhibit 3–2 describes the Iowa Women’s Health Study (IWHS). The data col- lected illustrate the various measures of disease frequency defined in this chapter.

Prevalence The term prevalence refers to the number of existing cases of a disease or health condition in a population at some designated time.1 As shown in Figure 3–3, prevalence is analogous to water that has collected in a pool at the base of a waterfall. Prevalence data provide an indication of the extent of a health problem and thus may have implications for the scope of health services needed in the community. Prevalence can be expressed as a number, a percentage, or number of cases per unit size of population. Consider three examples: The prevalence of diarrhea in a children’s camp on July 13 was 15, the prevalence of phenylke- tonuria-associated mental disabilities in institutions for the developmentally dis- abled was 15%, and the prevalence of obesity among women aged 55–69 years was 367 per 1,000. These examples illustrate that the designated time can be specified (e.g., one day) or unspecified. When the time period is unspecified,

51589_CH03_Printer.indd 113 09/02/13 12:19 PM

114 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

the iowa Women’s health Study

The IWHS is a longitudinal study of mortality and cancer occur- rence in older women.3,4 The state of Iowa was chosen as the site of this study because of the availability of cancer incidence and mor- tality data from the State Health Registry of Iowa. This registry is a participant in the National Cancer Institute’s Surveillance, Epide- miology, and End Results Program. The sample was selected from

a January 1985 current drivers list obtained from the Iowa Department of Transportation. The list contained the names of 195,294 women aged 55 to 69 and represented approximately 94% of the women in the state of Iowa in this age range.

In December 1985, a 50% random sample of the eligible women was selected, yielding 99,826 women with Iowa mailing addresses. A 16-page health history questionnaire was mailed on January 16, 1986, followed by a reminder postcard one week later and a follow-up letter four weeks later; a total of 41,837 women responded. Information was collected about basic demographics, medical history, reproductive history, personal and family history of cancer, usual dietary intake, smoking and exercise hab- its, and medication use. A paper tape measure also was provided along with detailed instructions for the subject to record selected body measure- ments: height, weight, and circumferences of the waist and hips.

The primary focus of the study was to determine whether distribution of body fat centrally (i.e., around the waist) rather than peripherally (i.e., on the hips) is associated with increased risk of cancer. The occurrence of can- cer was determined by record linkage with the State Health Registry. A com- puter program was used to match new cancer cases in the registry with study participants on name, ZIP code, birth date, and Social Security number. n

e x

h ib

it 3

–2

prevalence usually implies a particular point in time. More specifically, these examples refer to point prevalence.

A second type of prevalence measure is period prevalence, which denotes the total number of cases of a disease that exist during a specified period of time, for instance a week, month, or longer time interval. To determine the period prevalence, one must combine the number of cases at the beginning of the time interval (the point prevalence) with the new cases that occur during the inter- val. Because the denominator may have changed somewhat (the result of people

51589_CH03_Printer.indd 114 09/02/13 12:19 PM

d e f i n i t i o n s o f C o u n t , r a t i o , p r o p o r t i o n , a n d r a t e 115

Figure 3–3 Analogy of prevalence and incidence. The water flowing down the waterfall symbolizes incidence and the water collecting in the pool at the base symbolizes prevalence.

Point prevalence Number of persons ill

Total number in the group at a time point=

Example: In the IWHS, respondents were asked: “Do you smoke ciga- rettes now?” The total number in the group was 41,837. The total num- ber who responded yes to the smoking question was 6,234. Therefore, the prevalence of current smokers in the IWHS on January 16,1985, was 6,234/41,837. This result could be expressed as a percentage (14.9%) or as a frequency per 1,000 (149.0).

51589_CH03_Printer.indd 115 09/02/13 12:20 PM

116 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

entering or leaving during the period of observation), one typically refers to the average population. Note that for period prevalence, cases are counted even if they die, migrate, or recur as episodes during the period.

A second example is from the National Center for Health Statistics— the National Health and Examination Survey, United States, 2009–2010. Figure 3–4 shows the prevalence of obesity among adults 20 years of age or older. Obesity was defined as a body mass index (BMI) of 30 or greater. The prevalence (percentage) was 58.5% among non-Hispanic black women.

Technically speaking, both point and period prevalence are proportions. As such, they are dimensionless and should not be described as rates, a mistake that is commonly made. To illustrate the distinction between point and period preva- lence, consider as an example the issue of homelessness in the United States. The conditions surrounding homelessness present a serious public health prob- lem, particularly in the control of infectious diseases and the effect on home- less persons’ physical and mental health. Consequently, public officials have a legitimate need to estimate the magnitude of the problem, an issue that has produced intense debate. Surveys of currently homeless people pose extremely challenging methodologic difficulties that have led some authorities to believe that point prevalence may lead to serious underreporting. According to Link and colleagues, “The first problem is finding people who are currently homeless. Surveys may miss the so-called hidden homeless, who sleep in box cars, on the roofs of tenements, in campgrounds, or in other places that researchers cannot effectively search. [Even if located] . . . respondents may refuse to be interviewed or deliberately hide the fact that they are homeless.”5(p 1907) People who experi- ence relatively short or intermittent episodes of literal homelessness are likely

Period prevalence Number of persons ill

Average population during a time period=

Example: In the IWHS, women were asked: “Have you ever been diag- nosed by a physician as having any form of cancer, other than skin cancer?” Note that the question did not ask about current disease but rather about the lifetime history. Thus, it refers to period prevalence, the period being the entire life span. To calculate the period preva- lence, one needs to know the average population (still 41,837) and the number who responded yes to the question (2,293). Therefore, the period prevalence of cancer in the study population was 2,293/41,837, or 5.5%.

51589_CH03_Printer.indd 116 09/02/13 12:20 PM

d e f i n i t i o n s o f C o u n t , r a t i o , p r o p o r t i o n , a n d r a t e 117

to be missed in brief surveys. To address these problems, Link et al.5 conducted a national household telephone survey to provide lifetime and five-year period prevalence estimates. They found that 14% of the sample had ever been home- less, 4.6% in the last five years. Compared with previous estimates based on point prevalence, the investigators concluded that the magnitude of the problem was much greater than previous estimates had indicated.

Prevalence studies are useful in describing the health burden of a population and in allocation of health resources, such as facilities and personnel. The forego- ing data on the prevalence of smoking, obesity, and homelessness were illustra- tions. Also, epidemiologists use prevalence data to estimate the frequency of an exposure in a population. They can survey a sample of respondents in order to determine the types of exposures (e.g., use of drugs, medications, or other types of exposures) they have had; in other cases environmental researchers can make direct measures of toxic contaminants through environmental monitoring.

Figure 3–4 Prevalence of obesity* among adults aged ≥ 20 years, by race/ethnicity and sex—National Health and Nutrition Examination Survey, United States, 2009–2010. Source: Reproduced from Centers for Disease Control and Prevention. QuickStats. MMWR. Vol 61, No. 7, p. 130, February 24, 2012.

*Defined as a body mass index (weight [kg]/height [m]2) ≥ 30.

Men Women

Total§ Total§White, non-

Hispanic

Black, non-

Hispanic

Race/Ethnicity

† 95% confidence interval. § Includes other races (i.e., Asians and American Indians/Alaska natives) not shown seperately because of small sample sizes, which affect reliability of estimates.

P er

ce nt

ag e

White, non-

Hispanic

Black, non-

Hispanic

51589_CH03_Printer.indd 117 09/02/13 12:20 PM

118 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Typically, prevalence studies are not as helpful as other types of epidemiologic research designs for studies of etiology. Among several reasons, the most impor- tant is the possible influence of differential survival. That is, for a case to be included in a prevalence study, he or she would have had to survive the disease long enough to participate. Cases that died before participation would obviously be missed, resulting in a truncated sample of eligible cases. Risk factors for rap- idly fatal cases may be quite different from risk factors for less severe manifes- tations. One situation in which the use of prevalent cases may be justified for studies of disease etiology arises when a condition has an indefinite time of onset, such as occurs with mental disorders.2

Incidence Rate Incidence is defined as “[t]he number of instances of illnesses commencing, or of persons falling ill, during a given period in a specified population. More generally, the number of new health-related events in a defined population within a speci- fied period of time. It may be measured as a frequency count, a rate, or a propor- tion.”1 Incidence is a measure of the risk of a specified health-related event. (We will explain this concept later in the chapter.) In Figure 3–3 incidence is analo- gous to water flowing in the waterfall (new cases). An example of incidence mea- sured as a frequency is the number of new cases of HIV infection diagnosed in a population in a given year: A total of 164 HIV diagnoses were reported among American Indians or Alaska natives in the United States during 2009.

The term incidence rate describes the rate of development of a disease in a group over a certain time period; this period of time is included in the denomi- nator. An incidence rate (Exhibit 3–3) includes three important elements:

1. a numerator: the number of new cases 2. a denominator: the population at risk 3. time: the period during which the cases accrue

Number of New Cases The incidence rate uses the frequency of new cases in the numerator. This means that individuals who have a history of the disease are not included.

Population at Risk The denominator for incidence rates is the population at risk. One therefore should exclude individuals who have already developed the disease of inter- est (e.g., those who have had heart attacks) or are not capable of developing

51589_CH03_Printer.indd 118 09/02/13 12:20 PM

d e f i n i t i o n s o f C o u n t , r a t i o , p r o p o r t i o n , a n d r a t e 119

the disease. For example, if one wanted to calculate the rate of ovarian cancer in the IWHS, women who had had their ovaries removed (oophorectomized women) should be excluded from the cohort at risk. It is not uncommon, how- ever, to see some incidence rates based on the average population as the denomi- nator rather than the population at risk. This distinction really must be made for those infectious diseases that confer lifetime immunity against recurrence. Regarding chronic diseases to which most people appear to be susceptible, the distinction is less critical. The population at risk may include those exposed to a disease agent or unimmunized or debilitated people, or it may consist of an entire population (e.g., a county, a city, or a nation). The population at risk may represent special risk categories; occupational injury and illness incidence rates are calculated for full-time workers in various occupations, for example, because these are the populations at risk.

incidence rate

Incidence rate

Number ofnewcases over a time p= eeriod

Total population at risk duringthe same ttime period

Multiplier (e.g.,100,000)

The denominator consists of the population at risk (i.e., those who are at risk for contracting the disease).

Example: Calculate the incidence rate of postmenopausal breast cancer in the IWHS. The population at risk in this example would not include women who were still premenopausal (n = 569), women who had had their breasts surgically removed (n = 1,870), and women with a previous diagnosis of cancer (n = 2,293). Thus, the denominator is 37,105 women. After eight years of follow-up, 1,085 cases were identified through the State Health Registry. The incidence rate is therefore 1,085/37,105 per eight years. To express this rate per 100,000 population: Divide 1,085 by 37,105 (answer: 0.02924). This is the rate over an eight-year period. For the annual rate, divide this number by eight years (answer: 0.003655) and multiply by 100,000. n

Answer: 365.5 cases of postmenopausal breast cancer per 100,000 women per year.

e x

h ib

it 3

–3

51589_CH03_Printer.indd 119 09/02/13 12:20 PM

120 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Specification of a Time Period The definition of incidence entails the designation of a time period, such as a week, a month, a year, or a multiyear time period. To determine an incidence rate, one must be able to specify the date of onset for the condition during the time period. Some acute conditions (e.g., a severe stroke or an acute myocardial infarction) may have a readily identifiable time of onset. Other conditions (e.g., cancer) may have an indefinite time of onset, which is defined by the initial definitive diagnosis date for the disease.6

Attack Rate The attack rate (AR) is an alternative form of the incidence rate that is used when the nature of the disease or condition is such that a population is observed for a short time period, often as a result of specific exposure.2 In reporting outbreaks of salmonella infection or other foodborne types of gastroenteritis, epidemiolo- gists employ the AR. The formula for the AR is:

AR = Ill/(Ill + well) × 100 (during a time period)

Calculation example: a total of 87 people at a holiday dinner ate roast turkey. Among these persons, 63 who consumed roast turkey became ill; the remainder did not become ill.

AR (for the roast turkey) = 63/(63 + 24) × 100 = 72.4%

As shown in this formula, the numerator consists of people made ill as a result of exposure to the suspected agent, and the denominator consists of all people, whether well or ill, who were exposed to the agent during a time period. Strictly speaking, the AR is not a true rate because the time dimension is often uncertain or specified arbitrarily.

Although the AR often is used to measure the incidence of disease during acute infectious disease epidemics, it also may be used for the incidence of other conditions where the risk is limited to a short time period or the etiologic fac- tors operate only within certain age groups. An example is hypertrophic pyloric stenosis (a blockage from the stomach to the intestines), which occurs predomi- nantly in the first three months of life and is practically unknown after the age of six months.

51589_CH03_Printer.indd 120 09/02/13 12:20 PM

r i s k V e r s u s r a t e ; C u M u l a t i V e i n C i d e n C e 121

Risk Versus Rate; Cumulative Incidence

Epidemiologists have been known to use the terms risk and rate interchangeably. However, if pressed to explain the difference, they would be able (one hopes) to identify several key distinctions. First, risk is a statement of the probability or chance that an individual will develop a disease over a specified period, condi- tioned on that individual’s not dying from any other disease during the period.7 As such, risk ranges from 0 to 1 and is dimensionless. Statements of risk also require a specific reference period, for example, the five-year risk of developing asthma.

Cumulative Incidence Cumulative incidence refers to “[t]he number . . . of a group (cohort) of people who experience the onset of a health-related event during a specified time inter- val.”1 Cumulative incidence is used when all individuals in the population are thought to be at risk of the health-related event being investigated, as in a pro- spective cohort study in which the population is fixed. The cumulative incidence estimates the risk of a particular health-related outcome in the cohort. If it is possible to follow up every individual in the cohort during a given time period, then the cumulative incidence is the number of events that occur during that time period expressed relative to the denominator. (However, as we will describe later, a problem arises in determining cumulative incidence and incidence when individuals are observed for different periods of time.)

The illustration regarding the incidence of postmenopausal breast cancer in the IWHS is an example of a cumulative incidence. Because the population is fixed, no individuals are allowed to enter the denominator after the start of the observation period, and the numerator can include only individuals who were members of that fixed population. Calculation of cumulative incidence also requires that disease status be determined for everyone in the denominator. That is, once a group of individuals is selected for follow-up for disease occurrence, subsequent information about the occurrence of disease is obtained for everyone selected, which is difficult to achieve even in the best of circumstances. Most of the regions where we live and work contain dynamic populations; people move into and out of the area. Some individuals who were not in the study popula- tion at the baseline period may move into the region and become ill. Thus, the numerator has increased but the denominator has not. Conversely, if an indi- vidual moves away and then develops the disease, he or she would be counted in the denominator but not in the numerator. One solution to the problem of

51589_CH03_Printer.indd 121 09/02/13 12:20 PM

122 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

geographic mobility and loss to follow-up is to use rates as an indicator of risk. A simple perspective is that groups with high rates of disease are at greater risk than are groups with low rates of disease. The issue is a bit more complicated than that perspective (and beyond the scope of this text). The main caveat is that rates can be used to estimate risk only when the period of follow-up is short and the rate of disease over that interval is relatively constant. Thus, to estimate small risks, one simply multiplies the average rate times the duration of follow-up.8

Incidence Density The incidence density is “[t]he average person-time incidence rate.”1 This varia- tion in the incidence rate is calculated by using the person-time of observation as the denominator. Person-time “. . . is the sum of the periods of time at risk for each of the subjects. The most widely used measure is person-years.”1 Person- time is used when the amounts of time of observation of each of the subjects in the study varies instead of remaining constant for each subject.

Here is an example of how incidence density becomes useful. A special prob- lem occurs when a population or study group is under observation for different lengths of time. This may occur for a variety of reasons, including attrition or dropout, mortality, or development of the disease under study. An illustration is the calculation of the incidence of postmenopausal breast cancer in the IWHS. Although the study was able to identify all cancers diagnosed within the state, some women may have moved out of state after the initial questionnaire admin- istration. Any cancers diagnosed among these women would be unknown to the investigators. Other women died before the end of the follow-up period. In the previous calculation, we merely counted the number of cases over the 8-year period of follow-up (n = 1,042) and divided by the number of women at risk (n = 37,105). The implicit assumption of this calculation is that each of the 37,105 women was “observed” for the full 8-year period. Clearly, this could not be the case. To allow for varying periods of observation of the sub- jects, one uses a modification of the formula for the incidence rate in which the denominator becomes person-time of observation. Incidence density is defined in Exhibit 3–4.2 An example of how to calculate person-years, the most com- mon measure of person-time, is shown in Table 3–2.

In Table 3–2, person-years were derived simply by summing the product of each category of length of observation and the number of subjects in the cate- gory. A more difficult issue is how one actually determines the length of observa- tion for each individual. Visiting again the IWHS example, a computer program

51589_CH03_Printer.indd 122 09/02/13 12:20 PM

r i s k V e r s u s r a t e ; C u M u l a t i V e i n C i d e n C e 123

Table 3–2 Person-Years of Observation for Hypothetical Study Subjects in a 10-Year Heart Disease Research Project

A Number of Subjects

B Length of Observation (Years)

A ë B Person-Years

30 10 3000 10 9 90

7 8 56 2 7 14 1 1 1

Total 50 461

Number of health events (heart attacks) observed during the 10-year period: 5.

Incidence density = (5/461) × 100 = 1.08 per 100 person-years of observation.

incidence Density

Incidence density Number of new cases during the time period

Total person-time of observation =

When the period of observation is measured in years, the formula becomes:

Incidence density Number of new cases during the time period

Total person-years of observation =

Example: In the IWHS, the 37,105 women at risk for postmenopausal breast cancer contributed 276,453 person-years of follow-up. Because there were 1,085 incident cases, the rate of breast cancer using the incidence density method is 1,085/276,453 = 392.5 per 100,000 per year. Note that had each woman been followed for the entire eight-year period of follow-up, the total person-years would have been 296,840. Because the actual amount of follow-up was 20,000 person-years less than this, the estimated rate of breast cancer was higher (and more accurate) using the incidence density method. n

e x

h ib

it 3

–4

was used to tabulate, for each individual, the amount of time that elapsed from receipt of the mailed questionnaire until the occurrence of one of the following events (listed in order of priority): breast cancer diagnosis, death (if in Iowa), a move out of Iowa (if known through the National Change of Address Service), midpoint of interval between date of last contact and December 30, 1993, or

51589_CH03_Printer.indd 123 09/02/13 12:20 PM

124 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

midpoint of interval between date of last contact and date of death (for deaths that occurred out of Iowa, identified through the National Death Index). Women who did not experience any of these events were assumed to be alive in Iowa and contributed follow-up until December 30, 1993. This real-life example illus- trates that actual computation of person-years, although conceptually straight- forward, can be a fairly complicated procedure.

Interrelat ionship Between Prevalence and Incidence

Interrelationship: P ≅ ID

The prevalence (P) of a disease is proportional to the incidence rate (I) times the duration (D) of a disease.

For conditions of short duration and high incidence, one may infer from this formula that, when the duration of a disease becomes short and the incidence is high, the prevalence becomes similar to incidence. For diseases of short duration, cases recover rapidly or are fatal, eliminating the build-up of prevalent cases. In fact that is the case for infectious diseases of short duration, such as the com- mon cold.

Typically chronic diseases have a low incidence and, by definition, long dura- tion; as the duration of the disease increases, even though incidence is low or stable, the prevalence of the disease increases relative to incidence. An example is HIV/AIDS prevalence as shown in Figure 3–5. The line for HIV prevalence is much higher than the line for HIV incidence and shows an increasing trend. The explanation is that the prevalence of HIV is increasing gradually (about 1.1 million cases in 2006); however, the annual incidence of HIV in the United States has remained stable (slightly fewer than about 60,000 cases each year).

Figure 3–6 illustrates a second example of the relationship between incidence and prevalence. Suppose that there is an outbreak of meningococcal disease in a summer school class of 10 students. The frequency of the disease is recorded for 2 weeks. Individual cases plotted by the duration of each case for the period July 1–July 14 are shown in Figure 3–6. For the 10-day period (July 5–July 14), the period prevalence of meningococcal disease was 8/10; the point prevalence of disease on July 5 was 5/10. Because the disease in this example is one that can affect individuals more than once (no lifetime immunity after initial infection),

51589_CH03_Printer.indd 124 09/02/13 12:20 PM

i n t e r r e l a t i o n s h i p b e t w e e n p r e V a l e n C e a n d i n C i d e n C e 125

Figure 3–6 Outbreak of meningococcal infections in a summer school class of 10 students. Note: Students H, I, and J were not ill.

July 5 July 14

S tu

de nt

Day of the Week—SU M TU W TH F SA SU M TU W TH F SA

Duration of illness

Figure 3–5 HIV incidence and prevalence, United States, 1977–2006. Source: Adapted and reprinted from HIV and AIDS in the United States. Available at: http://www.cdc.gov/hiv/topics/surveillance/print/ united_states.htm. Accessed August 25, 2012.

HIV Prevalance (People Living with HIV/AIDS)

HIV Incidence (New HIV Infections)

Year

200,00

400,00

600,00

800,00

1,000,00

1,200,00

19 77

19 78

19 79

19 80

19 82

19 83

19 84

19 85

19 86

19 87

19 88

19 89

19 90

19 91

19 92

19 93

19 94

19 95

19 96

19 97

19 98

19 99

20 00

20 01

20 02

20 03

20 04

20 05

20 06

19 81

51589_CH03_Printer.indd 125 09/02/13 12:20 PM

126 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

the incidence rate of disease was 3/10. Note that on July 5 cases A, B, C, D, and F were existing cases of disease and were not included in the count for incidence; subsequently, case A was a recurrent case and should be counted once for inci- dence and twice for period prevalence. The measure of incidence would be more accurate if the cumulative duration of observation (person-days) was used in the denominator. If one was interested only in the first occurrence of meningococcal disease, then students A, B, C, D, and F would not have been included in the estimation of incidence, because they were prevalent cases on July 5. In that situ- ation, the incidence would have been 2/5.

Applicat ions of Incidence Data

It was noted earlier that prevalence data are useful for determining the extent of a disease (particularly chronic diseases) or health problem in the community. Prevalence data are not as helpful as incidence data for studies of etiology because of the possible influence of differential survival. (The prevalent cases may be the survivors who remain after the other cases died; consequently, the prevalent cases may represent an incomplete picture of the outcome variable.) Incidence data (e.g., cumulative incidence rates) help in research on the etiology of disease because they provide estimates of risk of developing the disease. Thus, incidence rates are considered to be fundamental tools in research that pursues the cau- sality of diseases. Note how the incidence rate of postmenopausal breast can- cer was calculated in the IWHS. Comparison of incidence rates in population groups that differ in exposures permits one to estimate the effects of exposure to a hypothesized factor of interest. This study design, known as a cohort study, differs from a prevalence study in that it selects participants who have a specific kind of exposure (e.g., exposure to a toxic chemical).

Crude Rates

The basic concept of a rate can be broken down into three general categories: crude rates, specific rates, and adjusted rates. Crude rates are summary rates based on the actual number of events in a population over a given time period. An example is the crude death rate, which approximates the proportion of a popula- tion that dies during a time period of interest.1 Refer to the study questions and exercises at the end of this chapter for calculation problems. Some of the more commonly used crude rates are presented in Exhibit 3–5. The definitions for measures of natality (statistics associated with births) come from Health, United States, 2010.9

51589_CH03_Printer.indd 126 09/02/13 12:20 PM

C r u d e r a t e s 127

examples of Crude rates: Overview of Measures that pertain to birth, Fertility, infant Mortality, and related phenomena

●● Crude birth rate: used to project population changes; it is affected by the number and age composition of women of childbearing age.

●● Fertility rate: used for comparisons of fertility among age, racial, and socioeconomic groups.

●● Infant mortality rate: used for international comparisons; a high rate indicates unmet health needs and poor environmental conditions.

●● Fetal death rate (and late fetal death rate): used to estimate the risk of death of the fetus associated with the stages of gestation.

●● Fetal death ratio: provides a measure of fetal wastage (loss) relative to the number of live births.

●● Neonatal mortality rate: reflects events happening after birth, primarily: 1. Congenital malformations 2. Prematurity (birth before gestation week 28) 3. Low birth weight (birth weight less than 2,500 g)

●● Postneonatal mortality rate: reflects environmental events, control of infectious diseases, and improvement in nutrition. Since 1950, neonatal mortality in the United States has declined; postneonatal mortality has not declined greatly.

●● Perinatal mortality rate: reflects events that occur during pregnancy and after birth; it combines mortality during the prenatal and post- natal periods.

●● Maternal mortality rate: reflects healthcare access and socioeconomic factors; it includes maternal deaths resulting from causes associated with pregnancy and puerperium (during and after childbirth). n

e x

h ib

it 3

–5

Birth Rate The crude birth rate refers to the number of live births during a specified period of time (e.g., one calendar year) per the resident population during the midpoint of the time period (expressed as rate per 1,000). The crude birth rate is a use- ful measure of population growth and is an index for comparison of developed and developing countries. The crude birth rate is generally higher in less devel- oped areas than in more developed areas of the world. As an illustration of this

51589_CH03_Printer.indd 127 09/02/13 12:20 PM

128 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

measure, Figure 3–7 presents birth rates categorized by age of mother for the United States for the years 1990–2009. The birth rate has trended upward for older women and downward for women in the youngest age group.

Crude birth rate

Number of live births within a given period

Population size at the middle of that period

1,000 population= ×

Sample calculation: 4,130,665 babies were born in the United States during 2009, when the U.S. population was 307,006,550. The birth rate was 4,130,665/307,006,550 = 13.5 per 1,000.

Figure 3–7 Birth rates by selected age of mother: United States, 1990–2009. Source: Reproduced from JA Martin, BE Hamilton, SJ Ventura, et al. Births: Final data for 2009, National Vital Statistics Reports, Vol 60, No 1, p. 6. Hyattsville, MD: National Center for Health Statistics; 2011.

20092005200019951990 1

R at

e pe

r 1,

00 0

w om

100

200

25–29 years

20–24 years 30–34 years

15–19 years

35–39 years

40–44 years

100

200

NOTE: Rates are plotted on a logarithmic scale. SOURCE: CDC/NCHS, National Vital Statistics System.

51589_CH03_Printer.indd 128 09/02/13 12:20 PM

C r u d e r a t e s 129

Fertility Rate Among the several types of fertility rates, one of the most noteworthy is the general fertility rate. This rate consists of the number of live births reported in an area during a given time interval (for example, during 1 year) divided by the number of women aged 15–44 years in that area. The population size for the number of women aged 15–44 years is assessed at the midpoint of the year. Sometimes the age range of 15–49 years is used. Figure 3–8 illustrates fertility rates compared

Figure 3–8 Live births and rates: United States, 1920–2009. Source: Reprinted from JA Martin, BE Hamilton, SJ Ventura, et al. Births: Final data for 2009, National Vital Statistics Reports, Vol 60, No 1, p. 3. Hyattsville, MD: National Center for Health Statistics; 2011.

00 1920

NOTE: Beginning with 1959, trend lines are based on registered live births; trend lines for 1920–1958 are based on live births adjusted for underregistration. SOURCE: CDC/NCHS, National Vital Statistics System.

1930 1940 1950 1960 1970

Rate

Number

R ate per 1,000 w

om en aged 15–44

B ir

th s

in m

ill io

1980 1990 2000 2009

120

160

200

General fertility rate

Number of live births within a year

Number of women age 15–44 years during the midpoint of the year

1,000 women aged 15–44

= ×

Sample calculation: During 2009, there were 61,948,144 women aged 15 to 44 in the United States. There were 4,130,665 live births. The general fertility rate was 4,130,665/61,948,144 = 66.7 per 1,000 women aged 15 to 44.

51589_CH03_Printer.indd 129 09/02/13 12:20 PM

130 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

with the number of live births for the United States from 1920 to 2009. (The general fertility rate is often referred to more generically as the fertility rate.)

A second type of fertility rate is the total fertility rate. This rate is “[t]he aver- age number of children that would be born if all women lived to the end of their childbearing years and bore children according to a given set of age-specific fertility rates.”1 In the United States, the total fertility rate was estimated to be 2.06 in 2012. This rate is close to the replacement fertility rate of 2.1, the rate at which the number of births is equivalent to the number of deaths; consequently, when its fertility rate is about 2.1, the United States does not have a net popula- tion gain due to births.

Fetal Mortality Fetal mortality is an issue of major public health significance, although often overlooked. The term fetal mortality is defined as “spontaneous intrauterine death at any time during pregnancy.”10(p 1) When such deaths occur during the later stages of pregnancy, they are sometimes referred to as stillbirths. Fetal mortality indices depend on estimation of fetal death after a certain number of weeks of gestation. In the following three definitions, the gestation time is stated or presumed. The fetal death rate is defined as the number of fetal deaths after 20 weeks or more gestation divided by the number of live births plus fetal deaths (after 20 weeks or more gestation). It is expressed as rate per 1,000 live births and fetal deaths. The late fetal death rate refers to fetal deaths after 28 weeks or more gestation. Both measures pertain to a calendar year.

The fetal death ratio refers to the number of fetal deaths after gestation of 20 weeks or more divided by the number of live births during a year. It is expressed as rate per 1,000 live births.

Fetal death rate (per 1,000 live births plus fetal deaths)

Number of fetal deaths after 20 weeks or more gestation

Number of live births number of fetal deaths after 20 weeks or more getation

1,000

Late fetal death rate (per 1,000 live births plus late fetal deaths)

Number of fetal deaths after 28 weeks or more gestation

Number of live births number of fetal deaths after 28 weeks or more getation

1,000

= +

51589_CH03_Printer.indd 130 09/02/13 12:20 PM

C r u d e r a t e s 131

See Figure 3–9 for comparisons of the fetal mortality rate and late fetal mortality rate in the United States between 1990 and 2005. The overall fetal death rate (fetal mortality plus late fetal mortality) declined by 17% between 1990 and 2003. The decline was attributable to decreases in late fetal deaths.10 The rate stabilized at 6.23 per 1,000 live births in 2003 and was nearly the same (6.22 per 1,000) in 2005. In comparison with other racial/ethnic groups, fetal death rates were high- est for non-Hispanic black women, due to the greater risk of preterm delivery.

Figure 3–9 Fetal mortality rates, by period of gestation: United States, 1990–2005. Source: Reproduced from MF MacDorman, S Kirmeyer. The Challenge of Fetal Mortality. NCHS Data Brief, No 16, April 2009.

1990 0

1995

20−27 weeks

28 weeks or more

R at

e pe

r 1,

00 0

liv e

bi rt

hs a

nd fe

ta l

de at

hs in

s pe

ci fie

d gr

ou p

Total

2000 2005

Fetal death ratio

Number of fetal deaths after 20 weeks or more gestation

Number of live births 1,000 (during a year)= ×

Sample calculation: During 1 year there were 134 fetal deaths with 20 weeks or more gestation and 10,000 live births. The fetal death ratio is (134/10,000) = 13.4 per 1,000. Note that the fetal death rate is (134/10,134) = 13.2 per 1,000, which is slightly lower than the fetal death ratio.

1,000 (during a year)

51589_CH03_Printer.indd 131 09/02/13 12:20 PM

132 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Infant Mortality Rate The infant mortality rate is obtained by dividing the number of infant deaths during a calendar year by the number of live births reported in the same year. The infant mortality rate measures the risk of dying during the first year of life among infants born alive. Note that not all infants who die in a calendar year are born in that year, which represents a source of error. Typically, however, the number of infant deaths from previous years’ births is balanced by an equal num- ber of deaths during the following year among the current year’s births. The following is the formula for the infant mortality rate:

Infant mortality

Number of infant deaths among infants aged 0–365 days

during the year

Number of live births during the year

1,000 live births= ×

Sample calculation: In the United States during 2007, there were 29,153 deaths among infants under 1 year of age and 4,316,233 live births. The infant mortality rate was (29,153/4,316,233) × 1,000 = 6.75 per 1,000 live births.

Infant mortality rates are highest among the least developed countries of the world (e.g., Afghanistan with 165 per 1,000 births in 2009) in comparison with some developing countries (e.g., India with 50 per 1,000), less developed coun- tries of Eastern Europe (e.g., Romania with 10 per 1,000), and developed market economies (e.g., Sweden with about 2 per 1,000).12

Figure 3–10 shows trends in U.S. infant mortality by race from 1940 to 1995 (part A) and from 1995 to 2004 (part B). Note how total infant mortality rates declined steadily until 2000 and have declined very little since then. Infant mortality rates vary greatly by race/ethnicity in the United States The rate for non-Hispanic blacks is approximately twice the rate for the United States as a whole (part C).

Figure 3–11 presents a comparison of the infant mortality rate of the United States with that reported by other industrialized nations. In 2007, the U.S. infant mortality rate exceeded that of many other nations. Some of the differ- ences observed between the United States and other developed/industrialized nations, may be artifactual (i.e., due to variations in the definition measurement and reporting of infant deaths). It is most likely, however, that the differences are associated with a high rate of preterm births in the United States.10,11

51589_CH03_Printer.indd 132 09/02/13 12:20 PM

C r u d e r a t e s 133

Figure 3–10 Infant mortality. Part A: Infant mortality rates by race: United States, 1950–1995. Part B: Infant mortality rates by race and ethnicity, 1995–2004. Part C: Infant mortality rates, by race and Hispanic origin of mother: United States, 2007. Sources: Part A: Modified from Anderson RN, Kochanek KD, and Murphy SL. Report of Final Monthly Statistics, 1995, Monthly Vital Statistics Report, Vol 45, No 11, Suppl 2, p. 12. Hyattsville, MD: National Center for Health Statistics; 1997. Part B: From Mathews TJ, MacDorman MF. Infant mortality statistics from the 2004 period linked birth/ infant death data set, National Vital Statistics Reports; Vol 55, No 15, p. 1. Hyattsville, MD: National Center for Health Statistics; 2007. Part C: MF MacDorman, TJ Mathews. Understanding Racial and Ethnic Disparities in U.S. Infant Mortality Rates. NCHS Data Brief, No 74, September 2011.

Race of mother 80 70 60 50

10 9 8 7 6 0

80 70 60 50

10 9 8 7 6 0 1940A

NOTES: Infant deaths are classified by race of decedent. For 1940−90, live births are classified by race of child and for 1980−94, by race of mother.

1950 1960 1970 1980 19901995

Black

All races

White

D ea

th s

un de

r 1

ye ar

p er

1 ,0

00 li

ve b

ir th

B 1995

Non-Hispanic black

American Indian or Alaskan Native

Total

Asian or Pacific Islander

Non-Hispanic white

Hispanic

1996 1997

R at

e pe

r 1,

00 0

liv e

bi rt

1998 1999 Year

2000 2001 2002 2003 2004

2007

In fa

nt m

or ta

lit y

ra te

p er

1 ,0

00 li

ve b

ir th

13.31

Non- Hispanic

black

American Indian

or Alaska native

Puerto Rican

Total Non- Hispanic

white

Mexican Cuban Asian or Pacific Islander

Central and

South American

9.22

7.71 6.75

5.63 5.42 5.18 4.78 4.57

51589_CH03_Printer.indd 133 09/02/13 12:20 PM

134 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Neonatal Mortality Rate The neonatal mortality rate measures risk of dying among newborn infants who are under the age of 28 days (0–27 days) for a given year. The formula is as follows:

Figure 3–11 International infant mortality rates, selected countries,* 2007. Source: Reproduced from U. S. Department of Health and Human Services, Health Resources and Services Administration, Maternal and Child Health Bureau. Child Health USA 2011, International Infant Mortality, p. 28.

*2007 data were not available for all Organization for Economic Co-operation and Development (OECD) countries.

1.8Luxembourg

Iceland

Sweden

Japan

Finland

Czech Republic

Ireland

Norway

Portugal

Greece

Austria

Italy

Spain

Germany

Switzerland

Belgium

Denmark

Netherlands

Australia

New Zealand

United Kingdom

Hungary

Poland

Slovak Republic

United States

Mexico

Turkey

2.0

2.5

2.6

2.7

3.1

3.4

3.6

3.7

3.9

4.0

4.1

4.2

4.8

5.9

6.0

6.1

15.7

20.7

6.8

Deaths per 1,000 live births

15 20 5

Postneonatal Mortality Rate A statistic that is related to the neonatal mortality rate is the postneonatal mortality rate. The postneonatal mortality rate measures risk of dying among older infants during a given year.

Neonatal mortality rate

= Number of infant deaths under 28 days of age

Number of live births 1,000 live births (during a year)× (during a year)

51589_CH03_Printer.indd 134 09/02/13 12:20 PM

C r u d e r a t e s 135

Figure 3–12 illustrates trends in infant mortality rates, neonatal mortality rates, and postneonatal mortality rates in the United States. All three mea- sures showed a declining trend after 1940. However, the infant mortality rate was higher than either the neonatal mortality rate or the postneonatal mortal- ity rate; in addition, both infant and neonatal mortality rates were higher than the postneonatal mortality rate. Between 1997 and 2007 the following mor- tality trends I occurred: infant mortality decreased by 7%; neonatal mortality decreased by 8%; and postneonatal mortality rate decreased by 5%.13 Between 2006 and 2007 neonatal mortality did not change significantly. In 2007 the neonatal mortality rate was 4.42. Postneonatal mortality showed a statistically significant increase over 2006 of 3.5% (from 2.24 to 2.34 per 1,000 live births, all races combined).14

Figure 3–12 Infant, neonatal, and postneonatal mortality rates: United States, 1940–2007. Source: Reproduced from Xu J, Kochanek KD, Murphy SL, et al. Deaths: Final Data for 2007, National Vital Statistics Reports. Vol 58, No 19, p. 13. Hyattsville, MD: National Center for Health Statistics, 2010.

Infant

Neonatal

Postneonatal

1940

NOTE: Rates are infant (under 1 year), neonatal (under 28 days), and postneonatal (28 days—11 months) deaths per 1,000 live births in specified group. SOURCE: CDC/NCHS, National Vital Statistics System, Mortality.

D ea

th s

pe r

1, 00

0 liv

e bi

rt hs

1950 1960 1970 1980 1990 2000 2007

Postneonatal mortality rate

= Number of infant deaths from 28 days to 365 days after birth

Number of live births neonatal deaths 1,000 live births

− ×

1,000 live births

51589_CH03_Printer.indd 135 09/02/13 12:20 PM

136 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Perinatal Mortality Two measures of perinatal mortality are the perinatal mortality rate and the perinatal mortality ratio. The perinatal period used in these measures captures late fetal deaths (stillbirths) plus infant deaths within 7 days of birth. Interna- tionally, perinatal mortality reflects variations in the care of mothers as well as their health and nutritional statuses and also is a quality metric for obstetrics and pediatrics.15 Approximately 3 million stillbirths and 3 million infant deaths (during the first 7 days of life) occur across the globe each year. According to the World Health Organization, “[t]he perinatal mortality rate is five times higher in developing than in developed regions: 10 deaths per 1,000 births in developed countries; 50 per 1,000 in developing regions and over 60 per 1,000 in least developed countries. It is highest in Africa, with 62 deaths per 1,000 births, and especially in middle and western Africa, which have rates as high as 75 and 76 per 1,000.”15(p 20) Figure 3–13 compares rates of perinatal mortality in world regions. The United States had a perinatal mortality rate of 7 per 1,000 in 2000. The formulas for the perinatal mortality rate and perinatal mortality ratio are:

Perinatal mortality rate

Number of late fetal deaths after 28 weeks or more gestation +

infant deaths within 7 days of birth

Number of live births + number of late fetal deaths

1,000 live births and fetal deaths

Perinatal mortality ratio

Number of late fetal deaths after 28 weeks or more gestation +

infant deaths within 7 days of birth

Number of live births 1,000 live births

Maternal Mortality Rate The maternal mortality rate is the number of maternal deaths ascribed to childbirth (i.e., pregnancy and puerperal causes) per 10,000 or 100,000 live births. Factors that affect maternal mortality include maternal age, socioeconomic status, nutritional status, and healthcare access. Figure 3–14 gives causes of maternal mortality. Direct causes include complications related to the puerperium (period after childbirth), eclampsia (a condition marked by convulsions following delivery), and hemorrhage.

51589_CH03_Printer.indd 136 09/02/13 12:20 PM

C r u d e r a t e s 137

Figure 3–13 Global estimates of perinatal mortality rates by geographical (United Nations) region and subregion, 2000. Source: Data from World Health Organization, Neonatal and perinatal mortality: country, regional and global estimates. Geneva: Switzerland.

42Oceania

North America

Latin America/Caribbean

Europe

Asia

Africa

U.S.

World

0 0 20 30 40

Rate

50 60 70

Figure 3–14 Leading causes of maternal mortality, 2007. Source: Reproduced from U. S. Department of Health and Human Services, Health Resources and Services Administration, Maternal and Child Health Bureau. Child Health USA 2011, International Infant Mortality, p. 29.

0.9

0.7

3.8

3.1

0.5

1 2 3

Maternal Deaths per 100,000 Live Births

4 5

1.5

2.2 Direct causes

Complications related to the puerperium

Eclampsia and pre-eclampsia

Other direct causes

Pregnancy with abortive outcome

Hemorrhage of pregnancy, childbirth, and placenta previa

Indirect causes

Unspecified causes

Maternal mortality rate (per 100,000 live births, including multiple births)

Number of deaths assigned to causes related to childbirth

Number of live births 100,000 live births (during a year)= ×

Maternal mortality rate (per 100,000 live births, including multiple births)

Number of deaths assigned to causes related to childbirth

Number of live births 100,000 live births (during a year)= ×

51589_CH03_Printer.indd 137 09/02/13 12:21 PM

138 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Specif ic Rates and Proport ional Mortal i ty Ratio

Specific rates are a type of rate based on a particular subgroup of the population defined, for example, in terms of race, age, or sex, or they may refer to the entire population but be specific for some single cause of death or illness. Although the crude rates described so far are important and useful summary measures of the occurrence of disease, they are not without limitations. A crude rate should be used with caution in making comparative statements about disease frequencies in populations. Observed differences between populations in crude rates of disease may be the result of systematic factors within the populations rather than true variations in rates. Systematic differences in sex or age distributions would affect observed rates. To correct for factors that may influence the make-up of popula- tions and in turn influence crude rates, one may construct specific and adjusted rates. Two other measures that have been defined previously also can be con- sidered specific measures: incidence and prevalence. That is, both are typically specific to a particular end point. Examples of specific rates are cause-specific rates and age-specific rates.

Cause-Specific Rate A cause-specific rate is “[a] rate that specifies events, such as deaths according to their cause.”1 An example of a cause-specific rate is the cause-specific mortality rate. As the name implies, it is the rate associated with a specific cause of death. Sample calculations are shown in Table 3–3. The number of deaths among the 25- to 34-year-old age group (population 39,872,598) due to human immu- nodeficiency virus (HIV) infection was 1,588 during 2003. The cause-specific mortality rate due to HIV was (1,588/39,872,598), or 4.0 per 100,000.

Cause-specific rate

Mortality (or frequency of a given disease)

Population size at midpoint of time period 100,000= ×

Age-Specific Rates An age-specific rate is defined as “[a] rate for a specified age group. The numer- ator and denominator refer to the same age group.”1 To calculate age-specific rates, one subdivides (or stratifies) a population into age groups, such as those

51589_CH03_Printer.indd 138 09/02/13 12:21 PM

s p e C i f i C r a t e s a n d p r o p o r t i o n a l M o r t a l i t y r a t i o 139

Table 3–3 The 10 Leading Causes of Death, 25–34 Years, All Races, Both Sexes, United States, 2003 (Number in Population Aged 25–34 Years = 39,872,598)

Rank Order Cause of Death Number

Proportional Mortality Ratio (%)

Cause-Specific Death Rate per 100,000

1 Accidents ( unintentional injuries)

12,541 30.4 31.5

2 Intentional self harm ( suicide) 5,065 12.3 12.7 3 Assault (homicide) 4,516 10.9 11.3 4 Malignant neoplasms 3,741 9.1 9.4 5 Diseases of the heart 3,250 7.9 8.2 6 Human immunodeficiency

virus (HIV) disease 1,588 3.8 4.0

7 Diabetes mellitus 657 1.6 1.6 8 Cerebrovascular diseases 583 1.4 1.5 9 Congenital malformations,

deformations, and chromosomal abnormalities

426 1.0 1.1

10 Influenza and pneumonia 373 0.9 0.9 All causes 41,300

Source: Adapted from Heron MP, Smith BL. Deaths: Leading Causes for 2003. National Vital Statistics Reports, Vol 55, No 10, p. 18. Hyattsville, MD: National Center for Health Statistics; 2007.

defined by 5- or 10-year intervals. Then, one divides the frequency of a disease in a particular age stratum by the total number of persons within that age stratum to find the age-specific rate. A similar procedure may be employed to calculate sex-specific rates. An example of an age-specific cancer mortality rate is shown in Exhibit 3–6. A second example of the calculation of age-specific mortality rates for the U.S. population is shown in Table 3–4. (Some age-specific death rates shown in Table 3–4 differ from published rates because of differences in estima- tion of population size and use of different intervals for age groups.)

In summary, this section has demonstrated how to calculate cause-specific and age-specific rates. It is also possible to define other varieties of specific rates (e.g., sex specific rates). All in all, specific rates are a much better indicator of risk than crude rates, especially for rates specific to defined subsets of the population (e.g., age, race, and sex specific). A disadvantage of specific rates is the difficultly in visualizing the “big picture” in those situations where specific rates for several factors are presented in complex tables. Table 3–5 shows the age-specific cancer incidence rates by sex, age group, and year of diagnosis. Most people would find it difficult to synthesize the data from a complex table, such as this one, and discern any specific trends. The numbers shown in the table are age-adjusted, a procedure that we will describe later in the chapter.

51589_CH03_Printer.indd 139 09/02/13 12:21 PM

140 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Proportional Mortality Ratio The proportional mortality ratio (PMR) is the number of deaths within a popula- tion due to a specific disease or cause divided by the total number of deaths in the population.

age-Specific rate (ri)

Age-specific rate: The number of cases per age group of population (during a specified time period).

Example:

iR Number of deaths among those aged 5–14 years

Number of persons who are aged 5–14 years (during time period)

100,000= ×

Sample calculation: In the United States during 2003, there were 1,651 deaths due to malignant neoplasms among the age group 5 to 14 years, and there were 40,968,637 persons in the same age group. The age-specific malig- nant neoplasm death rate in this age group is (1,651/40,968,637) = 4.0 per 100,000. n

e x

h ib

it 3

–6

Table 3–4 Method of Calculation of Age-Specific Death Rates

Age Group (Years) Number of Deaths

(Di) in 2003 Number in Population (Pi) as of July 1, 2003*

Age-Specific Rate (Ri) per 100,000

Under 1 28,025 4,003,606 700.0 1–4 4,965 15,765,673 31.5 5–14 6,954 40,968,637 17.0 15–24 33,568 41,206,163 81.5 25–34 41,300 39,872,598 103.6 35–44 89,461 44,370,594 201.6 45–54 176,781 40,804,599 433.2 55–64 262,519 27,899,736 940.9 65–74 413,497 18,337,044 2,255.0 75–84 703,024 12,868,672 5,463.1 85+ 687,852 4,713,467 14,593.3 Not stated 342 NA NA Totals 2,448,288 290,810,789 841.9**

* Estimated

** The crude mortality rate for the United States

Source: Data from Hoyert DL, Heron MP, Murphy SL, Kung H. Deaths: Final Data for 2003. National Vital Statistics Reports, Vol 54, No 13, pp. 23 and 112. Hyattsville, MD: National Center for Health Statistics; 2006.

51589_CH03_Printer.indd 140 09/02/13 12:21 PM

s p e C i f i C r a t e s a n d p r o p o r t i o n a l M o r t a l i t y r a t i o 141

T ab

le 3

–5

Tr en

ds i

n A

ge -S

pe ci

fi c

C an

ce r

In ci

de nc

e R

at es

* (B

y Se

x, A

ge G

ro up

, an

d Ye

ar o

f D

ia gn

os is

, SE

ER P

ro gr

am ,

20 00

–2 00

9; A

ll Si

te s

C om

bi ne

d, A

ll R

ac es

)

Ye ar

o f

D ia

gn os

Se x/

A ge

20 00

20 01

20 02

20 03

20 04

20 05

20 06

20 07

20 08

20 09

A ll

A ge

s <

1 6.

1 7.

1 5.

1 6.

1 7.

1 6.

1 7.

3 A

ge s

20 –4

14 6.

14 7.

14 5.

9 1

46 .0

8. 5

1 49

.3 1

50 .6

15 1.

9 15

1. 2

A ge

s 50

–6 4

6. 9

84 .1

1. 6

47 .1

8 44

.7 8

35 .4

46 .1

8 59

84 5.

9 8

34 .7

A ge

s 65

–7 4

2 02

5. 3

2 04

1. 9

2 01

4. 0

1 94

4. 3

1 94

1. 2

1 90

3. 2

1 93

2. 8

1 97

0. 1

19 19

.7 1

88 6.

2 A

ge s

75 +

2 46

3. 2

24 76

.9 2

45 7.

7 2

39 0.

1 24

04 .0

23 81

.4 2

36 7.

3 23

69 .3

2 31

8. 8

22 48

.4 M

al es

A ge

s <

1 6.

1 8.

1 6.

1 7.

1 8.

1 7.

1 8.

1 A

ge s

20 –4

11 3.

1 12

.1 1

10 .8

2. 3

3. 1

11 .3

1 12

.9 1

13 .4

1 10

.7 A

ge s

50 –6

9 62

.2 9

71 .7

3. 8

9 34

.7 9

33 .9

1. 2

0. 5

57 .3

9. 5

9 10

.7 A

ge s

65 –7

4 2

64 4.

9 26

75 .9

26 20

.4 2

51 4.

3 25

12 .8

2 42

7. 1

2 48

5. 2

2 53

8. 1

2 42

4. 9

23 56

.7 A

ge s

75 +

33 49

.4 3

36 4.

2 3

33 3.

3 3

22 8.

3 3

22 7.

3 3

18 2.

1 3

16 2.

8 31

69 .2

3 04

1. 9

2 92

3. 1

Fe m

al es

A ge

s <

1 5.

1 6.

1 5.

1 6.

1 5.

1 6.

1 5.

1 6.

4 A

ge s

20 –4

1 79

.1 1

81 .0

9. 6

1 81

.0 1

84 .8

1 86

1 87

.5 1

88 .6

0. 9

2. 2

A ge

s 50

–6 4

8. 9

8 04

7 97

.0 7

66 .6

7 62

.9 7

65 .9

59 .3

7 69

76 8.

76 4.

8 A

ge s

65 –7

4 1

52 1.

5 15

22 .6

1 51

4. 2

1 47

2. 3

14 65

.7 14

65 .0

14 69

.4 14

92 .0

14 92

.4 14

86 .6

A ge

s 75

+ 1

93 9.

3 1

94 7.

6 19

33 .4

1 88

4. 1

19 04

.9 18

87 .4

18 72

.2 18

68 .6

18 62

.7 1

82 2.

* R

at es

a re

p er

1 00

,0 00

a nd

a re

a ge

-a dj

us te

d to

t he

2 00

0 U

S St

an da

rd P

op ul

at io

n (1

9 ag

e gr

ou ps

— C

en su

s P2

5– 11

30 ).

So ur

ce : h

tt p:

// se

er .c

an ce

r. go

v/ fa

st st

at s/

. A cc

es se

d M

ay 2

2, 2

01 2.

51589_CH03_Printer.indd 141 09/02/13 12:21 PM

142 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Refer to Table 3–3 for a more detailed example of a PMR. In Table 3–3, the PMR is calculated according to the formula given above. For example, the proportional mortality ratio for HIV among the 25- to 34-year-old group was 3.8% (1,588/41,300). This PMR should be used with caution when compari- sons are made across populations, especially those that have different rates of total mortality. To illustrate, consider that two countries have identical death rates from cardiovascular disease (perhaps 5 per 100,000 per year) and that each country has exactly 1 million inhabitants. Therefore, one would expect 50 deaths from cardiovascular disease to occur in each country (5 per 100,000 per year × 1,000,000). Suppose further, however, that in country A the total death rate per 100,000 per year is 30 and that it is only 10 in country B. Therefore, the expected total number of deaths would be 300 in country A and only 100 in country B. When these data are used to construct a PMR, one sees that the proportion of deaths from cardiovascular disease is higher in country B (0.50) than in country A (0.17). The PMR is not a measure of the risk of dying of a particular cause. It merely indicates, within a population, the relative importance of a specific cause of death. For a health administrator, such information may be useful to determine priorities and planning. To an epidemiologist, such differ- ences may indicate an area for further study. For example, why does country A have such higher total mortality rates than country B? Is it merely because of dif- ferences in age structure? Is the difference a result of access to health care or cer- tain behavioral or lifestyle patterns associated with elevated mortality? The PMR should not be confused with a case fatality rate, which expresses the proportion of fatal cases among all cases of disease during a specific time period.

Table 3–6 presents a summary of unadjusted measures of morbidity and mor- tality discussed in this chapter. We provide this table to assist you with future review and reference to these measures.

PMR (%)

Mortality due to a specific cause during a time period

Mortality due to all causes during the same time period 100= ×

Sample calculation: In a certain community, there were 66 deaths due to coronary heart disease during a year and 200 deaths due to all causes in that year. The PMR is (66/200) × 100 = 33%.

51589_CH03_Printer.indd 142 09/02/13 12:21 PM

s p e C i f i C r a t e s a n d p r o p o r t i o n a l M o r t a l i t y r a t i o 143

T ab

le 3

–6

Su m

m ar

y of

U na

dj us

te d

M ea

su re

s of

M or

bi di

ty a

nd M

or ta

lit y

D is

cu ss

ed in

T hi

s C

ha pt

M ea

su re

N um

er at

or (

x) D

en om

in at

or (

y) Ex

pr es

se d

pe r

N um

be r

at R

is k

(1 0*

)

A ge

-s pe

ci fi

c ra

te (

e. g.

, m

or ta

lit y

ra te

) N

um be

r o f d

ea th

s am

on g

a sp

ec ifi

c ag

e gr

ou p

of th

e po

pu la

tio n

du rin

g a

sp ec

ifi ed

ti m

e pe

rio d

N um

be r

of p

er so

ns w

ho c

om pr

is e

th at

a ge

g ro

du rin

g th

e sa

m e

tim e

pe rio

d 10

0, 00

A tt

ac k

ra te

N um

be r

ill d

ur in

g a

tim e

pe rio

d N

um be

r ill

+ w

el l d

ur in

g th

e sa

m e

tim e

pe rio

d 10

0 (o

r pe

rc en

t) C

as e

fa ta

lit y

ra te

( di

sc us

se d

in C

ha pt

er 1

2) N

um be

r of

d ea

th s

du e

to d

is ea

se “

X” d

ur in

g a

gi ve

n tim

e in

te rv

al N

um be

r of

n ew

c as

es o

f t ha

t di

se as

e re

po rt

ed d

ur in

g th

e sa

m e

tim e

in te

rv al

10 0

(o r

pe rc

en t)

C au

se -s

pe ci

fi c

ra te

, ( e.

g. ,

ca us

e- sp

ec if

ic d

ea th

o r

m or

ta lit

y ra

te )

N um

be r

of d

ea th

s (m

or ta

lit y)

a ss

ig ne

d to

sp ec

if ic

c au

se d

ur in

g a

ye ar

Es tim

at ed

p op

ul at

io n

si ze

d ur

in g

th e

m id

po in

t of

at s

am e

ye ar

10 0,

00 0

C ru

de b

ir th

r at

e N

um be

r of

li ve

b ir

th s

w ith

in a

g iv

en p

er io

d Po

pu la

tio n

si ze

a t

th e

m id

po in

t of

t ha

t tim

e pe

rio d

1, 00

0 C

ru de

d ea

th r

at e

N um

be r

of d

ea th

s du

rin g

a gi

ve n

ye ar

R ef

er en

ce p

op ul

at io

n du

rin g

th e

m id

po in

t of

t he

y ea

r 10

0, 00

0 Fe

ta l d

ea th

r at

e N

um be

r of

fe ta

l d ea

th s

af te

r 20

w ee

ks o

r m

or e

ge st

at io

n du

rin g

a ye

ar N

um be

r of

li ve

b ir

th s

+ nu

m be

r of

fe ta

l d ea

th s

af te

r 20

w ee

ks o

r m

or e

ge st

at io

n du

rin g

th e

sa m

e ye

ar 1,

00 0

Fe ta

l d ea

th r

at io

N um

be r

of fe

ta l d

ea th

s af

te r

20 w

ee ks

o r

m or

e ge

st at

io n

du rin

g a

ye ar

N um

be r

of li

ve b

ir th

s du

rin g

th e

sa m

e ye

ar 1,

00 0

G en

er al

fe rt

ili ty

r at

e N

um be

r of

li ve

b ir

th s

w ith

in a

y ea

r N

um be

r of

w om

en a

ge d

15 –4

4 ye

ar s

du rin

g th

e m

id po

in t

of t

he y

ea r

1, 00

In ci

de nc

e de

ns ity

N um

be r

of n

ew c

as es

d ur

in g

a tim

e pe

rio d

To ta

l p er

so n-

ye ar

s of

o bs

er va

tio n

du rin

g th

e sa

m e

tim e

pe rio

d Va

rie s

.g .,

10 0,

00 0)

In ci

de nc

e ra

te N

um be

r of

n ew

c as

es d

ur in

g a

tim e

pe rio

d To

ta l p

op ul

at io

n at

r is

k du

rin g

th e

sa m

e tim

e pe

rio d

Va rie

(e .g

., 10

0, 00

0) In

fa nt

m or

ta lit

y ra

te N

um be

r of

in fa

nt d

ea th

s un

de r

1 ye

ar o

f a ge

–3 65

d ay

s) d

ur in

g a

ye ar

N um

be r

of li

ve b

ir th

s re

po rt

ed d

ur in

g th

at s

am e

ye ar

1, 00

La te

fe ta

l d ea

th r

at e

N um

be r

of fe

ta l d

ea th

s af

te r

28 w

ee ks

o r

m or

e ge

st at

io n

du rin

g a

ye ar

N um

be r

of li

ve b

ir th

s +

nu m

be r

of fe

ta l d

ea th

s af

te r

28 w

ee ks

o r

m or

e ge

st at

io n

du rin

g th

e sa

m e

ye ar

1, 00

M at

er na

l m or

ta lit

y ra

te N

um be

r of

d ea

th s

as si

gn ed

t o

ca us

es r

el at

ed t

o ch

ild bi

rt h

du rin

g a

ye ar

N um

be r

of li

ve b

ir th

s du

rin g

th e

sa m

e ye

ar 10

0, 00

N eo

na ta

l m or

ta lit

y ra

te N

um be

r of

in fa

nt d

ea th

s un

de r

28 d

ay s

of a

du rin

g a

ye ar

N um

be r

of li

ve b

ir th

s du

rin g

th e

sa m

e ye

ar 1,

00 0

Pe rin

at al

m or

ta lit

y ra

te N

um be

r of

la te

fe ta

l d ea

th s

af te

r 28

w ee

ks o

r m

or e

ge st

at io

n +

in fa

nt d

ea th

s w

ith in

7 d

ay s

bi rt

h du

rin g

a ye

N um

be r

of li

ve b

ir th

s +

nu m

be r

of la

te fe

ta l d

ea th

s du

rin g

th e

sa m

e ye

ar 1,

00 0

Pe rin

at al

m or

ta lit

y ra

tio N

um be

r of

la te

fe ta

l d ea

th s

af te

r 28

w ee

ks o

r m

or e

ge st

at io

n +

in fa

nt d

ea th

s w

ith in

7 d

ay s

bi rt

h du

rin g

a ye

N um

be r

of li

ve b

ir th

s du

rin g

th e

sa m

e ye

ar 1,

00 0

Po st

ne on

at al

m or

ta lit

y ra

te N

um be

r of

in fa

nt d

ea th

s fr

om 2

8 da

ys t

o 36

5 da

ys a

ft er

b ir

th d

ur in

g a

ye ar

N um

be r

of li

ve b

ir th

s −

ne on

at al

d ea

th s

du rin

g th

e sa

m e

ye ar

1, 00

Pr op

or tio

na l m

or ta

lit y

ra tio

M or

ta lit

y du

e to

a s

pe ci

fi c

ca us

e du

rin g

a ye

ar M

or ta

lit y

du e

to a

ll ca

us es

d ur

in g

th at

s am

e ye

ar 10

0 or

1 ,0

*N ot

e: T

he fo

re go

in g

m ea

su re

s ar

e ex

pr es

se d

in t

he g

en er

al fo

rm x — y

× 1

0n .

51589_CH03_Printer.indd 143 09/02/13 12:21 PM

144 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Adjusted Rates

Adjusted rates are summary measures of the rate of morbidity or mortality in a population in which statistical procedures have been applied to remove the effect of differences in composition of the various populations. A common factor for rate adjustment is age, which is probably the most important variable in risk of morbidity and mortality, although rates can be adjusted for other variables. Crude rates mask differences between populations that differ in age and thus are not satisfactory for comparing health outcomes in such populations.16 Members of older populations have a much greater risk of mortality than those in younger populations. Consequently, when a population is older the crude mortality rate will be higher than when the population is younger. Refer to Table 3–7.

The crude death rates (all ages) in Group A and Group B are 50 per 1,000 and 40 per 1,000, respectively; these rates suggest that Group A has a higher mortality rate than Group B. Next, we will examine the age-adjusted death rates for the same populations: These rates are 42 per 1,000 and 52 per 1,000. (For the time being, ignore the procedures for age adjustment; these will be described later.) Group A has a lower age-adjusted mortality rate than Group B because the population of Group A is older.

Figure 3–15 presents trends in U.S. crude and age-adjusted mortality rates between 1960 and 2007, a time interval during which the population has aged. Both rates have trended downward, with the age-adjusted rate declining much more steeply.

Now let’s examine methods for adjusting rates: Two methods for the adjust- ment of rates are the direct method and the indirect method. An easy way to remember how they differ is that direct and indirect refer to the source of the rates. The direct method may be used if age-specific death rates in a population to be standardized are known and a suitable standard population is available. The direct method is presented in Table 3–8. Note that each age-specific rate found in Table 3–4 is multiplied by the number of persons in the age group in the standard population. Before the year 2000, the U.S. population in 1940 was used as the standard; now the standard shown in Table 3–8 is the estimated number in the standard population in the year 2000. (See Exhibit 3–7 for infor- mation on the development of the year 2000 standard for age adjustment.) As indicated in the fourth column of Table 3–8, the result is the expected number of deaths in each age group, which is then summed across all age groups to determine the total number of expected deaths. The age-adjusted rate is the total expected number of deaths divided by the total estimated 2000 population times 100,000: [(2,286,926.31/274,633,642) × 100,000] = 832.7 per 100,000.

51589_CH03_Printer.indd 144 09/02/13 12:21 PM

a d j u s t e d r a t e s 145

T ab

le 3

–7

G ro

up C

om pa

ri so

n of

C ru

de a

nd A

ge -A

dj us

te d

D ea

th R

at es

G ro

up A

G ro

up B

A ge

D ea

th s

) P

op ul

at io

n (2

) R

at e1

)

W ei

gh te

d R

at e2

) D

ea th

(5 )

P op

ul at

io n

(6 )

R at

(7 )

W ei

gh te

d R

at e2

)

St an

da rd

op ul

at io

n (9

)

St an

da rd

ei gh

(1 0)

A ll

ag es

50 0

10 ,0

00 50

. . .

40 0

10 ,0

00 40

. . .

10 ,0

00 1.

0 0–

24 y

ea rs

2 0

1 ,0

00 20

18 0

6 ,0

00 30

9 3

,0 00

0. 3

25 –6

4 ye

ar s

12 0

3 ,0

00 40

1 2

15 0

3 ,0

00 50

15 3

,0 00

0. 3

65 y

ea rs

d ov

er 36

0 6

,0 00

60 2

4 7

0 1

,0 00

70 28

4 ,0

00 0.

A ge

-a dj

us te

d de

at h

ra te

1 . .

. . .

. 4

2 . .

. . .

. 52

. . .

C at

eg or

y no

t ap

pl ic

ab le

. 1

R at

e pe

r 1,

00 0

po pu

la tio

n. 2

Th e

w ei

gh te

d ra

te is

c al

cu la

te d

by m

ul tip

ly in

g th

e ag

e- sp

ec if

ic r

at e

by t

he s

ta nd

ar d

w ei

gh t.

3 Th

e st

an da

rd w

ei gh

t fo

r ea

ch a

ge g

ro up

is c

al cu

la te

d by

d iv

id in

g th

e st

an da

rd p

op ul

at io

n at

e ac

h ag

e by

t he

t ot

al s

ta nd

ar d

po pu

la tio

So ur

ce : F

ro m

A nd

er so

n R

N , R

os en

be rg

H M

. A ge

S ta

nd ar

di za

tio n

of D

ea th

R at

es : I

m pl

em en

ta tio

n of

t he

Y ea

r 20

00 S

ta nd

ar d,

N at

io na

l V ita

l S ta

tis tic

s R ep

or ts

, Vo

l 4 7,

N o

3, p

. 2 . N

at io

na l C

en te

r fo

r H

ea lth

S ta

tis tic

s; 1

99 8.

51589_CH03_Printer.indd 145 09/02/13 12:21 PM

146 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Figure 3–15 Crude and age-adjusted death rates: United States, 1960–2007. Source: Reproduced from Xu J, Kochanek KD, Murphy SL, et al. Deaths: Final Data for 2007, National Vital Statistics Reports. Vol 58, No 19, p. 4. Hyattsville, MD: National Center for Health Statistics, 2010.

Crude

1960

NOTE: Crude death rates are on an annual basis per 100,000 population; age-adjusted rates are per 100,000 U.S. standard population. SOURCE: CDC/NCHS, National Vital Statistics System, Mortality.

800

1,000

1,200

1,400

1970 1980 1990 2000 2007

Age adjusted R

at e

pe r

10 0,

00 0

po pu

la tio

the National Center for health Statistics adopts a New Standard population for age Standardization of Death rates

The crude death rate is a widely used measure of mortality. However, crude rates are influenced by the age composition of the population. As such, comparisons of crude death rates over time or between groups may be misleading if the populations being compared differ

in age composition. This is relevant, for example, in trend comparisons of U.S. mortality, given the aging of the U.S. population. . . . The crude death rate for the United States rose from 852.2 per 100,000 population to 880.0 during 1979 to 1995. This increase in the crude death rate was due to the increasing proportion of the U.S. population in older age groups that have higher death rates. Age standardization, often called age adjustment, is one of the key tools used to control for the changing age distribution of the population, and thereby to make meaningful death comparisons of vital

e x

h ib

it 3

–7

continues

51589_CH03_Printer.indd 146 09/02/13 12:21 PM

a d j u s t e d r a t e s 147

To summarize, direct adjustment requires the application of the observed rates of disease in a population to some standard population to derive an expected number (rate) of mortality. The same procedure would be followed for other populations that one might wish to compare. By standardizing the observed rates of disease in the populations being compared to the same reference population, one is thereby assured that any observed differences that remain are not simply a reflection of dif- ferences in population structure with respect to factors such as age, race, and sex.

A method of direct adjustment that achieves the same results as those reported in Table 3–8 uses year 2000 standard weights (refer to Table 3–9). From the previous discussion, you may have inferred the following relationship:

Pi i

R =

where Ri = the age-specific death rate for the i-th interval (row) in Table 3–4 and: Di = number of deaths in age interval i Pi = number of persons in age interval i at midyear

exhibit 3–7 continued

rates over time and between groups. In contrast to the rising crude death rate, the age-adjusted death rate for the United States dropped from 577.0 per 100,000 U.S. standard population to 503.9 during 1979 to 1995. This age-adjusted comparison is free from the confounding effect of changing age distribution and therefore better reflects the trend in U.S. mortality. To use age adjustment requires a standard population, which is a set of arbitrary population weights.

The new standard is based on the year 2000 population and beginning with data year 1999 will replace the existing standard based on the 1940 population. . . . Currently, at least three different standards are used among Department of Health and Human Services agencies. Implementation of the year 2000 standard will reduce confusion among data users and the burden on state and local agencies. Use of the year 2000 standard also will result in age-adjusted death rates that are substantially larger than those based on the 1940 standard. Further, the new standard will affect trends in age-adjusted rates for certain causes of death and will narrow race differentials in age-adjusted death rates. n

Source: Adapted from Anderson RN, Rosenberg HM. Age Standardization of Death Rates: Implementation of the Year 2000 Standard, National Vital Statistics Reports, Vol 47, No 3, p. 1. National Center for Health Statistics; 1998.

51589_CH03_Printer.indd 147 09/02/13 12:21 PM

148 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Table 3–8 Direct Method for Adjustment of Death Rates

Age Group (Years)

2003 Age-Specific Death Rate per 100,000†

Number in Standard Population, 2000*

Expected Number of Deaths

Under 1 700.0 3,794,901 26,564.08 1–4 31.5 15,191,619 4,784.22 5–14 17.0 39,976,619 6,785.62 15–24 81.5 38,076,743 31,018.66 25–34 103.6 37,233,437 38,566.36 35–44 201.6 44,659,185 90,042.86 45–54 433.2 37,030,152 160,428.66 55–64 940.9 23,961,506 225,462.73 65–74 2,255.0 18,135,514 408,952.54 75–84 5,463.1 12,314,793 672,765.23 85+ 14,593.3 4,259,173 621,555.36 Totals 274,633,642 2,286,926.31

† Age-specific death rates are from Table 3–4.

* Estimated

Age-adjusted rate per 100,000 = 832.7 (total expected number of deaths/estimated 2000 population) × 100,000 (2,286,926.31/274,633,642) × 100,000

Source: Data from Hoyert DL, Heron MP, Murphy SL, Kung H. Deaths: Final Data for 2003. National Vital Statistics Reports, Vol 54, No 13, p. 114. Hyattsville, MD: National Center for Health Statistics; 2006.

Table 3–9 Weighted Method for Direct Rate Adjustment

Age Group (Years)

Number in Standard Population, 2000* (Psi)

Standard Weight (Wsi) for 2000

Age-Specific Death Rate (Ri ), 2003† Wsi ⋅ Ri

Under 1 3,794,901 0.013818 700.0 9.6726 1–4 15,191,619 0.055316 31.5 1.7420 5–14 39,976,619 0.145563 17.0 2.4708 15–24 38,076,743 0.138646 81.5 11.2946 25–34 37,233,437 0.135575 103.6 14.0428 35–44 44,659,185 0.162164 201.6 13.7865 45–54 37,030,152 0.134835 433.2 15.4155 55–64 23,961,506 0.087249 940.9 18.0958 65–74 18,135,514 0.066035 2,255.0 148.9084 75–84 12,314,793 0.044841 5,463.1 244.9682 85+ 4,259,173 0.015509 14,593.3 266.3216 Totals ∑iPsi = 274,633,642 1.0 N/A ∑iWsi ⋅ Ri = 832.7

* Estimated

† From Table 3–4.

51589_CH03_Printer.indd 148 09/02/13 12:21 PM

a d j u s t e d r a t e s 149

We may assign standard weights (Wsi) to each interval according to the following formula:

W P

Psi si

si i

∑ =

where Wsi is the standard weight associated with the i-th interval of the year 2000 standard U.S. population and:

Psi = the population in the i-th age interval in the standard population Psi

i ∑ = total number in the standard population

Example (see Table 3–9): For age interval 1–4 years,

15,191,619

274,633,642 0.055316W = =

Then the age-adjusted death rate (AADR) is:

AADR W D

P W

si i

i i

si i= = =∑ ∑• • .R 832 7

The formula for AADR indicates that the year 2000 standard weights for each age group are multiplied by the age-specific death rates in that same row. These products are then summed to obtain the AADR (see Table 3–9). Note that the results for this method of standardization are the same as those reported in Table 3–8.

A second method of age adjustment is the indirect method, which may be used if age-specific death rates of the population for standardization are unknown or unstable (e.g., because the rates to be standardized are based on a small population). The stratum-specific rates of a larger population, such as that of the United States, are applied to the number of persons within each stratum of the population of interest to obtain the expected numbers of deaths. Thus, the indirect method of standardization does not require knowledge of the actual age- specific incidence or mortality rates among each age group for the population to be standardized. By applying the rates of disease from a standard population (in this example, the 2003 population) to the observed structure of the population of interest, one is left with an expected number of cases (or deaths) in the study population if the rates of disease were the same as in the standard population. One way to evaluate the result is to construct a standardized morbidity ratio or a standardized mortality ratio (SMR).

51589_CH03_Printer.indd 149 09/02/13 12:21 PM

150 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

If the observed and expected numbers are the same, the SMR would be 100% (1.0), indicating that the observed morbidity or mortality in the study popula- tion is not unusual. An SMR of 200% (2.0) is interpreted to mean that the death (or disease) rate in the study population is two times greater than expected.

A second example of the indirect method of adjustment is shown in Table 3–10. Note that the standard age-specific death rates for the year 2003 (which we will designate as the year for obtaining the standard population) from Table 3–4 were multiplied by the number in each age group of the population of interest to obtain the expected number of deaths. To calculate the SMR, the observed number of deaths was divided by the expected number. The crude mortality rate is 502/230,109 = 218.2 per 100,000. The SMR is (502/987.9) × 100 = 50.8%. From the SMR, one may conclude that the observed mortality in this population falls below expectations, because the SMR is less than 1.0 or 100%.

Note that construction of an SMR is not the only way to interpret the net effect of the indirect adjustment procedure. An alternative is to compute a mor- tality rate per 100,000 by using the expected number of deaths as the numerator,

Table 3–10 Illustration of Indirect Age Adjustment: Mortality Rate Calculation for a Fictitious Population of 230,109 Persons

Age (Years)

Number in Population of Interest

Death Rates (per 100,000) in Standard Population*

Expected Number of Deaths in Population of Interest

15–24 7,989 81.5 6.5 25–34 37,030 103.6 38.4 35–44 60,838 201.6 122.6 45–54 68,687 433.2 297.6 55–64 55,565 940.9 522.8 Totals 230,109 987.9

Total expected number of deaths = 987.9 Observed number of deaths in this population = 502

* Standard death rates are from Table 3–4.

SMR Observed deaths

Expected deaths 100= ×

Sample calculation: The number of observed deaths due to heart disease is 600 in a certain county during year 2014. The expected number of deaths is 1,000. The SMR = (600/1,000) × 100 = 60% (0.6).

51589_CH03_Printer.indd 150 09/02/13 12:21 PM

C o n C l u s i o n 151

rather than the observed number of deaths in the study population. If we wanted to focus on an outcome other than mortality, we could use the expected num- ber of morbid events as the numerator. In either case, the calculation would be based on the expected numbers derived from the standard population. Referring to the example in Table 3–10, the total population size was 230,109 and the total expected number of deaths was 987.9. The adjusted death rate would be 987.9/230,109 × 100,000 = 429.3 per 100,000 per year. In comparison, the unadjusted death rate was 502/230,109 or 218.2 per 100,000 per year.

It is important to be aware that the numeric magnitude of an SMR in this situation is a reflection of the standard population. That is, if one were to use the age distribution of the 1970 U.S. population instead of the 2003 U.S. popula- tion for age adjustment, the adjusted rates that one would find would be quite different. Accordingly, SMRs for different populations typically cannot be com- pared with one another unless the same standard population has been applied to them. In addition, SMRs sometimes can be misleading: As a summary index, the overall SMR can be equal to 1.0 across different populations being compared, yet there might still be important differences in mortality in various subgroups. Finally, the longer a population is followed, the less information the SMR pro- vides. Because it is expected that everyone in the population will die eventually, the SMR will tend to be equal to 1.0 over time.

Conclusion

This chapter defined several measures of disease frequency that are commonly employed in epidemiology. Counts or frequency data refer to the number of cases of a disease or other health phenomenon being studied. A ratio consists of a numerator and a denominator that express one number relative to another (e.g., the sex ratio). Prevalence is a measure of the existing number of cases of disease in a population at a point in time or over a specified period of time. A rate is defined as a proportion in which the numerator consists of the frequency of a disease during a period of time and the denominator is a unit size of popula- tion. Rates improve one’s ability to make comparisons of health indices across contrasting populations. Examples of rates include the crude mortality rate, inci- dence rates, and infant mortality rates. Other examples of rates discussed were the birth rate, fertility rate, and perinatal mortality rate. Specific rates are more precise indicators of risk than crude rates. It was noted that, to make compari- sons across populations, adjusted rates also may be used. Two techniques were presented on how to adjust rates. Finally, the chapter gave illustrations of how the SMR (an example of indirect adjustment) is used.

51589_CH03_Printer.indd 151 09/02/13 12:21 PM

152 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

Study Questions and Exercises

1. Define the following terms: a. crude death rate b. age-specific rate c. cause-specific rate d. proportional mortality ratio (PMR) e. maternal mortality rate f. infant mortality rate g. neonatal mortality rate h. fetal death rate and late fetal death rate i. fetal death ratio j. perinatal mortality rate k. postneonatal mortality rate l. crude birth rate

m. general fertility rate n. age-adjusted (standardized) rate o. direct method of adjustment p. indirect method of adjustment q. standardized mortality ratio (SMR)

2. Using Table 3A–1, calculate age-specific death rates for the category of malignant neoplasms of trachea, bronchus, and lung. What inferences can be made from the age-specific death rates for malignant neoplasms of trachea, bronchus, and lung?

Table 3A–1 Malignant Neoplasms of Trachea, Bronchus, and Lung Deaths by Age Group, United States, 2003

Age (Years) Population Malignant Neoplasms of Trachea,

Bronchus, and Lung* Deaths

25–34 39,872,598 154 35–44 44,370,594 2,478 45–54 40,804,599 12,374 55–64 27,899,736 30,956 65–74 18,337,044 49,386

* Includes ICD-10, 1992 codes C33–C34.

Sources: Data are from Hoyert DL, Heron MP, Murphy SL, Kung H. Deaths: Final Data for 2003, National Vital Statistics Reports, Vol 54, No 13, p. 30. Hyattsville, MD: National Center for Health Statistics; 2006; and from Heron MP, Smith BL. Deaths: Leading Causes for 2003, National Vital Statistics Reports, Vol 55, No 10, p. 92. Hyattsville, MD: National Center for Health Statistics; 2007.

51589_CH03_Printer.indd 152 09/02/13 12:21 PM

s t u d y Q u e s t i o n s a n d e x e r C i s e s 153

4. Refer to both Table 3A-2 and Table 3A–3. The total population in 2003 was 290,810,789 (males = 143,037,290; females = 147,773,499). For 2003, the total number of live births was 4,089,950.

3. Using Table 3A–2, calculate the following for the United States: the age-specific death rates and age- and sex-specific death rates per 100,000 (for age groups 20–24, 25–34, and 35–44 years). Note that there are nine calculations and answers. For example, the age- and sex-specific death rate for females aged 15–19 years is [(3,889/9,959,789) × 100,000].

Table 3A–2 Mortality by Selected Age Groups, Males and Females, United States, 2003

Males Females Total

Age (Years) Population Number of

Deaths Population Number of

Deaths

15–19 10,518,680 9,706 9,959,789 3,889 20,478,469 13,595 20–24 10,663,922 14,964 10,063,772 5,009 20,727,694 19,973 25–34 20,222,486 28,602 19,650,112 12,698 39,872,598 41,300 35–44 22,133,659 56,435 22,236,935 33,026 44,370,594 89,461 45–54 20,043,656 110,682 20,760,943 66,099 40,804,599 176,781

Sources: Data are from Heron MP, Smith BL. Deaths: Leading Causes for 2003, National Vital Statistics Reports, Vol 55, No 10, p. 92. Hyattsville, MD: National Center for Health Statistics; 2007; and from Hoyert DL, Heron MP, Murphy SL, Kung H. Deaths: Final Data for 2003, National Vital Statistics Reports, Vol 54, No 13, p. 21. Hyattsville, MD: National Center for Health Statistics; 2006.

Table 3A-3 Total Mortality from Selected Causes, Males and Females, United States, 2003

Cause of Death Males Females Total

All Causes 1,201,964 1,246,324 2,448,288 Accidents 70,532 38,745 109,277 Malignant Neoplasms 287,990 268,912 556,902 Alzheimer’s Disease 18,335 45,122 63,457 Infant Deaths 15,902 12,123 28,025 Maternal Deaths NA 495 495

Sources: Data are from Heron MP, Smith BL. Deaths: Leading Causes for 2003, National Vital Statistics Reports, Vol 55, No 10, p. 7–8. Hyattsville, MD: National Center for Health Statistics; 2007; and Hoyert DL, Heron MP, Murphy SL, Kung H. Deaths: Final Data for 2003, National Vital Statistics Reports, Vol 54, No 13, p. 101–102. Hyattsville, MD: National Center for Health Statistics; 2006.

51589_CH03_Printer.indd 153 09/02/13 12:21 PM

154 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

a. Calculate the crude death rates (per 100,000) and the cause-specific death rates (per 100,000) for accidents, malignant neoplasms, and Alzheimer’s disease. Repeat these calculations for males and females separately.

b. What are the PMRs (percent) for accidents, malignant neoplasms, and Alzheimer’s disease? Repeat these calculations for males and females separately.

c. Calculate the maternal mortality rate (per 100,000 live births). d. Calculate the infant mortality rate (per 1,000 live births). e. Calculate the crude birth rate (per 1,000 population). f. Calculate the general fertility rate (per 1,000 women aged 15–44

years). 5. The population of Metroville was 3,187,463 on June 30, 2013. During

the period January 1 through December 31, 2013, a total of 4,367 city residents were infected with HIV. During the same year, 768 new cases of HIV were reported. Calculate the prevalence per 100,000 population and incidence per 100,000 population.

6. Give definitions of the terms prevalence and incidence. What are appro- priate uses of prevalence and incidence data? State the relationships among prevalence, incidence, and duration of a disease.

7. Suppose that “X” represents the name of a disease. An epidemiologist conducts a survey of disease “X” in a population. The prevalence of disease “X” among women is 40/1,000 and among men is 20/1,000. Assuming that the data have been age adjusted, is it correct to conclude that women have twice the risk of disease “X” as men? Explain.

8. The following data regarding alcohol drinking status among persons in the United States were reported for 2005:

Number in thousands

All persons 18 years of age and older

Current regular alcoholic beverage drinkers

Male 104,919 59,300 Female 112,855 44,373

a. What is the sex ratio of male to female regular alcoholic beverage drinkers?

b. What proportion (percent) of regular alcoholic beverage drinkers are women?

c. What is the prevalence per 1,000 of regular alcoholic beverage drink- ing among men only, women only, and the total population aged 18 and older?

51589_CH03_Printer.indd 154 09/02/13 12:21 PM

r e f e r e n C e s 155

9. During 2005, the following statistics were reported regarding the frequency of diabetes, ulcers, kidney disease, and liver disease:

Diabetes 7% of adults had ever been told by their doctor that they had diabetes Ulcers 7% had ever been told by their doctor that they had an ulcer Kidney 2% had been told in the past 12 months that they had kidney disease Liver 1% had been told in the past 12 months that they had liver disease

Which of the foregoing statistics were stated as incidence data and which as prevalence data? a. Diabetes b. Ulcers c. Kidney disease d. Liver disease

10. The National Health Interview Survey reported the percent of respon- dents with a hearing problem by age group during 2005:

Age (years) Reporting a hearing problem, %

18–44 8.2 45–64 19.2 65–74 30.4 75+ 48.1

Would it be correct to state that the risk of hearing loss increases with age? Be sure to explain and defend your answer.

11. During January 1 through December 31, 2008, epidemiologists con- ducted a prevalence survey of type 2 diabetes; 500,000 cases were detected in a population of 10,000,000 persons. It was known that the incidence of diabetes in this population was 10 per 1,000. Estimate the percentage of the prevalent cases that were newly identified during the year.

12. The sex ratio for the entire United States was less than 100, indicating that there were more females than males. The sex ratio at birth exceeded 1.0, denoting a greater number of male births to female births. How could one account for the difference between the sex ratio for the United States and sex ratio at birth?

References 1. Porta M, ed. A Dictionary of Epidemiology, 5th ed. New York, NY: Oxford University

Press; 2008. 2. Hennekens CH, Buring JE. Epidemiology in Medicine. Boston, MA: Little,

Brown; 1987. 3. Folsom AR, Kaye SA, Sellers TA, et al. Body fat distribution and 5-year risk of death

in older women. JAMA. 1993;269:483–487.

51589_CH03_Printer.indd 155 09/02/13 12:21 PM

156 C h a p t e r 3 M e a s u r e s o f M o r b i d i t y a n d M o r ta l i t y

4. Sellers TA, Kushi LH, Potter JD, et al. Effect of family history, body-fat distribution, and reproductive factors on the risk of postmenopausal breast cancer. N Engl J Med. 1992;326:1323–1329.

5. Link BG, Susser E, Stueve A, Phelan J, Moore RE, Struening E. Lifetime and five-year prevalence of homelessness in the United States. Am J Public Health. 1994;84:1907–1912.

6. Mausner JS, Kramer S. Mausner & Bahn Epidemiology–An Introductory Text, 2nd ed. Philadelphia, PA: Saunders; 1985.

7. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic Research: Principles and Quantitative Methods. Belmont, CA: Lifetime Learning; 1982.

8. Rothman KJ. Modern Epidemiology. Boston, MA: Little, Brown; 1986. 9. National Center for Health Statistics. Health, United States, 2010. With Special Feature

on Death and Dying. Hyattsville, MD: National Center for Health Statistics; 2011. 10. MacDorman MF, Kirmeyer S. The challenge of fetal mortality. U.S. Department of

Health and Human Services, Centers for Disease Control and Prevention, National Center for Health Statistics. NCHS Data Brief No 16 April 2009.

11. U.S. Department of Health and Human Services, Health Resources and Services Administration, Maternal and Child Health Bureau. Child Health USA 2011. Rockville, Maryland: U.S. Department of Health and Human Services; 2011.

12. World Health Organization (WHO). World Health Statistics 2011. Geneva, Switzerland: World Health Organization; 2011.

13. National Center for Health Statistics. Health, United States, 2010. With Special Feature on Death and Dying. Hyattsville, MD: National Center for Health Statistics; 2011.

14. Xu J, Kochanek KD, Murphy SL, Tejada-Vera B. Final data for 2008. National Vital Statistics Reports. 58(19). Hyattsville, MD: National Center for Health Statistics; 2010.

15. World Health Organization. Neonatal and perinatal mortality: Country, regional and global estimates. Geneva, Switzerland: World Health Organization; 2006.

16. Anderson RN, Rosenberg HM. Age standardization of death rates: Implementation of the Year 2000 Standard. National Vital Statistics Reports. 47(3). Hyattsville, MD: National Center for Health Statistics; 1998.

51589_CH03_Printer.indd 156 09/02/13 12:21 PM

3chapte r

4chapte r

157

Descriptive Epidemiology: Person, Place, Time

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● state the three primary objectives of descriptive epidemiology ●● provide examples of the main subtypes of descriptive studies ●● list at least two characteristics of each person, place, and time, and

provide a rationale for why they are associated with variations in health and disease

●● characterize the differences between descriptive and analytic epidemiology

●● describe the difference between secular trends and cohort effects

CHAPTER OUTLINE

I. Introduction II. Characteristics of Persons

III. Characteristics of Place IV. Characteristics of Time V. Conclusion

VI. Study Questions and Exercises

51589_CH04_Printer.indd 157 09/02/13 12:34 PM

158 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Introduction

The basic premise of epidemiology is that disease does not occur randomly but rather in patterns that reflect the operation of underlying factors. This chapter surveys methods for describing disease patterns, which, in general, fall into one or more of three categories: person, place, or time. The category person encom- passes who is being affected: young versus old? males versus females? rich versus poor? migrants versus nonmigrants? more educated versus less educated? Place relates to where the problem is occurring: in cities more than in rural areas? some states more than others? in the United States versus other counties? at high altitudes versus low altitudes? where there is plentiful rainfall or little rainfall? in polluted areas more than unpolluted areas? Time refers to when the problem is occurring: Was there a sudden increase over a short period of time? Is the prob- lem greater in winter than in summer? Is the problem gradually increasing over long periods of time or increasing greatly over just a few years? The occurrence of disease with respect to the characteristics of person, place, and time is central to the field of descriptive epidemiology.

Epidemiologists need to describe patterns of disease occurrence carefully and accurately in order to discover etiologic clues. Throughout this chapter, the authors will identify descriptive characteristics (e.g., age, sex, and race—all person characteristics) that help to delineate patterns of disease and generate hypotheses regarding their underlying causes. The variables—person, place, and time—directly or indirectly relate to the occurrence of illnesses by affecting a wide range of exposures associated with lifestyle, behavioral patterns, healthcare access, and exposure to environmental hazards, to name a few examples. In illus- tration, being male is more likely to be associated with unintentional death or injury than being female; behavior patterns characteristic of a particular racial or ethnic group may affect subcultural levels of stress and methods for coping with social stresses. Sometimes race and ethnicity are referred to as social group membership factors that affect perceptions, interpretations, and reactions to the social and physical environment.1 Membership in specific social groups relates to the occurrence of exposures to health hazards and how people view, interpret, and respond to hazards. We shall observe also that combinations of variables, for example, age and sex, are noteworthy.

Descriptive Versus Analytic Epidemiology Epidemiologists distinguish between two broad categories of epidemiologic studies: descriptive and analytic. Descriptive studies characterize the amount and

51589_CH04_Printer.indd 158 09/02/13 12:34 PM

i n t r o D u C t i o n 159

distribution of disease within a population. Analytic studies, on the other hand, explore the determinants of disease—the causes of relatively high or low frequency of diseases in specific populations. Determinants are variables such as infectious agents, environmental exposures, and risky behaviors. Descriptive studies generally precede analytic studies: The former are used to identify any health problems that may exist, and the latter proceed to identify the cause(s) of the problem.

Objectives of Descriptive Epidemiology The authors identify three broad objectives of descriptive epidemiology; these are to:

1. permit evaluation of trends in health and disease and comparisons among countries and subgroups within countries; this objective includes monitoring of known diseases as well as the identification of emerging problems

2. provide a basis for planning, provision, and evaluation of health services; data needed for efficient allocation of resources often come from descrip- tive epidemiologic studies

3. identify problems to be studied by analytic methods and suggest areas that may be fruitful for investigation

The acquired immune deficiency syndrome (AIDS) epidemic illustrates how these objectives are implemented. Descriptive data on the epidemiology of AIDS provided an indication of the emergence of the epidemic (objective 1), were useful for allocation of hospital beds and treatment centers (objective 2), and spurred etiologic studies into why intravenous drug users and gay and bisexual men were more likely than other groups to develop the disease (objective 3).

Descriptive Studies and Epidemiologic Hypotheses The third objective relates to the use of descriptive epidemiology to aid in the cre- ation of hypotheses. “For any public health problem, the first step in the search for possible solutions is to formulate a reasonable and testable hypothesis.”2(p 112) “Hypotheses are suppositions that are tested by collecting facts that lead to their acceptance or rejection. They are not assumptions to be taken for granted, nei- ther are they beliefs that the investigator sets out to prove. They are ‘refutable predictions.’”3(p 40) Three common ways of stating hypotheses are as follows:3

1. Positive declaration (research hypothesis): The infant mortality rate is higher in one region than another.

51589_CH04_Printer.indd 159 09/02/13 12:34 PM

160 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

2. Negative declaration (null hypothesis): There is no difference between the infant mortality rates of two regions. As part of significance testing, statisticians use the term alternative hypothesis to signify that the null hypothesis is false.

3. Implicit question: To study the association between infant mortality and geographic region of residence.

Hypotheses should be made as explicit as possible and not left as implicit.

Mill’s Canons What is the source of hypotheses that guide epidemiologic research? The logical processes for deriving hypotheses are patterned after John Stuart Mill’s canons of inductive reasoning.4 The following are four of his canons:

1. the method of difference 2. the method of agreement 3. the method of concomitant variation 4. the method of residues

The method of difference All of the factors in two or more domains are the same except for a single factor. The frequency of a disease that varies across the two settings is hypothesized to result from variation in the single causative factor. The method of difference has been employed widely in epidemiologic research. It is similar to classic experi- mental design. A hypothetical example of its use would be the study of the role of physical activity in reducing morbidity from coronary heart disease (CHD). Suppose groups of workers in the same factory were compared when the factors of age, diet, socioeconomic status, and other variables were constant (equivalent for all groups). It is plausible that workers in a particular factory might have simi- lar sociodemographic and lifestyle characteristics but differ greatly in physical activity levels on the job. The hypothesis would be that differences in morbidity from CHD are due to level of physical activity, that is, sedentary workers are at greater risk of developing heart attacks than physically active workers.

The method of agreement A single factor is common to a variety of different settings. It is hypothesized that the common factor is a cause of the disease. Wherever air pollution is present, for example, the prevalence of chronic respiratory diseases, such as asthma and

51589_CH04_Printer.indd 160 09/02/13 12:34 PM

i n t r o D u C t i o n 161

emphysema, tends to increase. This observation leads to the hypothesis that air pollution, if present, is a contributing factor to lung diseases.

The method of concomitant variation The frequency of a disease varies according to the potency of a factor. This linked association suggests that the factor is the causative agent for the disease. An example confirmed by numerous studies is the direct relationship between the incidence of lung diseases (e.g., bronchitis, emphysema, and lung cancer) and the number of cigarettes smoked: The more cigarettes smoked, the greater the risk of incurring lung cancer and other lung diseases.

The method of residues The method of residues involves subtracting potential causal factors to deter- mine which individual factor or set of factors makes the greatest impact upon a dependent variable. In research on heart disease, statistical methods similar to the method of residues (e.g., multiple regression analysis) have been used to determine which of a number of risk factors may be associated with coronary attack or death from CHD. The individual contribution to CHD of one’s hered- ity, diet, stress level, amount of exercise, and blood lipid level can be quantified. One then can determine which factor has the greatest impact.

The method of analogy (additional criterion from MacMahon and Pugh5) The distribution of a disease of unknown etiology bears a pattern similar to that of a known disease, which epidemiologists have investigated more thor- oughly.5 This information suggests that the known and unknown diseases share certain similar causes. The field of infectious diseases provides many examples of the method of analogy. Diseases of unknown origin that spread during the summer and that are confined to certain geographic areas could be disseminated by a vector (e.g., mosquito). Another illustration is Legion- naires’ disease; the symptoms of Legionnaires’ disease were similar to those produced by an infectious respiratory disease agent of either viral or bacte- rial origin. Epidemiologists discovered that a bacterium causes Legionnaires’ disease, the bacterium lived in water (e.g., in air conditioning cooling tow- ers), and could become aerosolized, causing possible airborne transmission. According to MacMahon, one situation in which the method of analogy was applied (sometimes incorrectly) was in the hypothesized relationship between

51589_CH04_Printer.indd 161 09/02/13 12:34 PM

162 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

smoking and tuberculosis based on the analogy of smoking and lung cancer; a second was for the hypothesis of an infectious agent for multiple sclerosis (MS) because of the similarity of the geographic distribution of MS with that of polio.

Three Approaches to Descriptive Epidemiology The three approaches to descriptive epidemiology are case reports (counts), case series, and cross-sectional studies. Sometimes counts are helpful for mak- ing epidemiologic descriptions of diseases; usually they are more informative when expressed relative to a denominator (as a proportion). One could view groups of case reports (counts) as the simplest form of descriptive epidemiol- ogy. Astute clinical observations of unusual cases may spur additional investi- gations to determine whether large numbers of cases with similar presentations exist; also, case reports are a starting point for exploring underlying causal mechanisms and introducing preventive interventions. Here is an example of a case report:

Methylene chloride, a volatile chemical used in paint strippers, produces a highly toxic vapor. Workers who use methylene chloride-based strippers must wear protective equipment such as respirators. The Centers for Disease Control and Pre- vention reported a death associated with use of a chemical for stripping bathtubs. “In March 2010, the co-owner of a Michigan-based bathtub refinishing com- pany, aged 52 years, was refinishing a bathtub in an apartment bathroom that was approximately 5 feet by 8 feet . . . He was using an aircraft paint stripper product that contains 60–100% methylene chloride . . . The [bathroom] fan was off. The man wore latex gloves and did not wear respiratory protection . . . Approximately 90 minutes after the man began working on the tub, he did not answer a call to his cellular telephone. An apartment maintenance man entered the apartment to look for the man and found him behind the closed bathroom door, unresponsive, and slumped over the tub . . . [After treatment by emergency responders and transpor- tation to a hospital], the man was declared dead at the hospital.6

The second approach is a case series. Because of the difficulty in drawing firm conclusions from a single case report, one may wish to expand a single observation to a series of cases. Typically this series involves a summary of the characteristics of a consecutive listing of patients from one or more major clinical settings. From a set of observations one is able to generate summary measures to help distill typical features. An example is from the CDC, which reported a series of five cases of hantavirus pulmonary syndrome (HPS) among five pediat- ric patients during 2009.7 HPS is highly fatal condition associated with contact with rodents. [Three years later, an unusual hantavirus outbreak among visitors

51589_CH04_Printer.indd 162 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 163

to Yosemite National Park resulted in 9 confirmed cases (three of whom died) as of mid-September 2012.] CDC’s fives cases in 2009 were:

Case 1: (May 16, 2009)—a boy aged 6 years who lived in Colorado was hospitalized for a severe illness. Within 2 hours of admission to the hospital, the boy died from apparent cardiac failure secondary to shock. Rodent drop- pings and nesting materials were found under the boy’s bed and near his home. Case 2: (June 7, 2009)—an adolescent boy aged 14 years went to a Washington emergency department with a 5-day history of shortness of breath, chest pain, cough, and fever. The boy subsequently recovered. Rodent fecal contamination was found in a container of corn that the youth reported hand grinding. Case 3: (July 12, 2009)—a boy aged 6 years went to a Colorado emergency department with a 5-day history of fever and later recovered. The boy had been bitten on the finger by a mouse. Case 4: (July 12, 2009)—a girl aged 9 years who lived in Arizona went to a New Mexico hospital with chest pain and shortness of breath. The patient was hospitalized until August 5. Evidence of rodents was found in several houses that the girl had frequented. Case 5: (November 25, 2009)—an adolescent boy aged 13 years went to a Cali- fornia emergency department with a five-day history of fever and other symptoms. The patient recovered after a period of hospitalization. Mice had been trapped in the youth’s kitchen and garage approximately three months before disease onset.

The third major approach to descriptive epidemiology utilizes cross-sectional studies conducted at various times. These are surveys of the population to esti- mate the prevalence of a disease or exposures. One can sometimes use data from repeated cross-sectional surveys at different points in time to examine time trends in prevalence of disease or risk factors. The National Health Interview Survey operated by the CDC is an example of a cross-sectional study.

Characterist ics of Persons

Age Age is perhaps the most important factor to consider when one is describing the occurrence of virtually any disease or illness, because age-specific disease rates usually show greater variation than rates defined by almost any other personal attribute. For this reason, public health professionals often use age-specific rates when comparing the disease burden among populations. Trends in mortality

51589_CH04_Printer.indd 163 09/02/13 12:34 PM

164 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

from the leading causes of death fluctuated markedly according to age group.8 Data for 2007 reveal the following trends for specific causes:9

●● Unintentional injuries—the leading cause of death for persons aged 1–44 years

●● Homicide and suicide—important causes of death for the 1–44 years age group

●● Stroke, influenza, and pneumonia—among the 10 leading causes of death for the 1–44 years age group

●● Chronic diseases (e.g., cancer and heart disease)—leading causes of death among person aged 45 years and older

●● Human immunodeficiency virus (HIV)—among the leading causes in the 20–44 age group

The following sections present information regarding trends in mortality for specific age groups.

Childhood to early adolescence (1–14 years): For infants, developmen- tal problems such as congenital birth defects and immaturity are among the major causes of death. Incidence of some infectious and communicable diseases (e.g., otitis media (ear infections), measles, mumps, chicken pox, and meningo- coccal disease) tend to occur most commonly in childhood.

Figure 4–1 presents data on the burden of disease (number of cases) and rate per 100,000 for meningococcal disease for 2000–2009. The condition has a bimodal distribution with peaks in incidence among infants younger than 1 year of age and among adolescents (at about 18 years of age as shown in part B of the figure). The inset in part A (1989–1991 data) illustrates the decreasing incidence of meningococcal disease from early infancy to about age 2. During the 10-year interval, 2000–2009, about 16% of cases were among infants and 20% among adolescents and young adults (part A).

During 1997 to 2007, unintentional injuries (inappropriately called acci- dents) were the leading cause of death among persons aged 1–14 years. Other important causes of death were cancer, congenital malformations, and heart dis- ease. (See Figure 4–2, part A.)

Teenage years through young adulthood (15–24 years): In 2007, among teenagers and young adults, the three leading causes of death were unintentional injuries, homicide, and suicide. Cancer and heart disease were the fourth and fifth causes, respectively. (See Figure 4–2, part B.)

Additional health-related conditions that impinge upon teenagers are unplanned pregnancy; and tobacco use, substance abuse, and binge drinking affect

51589_CH04_Printer.indd 164 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 165

Figure 4–1 Meningococcal disease: Part A. Rates of meningococcal disease by age group and burden of disease, United States, 2000–2009. Inset: Incidence of meningococcal disease, by age group (0 to 23 months), in selected U.S. areas during 1989–1991. Part B. Projected rates of meningococcal disease [caused by the four most common serogroups] by age, United States, 2000–2009. Sources: Reproduced from Centers for Disease Control and Prevention. Vaccines and Immunizations. Fact sheet: Meningococcal Disease and Meningococcal Vaccine. Available at: http://www.cdc.gov/ vaccines/ vpd-vac/mening/vac-mening-fs.htm. Accessed August 26, 2012 and (Inset) Centers for Disease Control and Prevention. Laboratory-based surveillance for meningococcal disease in selected areas–United States, 1989–1991, MMWR, Vol 42, No SS-2, p 25, June 4, 1993.

Age Group (years)

Age (Months)

<1 > = 641 Part A

B u

rd en

o f

D is

ea se

( %

)

In ci

d en

ce (

ca se

s/ 10

0, 00

R at

e p

er 1

00 ,0

35 5

4.5

3.5

2.5

1.5

0.5

2–4 5–9 10–13 14–17

≤3 4–7

1989–1991 Data

8–11 12–15 16–19 20–23

18–24 25–64

Burden of Disease Rate per 100,000

11 0

0.5

12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

R at

e p

er 1

00 ,0

Age (years)Part B

51589_CH04_Printer.indd 165 09/02/13 12:34 PM

166 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Figure 4–2 Death rates for leading causes of death by age group: United States, 1997–2007. Source: Reproduced from National Center for Health Statistics. Health, United States, 2010: With Special Feature on Death and Dying. Hyattsville, MD, 2011, pp 37–40.

1997 1999 Year

D ea

th s

pe r

10 0,

00 0

po pu

la tio

2001 2003 2005 2007

8 Unintentional injuries

Homicide

Heart disease

Cancer Congenital malformations

Unintentional injuries

Homicide

Heart disease

Suicide Cancer

1997 1999

D ea

th s

pe r

10 0,

00 0

po pu

la tio

2001 2003 2005 2007

Year

Unintentional injuries

Homicide HIV

Heart disease

Suicide

Cancer

D ea

th s

pe r

10 0,

00 0

po pu

la tio

0 1997 1999

Year 2001 2003 2005 2007

Unintentional injuries Diabetes

StrokeCLRD

Heart disease

Cancer

300

200

100

1997 1999 Year

D ea

th s

pe r

10 0,

00 0

po pu

la tio

2001 2003 2005 2007 0

A Deaths, 1–14 Years of Age B Deaths, 15–24 Years of Age

C Deaths, 25–44 Years of Age D Deaths, 45–64 Years of Age

both teenagers and young adults. Alcohol, marijuana, and tobacco are the drugs of choice among youths aged 12–17 years.10 In addition, almost one-twelfth of young persons in this age range abuse prescription drugs such as opioids, central nervous system depressants, and stimulants.

Also, adverse lifestyle choices impinge upon this age group. Because many teenagers enjoy playing with the latest electronic gadgets, excessive “screen time” may lead to obesity, a risk factor for subsequent development of chronic diseases. The widespread availability of sugary beverages vended in giant containers and high fat foods leads young people to opt for these food items.

51589_CH04_Printer.indd 166 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 167

Adults (25–44 years): In this age group, unintentional injuries were the leading cause of death during 1997 to 2007. Three causes—cancer, heart disease, and sui- cide—were among the leading causes of mortality, as were homicide and HIV dis- ease. Among men, lung, brain, and colon cancers were the most common causes of cancer death. The leading causes among women were breast, lung, and cervical can- cers. Death rates for suicide and homicide were three times as high among men as among women. Since the 1990s, the death rates for HIV disease (the sixth leading cause of death in 2007) have declined by more than 50%. (See Figure 4–2, part C.)

Older adults (45–65): Chronic diseases such as heart disease and can- cer dominate as sources of morbidity and mortality after age 45. (See Figure 4–2, part D.) Although these conditions also impact younger persons, they are the leading causes of death among older adults. For example, cancer incidence increases with age, as demonstrated by the generally linear increase in age-spe- cific incidence rates after age 45 among all racial groups (Figure 4–3).

The elderly (65+): The five leading causes of death in 2007 were diseases of the heart, malignant neoplasms, cerebrovascular diseases, chronic lower respira- tory diseases (CLRD), and Alzheimer’s disease.8 Note that the apparent decline in cancer incidence (Figure 4–3) around age group 80–84 is somewhat decep- tive because the sizes of the numerators and denominators for the very elderly categories are smaller than for other age groups, resulting in unstable estimates. Unintentional injuries ranked number eight among the leading causes of death.

As a result of declining physical and mental health, some elderly persons experi- ence impairment of their ability to live independently. In 2010, approximately 4% of seniors aged 65–74 years and 11% aged 75 years and older suffered from limitations in activities of daily living (ADLs). (Refer to Figure 4–4.) The term ADL refers to “need[ing] the help of other persons with personal care needs, such as eating, bath- ing, dressing, or getting around inside [the] home . . . [b]ecause of a physical, mental, or emotional problem.”11(p 249) Limitations of instrumental activities of daily living (IADLs) were even more frequent. The definition of IADLs is “need[ing] the help of other persons in handling routine needs, such as everyday household chores, doing necessary business, shopping or getting around for other purposes . . . [b]ecause of a physical, mental, or emotional problem.”11(p 249) More than 6% of persons aged 65–74 years and 19% of persons aged 75 years and older had limitations in IADLs.

MacMahon and Pugh5 suggested four reasons for age associations. These are:

●● the validity of diagnoses across the life span ●● multimodality (e.g., bimodality) of trends ●● latency effects ●● action of the “human biologic clock”

51589_CH04_Printer.indd 167 09/02/13 12:34 PM

168 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

The validity of diagnoses across the life span may be affected by classification errors. Age-specific incidence rates may be inaccurate among older age groups, causing distortions in the shape of an age-incidence graph. Inaccuracies may result from the difficulty in fixing the exact cause of death among older individu- als, who may be afflicted concurrently with a number of sources of morbidity.

Some health conditions show multimodal age-specific incidence curves, mean- ing that there are several peaks and declines in the frequency of the diseases at

Figure 4–3 Age-specific (crude) SEER incidence rates for all cancer sites by race and sex during 1992–2008 in the United States. Source: Reproduced from Fast Stats: An interactive tool for access to SEER cancer statistics. Surveillance Research Program, National Cancer Institute. Available at: http://seer.cancer.gov/faststats. Accessed August 26, 2012.

3500

3000

2500

2000

1500

1000

500

<1 1– 4

5– 9

10 –1

15 –1

20 –2

25 –2

30 –3

35 –3

40 –4

45 –4

50 –5

55 –5

60 –6

4 65

–6 9

70 –7

75 –7

80 –8

4 85

White Black American Indian/Alaska Native Non-Hispanic White

Asian/Pacific Islander Hispanic

Age at Diagnosis

R at

e pe

r 10

0, 00

51589_CH04_Printer.indd 168 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 169

various ages. Two examples are meningococcal disease (discussed previously) and tuberculosis (TB), which has two peaks, one between age 0 and 4 years and another around age 20 to 29 years. A third example is Hodgkin’s disease, which shows a peak in the mid-20s and another in the early 70s.5 Bimodal (two-peak) distributions for these conditions may suggest two different causal mechanisms. For example, in the case of TB, the increase in prevalence of the disease in the early years of life may be due to the increased susceptibility of children to infectious dis- eases, and the other peak during young adulthood may reflect the increased social interaction of individuals at this age or change in immune status due to puberty.

Another explanation for age effects on mortality is that they reflect the long latency period between environmental exposures and the development of cer- tain diseases. An example of a long latency period is the passage of many years

Figure 4–4 Percentage of adults with activity limitations, by age group and type of limitation–National Health Interview Survey, United States, 2010.* Source: Reproduced from Centers for Disease Control and Prevention, MMWR 2012:61(14): 249.

18–44 yrs 45–64 yrs 65–74 yrs >–75 yrs

Limitations in instrumental activities of daily living (IADLs)¶

Type of limitation

† 95% confidence interval. § Limitations in ADLs are based on response to the question, “Because of a physical, mental, or emotional problem, does [person] need the help of other persons with personal care needs, such as eating, bathing, dressing, or getting around inside this home?” ¶ Limitations in IADLs are based on response to the question, “Because of a physical, mental, or emotional problem, does [person] need the help of other persons in handling routine needs, such as everyday household chores, doing necessary business, shopping, or getting around for other purposes?”

Limitations in activities of daily living (ADLs)§

P er

ce nt

ag e

0 †

* Estimates are based on household interviews of a sample of the civilian non- institutionalized U.S. population. Persons with unknown limitation status were excluded from the denominators.

51589_CH04_Printer.indd 169 09/02/13 12:34 PM

170 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

between initial exposure to a potential carcinogen and the subsequent appearance of cancer later in life. Furthermore, older individuals in comparison with younger persons have had a greater opportunity to be exposed to multiple potential car- cinogens. The additive effects of such exposures would be more likely to affect older persons than younger individuals.

The “human biologic clock” phenomenon refers to an endogenous process associated with increased vulnerability to disease. For example, as a component of the aging process the immune system may wane, producing increased tissue susceptibility to disease. Another manifestation of the biologic clock is the trig- gering of conditions that are believed to have a genetic basis (e.g., Alzheimer’s disease, which usually occurs among older persons).

Other causes (not specified by MacMahon and Pugh) of age-related changes in rates of morbidity and mortality are related to life cycle and behavioral phe- nomena. As noted previously, unintentional injuries, homicide, and suicide as causes of mortality differ greatly in importance according to age group; varia- tions in these three causes of death are influenced by factors such as personal behavior and risk taking, especially among the young. Lifestyle influences the occurrence of diabetes and other chronic diseases, many of which are believed to have significant behavioral components. Some aging-associated problems, which impact the far end of the age distribution, illustrate life cycle phenomena. As the elderly population continues to increase, more epidemiologic studies will be needed on the topics of the “retirement syndrome,” the bereavement process, falls, and other behaviorally related health issues among the aged.

Sex/Gender Numerous epidemiologic studies have shown sex differences in a wide scope of health phenomena, including mortality. The following discussion presents data on sex differences in mortality. With the exception of some calendar years, the population age-adjusted death rate has declined in the United States since 1980.12 Males generally have higher all-cause age-specific and age-adjusted mor- tality rates than females from birth to age 85 and older;13 the ratio of male to female age-specific death rates in 2007 was 1.4 to 1.14

A classic study noted that male versus female morbidity differences were the reverse of the differences for mortality—females were reported to have higher age-standardized morbidity rates for acute conditions, chronic conditions, and disability due to acute conditions.14 This phenomenon is known as the female paradox. Women suffer from higher rates of pain, some respiratory ailments such as asthma and lung difficulties not induced by cancer, and depression. Problems

51589_CH04_Printer.indd 170 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 171

that are more common among men are hearing impairment, smoking-associated conditions, and cardiovascular diseases.15 Men who are affected by the same chronic diseases (e.g., lung cancer, cardiovascular disease, and diabetes) as women are more likely to develop severe forms of these conditions and die from them. Other research has suggested that sex differences in morbidity have nar- rowed and some conditions, such as hypertension and cancer, produce increased morbidity rates among men in comparison with women.16

Speculations regarding the sources of sex differences in morbidity and mortal- ity are fascinating and capable of inspiring heated debate. An interesting ques- tion concerns the extent to which sex differences in mortality will narrow as the lifestyle, employment, and health-related behaviors of women become more similar to those of men. Specific research studies have investigated genetic and environmental factors, differentials in exposure to stress, reporting of illness, and the effects of women’s changing role in society upon mortality. Waldron’s17 ven- erable research attributed higher male mortality to greater frequency of smoking, a greater prevalence of the coronary-prone behavior pattern, higher suicide and motor vehicle accident rates, as well as risky behavioral patterns that are expected of and condoned among men. Consider the example of lung cancer mortality. Data for this cause of mortality, especially between 1975 and 1990, show that it increased among women much faster than among men, supporting the view that certain behavioral and lifestyle variables (i.e., smoking behavior) may relate to male/female lung cancer mortality differences (Figures 4–5 and 4–6). A Dan- ish study isolated smoking as the major factor that explained sex differences in mortality in that country.18

Coronary heart disease (CHD) is the leading cause of death for both men and women. However, sex differences in mortality from CHD persist between men and women, even when both have high-risk factor status for serum cholesterol, blood pressure, and smoking. This finding implicates important biologic param- eters as the basis for the observed differences (e.g., differences in hormonal pro- files). Production of estrogen changes during menopause; the result is that heart disease typically does not leave its mark on women until after age 60. Among women, endogenous estradiol is strongly implicated in cardiovascular changes that are similar to the effects of exercise. These effects cause the “jogging female heart” that may account for the lower incidence of cardiovascular disease before menopause and postmenopausal increases in rates of cardiac disease.19

Many women are unaware of the fact that they may be at high risk of cardiac disease.20 Consequently, women may not be alert for symptoms of CHD, caus- ing delay in seeking treatment when symptoms occur. Women also may resist

51589_CH04_Printer.indd 171 09/02/13 12:34 PM

172 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Figure 4–5 Age-adjusted cancer death rates:* males by site, United States, 1930–2008. Source: Reproduced from American Cancer Society, Cancer Facts & Figures 2012. Atlanta: American Cancer Society, Inc.; 2012, p 2.

100

0 1930

Stomach Colon & rectum Prostate

Lung & bronchus

PancreasLeukemia

Liver

1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

R at

e pe

r 10

0, 00

0 m

al e

po pu

la tio

Note: Due to changes in ICD coding, numerator information has changed over time. Rates for cancer of the liver, lung, and colon and rectum are affected by these coding changes.

*Per 100,000 age adjusted to the 2000 US standard population

Figure 4–6 Age-adjusted cancer death rates:* females by site, United States, 1930–2008. Source: Reproduced from American Cancer Society, Cancer Facts & Figures 2012. Atlanta: American Cancer Society, Inc.; 2012, p 3.

Colon & rectum

100

0 1930

Stomach

Lung & bronchus

Pancreas

Breast

1935 1940 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005

R at

e pe

r 10

0, 00

0 fe

m al

e po

pu la

tio n.

Ovary

Uterus†

†Uterus cencer death rates are for uterine cervix and uterine corpus combined. Note: Due to changes in ICD coding, numerator information has changed over time. Rates for cancer of the lung and brounchus, colon and rectum, and ovary are affected by these coding changes.

*Per 100,000 age adjusted to the 2000 US standard population.

51589_CH04_Printer.indd 172 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 173

lifestyle changes such as increased activity level and consumption of low-fat food. Concern also has been raised about poorer care for women who present with cardiac disease. Although women who have myocardial infarctions may receive slightly different treatment from men in some respects (e.g., lower use of aspirin and greater use of angiotensin-converting enzyme inhibitors), multivariate analy- ses of data from the Cooperative Cardiovascular Project and the National Heart Failure Project as well as other databases did not disclose major gender biases among patients 65 years of age and older.21

In some regions of the United States, particularly the economically disadvan- taged areas, minority women face a higher burden of morbidity from chronic dis- eases than men. For example, minority women who live in Los Angeles County confront higher rates of diabetes, hypertension, and elevated cholesterol. During 2005, almost one-half of all women in the county reported little physical activity. The frequency of obesity was high, particularly among Latinas (more than 25%) and African Americans (about 33%).22

Healthy People 2020 seeks to eliminate health disparities and health ineq- uities among persons who are lesbian, gay, bisexual, or transgender (LGBT).23 According to Healthy People 2020’s section on topics and objectives, “[r]esearch suggests that LGBT individuals face health disparities linked to societal stigma, discrimination, and denial of their civil and human rights. Discrimination against LGBT persons has been associated with high rates of psychiatric disor- ders, substance abuse, and suicide. Experiences of violence and victimization are frequent for LGBT individuals, and have long-lasting effects on the individual and the community. Personal, family, and social acceptance of sexual orienta- tion and gender identity affects the mental health and personal safety of LGBT individuals.”24

Marital Status Marital status includes the following categories: single or nonmarried (never married, divorced, separated, and widowed), married, and living with a partner. In general, epidemiologic research has shown that married individuals, especially men, have lower rates of morbidity and mortality than those who are single, divorced, or widowed. Also, data suggest that among older women, divorce and separation are associated with adverse health outcomes such as physical impair- ments.25 Schoenborn reported that “Regardless of population subgroup (age, sex, race, Hispanic origin, education, income, or nativity) or health indicator (fair or poor health, limitations in activities, low back pain, headaches, serious psychological distress, smoking, or leisure-time physical inactivity), married

51589_CH04_Printer.indd 173 09/02/13 12:34 PM

174 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

adults were generally found to be healthier than adults in other marital status categories.”26(p 1) The exception was that among the various categories of marital status, never married adults were least likely to be overweight or obese; being married was associated with obesity, especially among men.26

An influential analysis of U.S. nationwide trends showed lower rates of mor- tality from chronic diseases among married individuals for coronary diseases and many forms of cancer as well as suicide, motor vehicle crashes, and some infectious diseases.27 Swedish data indicated that being unmarried or divorced was linked with higher rates of mortality than those observed among married persons.28 Among Danish younger males, marital breakups, divorce, and wid- owhood were reported to be independent predictors of mortality.29 A 2007 Jap- anese study reported similar mortality trends: In the Japan Collaborative Cohort Study (a prospective study of 94,062 Japanese men and women aged 40–79), single individuals experienced a higher mortality risk than married persons. Fur- thermore, divorce and widowhood affected men and women differently; men had higher mortality risks than women if divorced or widowed.30

Another dimension of marital status is its relationship with suicide rates, which appear to be elevated among widowed persons, especially young men. In a national study of suicide in the United States, investigators reported a 17-fold elevation in suicide rates among widowed white men who were aged 20–34 years and a 9-fold elevation among young African-American men in comparison with married men in the same age group; widows did not have similarly elevated sui- cide rates.31

One of the correlates of suicide is depressed mood, which is characterized by the sense that “everything is an effort” as well as feelings of sadness, hopelessness, and worthlessness. The National Health interview Survey (2010) found that these four aspects of depressed mood (all or most of the time) were more com- mon among widowed individuals than among persons who were married, living with a partner, never married, or divorced/separated. In comparison, married persons had the lowest percentages of depressed mood.32 (Refer to Figure 4–7.)

Schottenfeld33 presented data on the risk of developing breast cancer among single women compared with ever-married women and for ever-married nul- liparous women compared with parous married women. Married women had a reduced risk of breast cancer mortality in comparison with single women, and among all married women childbirth slightly reduced the risk. Following Schottenfeld’s groundbreaking research, other investigators reported that even controlling for stage of cancer diagnosis, married women had lower breast cancer mortality rates than nonmarried women.34

51589_CH04_Printer.indd 174 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 175

A theoretical account of the action of marital status upon health posits that marriage may operate as either a protective or a selective factor.35 The protective hypothesis suggests that marriage makes a positive contribution to health by influencing lifestyle factors, providing mutual psychological and social support, and increasing available financial resources. According to the marital selection model, physically attractive individuals are more likely to compete for a part- ner successfully and are healthier than those persons who never marry, resulting in the lower morbidity and mortality rates observed among married persons. Furthermore, the model proposes that less healthy individuals, if married, are more predisposed than healthy persons to gravitate to nonmarried status. In summary, the marital environment and factors associated with marriage appar- ently reduce the risk of death and, therefore, should be considered possible sources of differences in disease rates.35 Currently, perceptions of marital status have changed within the American society, lessening the stigma that was once

Figure 4–7 Age-adjusted percentages of feelings of sadness, hopelessness, worthlessness, or that everything is an effort (all or most of the time) among persons aged 18 years and over, by selected characteristics: United States, 2000. Source: Reproduced from Schiller JS, Lucas JW, Ward BW, Peregoy JA. Summary health statistics for US adults: National Health Interview Survey, 2010. National Center for Health Statistics. Vital Health Stat 10 (252). 2012.

8.2

7.3

9.5

9.8

Married

Widowed

Divorced/separated

Never married

Living with partner

4.61.31.5

3.3

4.75.7

4.5

2.6

2.02.6

3.5

6.1

11.0

2.2

Sadness Hopelessness Worthlessness Everything is an effort

2.1

1.6

51589_CH04_Printer.indd 175 09/02/13 12:34 PM

176 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

associated with divorce and living together without being officially married. These changing perceptions may impact associations between marital status and health in the future.

During the final three decades of the 20th century, differentials in the self- rated health status of married and never-married persons tended to narrow. However, in comparison with married persons, the health of widowed, divorced, and separated individuals decreased. These effects appeared to be more salient for women than for men.36

Race and Ethnicity Increasingly, with respect to race and ethnicity, the United States is becoming more diverse than at any time in history (Figure 4–8). Race and ethnicity are, to some extent, ambiguous characteristics that tend to overlap with nativity and

Figure 4–8 Racial and ethnic diversity. Source: Reprinted with permission from Nebraska Department of Health and Human Services, Office of Health Disparities and Health Equity. Improving health outcomes for Nebraska’s Culturally Diverse Populations. Available at: http://dhhs.ne.gov/publichealth/ Pages/healthdisparities_index.aspx. Accessed August 26, 2012.

51589_CH04_Printer.indd 176 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 177

religion. Scientists have proposed that race is a social and cultural construct rather than a biological construct.37 In Census 2000, the U.S. Bureau of the Census classified race into five major categories: white; black or African American; American Indian and Alaska Native; Asian; and Native Hawaiian and other Pacific Islander. To a degree, race tends to be synonymous with ethnicity because people who come from a particular racial stock also may have a com- mon ethnic and cultural identification. Also, assignment of some individuals to a particular racial classification on the basis of observed characteristics may be dif- ficult. Often, one must ask the respondent to elect the racial group with which he or she identifies. Classification of persons of mixed racial parentage also may be problematic.38 The 2000 Census allowed respondents to check a multiracial category, which was used for the first time. Changes in the definitions of racial categories affect the denominators (i.e., the numbers in a particular racial sub- group) of rates used to track various health outcomes and the consequent assess- ments of unmet needs and social inequalities in health.39 In general, the 2010 Census continued with this classification scheme, as discussed in Exhibit 4–1.

Overview of race and hispanic Origin, census 2010

UNDERSTANDING RACE AND HISPANIC ORIGIN DATA FROM THE 2010 CENSUS

The 2010 Census used established federal standards to collect and present data on race and Hispanic origin.

For the 2010 Census, the questions on race and Hispanic origin were asked of individuals living in the United States (see Figure 4–8). An individual’s responses to the race question and to the Hispanic origin question were based upon self-identification. The U.S. Census Bureau collects race and Hispanic origin information following the guidance of the U.S. Office of Management and Budget’s (OMB) 1997 Revisions to the Standards for the Classification of Federal Data on Race and Ethnicity. These federal standards mandate that race and Hispanic origin (ethnicity) are separate and distinct concepts and that when collecting these data via self-identification, two different questions must be used.

e x

h ib

it 4

–1

continues

51589_CH04_Printer.indd 177 09/02/13 12:34 PM

178 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

NOTE: Please answer BOTH Question 5 about Hispanic origin and Question 6 about race. For this census, Hispanic origins are not races.

5. Is this person of Hispanic, Latino, or Spanish origin? No, not of Hispanic, Latino, or Spanish origin Yes, Mexican, Mexican Am., Chicano Yes, Puerto Rican Yes, Cuban Yes, another Hispanic, or Spanish origin—Print origin, for example, Argentinian, Colombian, Dominican, Nicaraguan, Salvadoran, Spaniard, and so on.

6. What is this person’s race? Mark ✗ one or more boxes. White Black, African Am., or Negro American Indian or Alaska Native—Print name of enrolled or principal tribe. Asian Indian Japanese Native Hawaiian Chinese Korean Guamanian or Chamorro Filipino Vietnamese Samoan Other Asian—Print race, for example, Hmong, Laotian, Thai, Pakistani, Cambodian, and so on.

Other Pacific Islander—Print race, for example, Fijian, Tongan, and so on.

Some other race—Print race.

Figure 4–8 Reproduction of the questions on Hispanic origin and race from the 2010 census. Source: Adapted and reprinted from U.S. Census Bureau, 2010 Census questionnaire.

Hispanic Origin

The OMB definition of Hispanic or Latino origin used in the 2010 Census is presented in the text box “Definition of Hispanic or Latino Origin Used in the 2010 Census.” OMB requires federal agencies to use a minimum of two ethnicities: Hispanic or Latino and Not Hispanic or Latino. Hispanic origin can be viewed as the heritage, nationality group, lineage, or country of birth of the person or the person’s parents or ancestors before their arrival in the United States. People who identify their origin as Hispanic, Latino, or Spanish may be any race.

The 2010 Census question on Hispanic origin included five sepa- rate response categories and one area where respondents could write-in a specific Hispanic origin group. The first response category is intended

exhibit 4–1 continued

continues

51589_CH04_Printer.indd 178 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 179

for respondents who do not identify as Hispanic. The remaining response categories (“Mexican, Mexican Am., or Chicano”; “Puerto Rican”; “Cuban”; and “Another Hispanic, Latino, or Spanish origin”) and write-in answers can be combined to create the OMB category of Hispanic.

Race

The OMB definitions of the race categories used in the 2010 Census, plus the Census Bureau’s definition of Some Other Race, are pre- sented in the text box “Definition of Race Categories Used in the 2010 Census.” Starting in 1997, OMB required federal agencies to use a minimum of five race categories: White, Black or African American, American Indian or Alaska Native, Asian, and Native Hawaiian or Other Pacific Islander. For respondents unable to identify with any of these five race categories, OMB approved the Census Bureau’s inclu- sion of a sixth category—Some Other Race—on the 2000 and 2010 Census questionnaires.

Data on race have been collected since the first U.S. decennial census in 1790. For the first time in Census 2000, individuals were presented with the option to self-identify with more than one race and this continued with the 2010 Census, as prescribed by OMB. There are 57 possible multiple race combinations involving the five OMB race categories and Some Other Race.

The 2010 Census question on race included 15 separate response cat- egories and three areas where respondents could write-in detailed informa- tion about their race. The response categories and write-in answers can be combined to create the five minimum OMB race categories plus Some Other Race. In addition to White, Black or African American, American

exhibit 4–1 continued

“Hispanic or Latino” refers to a person of Cuban, Mexican, Puerto Rican, South or Central American, or other Spanish culture or origin regardless of race. n

Definition of Hispanic or Latino Origin Used in the 2010 Census

continues

51589_CH04_Printer.indd 179 09/02/13 12:34 PM

180 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Indian and Alaska Native, and Some Other Race, 7 of the 15 response categories are Asian groups and 4 are Native Hawaiian and Other Pacific Islander groups.

Definition of Race Categories Used in the 2010 Census

“White” refers to a person having origins in any of the original peoples of Europe, the Middle East, or North Africa. It includes people who indicated their race(s) as “White” or reported entries such as Irish, German, Italian, Lebanese, Arab, Moroccan, or Caucasian.

“Black or African American” refers to a person having origins in any of the Black racial groups of Africa. It includes people who indicated their race(s) as “Black, African American, or Negro” or reported entries such as African American, Kenyan, Nigerian, or Haitian.

“American Indian or Alaska Native” refers to a person having origins in any of the original peoples of North and South America (including Central America) and who maintains tribal affiliation or community attachment. This category includes people who indicated their race(s) as “American Indian or Alaska Native” or reported their enrolled or principal tribe, such as Navajo, Blackfeet, Inupiat, Yup’ik, or Central American or South American Indian groups.

“Asian” refers to a person having origins in any of the original peoples of the Far East, Southeast Asia, or the Indian subcontinent, including, for example, Cambodia, China, India, Japan, Korea, Malaysia, Pakistan, the Philippine Islands, Thailand, and Vietnam. It includes people who indi- cated their race(s) as “Asian” or reported entries such as “Asian Indian,” “Chinese,” “Filipino,” “Korean,” “Japanese,” “Vietnamese,” and “Other Asian” or provided other detailed Asian responses.

“Native Hawaiian or Other Pacific Islander” refers to a person having origins in any of the original peoples of Hawaii, Guam, Samoa, or other Pacific Islands. It includes people who indicated their race(s) as “Pacific Islander” or reported entries such as “Native Hawaiian,” “Guamanian or Chamorro,” “Samoan,” and “Other Pacific Islander” or provided other detailed Pacific Islander responses.

exhibit 4–1 continued

continues

51589_CH04_Printer.indd 180 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 181

“Some Other Race” includes all other responses not included in the White, Black or African American, American Indian or Alaska Native, Asian, and Native Hawaiian or Other Pacific Islander race categories described above. Respondents reporting entries such as multiracial, mixed, interracial, or a Hispanic or Latino group (for example, Mexican, Puerto Rican, Cuban, or Spanish) in response to the race question are included in this category.

RACE AND HISPANIC ORIGIN IN THE 2010 CENSUS Data from the 2010 Census provide insights to our racially and ethni- cally diverse nation. According to the 2010 Census, 308.7 million people resided in the United States on April 1, 2010—an increase of 27.3 million people, or 9.7 percent, between 2000 and 2010. The vast majority of the growth in the total population came from increases in those who reported their race(s) as something other than White alone and those who reported their ethnicity as Hispanic or Latino.

More than half of the growth in the total population of the United States between 2000 and 2010 was due to the increase in the Hispanic population.

In 2010, there were 50.5 million Hispanics in the United States, compos- ing 16% of the total population (see Table 1–1). Between 2000 and 2010, the Hispanic population grew by 43%—rising from 35.3 million in 2000, when this group made up 13% of the total population. The Hispanic popu- lation increased by 15.2 million between 2000 and 2010, accounting for over half of the 27.3 million increase in the total population of the United States.

The non-Hispanic population grew relatively slower over the decade, about 5%. Within the non-Hispanic population, the number of people who reported their race as White alone grew even slower between 2000 and 2010 (1%). While the non-Hispanic White alone population increased numerically from 194.6 million to 196.8 million over the 10-year period, its proportion of the total population declined from 69% to 64%. (Refer to Table 1-1.) n

exhibit 4–1 continued

continues

51589_CH04_Printer.indd 181 09/02/13 12:34 PM

182 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Assuming that researchers have created valid measures of race, this variable does have implications for health disparities (differences in incidence and preva- lence of disease by race/ethnicity), as numerous epidemiologic studies have deter- mined. Noteworthy variations in the rates of disease and risk factors for disease have been identified by using race as a variable in epidemiologic and public health research.40 Socioeconomic status and migration history appear to be important influences in health disparities among racial groups.41 Further definitions and explanations of the race questions used in Census 2010 are in Exhibit 4–1.

Overall trends in mortality according to race and ethnicity Figure 4–9 illustrates distributions in the leading causes of mortality by race and ethnicity. Refer to Figure 4–9 for information on the percentage distributions

Table 1–1 Population by Hispanic or Latino Origin and by Race for the United States: 2000 and 2010

Hispanic or Latino origin and race

2000 2010

Number

Percentage of total

population Number

Percentage of total

population

Total population 281,421,906 100.0 308,745,538 100.0 Hispanic or Latino 35,305,818 12.5 50,477,594 16.3 Non-Hispanic or Latino 246,116,088 87.5 258,267,944 83.7 White alone 194,552,774 69.1 196,817,552 63.7 RACE Total population 281,421.906 100.0 308,745,538 100.0 One Race 274,595,678 97.6 299,736,465 97.1 White 211,460,626 75.1 223,553,265 72.4 Black or African American 34,658,190 12.3 38,929,319 12.6 American Indian/Alaska Native 2,475,956 0.9 2,932,248 0.9 Asian 10,242,998 3.6 14,674,252 4.8 Native Hawaiian/Pacific

Islander 398,835 0.1 540,013 0.2

Some other race 15,359,073 5.5 19,107,368 6.2 Two or more races 6,826,228 2.4 9,009,073 2.9

Source: Adapted and reprinted from United States Census Bureau. Overview of Race and Hispanic Origin: 2010. 2010 Census Briefs. March 2011.

Source: Reproduced from U.S. Census Bureau. Available at: http://www.census.gov/prod/ cen2010/briefs/c2010br-02.pdf. Accessed March 5, 2011.

exhibit 4–1 continued

51589_CH04_Printer.indd 182 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 183

of the 10 leading causes of death by race and ethnicity. Highlights of the distributions of the major causes of death are:8

●● Heart disease—leading cause of death among white, black, and American Indian or Alaskan Native populations.

●● Cancer—leading cause of death among the Asian or Pacific Islander (API) population.

●● Chronic lower respiratory disease (CLRD)—third most common cause of death for the white population.

●● Stroke—third most common cause of death for the black and API populations and fourth for the white population.

Figure 4–9 Percent distribution of the 10 leading causes of death, by race and ethnicity: United States, 2007. Source: Reproduce from Heron M. Deaths: Leading causes for 2007. National vital statistics reports. Vol 59, no 8. Hyattsville, MD: National Center for Health Statistics. 2011.

White

Other 23.1

Other 25.3

Heart disease 25.6

Heart disease 24.6

Cancer 23.3

Cancer 22.1

CLRD 5.7

CLRD

HIV

Septicemia

Stroke 5.5

Other 26.4

Other 21.9

5.5 4.9

4.3

4.1

2.7 2.0 1.9

Heart disease 18.4

Cancer 17.8

Cancer 27.0

Heart disease 23.2

CLRD

Suicide

Stroke 5.9

Stroke 7.9

Unintentional injuries

11.8

Unintentional injuries

Alzheimer’s disease

Diabetes

Diabetes Chronic

liver disease and cirrhosis

Stroke

Homicide

Influenza and Pneumonia

Kidney disease

Alzheimer’s disease

Kidney disease

Suicide

5.1

3.3

2.7 2.2

1.8 1.5

2.2 2.2 2.7

2.9 3.1

4.3

1.6 2.0

2.9 3.8

4.8

4.7

Black

American Indian or Alaska Native Asian or Pacific Islander

51589_CH04_Printer.indd 183 09/02/13 12:34 PM

184 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

African Americans According to a study of differential mortality in the United States, African Ameri- cans had the highest mortality of any of several racial groups examined.27 In 2007, the African-American population experienced an age-adjusted death rate that was 1.3 times that of the white population.12 African Americans in the United States are one of the groups that are afflicted by disparities with respect to many health conditions. Persons who self-identify as non-Hispanic black carry a greater bur- den from mortality and morbidity as well as injury and disability in comparison with non-Hispanic whites.42 Included in the 10 leading causes of death among non-Hispanic blacks are homicide, HIV, and septicemia, whereas these condi- tions are not among the 10 leading causes of death among non- Hispanic whites. Figure 4–10, which portrays AIDS cases among five racial/ethnic groups, shows that African Americans had the greatest number of AIDS diagnoses in 2009; also they had the highest HIV incidence rate (44.27 per 100,000).43

The age-adjusted incidence of certain forms of cancer is much higher among African Americans than among whites. In comparison with four other racial/ ethnic groups (whites, American Indians/Alaska Natives, Asians/Pacific Islanders,

Figure 4–10 Human immunodeficiency virus diagnoses. Percentage of diagnosed cases, by race/ethnicity–United States, 2009. Source: Reproduced from Centers for Disease Control and Prevention, Summary of notifiable diseases– United States, 2009, MMWR. Vol 58, No 53, p 63, 2011.

Asian/Pacific Islander, non-Hispanic (1.6%) Multiple races (1.0%)

American Indian/Alaska Native, non-Hispanic (0.4%)

Which group is most affected?

HIV—2009

Hispanic (18.8%)

White, non-Hispanic (29.7%)

Black, non-Hispanic (48.5%)

51589_CH04_Printer.indd 184 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 185

and Hispanics), African Americans have the highest age-adjusted female breast cancer death rate.44 Possible explanations for this disparity include differences in access to and quality of mammography services as well as access to treat- ment for breast cancer.45 African-American males have twice the death rate for prostate cancer as white males, a finding that points to a need for improvement in early detection and treatment of prostate cancer among the former group. Figure 4–11 presents life expectancy by race and sex in the United States between 1971 and 2007. African-American males had the lowest life expectancy rates of the four groups shown; nevertheless, differences in life expectancy between the African-American and white populations have tended to narrow over time.

1971 1975 1979 1983 1987 1991 1995 1999 20072003 Year

Black male

Black female

White male

White female

70A ge

in y

ea rs

Figure 4–11 Life expectancy at birth, by race and sex: United States, 1970–2007. Source: Reproduced from Arias E. United States life tables, 2007. National vital statistics reports. Vol 59, No 9. Hyattsville, MD: National Center for Health Statistics. 2011.

51589_CH04_Printer.indd 185 09/02/13 12:34 PM

186 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

The prevalence of hypertension is substantially higher among African Americans than among whites. The National Health and Nutrition Examina- tion Surveys for 1992 to 2002 obtained rates of hypertension among African- American and white adults 20 years of age and older of 40.5% and 27.4%, respectively.46 African Americans develop hypertension at younger ages and suffer from more severe health consequences than whites.47 Mortality from hypertension-related conditions is higher among African Americans than among whites. For example, the 2003 deaths rates from hypertensive heart disease among African Americans and whites were 17.9 versus 8.8 deaths per 100,000, respectively. The death rates from hypertension, including hypertensive renal disease, were 11.8 and 7.2 deaths per 100,000, respectively.8 A number of fac- tors could account for increased rates of hypertension among African Americans, including dietary factors (e.g., low consumption of fruits and vegetables), expo- sure to stress, reduced social support, higher rates of obesity, and lack of partici- pation in cardiovascular risk reduction programs.

American Indians/Alaska Natives American Indian/Alaska Native (AI/AN) adults have high rates of chronic dis- eases, adverse birth outcomes, and infectious diseases such as TB and hepatitis A in comparison with the general U.S. population. AI/ANs also have decreased life expectancy.48 During the late 20th century, hospitalizations for infectious dis- eases represented nearly one-quarter of all hospitalizations among elderly AI/ AN adults. Although the rates of such hospitalizations increased slightly in the United States between 1990 and 2002, those rates increased even more (by about one-fifth) in Alaska and the Southwest.49 In comparison with AI/AN females, AI/AN males bear a heavier burden from many illnesses and tend to underutilize healthcare services.50

Knowler et al.51 published a seminal study of the incidence and prevalence of diabetes mellitus in nearly 4,000 members of the Pima tribe (a group of North American Indians native to Arizona) aged 5 years and older over a 10-year period. The investigators reported a diabetes prevalence of about 21%, adjusting for age and sex; the incidence rate was about 26 cases per 1,000. Diabetes incidence was about 19 times greater than that of a predominantly white comparison popula- tion in Rochester, Minnesota.

During 1975–1984, the Pima Indians who resided in the Gila River Indian community had a death rate of 1.9 times that for all races in the United States. Among men aged 25–34 years, the Pima death rate was 6.6 times that for all races

51589_CH04_Printer.indd 186 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 187

in the United States. Diseases of the heart and malignant neoplasms accounted for 59% of the U.S. deaths in 1980, but only 19% for the American Indian community. The age- and sex-adjusted mortality rate was 5.9 times the rate for all races in the United States for accidents (unintentional injuries), 6.5 times for cirrhosis of the liver, 7.4 times for homicide, 4.3 times for suicide, and 11.9 times for diabetes. TB and coccidioidomycosis also were important causes of death in the Pima, for whom infectious disease was the 10th leading cause of death among all causes of death.52 As a group, the general AI/AN population in 2002 had an age-adjusted TB case rate that was twice that of the U.S. population and seven times that of the non-Hispanic white population.53

The number of AI/ANs who live in cities is increasing. Urban AI/ANs expe- rience disparities in health and socioeconomic characteristics when contrasted with the general U.S. population.54 Among women, these disparities include poorer birth outcomes due to inadequate prenatal care and sudden infant death syndrome associated with alcohol consumption. As a group, AI/ANs tend to be poorer, have lower rates of college graduation, and have higher levels of unem- ployment than other urban residents.

Asians/Pacific Islanders The Japanese, in comparison with other racial groups in the United States, have low mortality rates: one-third the rates for whites of both genders.27 Japanese culture seems to afford a protective influence that results in lower mortality, especially from chronic diseases such as CHD and cancer. The orientation of the Japanese culture even in our age of industrialization is toward conformity and group consensus rather than toward the “rugged individualism” and competi- tiveness that pervade the American culture.55 Degree of acculturation to Japan was related to low rates of CHD mortality.55 Acculturation is defined as modi- fications that individuals or groups undergo when they come into contact with another culture.56

According to Marmot,57 “Among industrialised countries, Japan is remark- able for its low rate of ischaemic heart disease. It is unlikely to be the result of some genetically-determined protection, as Japanese migrants to the USA lose this apparent protection.”57(p 378) The Honolulu Heart Study prospectively fol- lowed a large population of men of Japanese ancestry who resided on the island of Oahu at the beginning of the study.58,59 Various measures of the degree of early exposure to Japanese culture were collected, including birthplace in Japan; total number of years of residence in that country; ability to read, write, and

51589_CH04_Printer.indd 187 09/02/13 12:34 PM

188 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

speak Japanese; and a preference for the Japanese diet. After adjusting for the influence of well-established risk factors for CHD (age, serum cholesterol, sys- tolic blood pressure, and cigarette smoking), there was a gradient in incidence of CHD across variables related to identification with the Japanese culture. For example, men who could read and write Japanese well had an incidence rate about half that reported for those who could neither read nor write Japanese. Other studies of Japanese men living in Japan, Hawaii, and California have shown an increasing gradient in mortality, prevalence, and incidence of CHD from Japan to Hawaii to California. Observed lower rates of CHD in Japan in comparison with the United States have been attributed to a low-fat diet among the Japanese and to institutionalized stress-reducing strategies (e.g., community bonds and group cohesion) within Japanese society.60

Studies of acculturation among the Japanese provide evidence that envi- ronmental and behavioral factors influence chronic disease rates and provide a rationale for intervention and prevention of chronic disease.61 The Japanese who migrated shared a common ethnic background. After migrating to diverse geo- graphic and cultural locales, they experienced a shift in rates of chronic disease to rates more similar to those found in the host countries. This finding among the Japanese is consonant with the acculturation hypothesis, which proposes that as immigrants become acculturated to a host country, their health profiles tend to converge with that of the native-born population.

In the case of the United States, the originating culture of migrants sometimes affords protection against morbidity and mortality. This protective effect may be a function of health-related behaviors associated with a culture. During the late 20th century, foreign-born persons lived about 2 to 4 years longer than the native-born U.S. population; however, the risk of disability and chronic condi- tions grew as immigrants lived in the United States for longer periods of time.62

Some Asian groups have high rates of smoking when compared with the gen- eral U.S. population. Among all Asians, rates of cigarette smoking tend to be higher among men than among women. One Asian group that is thought to have high smoking rates is Cambodian Americans, who have rates as high as 70%; in addition, the frequency of smoking among Cambodian American men is reported to be three to four times the frequency among women.63

Asians had the highest tuberculosis (TB) rates of five racial and ethnic groups, although the incidence declined among Asians by 6.5% between 2010 and 2011 to 3.4 per 100,000. The case rate for Asians was 21.4 per 100,000 in 2011; this case rate was almost 25 times that of non-Hispanic whites. Figure 4–12 shows TB incidence by race/ethnicity between 1999 and 2009.64

51589_CH04_Printer.indd 188 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 189

Hispanics/Latinos In the United States, Hispanic/Latino populations include the major groups— Mexican Americans, Puerto Ricans, and Cubans—as well as other groups that have migrated to the United States from Latin America. In comparison with other non-Hispanics, Hispanics/Latinos have unique morbidity and mortal- ity profiles. Although Mexican Americans represent one of the dominant eth- nic minorities in the southwestern United States, this group has been relatively under-researched for prevalence of hypertension, CHD, diabetes, and other chronic diseases. The first special population survey of Hispanics in the United States was the Hispanic Health and Nutrition Examination Survey (HHANES). Conducted by the National Center for Health Statistics, HHANES assessed the health and nutritional status of Mexican Americans, mainland Puerto Ricans, and Cuban Americans.65 An entire supplementary issue of the American Journal of Public Health covered findings from HHANES.66 As more research is con- ducted among Latinos, it is becoming apparent that they are highly diverse and should be studied as distinct subpopulations (e.g., Cuban Americans or Salva- doran Americans). Low rates of CHD among Mexican Americans may be due to

Figure 4–12 Tuberculosis incidence* by race/ethnicity: United States, 1995–2009. Source: Reproduced from Centers for Disease Control and Prevention, Summary of notifiable diseases–United States, 2011, MMWR. Vol 58, No 53, p 78, 2011.

In ci

de nc

e 40

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Year

Black, non-Hispanic American Indian/Alaska Native, non-Hispanic Hispanic White, non-Hispanic Asian/Pacific Islander, non-Hispanic

*Per 100,000 population.

51589_CH04_Printer.indd 189 09/02/13 12:34 PM

190 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

cultural factors, such as dietary preferences and the availability of social support mechanisms found in large and extended family systems. A study of CHD among Puerto Ricans reported a low prevalence of this condition.67

A major epidemiologic investigation of diabetes and other cardiovascular risk factors in Mexican Americans and non-Hispanic whites is the San Antonio Heart Study.68 Among the findings were the high prevalence of obesity and noninsulin- dependent diabetes mellitus among the Mexican-American population. Despite a higher prevalence of diabetes mellitus and other risk factors for chronic disease, Hispanics/Latinos in the United States have a lower mortality rate than non- Latino whites. The mortality differential is sometimes referred to as the Hispanic mortality paradox (refer to the text box).

In the United States, Hispanics have a lower mortality rate than non-Hispanic whites and African Americans: 28.5% lower than the rate for the non-Hispanic white population and 44.7% lower than that of the non-Hispanic black population.12 This differential in mortality is sur- prising, because Hispanics as a group tend to have a less-advantaged socioeconomic profile than non-Hispanic whites.

The cause-specific mortality of Hispanics differs from that of non- Hispanic whites, who experience more years of potential life lost due to lung cancer. In contrast, Latinos have more years of life lost to HIV and diabetes among both men and women and to homicide and liver disease among men.69

How is it possible for Hispanics to have lower overall mortality rates than non-Hispanic whites? Part of the difference is due to underreport- ing of deaths among Hispanics. Another proposed explanation for the Hispanic mortality paradox is the “salmon bias effect,” where persons of Hispanic heritage who have immigrated to the United States may return to their original countries, where they die. Thus, they become “statisti- cally immortal.”70

Research has examined mortality differences among Hispanic sub- groups that are able to migrate, those that are restricted from migrating,

Hispanic (Latino) Mortality Paradox

continues

51589_CH04_Printer.indd 190 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 191

and those from territories of the United States. For example, Mexican Americans are able to return to their homeland, whereas Cuban Americans are limited in their ability to return to their country of birth. Mortality statistics for Puerto Ricans are included with U.S. national mortality sta- tistics. Thus, it is most feasible to test the salmon bias hypothesis among Mexican Americans. Support for the “salmon bias effect” is suggested for Mexican Americans, although the effect is by no means clear-cut and requires further research.70,71 Nevertheless, Cuban Americans and Puerto Ricans also maintain lower mortality rates than non-Hispanic whites, a difference that has not yet been fully explained.70,72

At present, the reasons for the Hispanic mortality paradox have not been elucidated fully. Two other possible explanations are the healthy migrant effect (discussed further in the next section) and the accultura- tion hypothesis, which suggests that the cultural orientation of Hispanics is associated with protective health behaviors. Later, these protective behaviors wane as Hispanics become increasingly acculturated to the United States.73 n

Figure 4–13 presents the distribution of the 10 leading causes of death among Hispanics (2007 data). Heart disease was the leading cause of death, followed by cancer in second place. Unintentional injuries, stroke, and diabetes were the third, fourth, and fifth leading causes of death, respectively.

Although in 2007 cancer was the second leading cause of deaths among the Hispanic population, and Hispanics have higher rates of some forms of can- cer, they have lower rates of participation in cancer screening programs. Dur- ing 2004 through 2008, the rate of cervical carcinoma among Hispanic women was 11.3 per 100,000 in comparison with 7.4 per 100,000 for non-Hispanic women. Despite this higher rate, Hispanic females had lower rates of screening for cervical cancer than did other racial and ethnic groups.

In New York City schools, surveys of schoolchildren found that Hispanics (among four racial/ethnic groups) had the highest prevalence of obesity (defined as a body mass index [BMI] above the 95th percentile). The surveys were con- ducted as part of a fitness program among kindergarten through eighth grade classes during each of five school years. Figure 4–14 shows the results of the surveys: Even though levels of obesity declined slightly over the five-year period, they were consistently higher among Hispanics.

51589_CH04_Printer.indd 191 09/02/13 12:34 PM

192 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Figure 4–13 Percent distribution of the 10 leading causes of death, Hispanics: United States, 2007. Source: Reproduced from Heron M. Deaths: Leading causes for 2007. National vital statistics reports. Vol 59, No 8. Hyattsville MD: National Center for Health Statistics. 2011.

Heart disease 21.4

Cancer 20.4

8.7 Stroke

5.2

4.7 2.9

2.6 2.6

2.2 2.0

Influenza and pneumonia

Certain perinatal conditions

Homicide

CLRD

Chronic liver disease and cirrhosis

Diabetes Unintentional injuries

Other 27.3

Nativity and Migration Nativity refers to the place of origin of the individual or his or her relatives. A common subdivision used in epidemiology is foreign-born or native-born. Thus, nativity is inextricably tied to migration because foreign-born persons have immigrated to their host country. As a result, nativity and migration fre- quently overlap the epidemiologic categories of race, ethnicity, and religion, because streams of immigrants may bring their cultural and religious practices to the new land or may comprise a common racial background.

The phenomenon of migration meets the criteria for a natural experiment in which the effects of change from one environment can be studied. For example, migration research has examined various health dimensions, including stress, acculturation, chronic disease, and infectious disease. Classic epidemiologic research conducted in the late 1930s examined rates of admission to mental hos- pitals in New York State. Admission rates were higher among foreign-born than native-born persons, suggesting that foreign-born individuals may experience stresses associated with migration to a new environment.5

51589_CH04_Printer.indd 192 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 193

In the United States, rates of some communicable diseases (e.g., tuberculosis) are higher among foreign-born persons than among U.S.-born persons. (Refer to Figure 4–15.) During the last two decades (before 2010), the TB rate among foreign-born persons as well as among U.S.-born persons has trended downward. The number of TB cases among foreign-born persons has tended to remain sta- ble, whereas the number of cases has declined among U.S.-born persons.

Another impact of migration has been to modify the profile of infectious diseases seen in public health departments throughout the developed world. Some immigrants from Southeast Asia and Mexico may import “Third World” diseases to the United States. For example, local health departments in Southern California have found that intestinal parasites, malaria, and certain other tropical diseases may occur among newly arrived immigrants from endemic areas in devel- oping countries; the same conditions are rare in the resident U.S. population. Likewise, the number of cases of Hansen’s disease (leprosy) rose dramatically along with increased immigration from Southeast Asia between 1978 and 1988. Figure 4–16 shows data on the number of reported cases of leprosy in the United States between 1970 and 2009.

Figure 4–14 Prevalence of obesity among public school children in grades K-8 who were aged 5–14 years, by school year and selected characteristics–New York City, 2006–07 to 2010–11 school years. Source: Data from Centers for Disease Control and Prevention, Obesity in K-8 students—New York City, 2006–07 to 2010–11 school years. MMWR. Vol 60, No 49, p 1675, 2011.

28.0

26.0 P

er ce

nt ag

24.0

22.0

20.0

18.0

16.0

14.0

12.0

10.0 2006–07 2007–08 2008–09 2009–10 2010–11

26.5 26.0 25.4 25.7 25.6

20.9

15.4

13.4

21.121.221.3

17.6 16.9

21.1

16.1 16.1

13.513.213.7 14.5

Hispanic BlackAsian/Pacific Islander White

51589_CH04_Printer.indd 193 09/02/13 12:34 PM

194 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Figure 4–16 Hansen’s disease (leprosy). Number of reported cases, by year: United States, 1970–2009. Source: Reproduced from Centers for Disease Control and Prevention, Summary of notifiable diseases–United States, 2005, MMWR. Vol 54, No 53, p 55, 2007; and from Centers for Disease Control and Prevention, Summary of notifiable diseases–Unite States, 2009, MMWR. Vol 58, No 53, p 60, 2011.

1970

400

360

320

280

N um

be r

200

240

160

120

0 1975 1980

Influx of refugees from Cambodia, Laos, and Vietnam,

1978–1988

1985 1990 Year

1995 2000 2005 2009

Figure 4–15 Number and rate of tuberculosis (TB) cases among U.S.-born and foreign-born persons, by year reported–United States, 1993–2011. Source: Reproduced from Centers for Disease Control and Prevention, Trends in tuberculosis–United States, 2011, MMWR. Vol 61, No 11, p 183, 2012.

N um

be r

(in th

ou sa

nd s)

0 1993 1995 1997 1999 2001 2003

Year

2005 2007 2009 2011 0

R ate (per 100,000 population)

TB rate among foreign-born persons TB rate among U.S.-born persons

No. of TB cases among U.S.-born persons No. of TB cases among foreign-born persons

51589_CH04_Printer.indd 194 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 195

Another impact of migration of people from developing countries has been the need to establish specialized TB and nutritional screening programs for ref- ugees and the need to reeducate physicians with respect to formerly uncom- mon (in the United States) tropical diseases. The inadequate immunization status of some migrants and refugees with respect to measles and other vaccine- preventable diseases has hampered the efforts of health officials to eradicate these conditions in the United States.

Researchers who strive to examine the health of migrants are confronted with significant methodologic challenges. One is the difficulty in separating environmental influences in the host country from selective factors opera- tive among those who choose to migrate. The term healthy migrant effect acknowledges the observation that healthier, younger persons usually form the majority of migrants. Nevertheless, despite methodologic difficulties, such as the healthy migrant effect, migration research is a fascinating area that has already yielded many intriguing findings. For a review, refer to Friis, Yngve, and Persson.74

Religion Religious beliefs also may be salient for rates of morbidity and mortality that are observed in the population. For example, adherents to some religious denomina- tions that prescribe particular lifestyles may demonstrate characteristic morbid- ity or mortality profiles. In addition, religiosity itself is believed to have health impacts. A thoughtful review of the subject of religion and health pointed out that the epidemiologic literature on this topic is extensive, with several hun- dred epidemiologic studies reporting statistically significant associations between religious indicators and morbidity and mortality.75 While there is definitive evidence of an association among these dimensions, the question remains unan- swered as to whether the association is causal. For some patients who are con- fronting a chronic disease, religion and spirituality may improve the quality of their lives. Healthcare providers can reinforce the coping skills of cancer patients by recognizing their religious needs.76,77 We provide several examples of work in this area that examine the occurrence of morbidity and mortality among popula- tion subgroups classified by religious membership.

The Seventh-Day Adventist church endorses a lacto-ovo vegetarian diet, which consists of meat, poultry, or fish less than once per week with no restric- tion on egg or dairy consumption. Members are encouraged to abstain from alcohol, tobacco, and pork products.78 The low rates of CHD observed among this religious group suggest that the corresponding lifestyle has health benefits.

51589_CH04_Printer.indd 195 09/02/13 12:34 PM

196 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Seventh-Day Adventists also have been reported to have low mortality rates from other chronic diseases. Armstrong et al.79 studied Seventh-Day Adventists in western Australia and found that mean systolic and diastolic blood pressures were significantly lower than those of the comparison population. Phillips80 reported findings of reduced risk of cancer and other chronic diseases among the Seventh-Day Adventist population in the United States. The latter research was conducted as part of the Adventist Mortality Study and the Adventist Health Study 1 (1974–1988).

The Adventist Health Study 2 was initiated in 2002 for the purpose of obtaining information on the association of dietary and other lifestyle practices with cancer risk.81 A preliminary analysis using these data compared black and white Adventist respondents with regard to disease and lifestyle characteristics. Although the study found similarities and discrepancies (sometimes favoring whites and other times favoring blacks) between the two racial groups with respect to health characteristics, the health profile of black Adventists tended to be superior to that of blacks nationally.82

Jarvis83 summarized findings on Mormons and other groups and presented data on Mormon mortality rates in Canada. He noted that, compared with the general U.S. population, Canadian Mormons have lower incidence and mortal- ity rates due to cancer and other diseases. The study of mortality rates among Mormons in Alberta, Canada, generally confirmed mortality findings for the United States. Jarvis speculated that mortality differences may be due to restric- tions on the intake of coffee, tea, and meats, and lifestyle variables related to physical fitness, social support, and a stress-reducing religious ideology. Data on Mormons in Alameda County, California, corroborate findings from previous descriptive studies of an unusually low risk for cancer.84

Socioeconomic Status Social class variations in health have been observed, formally and informally, since the beginning of organized society. The relationship between socioeco- nomic status (SES) and health is remarkably consistent, having been dem- onstrated for a wide range of health outcomes and confirmed by a massive body of evidence.85 Health outcomes related to SES include impaired cog- nitive functioning (e.g., mild mental retardation and low SES status), men- tal disorders, infections, disability, and mortality. Particularly noteworthy are social class differences in mortality, with persons lower in the social hierarchy having higher mortality rates than do persons in upper levels.86 Link et al.

51589_CH04_Printer.indd 196 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 197

wrote, “This dynamic connection between SES and risk factors has led to the observation of a persistent association between SES and mortality. We refer to social conditions whose association with mortality persists in this manner as ‘fundamental social causes’ of inequalities in health.”87(p 377) Enduring low income is especially consequential for mortality.88 Berkman and Syme89 con- cluded that low social class standing is related to excess mortality, morbidity, and disability rates. Some of the more obvious explanations for the negative health effects of low social class membership are poor housing, crowding, racial disadvantages, low income, poor education, and unemployment. A complex web of factors includes exposure to environmental and work-related hazards, both material and social deprivation, lack of access to health care, and negative lifestyle. Socioeconomic factors also may play a role in the association of race and ethnicity with health.90, 41

There is a striking consistency in the distribution of mortality and morbidity between social groups. The more advantaged groups, whether expressed in terms of income, education, social class or ethnicity, tend to have better health than the other members of their societies. The distribution is not bipolar (advantaged vs. the rest) but graded, so that each change in the level of advantage or disadvantage is in general associated with a change in health.91(p 903)

Measurement Much of the terminology of social class that is used in epidemiologic research has been derived from sociology, in particular the branch of sociology dealing with social class and social stratification. Social class also is related to ethnicity, race, religion, and nativity. This is because some ethnic and other minority groups often occupy the lowest social class rankings in the United States. There are sev- eral different measures of social class that draw upon the individual’s economic position in society. Such measures include the prestige of the individual’s occupa- tional or social position, educational attainment, income, or combined indices of two or more of these variables. Occupational prestige is often employed as a mea- sure of social class. For example, learned professionals (e.g., physicians, college professors, lawyers, and similar occupational groups) are accorded the highest occupational prestige, and other occupations are ranked below them. A measure of occupational prestige derived by the British Registrar General has five levels of occupational prestige. The U.S. Bureau of the Census has derived a ranked measure of occupational status that has more levels with finer categories within each of the major levels. Some measures of social class represent a composite of

51589_CH04_Printer.indd 197 09/02/13 12:34 PM

198 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

variables, including occupation, education, and income.92 Two approaches are illustrative of the work in this field. Hollingshead and Redlich93 derived a two- factor measure of social class that combines level of education with occupational prestige (see Exhibit 4–2). Duncan94 developed a three-factor SES index that has been used in epidemiologic research.

Problems arise in the assignment of social class to unemployed or retired per- sons and to students, who may ultimately occupy a high social class position in society upon graduation but who temporarily have low income and occupational prestige. It is difficult to assign social class ranking to a family when both mother and father work and have occupations that are disparate in occupational prestige. People who occupy the same category of occupational prestige may be quite diverse in income and other characteristics.

Measurement of income also is fraught with difficulty. The individual may not want to reveal income information or may not actually know the precise income of the family, as in the case of children, when it is necessary for the researcher

Socioeconomic Status and Mental illness Survey of New haven, connecticut

Hollingshead and Redlich82 classified New Haven, Connecticut, into five social class levels according to prestige of occupation, edu- cation, and address. These were some of the findings of the study:

●● There was a strong inverse association between social class and the likelihood of being a mental patient under treatment.

●● With respect to severity of mental illness, upper socioeconomic individuals were more likely to be neurotics, whereas lower socio- economic individuals were more likely to be psychotics (i.e., less severe forms of mental illness occurred in the upper social class strata, and the highest incidence of schizophrenia was found in the lowest social classes).

●● The type of treatment varied by SES ranking. It was more com- mon for upper socioeconomic individuals to receive treatment from a psychiatrist, whereas lower SES individuals were treated in state and public hospitals, where they received organic modes of treatment, such as shock therapy. n

e x

h ib

it 4

–2

51589_CH04_Printer.indd 198 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 199

to measure their social class. Also, two workers who have low-status occupations may have a combined family income that is higher than the total family income of one professional worker. The correct method for assigning social class in these situations may not be readily discernible.

Education (measured by number of years of formal schooling completed) is another component of SES. Higher levels of education, in contrast to income or occupation, appear to be the strongest and most important predictor of positive health status.92

Findings Despite the unreliability of measures of social class, studies of the association between social class and health have yielded noteworthy findings; thus, social class usually should be considered when one is evaluating the occurrence of dis- ease. Among the major illustrations of the association between social class and health are the findings on the frequency of mental illness. Exhibit 4–2 concerns one of the most noteworthy studies in this field, carried out by Hollingshead and Redlich,93 who surveyed New Haven, Connecticut, half a century ago, and reported that as SES increased, the severity of mental illness decreased.

Dunham95 and others proposed two alternative hypotheses for the finding of highest incidence of severe mental illness among the lowest social classes. One, the social causation explanation (known as the “breeder hypothesis”) suggested that conditions arising from membership in the low social class groups produced schizophrenia and other mental illnesses.96 However, an equally plausible expla- nation was the “downward drift hypothesis,” which stated that the clustering of psychosis was an artifact of drift of schizophrenics to impoverished areas of a city. Murphy et al.97 indicated that during the 1950s and 1960s the prevalence of depression was significantly and persistently higher in the low SES popula- tion than at other socioeconomic levels. Stresses associated with poverty may be linked to depression, which in turn may be associated with subsequent down- ward social mobility. Downward drift is consonant with the view that the con- centration of depressed people at the lower end of the social hierarchy may result from handicapping aspects of the illness. Although epidemiologic research that shows variation in mental disorders by position in social structures has gener- ated much excitement, additional work is needed to develop adequate theories to explain these findings.95

Low social class standing correlates with increased rates of infectious disease, including TB, rheumatic fever, influenza, pneumonia, and other respiratory

51589_CH04_Printer.indd 199 09/02/13 12:34 PM

200 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

diseases.89 It is reasonable to attribute increased rates of these conditions to overcrowding, increased exposure to infection, lack of medical care, nutritional deficiencies, and poor sanitary conditions.

In comparison with upper socioeconomic groups, lower socioeconomic groups have higher infant mortality rates and overall mortality rates and lower life expectancy.98 The influence of psychosocial and behavioral factors on dif- ferential health outcomes associated with varying SES levels is not fully under- stood.99 Social class differences in mortality and morbidity have persisted over time, even with overall reductions in infant mortality and infectious diseases.89 Life expectancies, based on data from the National Longitudinal Mortality Study for 1979 to 1985, were estimated for white men and white women by education, family income, and employment status. Life expectancy varied directly with amount of schooling and family income.100 Similarly, more recent data from the National Occupational Mortality Surveillance Program for 1984 through 1997 demonstrated an inverse gradient between SES and mortality among employed persons in the United States.101 Comparative international studies found socioeconomic inequalities in cardiovascular disease mortality, with higher levels among persons who were less educated or had lower occu- pational classifications.102 Although inadequate medical care and exposure to environmental hazards may account for some of the social differences in mor- bidity and mortality, other factors also may be relevant, such as exposure to stressful life events, stresses associated with social and cultural mobility, poverty, and health behaviors including smoking. Inequalities in mortality by level of socioeconomic status have continued to increase rapidly in the United States from the 1990s until the early 21st century. During this period, educated per- sons (white and black men and white women) have experienced reductions in mortality, while at the same time the least educated members of society have demonstrated a worsening trend.103

When the effects of poverty and limited access to health care are removed, infant mortality rates among African Americans improve. In illustration, researchers examined infant mortality rates among the dependents of African American military personnel. These personnel had guaranteed access to health care and tended to have levels of family income and education that were higher than those of the U.S. African American population. The infant mortality rates in this study group were somewhat lower than those for the general U.S. popu- lation.104 For persons who were younger than 65 years of age, mortality rates were lower among those with higher family incomes for both African Americans

51589_CH04_Printer.indd 200 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p e r s o n s 201

and whites and for both men and women. However, at each level of income, African Americans had higher mortality rates than whites. Higher levels of family income also were associated with lower death rates from cardiovascular diseases and cancer.105

Wide differentials in cancer survival were observed among socioeconomic groups in England and Wales.106 Lower socioeconomic groups tended to have a larger proportion of cancers with poor prognoses in comparison with upper socioeconomic groups. Poor survival rates among the lower socioeconomic groups might have been due to delay in seeking health care. Health system barriers, such as lack of access to care and the financial burden of diabetes care in the United States, may affect the health of insulin-dependent diabetes mellitus patients, which increases mortality rates among people 25–37 years of age.107

Among the other conditions that vary by social class is a specific form of men- tal retardation: mild mental retardation (IQ 60 through 75). Research reported a gradient in the frequency of mild mental retardation by social class; low social class groups had the highest prevalence of mild retardation. More severe forms of mental retardation tended to be more uniformly distributed across social classes.108 Because relatively few recent investigations have been reported on SES as a correlate of mental retardation, this topic merits additional research.

Previously we noted that components of SES include income and educa- tion. Both variables are linked to self-perception of health (i.e., whether indi- viduals perceive their own health as “excellent or very good,” “good,” or “fair or poor”). Data from the 2010 National Health Interview Survey indicated a strong association between overall self-perception of health and both educa- tion and family income. (Refer to Figure 4–17, Part A.) A total of 74.1% of persons who had a bachelor’s degree or higher reported their current health status as excellent or very good in comparison with 38.2% of persons who had less than a high school diploma. With respect to income, 76.3% of persons who had a family income of $100,000 or more reported excellent or very good health in comparison with 57.6% of persons who had an income of $49,999 or lower (Figure 4–17, Part B).32 One possible explanation for this associa- tion is access to preventive health care. Prompt attention to acute illnesses and worsening symptoms of chronic conditions reduces the need for later treat- ment in a hospital. Consequently, affluent persons are more likely than less economically advantaged individuals to obtain early treatments that result in higher levels of positive health status.

51589_CH04_Printer.indd 201 09/02/13 12:34 PM

202 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Excellent or very good

Good

Fair or poor

16.132.6

34.2 27.6

100%80%60%40%20%0%

Less than a high school diploma

High school diploma

Some college

Bachelor’s degree or higher

38.2

51.2

59.8 28.2 12.0

5.920.174.1

Part A

Figure 4–17 Age-adjusted percent distributions of health status among persons aged 18 years and over by education (part A) and income (part B): United States, 2010. Source: Reproduced from Schiller JS, Lucas JW, Ward BW, Peregoy JA. Summary health statistics for US adults: National Health Interview Survey, 2010. National Center for Health Statistics. Vital Health Stat 10 (252). 2012.

Excellent or very good

Good

Fair or poor

12.130.357.6

63.9 26.7 9.4

6.323.869.8

76.3$100,000 or more

$75,000–$99,999

$50,000–$74,999

$35,000–$49,999

19.1 4.6

100%80%60%40%20%0%

Part B

51589_CH04_Printer.indd 202 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p l a C e 203

Characterist ics of Place

International Comparisons of Disease Frequency

International Geographic (within-country) variations Urban/rural differences Localized occurrence of disease n

Types of Place Comparisons

For information on international trends in morbidity and mortality, one should consult organizations such as the World Health Organization (WHO) and the Organization for Economic Cooperation and Development (OECD). The major source of information about international variations in rates of disease, WHO and affiliated organizations sponsor and conduct ongoing infectious and chronic disease surveillance. WHO’s statistical studies portray international variations in infectious and communicable diseases, malnutrition, infant mortality, suicide, and other conditions. As might be expected, both infectious and chronic diseases show great variation from one country to another. Some of these differences may be attributed to climate, cultural factors, national dietary habits, and access to health care.

The OECD compiles data used to compute international comparisons in life expectancy, which varies greatly from one country to another. (Refer to Figure 4–18 for 2004 data.) Among the selected countries shown in Figure 4–18, the United States ranked in the bottom half for both males (life expectancy = 75.2 years) and females (life expectancy = 80.4 years). Japan reported the highest life expectancy (78.6 and 85.6 years, respectively). The Russian Federation had the lowest life expectancy (59.1 and 72.4 years, respectively).

CHD, hypertension, stroke, and diabetes are among the “diseases of afflu- ence,” conditions formerly confined primarily to the developed world but now occurring more frequently in developing regions as living standards improve. Deaths from noncommunicable diseases (e.g., cardiovascular diseases, diabetes,

51589_CH04_Printer.indd 203 09/02/13 12:34 PM

204 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Figure 4–18 Life expectancy ranking* at birth,† by sex in selected countries and territories, 2004.§¶ Source: Reproduced from Centers for Disease Control and Prevention, MMWR. Vol 57, No 13, p 346, 2008.

Female Country/Territory Male

85.6 84.7 83.8 83.7 83.7 83.0 82.7 82.6 82.5 82.3 82.3 82.3 82.2 82.1 82.0 81.7 81.5 81.5 81.4 81.4 81.1 81.0 80.8 80.7 80.7 80.6 80.4 79.9 79.8 79.3 79.2 79.0

Life expectancy (yrs)

77.8 76.9 76.3 75.6 72.4

90 9080 8070 7060 6050 50

78.6 79.0 76.7 78.6 77.2 78.1 48.4 77.8 76.8 77.5 75.3 74.1 77.9 76.4 77.1 77.5 75.6 76.6 75.7 76.9 76.8 74.5 76.0 76.4 75.8 74.0 75.2 75.2 75.4 74.2 70.7 72.6 70.3 68.6 69.1 68.3 59.1

Japan Hong Kong

France Switzerland

Spain Australia Sweden Canada

Italy Puerto Rico

Norway Finland Israel

Australia Singapore

New Zealand Greece Belgium

Netherlands Germany

England & Wales Portugal

Northern Ireland Ireland

Costa Rica Chile

United States Denmark

Cuba Scotland Poland

Czech Republic Slovakia Hungary Bulgaria Romania

Russian Federation

cancer, and chronic respiratory diseases) are expected to increase worldwide by 15% between 2010 and 2020.109 Cardiovascular diseases were the leading cause of death worldwide. Countries that have high economic standards tend to foster behavioral and lifestyle factors implicated in CHD and other chronic diseases: lack of exercise, rich diets, and use of tobacco products. Due to preventive efforts

* Rankings are from the highest to lowest female life expectancy at birth. † Life expectancy represents the average number of years that a group of infants would live if the infants were to experience throughout life age-specific death rates present at birth. § Countries and territories were selected based on quality of data, high life expectancy, and a population of at least 1 million population. Differences in life expectancy reflect differences in reporting methods, which can vary by country, and actual differences in mortality rates. ¶ Most recent data available. Data for Ireland and Italy are for 2003.

51589_CH04_Printer.indd 204 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p l a C e 205

that aim to modify these behavioral and lifestyle factors, CHD mortality in the United States is on the decline. Nevertheless, CHD mortality is the leading cause of death in the United States.

Cancer rates are increasing worldwide, especially as the population ages.109 Types of cancer mortality and morbidity vary according to income levels of countries. Cancer ranked second among the leading causes of death globally. The leading causes of cancer death are lung, breast, colorectal, stomach, and liver cancers. Lung cancer among men and breast cancer among women are the most common causes of cancer deaths in high-income countries. In low and middle income countries cancer mortality varies according to risk factors such as exposure to the human papillomavirus. For example, cervical cancer causes the greatest number of cancer deaths in sub-Saharan Africa.

The most commonly diagnosed cancers in upper-middle-income and high- income countries are prostate cancer and breast cancer. In lower-middle-income countries (e.g., China and India), the most common forms of cancer among males are lung, stomach, and liver cancer; among females, they are breast, cer- vical, and lung cancer. For low-income countries, lung and breast cancers are among the most commonly diagnosed cancers. High lung cancer rates occur in parts of the developing world where smoking is common, although exposure to environmental pollution may play a role.

Globally, the fourth most commonly diagnosed form of cancer is stomach cancer.110 The Republic of Korea, Mongolia, Japan, and China are the four coun- tries worldwide with the highest incidence of stomach cancer. Infection with Helicobacter pylori is one of the causes of stomach cancer. Risk factors for stom- ach cancer include consumption of preserved, cured, or salted foods. Decline in stomach cancer rates in some areas might be achieved through improved diet (e.g., greater consumption of fresh vegetables).

There are numerous examples of international variations in infectious diseases and related conditions. Schistosomiasis (an infection caused by blood flukes transmitted by contact with water from infected rivers and lakes) is endemic to the Nile River area of Africa and to sections of Latin America but rarely occurs in the United States unless it is imported into the country from an endemic area. Yaws (a contagious skin disease caused by the bacterium Treponema pallidum) tends to be localized in tropical climates and does not ordinarily occur in the temperate climate of the United States. Countries in tropical Africa account for more than 80% of all clinical cases of parasitic infections and more than 90% of all parasite carriers.111 Approximately 800,000 children died from malaria in 2000 in malaria-affected regions of Africa.112 Exacerbations of malaria outbreaks

51589_CH04_Printer.indd 205 09/02/13 12:34 PM

206 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

follow major ecologic or social changes, such as agricultural or other economic exploitation of jungle areas or sociopolitical unrest.

Communicable diseases, maternal and perinatal conditions, and nutri- tional deficiencies accounted for 7.0% of all deaths in high-income countries but—alarmingly—36.4% of deaths in lower- and middle-income nations.113 Deaths from infections hamper the efforts of developing nations to advance economically. HIV/AIDS takes a major toll in these countries. Some types of communicable diseases that show international variation are cutaneous leish- maniasis, Chagas disease, dengue fever, malaria, and cholera. Zoonotic diseases (transmitted from animal hosts to humans) vary greatly from one country to another. One example is vampire bat rabies, which increased during the 1980s as a cause of human death in Peru and Brazil.114 Another example is bovine spongiform encephalopathy (BSE; also known as mad cow disease), which at first was confined largely to the United Kingdom, then spread to Europe and later to Japan, Canada, and the United States. BSE has been linked to variant Creutzfeldt-Jakob disease in humans (refer to case study on BSE).

Suicide rates for selected countries show marked differences between the lowest and highest ranked countries. Low rates are reported in Mexico, Greece, and Italy; the United States and Canada fall in the middle range; and Belgium, Switzerland, Finland, and Hungary have the highest rates.115

Infant mortality rates demonstrate substantial variations from country to country. Central African nations have the highest rates of infant mortality fol- lowed by North Africa, the Middle East, and India. The lowest rates exist in Japan, the Scandinavian countries, and France. The United States, Great Britain, Canada, and Australia all have higher rates of infant mortality than the foregoing countries. Social factors, education, and availability of medical care may account, in part, for the international variation in infant mortality.

Many countries, especially those in Africa, have had drastic reductions of their budgets for health services in the past decade, creating a wide dis- crepancy in health status between those countries and the developed world. For example, the life expectancy (2011) in Chad is estimated at 48.7 years, in contrast with the European country Monaco (89.7 years).116 In develop- ing countries, high population growth also reduces the available resources for healthcare and prevention programs and at the same time increases the potential for spread of infection through crowding. Environmentally related health problems and adverse impacts of industrialization, urbanization, and slum growth will challenge the healthcare resources of developing countries in the future.

51589_CH04_Printer.indd 206 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p l a C e 207

A disorder of cattle, BSE is a progressive neurological condition that results from infection by an unusual transmissible agent called a prion. During the early 1990s, a BSE epizootic was first observed in the United Kingdom (UK) and later appeared elsewhere in Europe and around the globe. The fast-moving cattle outbreak in the United Kingdom mobilized public health authorities internationally and threatened travelers and others who were advised not to consume beef products of UK origin. BSE was linked causally with the human condition: variant Creutzfeldt- Jakob disease (vCJD).

The BSE epizootic in the United Kingdom peaked in January 1993 at almost 1,000 new cases per week. Over the next 17 years, the annual numbers of BSE cases dropped sharply. Cumulatively, through the end of 2010, more than 184,500 cases of BSE had been confirmed in the UK alone in more than 35,000 herds.

Evidence suggests that the UK outbreak originated as a result of feeding cattle meat-and-bone meal that contained BSE-infected products from either a spontaneously occurring case of BSE or from scrapie-infected sheep products (scrapie, a prion disease of sheep). The outbreak was then amplified and spread throughout the UK cattle industry by feeding rendered, prion-infected, bovine meat-and-bone meal to young calves.

Through February 2011, BSE surveillance has identified 22 cases in North America: three BSE cases in the United States and 19 in Canada. The United States and Canada ban feeds for animals and cattle that might contain tissues infected with BSE.

Epidemiologic findings suggest a causal association between the BSE outbreak in cattle and a new human prion disease called variant Creutzfeldt-Jakob disease (vCJD). First reported from the United Kingdom in 1996, vCJD is a rare, degenerative, fatal disorder in humans. The disorder produces confusion, progressive dementia, and loss of motor control.a

Case 1: Mad Cow Disease (Bovine Spongiform Encephalopathy—BSE) and Variant Creutzfeldt-Jakob

Disease (vCJD) Spread Internationally

continues

51589_CH04_Printer.indd 207 09/02/13 12:34 PM

208 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Within-Country Geographic Variation in Rates of Disease Many of the countries of Europe and North America have substantial variations in climate, geology, latitude, environmental pollution, and concentrations of eth- nic and racial stock, all of which may be related to differences in frequency of disease. This section focuses on the experiences of the United States. As an exam- ple, consider life expectancy in the United States. When researchers compare life expectancy among U.S. counties, substantial disparities exist. Also noteworthy are comparisons between U.S. counties and other countries. Shown in an inter- national context, life expectancies in some U.S. communities is very much below the levels reported in other high-income countries.117

Examples of within-country comparisons in the United States are case rates by region (e.g., Pacific, Mountain, Central, and Atlantic). Sometimes, compari- sons in rates are made by states, or, if fine comparisons are to be made across

From 1995 through mid-August 2006, a total of 195 human cases of vCJD were reported worldwide, 162 in the United Kingdom, 20 in France, 4 in Ireland, 2 in the United States, and 1 each in Canada, Japan, the Netherlands, Portugal, Saudi Arabia, and Spain. Seven of the non-UK case-patients were most likely exposed to the BSE agent in the United Kingdom because of their having resided there during a key exposure period of the UK population to the BSE agent. These latter case-patients were those from Canada, Japan, the United States, 1 of the 20 from France, and 2 of the 4 from Ireland. As of 2012, the total had increased to 217 persons from 11 countries with the majority from the United Kingdom and France; a total of 3 cases were from the United States. n

aHeymann DL, ed. Control of Communicable Diseases Manual, 19th ed. Washington, DC: American Public Health Association; 2008.

Source: Portions adapted and reprinted from:

Centers for Disease Control and Prevention. vCJD (Variant Creutzfeldt-Jakob Disease). Risk for travelers. http://www.cdc.gov/ncidod/dvrd/vcjd/risk_travelers.htm. Accessed July 30, 2012.

Centers for Disease Control and Prevention. BSE (bovine spongiform encephalopathy, or mad cow disease). http://www.cdc.gov/ncidod/dvrd/bse/index.htm. Accessed July 30, 2012.

Centers for Disease Control and Prevention. Fact sheet: variant Creutzfeldt-Jakob disease. http://www.cdc.gov/ncidod/dvrd/vcjd/factsheet_nvcjd.htm. Accessed July 30, 2012.

Case 1 continued

51589_CH04_Printer.indd 208 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p l a C e 209

the United States, rates may be calculated by counties and census tracts. In the United States, chronic diseases (e.g., some forms of cancer and multiple sclero- sis) and infectious diseases (e.g., intestinal parasites, influenza, AIDS, and many others) show variation in frequency across the country.

Cancer mortality A classic review published in 1975 demonstrated that cancer varied geographi- cally (e.g., high death rates due to leukemias were concentrated in the upper Midwest).118 Currently available data seems to confirm this finding. Also in 2008, age-adjusted death rates for leukemias tended to be highest in the upper Midwest (e.g., Nebraska and Michigan as well as in many states that bordered Canada).119 One southern state, Louisiana, had elevated death rates for leukemia in 2008.

The 1975 review suggested that mortality from malignant melanoma of the skin showed a relationship with latitude, the lowest rates occurring in the northern tier of the country and the highest rates being concentrated along the Sunbelt.118 This observation remains plausible as exposure to ultraviolet (UV) light from the sun is a risk factor for skin cancer. The UV index is higher in the southern latitudes of the United States than in the northern latitudes.

Multiple sclerosis Another classic study asserted that multiple sclerosis varied according to latitude in the United States.120 Rates of MS ranged from 15 per 100,000 in the South (Charleston, New Orleans, and Houston) to intermediate rates in Denver to high rates in Rochester, Minnesota (more than 40 per 100,000).121

Infectious, vector-borne, and parasitic diseases Infectious, vector-borne, and parasitic diseases show variations in frequency regionally and from state to state. Table 4–2 provides eight examples (not an exhaustive list) of variations by region and state.

Among diseases cited in Table 4–2 are giardiasis, influenza, and HIV infec- tions, which are described in the next section.

●● Giardiasis (agent, Giardia lamblia) is a gastrointestinal illness transmitted via the fecal-oral route (e.g., ingestion of food or water that has been con- taminated with feces). Campers may become infected when they drink untreated water from streams. Sometimes, recreational swimming is asso- ciated with giardiasis. Although the condition occurs over much of the

51589_CH04_Printer.indd 209 09/02/13 12:34 PM

210 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

United States, the greatest number of cases tends to be concentrated in the northern states, for example, Minnesota and Maine.122

●● Influenza waxes and wanes in differing geographic sections of the United States throughout the flu season. The pattern of outbreaks changes from year to year and is also related to the influenza virus that is circulating during a particular year. Influenza epidemics remain an important cause of hospitalization and also are a cause of mortality as well as a factor that exacerbates health problems among the elderly and persons with chronic diseases.123

●● HIV incidence varies considerably within the United States. The high- est reported rates (greater than 15.0 per 100,000 population, data for 2009) tended to be concentrated in some northeastern and southeastern states. Approximately, Figure 4–19 depicts variations in AIDS cases in the United States during 2009. Early in the present century, the frequency of HIV infections and AIDS cases remained approximately constant in most regions of the United States. At the same time, an epidemic of AIDS occurred in the Deep South, where cases were more common among African Americans, women, and rural inhabitants than among similar groups in other regions.124 Adding to the burden of AIDS in the Deep South are high poverty rates and low health insurance levels, which limit the availability of treatment and prevention programs.

Table 4–2 Examples of Infectious, Vector-borne, and Parasitic Diseases That Vary by Region and State in the United States

Condition Example Regions/states with highest incidence

Arboviral diseases West Nile virus Mississippi, South Dakota, Wyoming, Colorado, Nebraska

Coccidioidomycosis Coccidioidomycosis Arizona, California HIV/AIDS HIV virus diagnosis Southeast and Northeast Intestinal parasites Giardiasis Northern states (e.g., Minnesota, Maine) Seasonal influenza

(2010–11 flu season) Influenza A viruses;

influenza B viruses Predominant virus (A or B) varies by

region during season Vector-borne diseases

(tick vectors) Ehrlichiosis Upper Midwest and coastal New

England Vector-borne diseases

(tick vectors) Lyme disease About 90% of confirmed cases reported

from the northeastern and upper midwestern U.S.

Source: Data from Centers for Disease Control and Prevention. Summary of notifiable diseases—United States, 2009. MMWR. 2011;58(53):1–100.

51589_CH04_Printer.indd 210 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p l a C e 211

Urban/Rural Differences in Disease Rates According to the National Center for Health Statistics, “[h]ealth differences between urban and rural communities have long been recognized. These health variations reflect, in part, differences in underlying demographic, economic, physical, social, and environmental community characteristics, as well as the availability and nature of healthcare resources.”125(p 1) The U.S. Bureau of the Census defines the terms urban and rural by relating them to metropolitan statistical areas (MSAs) and census tracts. MSAs (formerly known as standard metropolitan statistical areas) are geographic areas of the United States established by the Bureau of the Census to provide a distinction between metropolitan and nonmetropolitan areas by type of residence, industrial concentration, and population concentration. According to the NCHS, “The general concept of a metropolitan area is one of a large population nucleus together with adjacent communities that have a high degree of economic and social integration with that nucleus.”126(p 3)

Figure 4–19 Human immunodeficiency virus diagnosis rates*–United States and U.S. territories, 2009. Source: Reproduced from Centers for Disease Control and Prevention, Summary of notifiable diseases–United States, 2011, MMWR. Vol 58, No 53, p 63, 2011.

CNMI GU PR

0.0–4.9

5.0–9.9 10.0–14.9 >–15.0

*Per 100,000 population.

51589_CH04_Printer.indd 211 09/02/13 12:34 PM

212 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

The NCHS urban-rural classic classification scheme for counties has six levels:125

●● Large metro, central—example, the entire population of the largest principal city of the MSA.

●● Large metro, fringe—counties located in an MSA and with 1 million or more population.

●● Medium metro—population of 250,000–999,999 in an MSA. ●● Small metro—population of 50,000–249,999 in an MSA. ●● Nonmetropolitan—micropolitan (in an MSA). ●● Nonmetropolitan—noncore (not in an MSA).

Census tracts are small geographic subdivisions of cities, counties, and adjacent areas. The average tract has about 4,000 residents and is designed to provide a degree of uniformity of population economic status and living condi- tions within each tract.

Urban and rural sections of the United States both have characteristic risks for morbidity and mortality. Urban diseases and causes of mortality are those that are more likely to be spread by person-to-person contact, crowding, and poverty, or to be associated with urban pollution. Lead poisoning has been associated with inadequate housing and is found in inner city areas, where increased expo- sure of children occurs through ingestion of lead-based paints. Although such paints are now outlawed for interior residential use, exposure may still be high in low-income urban areas.

Studies published in the 1990s reported that mortality rates due to atheroscle- rotic heart disease, TB, and cirrhosis of the liver were higher in urban areas than in rural areas. Urban areas also showed higher rates of bladder, lung, larynx, liver, and oral cancer and cancer of the pharynx and cervix, whereas rural areas dem- onstrated excesses in cancer of the lip in both sexes and cancer of the eye among men.127 Among African Americans, data for the prevalence, incidence, and mor- tality of CHD reflected higher rates for urban residents than for rural residents.128

Among all racial and ethnic groups, the residents of rural areas are affected by unique environmental and cultural factors that could reinforce unhealth- ful behaviors.129 Health and economic status of rural residents vary according to the region of the country in which they live: Southern rural inhabitants are poorer, smoke more frequently, and are more physically inactive; consequently, their rate of ischemic heart disease mortality is higher. Western rural inhabitants are afflicted more frequently by alcohol abuse and suicide. Northeastern rural dwellers endure a greater frequency of tooth loss. Other research has indicated that residence on farms was protective for the mental health of rural women of

51589_CH04_Printer.indd 212 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p l a C e 213

childbearing age, although this finding may be an artifact of reduced access to mental health services among persons who live in isolated rural areas.130

A number of studies have examined the health status of minority populations found in rural areas of the United States. Racial/ethnic minority populations in rural areas are reported to experience more disadvantages caused by health dis- parities and lack of access to health care than similar minority groups in urban areas.131

In summary, the following trends characterized morbidity and mortality among urban and rural residents for 2005–2007:125

●● Infant, child, and young adult mortality—increased with decreasing urbanization.

●● Adult mortality—lowest in the fringe counties of large metro areas and increased with decreasing urbanization.

●● Homicide—large central metropolitan counties had the highest rates. ●● Cerebrovascular disease mortality—increased with decreasing

urbanization. ●● Poor or fair health status—increased with decreasing urbanization; self-

reported health status was highest in large fringe metro counties. ●● Health insurance—the percentage of no health insurance coverage

increased with decreasing urbanization; the percentage of individuals who had no coverage was lowest in large fringe metro counties.

Localized Place Comparisons A local outbreak of a disease or localized elevated morbidity rate may be due to the unique environmental or social conditions found in a particular area of interest. Fluorosis, a disease resulting in mottled teeth, is most common in those areas of the world and the United States where there are naturally-elevated fluo- ride levels in the water. Goiter, associated with iodine deficiency in the diet, was historically more common in landlocked areas of the United States, where seafood was not consumed, although this problem has been greatly alleviated by the introduction of iodized salt.

Localized concentrations of ionizing radiation have been studied in relation to cancer incidence. Ohio communities with high risk of exposure to radon had higher rates than the rest of Ohio of all cancers in general and cancer of the respiratory system in particular.132 In Ontario, Canada, gold miners had excess mortality from carcinoma of the lung. This excess mortality was linked to exposure to radon decay products, arsenic, and high dust concentrations.133

51589_CH04_Printer.indd 213 09/02/13 12:34 PM

214 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Local geologic formations may affect water hardness, which, some studies suggest, is a protective factor against heart disease deaths. For 1969–1983, one study reported an east–west regional gradient in cardiovascular mortality within seven counties in central Sweden, supporting other reports that have suggested water hardness to be inversely related to cardiovascular mortality.134 Variation in water hardness accounted for 41% of the variation in the ischemic heart dis- ease mortality rate and 14% of the variation in the stroke mortality rate. A sec- ond Swedish study attributed variation in mortality rates for coronary disease to exposure to cold weather, which was positively associated with heart disease and negatively associated with water hardness.135 However, the evidence for this hypothesis has been both positive and negative.

Geographic Information Systems Increasingly, public health practitioners are using geographic information sys- tems (GIS) as a method to provide a spatial perspective on the geographic dis- tribution of health conditions. Although GIS has existed for some time, newer software has increased the ease of application of GIS methods by nonexperts.136 As a result, GIS has generated much excitement in the epidemiology community with respect to fresh applications. It is an advance that has enormous possibilities for defining new methods of comprehending data.137

A GIS is defined as “. . . a constellation of computer hardware and software that integrates maps and graphics with a database related to a defined geographi- cal space . . . The geographical data may be spatial or descriptive in nature. A GIS can be defined as an integrated set of tools within an automated system capable of collecting, storing, handling, analyzing, and displaying geographically referenced information.”138(p 1) A GIS contains a database, maps, and a method to link these elements.136 “GIS is, at its heart, a simple extension of statistical analyses that join epidemiological, sociological, clinical, and economic data with references to space. A GIS system does not create data but merely relates data using a system of references that describe spatial relationships.”137(p 183)

Although many uses exist, an important application of GIS is to map loca- tions that have higher disease occurrence or mortality risk than do other areas. Planners then are able to target these high-risk areas with appropriate social and health interventions. Subsequently, GIS facilitates the assessment of the impact of these interventions, which presumably would result in reductions of dis- ease occurrence that could be visualized on a map. Thus, it is apparent that the use of GIS in epidemiology follows from the heritage of John Snow, who pro- duced a map of the cholera outbreak in Broad Street, Golden Square, London.

51589_CH04_Printer.indd 214 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f p l a C e 215

At present, epidemiologists use GIS for many purposes, including research and planning in environmental health, infectious disease outbreaks, and policy evalu- ation.136 Table 4–3 provides examples of applications of GIS.

To obtain more information about GIS, navigate the CDC website, “Epi Info™ 7” (http://www.cdc.gov/epiinfo/; accessed May 26, 2012). The CDC website describes Epi Info™ freeware developed at the CDC. The Epi Map module (a component of Epi Info™) enables users to develop geographic maps. In conclusion, Figure 4–20 presents an example of a GIS map of infant mortal- ity rates in the U.S. state of Idaho. This map is called a choropleth map, defined as a map that represents disease rates (or other numerical data) for a group of regions by different degrees of shading.

Reasons for Place Variation in Disease Some place variations in disease are related to regional differences in detection and reporting of diseases. Other variations can be explained by the social and demographic composition of the population, gene and environment, and the influence of local environmental factors such as climate, pollution, and naturally occurring carcinogens and other chemicals/elements.

Concentration or clustering of racial, ethnic, or religious groups within a specific geographic area may result in higher or lower rates of diseases, depend- ing upon the lifestyle and behaviors of the particular religious or ethnic group.

Table 4–3 Representative Applications of Geographic Information Systems (GIS)

Mapping of environmental health risks Pesticide pollution of groundwater Drifting of crop pesticides Average air pollution concentrations Exposure to elevated levels of magnetic fields Lead hazards

Portraying the geographic distribution of infectious diseases Bovine spongiform encephalopathy (BSE) in Europe Surveillance of disease outbreaks (e.g., Lyme disease, hepatitis C) Mapping of water-borne, vector-borne, and tropical diseases

Health policy/planning Risk assessment Intervention evaluation Health services needs assessment Hospital accessibility

51589_CH04_Printer.indd 215 09/02/13 12:34 PM

216 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

The Seventh-Day Adventists, who espouse vegetarianism, are concentrated, among other places, in parts of the Los Angeles basin, and the rates for CHD tend to be low in these corresponding geographic areas. Similarly, low rates of cancer tend to be found in areas where a large proportion of the residents are Mormons, possibly because their religious beliefs advocate avoidance of stimu- lants, tobacco, and alcohol.

The genetic characteristics of the population may interact with the envi- ronment, suggesting a dynamic interplay between noxious environmental fac- tors and genetic makeup. An example of gene/environment interaction is the increased prevalence of the sickle-cell gene among people who live in sections of Africa that have high malaria rates.5 The sickle-cell trait is a genetic mutation that confers a selection advantage in areas where malaria is endemic. Tay-Sachs disease is especially common among persons of Jewish extraction and Eastern

Figure 4–20 GIS map of infant mortality rates in Idaho, United States. Source: Reproduced from Idaho Department of Health and Welfare, Bureau of Vital Statistics. Available at: http://inside.uidaho.edu/data/statewide/ esri/idtm/atlas/infamr98_id_esri.gif. Accessed: March 31, 2003.

Rates Are Number of Infant Deaths per 1000 Live Births, Idaho by County 1998

N/A

2.00–6.40

6.41–12.00

12.01–21.10

21.11–36.40

miles

kilometers March 2002

0 20 40

0 30 60

51589_CH04_Printer.indd 216 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f t i m e 217

European origin. It is now more widely distributed around the world as a result of the migration of the descendants of the original carriers.

Place variations in rates of disease may reflect the influence of climate (e.g., temperature and humidity) or environmental factors (e.g., the presence of environmental carcinogens). Certain geographic areas that have mild or tropical climates permit the survival of pathogenic organisms. Trypanosomiasis (African sleeping sickness) survives only in an environment that has the tsetse fly. Yaws and Hansen’s disease are found primarily in the tropics. As a consequence of global warming, disease vectors that survive only in mild climates may be able to move northward. Ectoparasites are more common in temperate climates because people wear many layers of clothing that may harbor these organisms. Naturally occurring or human-made chemical agents in particular geographic areas may be associated with the development of cancers or other diseases. For example, the concentration of fallout from U.S. nuclear testing has elicited concerns about the health effects of exposure to ionizing radiation.

In summary, Hutt and Burkitt stated:

The disease pattern in any country or geographical region is dependent on the constellation of environmental factors that affect each member of the population from birth to the grave. Within a particular geographical situation the response of individuals to any noxious influences may be modified by their genetic make-up. In general terms exogenous factors which play a role in the causation of disease can be categorized into one of four groups: physical agents, chemical substances, biological agents (which include all infective organisms), and nutritional factors. These are determined by the geographical features of the region, the cultural life of population groups living in the area, the socio-economic status of these groups, and, in certain situations, by specific occupational hazards. Often, the individual’s or group’s experience of specific factors is determined by a combination of geo- graphical, cultural, and socio-economic influences; this particularly applies to the type and quantity of food eaten.120(p 3)

Characterist ics of Time

Characteristics of time encompass cyclic fluctuations, point epidemics, secular time trends, and clustering. Variations in the pattern of disease associated with time often permit important insights into the pathogenesis of disease or the rec- ognition of emerging epidemics. Just as important, when one compares measures of disease frequency between two populations or within a population over time, the timing of data collection may need to be considered if there are seasonal or cyclic variations in the rate of disease.

51589_CH04_Printer.indd 217 09/02/13 12:34 PM

218 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

Cyclic Fluctuations/Seasonal Trends Cyclic fluctuations are increases and decreases in the frequency of diseases and health conditions over a period of years or within each year. Often these fluctua- tions reflect seasonal trends. In a comprehensive review of seasonality of infec- tious diseases, the author noted that “[t]he recognition of seasonal patterns in infectious disease occurrence dates back at least as far as the Hippocratic era, but mechanisms underlying seasonality of person to person transmitted diseases are not well understood.”139 (p 1)

Birth rates are an example of a phenomenon that conforms to a seasonal trend, increasing in the early summer.140 Another example is the occurrence of depressive symptoms, which peaked during the months of April through May in a Belgian study.141 Influenza, drownings, unintentional injuries, and mortal- ity from heart attacks manifest seasonal variations within each year. Analysis of data from a community registry of heart disease found that fatal and nonfatal coronary events in an Australian population were 20–40% more likely to occur in winter and spring than at other times of the year.142 Seasonal variations may be caused by seasonal changes in the behavior of persons that place them at greater risk for certain diseases, changes in exposure to infectious or environmen- tal agents, or endogenous biologic factors.

Pneumonia-influenza deaths in the United States demonstrate cyclic fluc- tuations, showing both annual peaks and periodic epidemics every few years. Seasonal increases in flu begin during the cold winter months of the year, peak in February, decrease in March and April, and then reach a minimum in June. Meningococcal disease is another condition that varies by season, appar- ently peaking in the winter and declining in the late summer. Rotavirus infec- tion (a form of viral gastroenteritis that occurs more commonly among young children) is still another condition that demonstrates seasonality. Refer to Figure 4–21 for seasonal trends for 2000–2009.

Many diseases demonstrate cyclic increases and decreases related to changes in lifestyle of the host, seasonal climatic changes, and virulence of the infectious agent for a communicable disease. Heart disease mortality peaks during the win- ter months, when sedentary men are suddenly required to free their automobiles from the aftermath of a snowstorm. Colds increase in frequency when people spend more time indoors and are in close contact with one another, whereas unintentional injuries tend to peak during the summer, and Rocky Mountain spotted fever increases in the spring, when the ticks that carry the rickettsia bacteria become more active. Reported malaria cases in the Americas and some Asian countries show marked seasonality related to cyclic occurrence of heavy rains, leading to occasional epidemics or serious exacerbations of endemicity.143

51589_CH04_Printer.indd 218 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f t i m e 219

Other examples of health phenomena that may show cyclic variation are responses of persons to temporary stressors. There may be an association between plasma lipid and lipoprotein levels among tax accountants during the tax season and among students at examination time. However, research has not shown such variation consistently.144

Common Source and Point Epidemics A common source epidemic is an outbreak “due to exposure of a group of per- sons to a noxious influence that is common to the individuals in the group.”145 A point epidemic may indicate the response of a group of people circumscribed in place to a common source of infection, contamination, or other etiologic factor to which they were exposed almost simultaneously.5 For an infectious condition, a point epidemic occurs within one incubation period for the dis- ease. When an outbreak lasts longer than the time span of a single incubation period and is caused by a common source of exposure, the outbreak is called

Figure 4–21 Percentage of rotavirus tests with positive results, by surveillance week–participating laboratories, National Respiratory and Enteric Virus Surveillance System, United States, July 2000–June 2009.* Source: Reproduced from Centers for Disease Control and Prevention, MMWR. Vol 58, No 41, p 1148, 2009.

2000–06 range†

2000–06 median

Peak activity Offset 2000–06

2007–08 2008–09

Onset§

2007–08 season

2008–09 season

30% p

os iti

27 Surveillance week

†Range created using the maximum or minimum percentage of rotavirus positive tests for each surveillance week during 2000–2006. Maximums and minimums for each week might have occurred during any of the six rotavirus seasons during 2000–2006. §The onset of rotavirus season was defined as the first of 2 consecutive weeks during which the percentage of stool specimens testing positive for rotavirus was >–10%; offset was defined as the last of 2 consecutive weeks during which the percentage of stool specimens testing positive for rotavirus was . >–10%.

29 31 33 35 37 39 41 43 45 47 49 51 1 3 5 7 9 11 13 15 17 19 21 23 25 27 0

*A median of 67 laboratories (range: 62–72) contributed rotavirus testing data to NREVSS during July 2000–June 2009.

51589_CH04_Printer.indd 219 09/02/13 12:34 PM

220 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

a continuous common source epidemic. This is the type of outbreak that occurred during the 1854 cholera epidemic in London. Acute infectious diseases and enteric infections sometimes are patterned according to common source or point source epidemics; sometimes mass illnesses due to exposure to chemical agents and noxious gases are point source epidemics. Figure 4–22 illustrates time clustering of cases of influenza-related illness at a residential facility dur- ing an outbreak reported to the CDC. Note that the greatest number of cases occurred on February 17, 2011, and that there was variation in the date of onset (e.g., February 10– February 30). Figure 4–21 is typical of the distribu- tion (marked by a rapid increase and subsequent decline) of outbreaks of acute infectious disease, foodborne illness, and acute responses to toxic substances.

Secular Time Trends Secular trends refer to gradual changes in the frequency of a disease over long time periods, as illustrated by changes in the rates of chronic diseases. For example, although heart disease was the leading cause of death in the United States from 1970 to 1988, the age-adjusted death rate for this cause declined by 34%; the decrease was 37% for white men and 24% for African-American men.7 These trends may reflect the long-term impact of public health programs, diet improvements, and better treatment as well as unknown factors. With more

Figure 4–22 Number of cases of influenza-related illness (n = 76) at a residential facility, by illness onset date and severity–Ohio, 2011. Source: Reproduced from Centers for Disease Control and Prevention, MMWR. Vol 58, Nos 51 & 52, p 1730, 2012.

25 MarFeb

*Cases were defined as severe if the patient was hospitalized or died.

3191371 0

Illness onset date

N o.

o f c

as es

Not severe (suspected)

Severe (suspected)*

Severe (confirmed)*

51589_CH04_Printer.indd 220 09/02/13 12:34 PM

C h a r a C t e r i s t i C s o f t i m e 221

women starting to smoke, especially teenagers and minority women, there was a secular increase in lung cancer mortality in the mid-1960s until the 1990s when the lung cancer mortality rates leveled off. (Refer to Figure 4–6.) Breast cancer mortality rates reflected secular decreases from 1995 to 2004 among African Americans and whites (Figure 4–23).

Clustering Case clustering refers to an unusual aggregation of health events grouped together in space or time. Examples of infectious disease clustering are the cholera epi- demic in London in the 1849 (reported by John Snow) and the outbreak of Legionnaires’s disease in the late 1970s. Examples of noninfectious disease

Figure 4–23 Age-adjusted total U.S. mortality rates for breast cancer, all ages, females for 1995–2004 by “expanded” race age-adjusted to the 2000 U.S. standard population. Source: Data from Surveillence, Epidemiology, and End Results (SEER) Program (www.seer.cancer.gov) SEER*Stat Database: Mortality—All COD, Public-Use with State, Total U.S. (1990–2004), National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, released April 2007, Underlying mortality data provided by NCHS (www.cdc.gov/nchs).

Black American Indian/Alaska Native

Hispanic

White

Asian/Pacific Islander

R at

e pe

r 10

0, 00

1995 1996 1997 1998 1999 2002 2003 200420012000 Year of death

51589_CH04_Printer.indd 221 09/02/13 12:34 PM

222 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

clustering include the development of angiosarcoma among workers exposed to vinyl chloride and adenocarcinoma of the vagina among daughters whose moth- ers were prescribed diethylstilbestrol.146 Other conditions that have been investi- gated for clustering include asthma, asbestos-related lung diseases, suicides, and leukemia and other cancers. Space and time clustering is one type of epidemio- logic evidence that might suggest an association between common exposure to an etiologic agent and development of morbidity and mortality. An illustration is a common source epidemic as described earlier in the chapter. Among the problems surrounding the study of clusters are the fact that health events that show clustering are usually rare (e.g., certain types of cancer), producing a small number of cases, and that some clusters may occur by chance alone.

Temporal clustering Postvaccination reactions (adverse reactions to vaccines), such as the develop- ment of jaundice among military personnel vaccinated for yellow fever5 and the development of puerperal psychoses, illustrate temporal clustering. Postpartum depression, ranging from the “blues” to more severe psychotic episodes, occurs in up to 80% of women within a few days after childbirth and may continue for several months or longer.147

Spatial clustering Concentration of cases of disease in a particular geographic area is the defini- tion of spatial clustering. Hodgkin’s disease, a condition of unknown etiology, is thought to have an infectious component. This possibility could be evaluated by a formal assessment of geographic and temporal variation in the incidence of Hodgkin’s disease. One investigation found that, among cases of Hodgkin’s disease diagnosed in those older than age 40 in Washington State, there was evidence that cases lived closer together than expected as young children and teenagers.148

As another example, six childhood leukemia cases were diagnosed in a small community in northern Germany.149 Using age-specific population and inci- dence data, this occurrence translated to a 460-fold standardized incidence ratio within a 5-km radius. Reports such as these can be quite alarming to the general public. Although they can be of significance, it is extremely important that they be put in perspective using appropriate epidemiologic methods. In particular, definition of the geographic area under study is critical.

Colleagues Julie Ross, PhD and Leslie Robison, PhD liken the situation to drawing a bull’s-eye around arrows that have already been shot. For example,

51589_CH04_Printer.indd 222 09/02/13 12:34 PM

s t u D y Q u e s t i o n s a n D e x e r C i s e s 223

suppose you are standing over a large enclosed circle with a jar full of 100 marbles. You dump the jar into the circle and note where the marbles land. Invariably, there will be some areas within the circle that will be dense with marbles and some areas that will be sparse. If you picked up all the marbles and repeated the process over and over again, you would note that the marble pattern would always be random and unlikely ever to appear evenly spaced. This description is analogous, perhaps, to clusters of disease, such as the childhood cancer examples. That is, marbles (cases) may fall within a small defined area (city) within a large geographic area (county, state, or country), and one could easily identify a cluster just by drawing a bull’s-eye around these cases. In order to evaluate such clusters properly, one must consider the phenomena in the context of a larger geographic area and the temporal occurrence of disease.

Conclusion

This chapter covered descriptive epidemiology: person, place, and time. We have seen, for example, that age and sex are among the most fundamental attributes associated with the distribution of health and illness in populations. Secular trends are among the more important time variables. An example of a secular trend is the decline in heart disease mortality in the United States, although it remains the leading cause of mortality. Among women, lung cancer mortal- ity showed a disturbing increase and became the leading cause of female cancer mortality in the latter third of the 20th century; currently, lung cancer mor- tality retains its same rank, although the increasing trend has leveled off. The data for person variables (e.g., age and sex) and place and time variables suggest hypotheses derived from descriptive epidemiologic studies; these hypotheses can be tested in analytic epidemiologic studies. For example, particularly low or high disease rates in a circumscribed geographic area may suggest interaction between ethnicity and local environmental conditions, a hypothesis that could be fol- lowed up in etiologic research.

Study Questions and Exercises

1. Compare and contrast the key features of descriptive epidemiology with those of analytic epidemiology.

2. Describe the relevance of descriptive epidemiology to the study of dis- ease. How do descriptive studies promote hypothesis formation? What is the relevance of Mill’s canons to descriptive epidemiology?

51589_CH04_Printer.indd 223 09/02/13 12:34 PM

224 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

3. Give definitions and examples of any three categories of descriptive epidemiology and apply them to your everyday experiences.

4. Review the section on age and compare mortality rates by age. Give some examples of age associations found in epidemiologic research. What explanations have been proposed to account for them? Discuss the possibility of combined age and gender effects.

5. How and why do mortality and morbidity differ by sex? Relate your discussion to the female paradox.

6. How do protective and selective factors increase or decrease the risk for disease based on marital status? How would one account for the differen- tial effect of marital status among men and women?

7. How do cultural practices and religious beliefs account for mortality and morbidity differences? Find some examples in the popular media to sup- port your view.

8. Describe procedures for measurement of race, ethnicity, and social class. What methodologic pitfalls are inherent in the ascertainment of these characteristics?

9. What health effects do variations in race, ethnicity, and social class have? Provide an analysis of the linkages among race, ethnicity, and social class and health disparities.

10. Select a health problem or disease with which you are familiar. Describe the occurrence in terms of person, place, and time.

11. To what extent are rates of common health problems similar or differ- ent across different geographic areas of the United States? How might demographic variables be linked to geographic variations in disease?

12. What are examples of the differences among international, regional (within-country), urban-rural, and localized patterns in disease? What factors may be linked to these differences?

13. Explain what is meant by a geographic information system (GIS). Describe the applications of GIS in epidemiology.

14. What time trends would characterize the occurrence of an influenza epidemic? How do epidemiologists account for seasonal variations in meningococcal disease and other communicable diseases, as well as for other health conditions?

15. What is meant by case clustering? Give some noteworthy examples; dis- tinguish between temporal and spatial clustering. What considerations bear upon the interpretation of spatial clustering?

16. Complete the project found in Appendix 4 at the end of this chapter.

51589_CH04_Printer.indd 224 09/02/13 12:34 PM

r e f e r e n C e s 225

References 1. Suchman EA. Health attitudes and behavior. Arch Environ Health. 1970;20:105–110. 2. Hennekins CH, Buring JE. Epidemiology in Medicine. Boston, MA: Little, Brown;

1987. 3. Abramson JH. Survey Methods in Community Medicine. 4th ed. New York: Churchill

Livingstone; 1990. 4. Mill JS. A System of Logic. London, UK: Parker, Son & Bowin; 1856. 5. MacMahon B, Pugh TF. Epidemiology Principles and Methods. Boston, MA: Little,

Brown; 1970. 6. Centers for Disease Control and Prevention. Fatal exposure to methylene chloride

among bathtub refinishers—United States, 2000–2011, MMWR. 2012;61(7): 119–120.

7. Centers for Disease Control and Prevention. Hantavirus pulmonary syndrome in five pediatric patients—four states, 2009. MMWR. 2009;58(50):1409–1410.

8. Heron MP. Deaths: Leading causes for 2007. National Vital Statistics Reports. 2011;59(8). Hyattsville, MD: National Center for Health Statistics.

9. National Center for Health Statistics. Health, United States, 2010: with Special Feature on Death and Dying. Hyattsville, MD; 2011.

10. National Institutes of Health, National Institute on Drug Abuse. Facts on prescription and over-the-counter drugs. http://teens.drugabuse.gov/peerx/pdf/PEERx_Toolkit_ FactSheets_RxDrugs.pdf. Accessed July 31, 2012.

11. Centers for Disease Control and Prevention. Percentage of adults with activity lim- itations, by age group and type of limitation—National Health Interview Survey, United States, 2010. MMWR. 2012;61(14):249.

12. Xu J, Kochanek KD, Murphy SL, Tejada-Vera B. Deaths: final data for 2007. National Vital Statistics Reports. 2010:58(19). Hyattsville, MD: National Center for Health Statistics

13. National Center for Health Statistics. Health, United States, 2011: with Special Feature of Socioeconomic Status and Health. Hyattsville, MD: 2012.

14. Verbrugge LM. Females and illness: recent trends in sex differences in the United States. J Health Soc Behav. 1976;17:387–403.

15. Yin S. Gender disparities in health and mortality. Population Reference Bureau; November 2007. www.prb.org/Articles/2007/genderdisparities.aspx?p=1. Accessed July 24, 2012.

16. Williams DR. The health of men: structured inequalities and opportunities. Am J Public Health. 2003;93:724–731.

17. Waldron I. Why do women live longer than men? Soc Sci Med. 1976;10:349–362. 18. Helwig-Larsen K, Juel K. Sex differences in mortality in Denmark during half a

century, 1943–92. Scand J Public Health. 2000;28:214–221. 19. Eskes T, Haanen C. Why do women live longer than men? Eur J Obstet Gynecol

Reprod Biol. 2007;133:126–133. 20. Hammond J, Salamonson Y, Davidson P, et al. Why do women underestimate the

risk of cardiac disease? A literature review. Aust Crit Care. 2007;20:53–59. 21. Gold LD, Krumholz HM. Gender differences in treatment of heart failure and

acute myocardial infarction: A question of quality or epidemiology? Cardiol Rev. 2006;14:180–186.

51589_CH04_Printer.indd 225 09/02/13 12:34 PM

226 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

22. Rosenblatt S. Minority women in L.A. County found to have higher rates of chronic disease. Los Angeles Times. Section B3, May 24, 2007. http://www.latimes.com/news. Accessed October 19, 2007.

23. U.S. Department of Health and Human Services, Office of Disease Prevention And Health Promotion. Healthy People 2020. ODPHP publication no. B0132. Washington, DC: U.S. Department of Health and Human Services.

24. HealthyPeople.gov. 2020 Topics and Objectives, Lesbian, Gay, Bisexual, and Transgender Health. http://www.healthypeople.gov/2020/topicsobjectives2020/ overview.aspx?topicid=25. Accessed July 31, 2012.

25. Waldron I, Weiss CC, Hughes ME. Marital status effects on health: are there dif- ferences between never married women and divorced and separated women? Soc Sci Med. 1997;45:1387–1397.

26. Schoenborn CA. Marital status and health: United States, 1999–2002. Advance Data from Vital and Health Statistics. December 15, 2004; Number 351. Centers for Disease Control and Prevention.

27. Kitagawa EM, Hauser PM. Differential Mortality in the United States. Cambridge, MA: Harvard University Press; 1973.

28. Nilsson PM, Nilsson J-Å, Östergren P-O, Berglund G. Social mobility, marital status, and mortality risk in an adult life course perspective: The Malmö Preventive Project. Scand J Public Health. 2005;33:412–423.

29. Lund R, Holstein BE, Osler M. Marital history from age 15 to 40 years and subsequent 10-year mortality: a longitudinal study of Danish males born in 1953. Int J Epidemiol. 2004;33:389–397.

30. Ikeda A, Hiroyasu I, Hideaki T, et al. Marital status and mortality among Japanese men and women: the Japan Collaborative Cohort Study. BMC Public Health. 2007;7:73.

31. Luoma JB, Pearson JL. Suicide and marital status in the United States, 1991–1996: Is widowhood a risk factor? Am J Public Health. 2002;92:1518–1522.

32. Schiller JS, Lucas JW, Ward BW, Pergoy JA. Summary health statistics for U.S. adults: National Health Interview Survey, 2010. National Center for Health Statistics. Vital Health Stat. 10(252). 2012

33. Schottenfeld D. Patient risk factors and the detection of early cancer. Prev Med. 1972;1:335–351.

34. Osborne C, Ostir GV, Du X, et al. The influence of marital status on the stage at diagnosis, treatment, and survival of older women with breast cancer. Breast Cancer Res Treat. 2005;93:41–47.

35. Syme SL. Behavioral factors associated with the etiology of physical disease: a social epidemiological approach. Am J Public Health. 1974;64:1043–1045.

36. Liu H, Umberson DJ. The times they are a changin’: marital status and health dif- ferentials from 1972 to 2003. J Health Soc Behav. 2008;49(3):239–253.

37. Fine MJ, Ibrahim SA, Thomas SB. The role of race and genetics in health disparities research (editorial). Am J Public Health. 2005;95:2125–2128.

38. McKenney NR, Bennett CE. Issues regarding data on race and ethnicity: the Census Bureau experience. Public Health Rep. 1994;109:16–25.

39. Krieger N. Editorial: counting accountably: implications of the new approaches to classifying race/ethnicity in the 2000 census. Am J Public Health. 2000;90: 1687–1689.

51589_CH04_Printer.indd 226 09/02/13 12:34 PM

r e f e r e n C e s 227

40. Kaufman JS, Cooper RS. Commentary: considerations for use of racial/ethnic classification in etiologic research. Am J Epidemiol. 2001;154:291–298.

41. Williams DR, Mohammed SA, Leavell J, Collins C. Race, socioeconomic status and health: complexities, ongoing challenges and research opportunities. Ann N Y Acad Sci. 2010;1186:69–101.

42. Centers for Disease Control and Prevention. Health disparities experienced by black or African Americans—United States. MMWR. 2005;54:1–3.

43. Centers for Disease Control and Prevention. Summary of notifiable diseases—United States, 2009. MMWR. 58(53):39.

44. Centers for Disease Control and Prevention. Recent trends in mortality rates for four major cancers, by sex and race/ethnicity—United States, 1990–1998. MMWR. 2002;51:49–53.

45. Hirschman J, Whitman S, Ansell D. The black:white disparity in breast cancer mortality: the example of Chicago. Cancer Causes Control. 2007;18:323–333.

46. Centers for Disease Control and Prevention. Racial/ethnic disparities in prevalence, treatment, and control of hypertension—United States, 1999–2002. MMWR. 2005;54:7–9.

47. Flack JM, Ferdinand KC, Nasser SA. Epidemiology of hypertension and cardiovascu- lar disease in African Americans. J Clin Hypertens. 2003;5(1 Suppl 1):5–11.

48. Baldwin L-M, Grossman DC, Casey S, et al. Perinatal and infant health among rural and urban American Indians/Alaska Natives. Am J Public Health. 2002;92:1491–1497.

49. Holman RC, Curns AT, Singleton RJ, et al. Infectious disease hospitalizations among older American Indian and Alaska Native adults. Public Health Rep. 2006;121:674–683.

50. Rhoades ER. The health status of American Indian and Alaska Native males. Am J Public Health. 2003;93:774–778.

51. Knowler WC, Bennett PH, Hamman RF, et al. Diabetes incidence and prevalence in Pima Indians: a 19-fold greater incidence than in Rochester, Minnesota. Am J Epidemiol. 1978;108:497–505.

52. Sievers ML, Nielson RG, Bennett PH. Adverse mortality experience of a southwest- ern American Indian community: overall death rates and underlying causes of death in Pima Indians. J Clin Epidemiol. 1990;43:1231–1242.

53. Schneider E. Tuberculosis among American Indians and Alaska Natives in the United States, 1993–2002. Am J Public Health. 2005;95:873–880.

54. Castor ML, Smyser MS, Taualii MM, et al. A nationwide population-based study identifying health disparities between American Indians/Alaska Natives and the general populations living in select urban counties. Am J Public Health. 2006;96: 1478–1484.

55. Yano K, Blackwelder WC, Kagan A, et al. Childhood cultural experience and the incidence of coronary heart disease in Hawaii Japanese men. Am J Epidemiol. 1979;109:440–450.

56. Williams CL, Berry JW. Primary prevention of acculturative stress among refugees. Am Psychol. 1991;46:632–641.

57. Marmot MG. Stress, social and cultural variations in heart disease. J Psychosom Res. 1983;27:377–384.

58. Syme SL, Marmot MG, Kagan A, Kato H, Rhoads G. Epidemiologic studies of coro- nary heart disease and stroke in Japanese men living in Japan, Hawaii and California: introduction. Am J Epidemiol. 1975;102:477–480.

51589_CH04_Printer.indd 227 09/02/13 12:34 PM

228 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

59. Marmot MG, Syme SL. Acculturation and coronary heart disease in Japanese-Americans. Am J Epidemiol. 1976;104:225–247.

60. Matsumoto YS. Social stress and coronary heart disease in Japan: a hypothesis. Milbank Mem Fund Q Health Soc. 1970;48:9–36.

61. Benfante R. Studies of cardiovascular disease and cause-specific mortality trends in Japanese-American men living in Hawaii and risk factor comparisons with other Japanese populations in the Pacific region: a review. Hum Biol. 1992;64:791–805.

62. Singh GK, Miller BA. Health, life expectancy, and mortality patterns among immigrant populations in the United States. Can J Public Health. 2004;95:114–121.

63. Friis RH, Forouzesh M, Chhim HS, et al. Sociocultural determinants of tobacco use among Cambodian Americans. Health Educ Res. 2006;21:355–365.

64. Centers for Disease Control and Prevention. Trends in tuberculosis—United States, 2011. MMWR. 2012;61:181–197.

65. Delgado JL, Johnson CL, Roy I, Trevino FM. Hispanic Health and Nutrition Examination Survey: methodological considerations. Am J Public Health. 1990;80(suppl):6–10.

66. Trevino FM, ed. Hispanic Health and Nutrition Examination Survey, 1982–84. Am J Public Health. 1990;80(suppl).

67. Garcia-Palmieri MR, Costas R Jr, Cruz-Vidal M, et al. Urban-rural differences in coronary heart disease in a low incidence area: the Puerto Rico Heart Study. Am J Epidemiol. 1978;107:206–215.

68. Hazuda HP, Haffner SM, Stern MP, Eifler CW. Effects of acculturation and socio- economic status on obesity and diabetes in Mexican Americans. Am J Epidemiol. 1988;128:1289–1301.

69. Wong MD, Tagawa T, Hsieh HJ, et al. Differences in cause-specific mortality between Latino and white adults. Med Care. 2005;43:1058–1062.

70. Abraido-Lanza AF, Dohrenwend BP, Ng-Mak DS, Turner JB. The Latino mortality paradox: a test of the “salmon bias” and healthy migrant hypotheses. Am J Public Health. 1999;89:1543–1548.

71. Markides KS, Eschbach K. Aging, migration, and mortality: current status of research on the Hispanic paradox. J Gerontol B Psychol Sci Soc Sci. 2005;60(Spec No 2):68–75.

72. Palloni A, Arias E. Paradox lost: explaining the Hispanic adult mortality advantage. Demography. 2004;41:385–415.

73. Morales LS, Kingston RS, Valdez RO, et al. Socioeconomic, cultural, and behav- ioral factors affecting Hispanic health outcomes. J Health Care Poor Underserved. 2002;13:477–503.

74. Friis R, Yngve A, Persson V. Review of social epidemiologic research on migrants’ health: findings, methodological cautions, and theoretical perspectives. Scand J Soc Med. 1998;26:173–180.

75. Levin JS. Religion and health: is there an association, is it valid, and is it causal? Soc Sci Med. 1994;38:1475–1482.

76. Büssing A, Ostermann T, Matthiessen PF. Search for meaningful support and the meaning of illness in German cancer patients. Anticancer Res. 2005;25:1449–1455.

77. Balboni TA, Vanderwerker LC, Block SD, et al. Religiousness and spiritual support among advanced cancer patients and associations with end-of-life treatment prefer- ences and quality of life. J Clin Oncol. 2007;25:555–560.

78. Fraser GE, Strahan TM, Sabate J, et al. Effects of traditional coronary risk factors on rates of incident coronary events in a low-risk population: the Adventist Health Study. Circulation. 1992;86:406–413.

51589_CH04_Printer.indd 228 09/02/13 12:34 PM

r e f e r e n C e s 229

79. Armstrong B, van Merwyk AJ, Coates H. Blood pressure in Seventh-Day Adventist vegetarians. Am J Epidemiol. 1977;105:444–449.

80. Phillips RL. Role of life-style and dietary habits in risk of cancer among Seventh-Day Adventists. Cancer Res. 1975;35:3515–3522.

81. Butler TL, Fraser GE, Beeson WL, et al. Cohort profile: the Adventist Health Study-2 (AHS-2). Int J Epidemiol. Aug. 27, 2007;1–6 [Epub ahead of print].

82. Montgomery S, Herring P, Yancey A, et al. Comparing self-reported disease outcomes, diet, and lifestyles in a national cohort of black and white Seventh-day Adventists. Prev Chronic Dis. [serial online] 2007;4(3). http://www.cdc.gov/pcd/issues/2007/ jul/06_0103.htm. Accessed July 24, 2012.

83. Jarvis GK. Mormon mortality rates in Canada. Soc Biol. 1977;24:294–302. 84. Enstrom JE. Health practices and cancer mortality among active California Mormons.

J Natl Cancer Inst. 1989;81:1807–1814. 85. Guralnik JM, Leveille SG. Annotation: race, ethnicity, and health outcomes—

unraveling the mediating role of socioeconomic status. Am J Public Health. 1997;87: 728–729.

86. Gregorio DI, Walsh SJ, Paturzo D. The effects of occupational-based social position on mortality in a large American cohort. Am J Public Health. 1997;87:1472–1475.

87. Link BG, Northridge ME, Phelan JC, et al. Social epidemiology and the fundamental cause concept: on the structuring of effective cancer screens by socioeconomic status. Milbank Q. 1998;76:375–402.

88. McDonough P, Duncan GJ, Williams D, House J. Income dynamics and adult mortality in the United States, 1972 through 1989. Am J Public Health. 1997;87: 1476–1482.

89. Berkman LF, Syme SL. Social networks, host resistance, and mortality: a nine-year follow-up study of Alameda County residents. Am J Epidemiol. 1979;109:186–204.

90. Kington RS, Smith JP. Socioeconomic status and racial and ethnic differences in func- tional status associated with chronic diseases. Am J Public Health. 1997;87:805–810.

91. Blane D. Social determinants of health—socioeconomic status, social class, and ethnicity (editorial). Am J Public Health. 1995;85:903–904.

92. Winkelby MA, Jatulis DE, Frank E, Fortmann SP. Socioeconomic status and health: how education, income, and occupation contribute to risk factors for cardiovascular disease. Am J Public Health. 1992;82:816–820.

93. Hollingshead A, Redlich F. Social Class and Mental Illness. New Haven, CT: Yale University Press; 1958.

94. Duncan OD. A socioeconomic index for all occupations. In: Reiss AJ Jr, Duncan OD, Hatt PK, North CC, eds. Occupations and Social Status. New York: Free Press; 1961:109–138.

95. Dunham HW. Society, culture, and mental disorder. Arch Gen Psychiatry. 1976;33:147–156.

96. Mausner JS, Kramer S. Epidemiology: An Introductory Text. 2nd ed. Philadelphia, PA: Saunders; 1985.

97. Murphy JM, Olivier DC, Monson RR, et al. Depression and anxiety in relation to social status: a prospective epidemiologic study. Arch Gen Psychiatry. 1991;48: 223–229.

98. Antonovsky A. Social class, life expectancy and overall mortality. Milbank Mem Fund Q Health Soc. 1967;45:31–73.

99. Adler NE, Boyce T, Chesney MA, et al. Socioeconomic status and health: the challenge of the gradient. Am Psychol. 1994;49:15–24.

51589_CH04_Printer.indd 229 09/02/13 12:34 PM

230 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

100. Rogot E, Sorlie PD, Johnson NJ. Life expectancy by employment status, income, and education in the National Longitudinal Mortality Study. Public Health Rep. 1992;107:457–461.

101. Steenland K, Hu S, Walker J. All-cause and cause-specific mortality by socioeco- nomic status among employed persons in 27 US states, 1984–1997. Am J Public Health. 2004;94:1037–1042.

102. Mackenbach JP, Bos V, Andersen O, et al. Widening socioeconomic inequalities in mortality in six Western European countries. Int J Epidemiol. 2003;32:830–837.

103. Jemal A, Ward E, Anderson RN, et al. Widening of socioeconomic inequalities in U.S. death rates, 1993-2001. PLoS ONE. 2008;3(5):e2181.

104. Rawlings JS, Weir MR. Race- and rank-specific infant mortality in a U.S. military population. Am J Dis Child. 1992;146:313–316.

105. Sorlie P, Rogot E, Anderson R, et al. Black-white mortality differences by family income. Lancet. 1992;340:346–350.

106. Kogevinas M, Marmot MG, Fox AJ, Goldblatt PO. Socioeconomic differences in cancer survival. J Epidemiol Community Health. 1991;45:216–219.

107. Songer TJ, DeBerry J, La Porte RE, Tuomilehto J. International comparisons of IDDM mortality. Diabetes Care. 1992;15(suppl 1):15–21.

108. Birch HG, Richardson SA, Baird D, et al. Mental Subnormality in the Community: A Clinical and Epidemiologic Study. Baltimore, MD: Williams & Wilkins; 1970.

109. World Health Organization (WHO). Global status report on noncommunicable diseases 2010. Geneva, Switzerland: WHO; 2011.

110. World Cancer Research Fund. Stomach cancer. http://www.wcrf.org/cancer_ statistics/stomach_cancer_statistics.php. Accessed July 25, 2012.

111. Division of Control of Tropical Diseases, World Health Organization. World malaria situation 1990. World Health Stat Q. 1992;45:257–266.

112. Rowe AK, Rowe SY, Snow RW, et al. The burden of malaria mortality among African children in the year 2000. Int J Epidemiol. 2006;35:691–704.

113. Lopez AD, Mathers CD, Ezzati M, et al. Global Burden of Disease and Risk Factors. Washington, DC and New York, NY: The World Bank/Oxford University Press; 2006.

114. Meslin FX. Surveillance and control of emerging zoonoses. World Health Stat Q. 1992;45:200–207.

115. Diekstra RFW, Gulbinat W. The epidemiology of suicidal behaviour: a review of three continents. World Health Stat Q. 1993;46:56–57.

116. Central Intelligence Agency. The World Factbook. https://www.cia.gov/library/ publications/the-world-factbook/rankorder/2102rank.html. Accessed July 25, 2012.

117. Kulkarni S, Levin-Rector A, Ezzati M, Murray CJL. Falling behind: life expectancy in US counties from 2000 to 2007 in an international context. Popul Health Metr. 2011;9(1):16.

118. Mason TJ, McKay FW, Hoover R, et al. Atlas of Cancer Mortality in U.S. Counties: 1950–1969. Department of Health, Education and Welfare publication (NIH) 75–780. Washington, DC: Government Printing Office; 1975.

119. Centers for Disease Control and Prevention. Interactive Cancer Atlas. http://www.cdc .gov/Features/CancerAtlas/. Accessed July 30, 2012.

120. Lauer K. The risk of multiple sclerosis in the USA in relation to sociogeographic features: a factor-analytic study. J Clin Epidemiol. 1994;47:43–48.

51589_CH04_Printer.indd 230 09/02/13 12:34 PM

r e f e r e n C e s 231

121. Hutt MSR, Burkitt DP. The Geography of Non-Infectious Disease. Oxford, UK: Oxford University Press; 1986.

122. Yoder JS, Hurral C, Beach MJ. Giardiasis surveillance—United States, 2006–2008. MMWR Surveill Summ. 2010;59:15–25.

123. Centers for Disease Control and Prevention. Update: influenza activity–United States, 2010–11 season. MMWR. 2011;60(21):705–712.

124. Reif S, Geonnotti KL, Whetten K. HIV infection and AIDS in the Deep South. Am J Public Health. 2006;96:970–973.

125. Ingram DD, Franco SJ. NCHS urban-rural classification scheme for counties. National Center for Health Statistics. Vital Health Stat. 2012:2 (154).

126. Collins JG, Hendershot GE. Health Characteristics of Large Metropolitan Statistical Areas: United States, 1988–89. Department of Health and Human Services publica- tion (PHS)93–1515. Hyattsville, MD: National Center for Health Statistics; 1993.

127. Doll R. Urban and rural factors in the aetiology of cancer. Int J Cancer. 1991;47: 803–810.

128. Keil JE, Saunders Jr. DE. Urban and rural differences in cardiovascular disease in blacks. In: Saunders E, ed. Cardiovascular Diseases in Blacks. Philadelphia, PA: Davis; 1991:17–28.

129. Hartley D. Rural health disparities, population health, and rural culture. Am J Public Health. 2004;94:1675–1678.

130. Hillemeier MM, Weisman CS, Chase GA, Dyer AM. Mental health status among rural women of reproductive age: findings from the Central Pennsylvania women’s health study. Am J Public Health. 2008;98(7):1271–1279.

131. Probst JC, Moore CG, Glover SH, et al. Person and place: The compounding effects of race/ethnicity and rurality on health. Am J Public Health. 2004;94:1695–1703.

132. Dzik AJ. Differences in cancer mortality rates in Ohio communities with respect to uraniferous geology. Ohio Med. 1989;85:566–568.

133. Kusiak RA, Springer J, Ritchie AC, Muller J. Carcinoma of the lung in Ontario gold miners: possible aetiological factors. Br J Ind Med. 1991;48:808–817.

134. Nerbrand C, Svardsudd K, Ek J, Tibblin G. Cardiovascular mortality and morbidity in seven counties in Sweden in relation to water hardness and geological settings. The project: myocardial infarction in mid-Sweden. Eur Heart J. 1992;13:721–727.

135. Gyllerup S, Lanke J, Lindholm LH, Scherstén B. Water hardness does not contribute substantially to the high coronary mortality in cold regions of Sweden. J Intern Med. 1991;230:487–492.

136. Clarke KC, McLafferty SL, Tempalski BJ. On epidemiology and geographic infor- mation systems: a review and discussion of future directions. Emerging Infectious Diseases. 1996;2(2):1–11.

137. Ricketts TC. Geographic information systems and public health. Annu Rev Public Health. 2003;24:18.1–18.6.

138. Pan American Health Organization. Use of Geographic Information Systems in Epidemiology (GIS-Epi). Epidemiol Bull. 1996;17:1–6. http://www.paho.org/ english/dd/ais/EB_v17n1.pdf. Accessed July 24, 2012.

139. Fisman DN. Seasonality of infectious diseases. Annual Review of Public Health. 2007;28:127-43

140. Russell D, Douglas AS, Allan TM. Changing seasonality of birth—a possible environmental effect. J Epidemiol Community Health. 1993;47:362–367.

51589_CH04_Printer.indd 231 09/02/13 12:34 PM

232 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e p r o j e C t : D e s C r i p t i v e e p i D e m i o l o g y

141. Maes M, Meltzer HY, Suy E, De Meyer F. Seasonality in severity of depression: relationships to suicide and homicide occurrence. Acta Psychiatr Scand. 1993;88: 156–161.

142. Enquselassie F, Dobson AJ, Alexander HM, Steele PL. Seasons, temperature and coronary disease. Int J Epidemiol. 1993;22:632–636.

143. Division of Control of Tropical Diseases, World Health Organization. World malaria situation, 1988. World Health Stat Q. 1990;43:68–79.

144. Niaura R, Herbert PN, Saritelli AL, et al. Lipid and lipoprotein responses to episodic occupational and academic stress. Arch Intern Med. 1991;151:2172–2179.

145. Porta M. A Dictionary of Epidemiology. 5th ed. New York, NY: Oxford University Press; 2008.

146. Centers for Disease Control and Prevention. Guidelines for investigating clusters of health events. MMWR. 1990;39(Recomm Rep-11):1–23.

147. Romito P. Postpartum depression and the experience of motherhood. Acta Obstet Gynecol Scand. 1990;69(suppl 154):7–19.

148. Ross A, Davis S. Point pattern analysis of the spatial proximity of residences prior to diagnosis of persons with Hodgkin’s disease. Am J Epidemiol. 1990;132 (suppl 1):53–62.

149. Hoffman W, Diekmann H, Schmitz-Feuerhake I. A cluster of childhood leukemia near a nuclear reactor in north Germany. Arch Environ Health. 1997;52:275–280.

51589_CH04_Printer.indd 232 09/02/13 12:34 PM

p r o j e C t : D e s C r i p t i v e e p i D e m i o l o g y 233

appenDix

4 Project : Descript ive Epidemiology of a Selected Health Problem

Select a health problem to explore in detail by using a descriptive epidemiologic approach; for examples of health problems or diseases that might be studied, refer to Exhibit 5–4. You can also access the Internet addresses shown in Exhibit 5–1 in order to develop a reference list.

The objectives of this exercise are as follows:

●● to gain experience in describing and analyzing the distribution of a health disorder in a population

●● to become familiar with various sources of data for the epidemiologic description of a health disorder

●● to enhance the ability to make sound epidemiologic judgments related to public health problems

Examples of sources of data:

●● morbidity and mortality reports (vital statistics): World Health Organization and international reports; U.S., federal, state, and local annual and periodic reports

●● current literature on the selected health problem ●● reports of special surveys

Model for organization of paper:

1. Define the problem (nature, extent, significance, etc.). 2. Describe the agent. 3. Describe the condition (briefly). 4. Examine the above sources for data on morbidity and mortality in the

selected health problem.

51589_CH04_Printer.indd 233 09/02/13 12:34 PM

234 Ch a p t e r 4 De s C r i p t i v e ep i D e m i o l o g y: pe r s o n, pl a C e, t i m e

5. Summarize these data on the distribution of the selected health problem according to the following factors, using tables, graphs, or other illustra- tions whenever possible. A. Host characteristics

1. Age 2. Sex 3. Nativity 4. Marital status 5. Ethnic group

B. Environmental attributes 1. Geographic areas 2. Social and economic factors

a. Income b. Housing

3. Occupation 4. Education

C. Temporal variation 1. Secular 2. Cyclic 3. Seasonal 4. Epidemic

D. Any additional characteristic that contributes to an epidemiologic description of the disease

6. Summarize any current hypotheses that have been proposed to explain the observed distribution.

7. List the principal gaps in knowledge about the distribution of the health problem.

8. Suggest areas for further epidemiologic research. 9. Critically appraise the data as a whole; consult primary sources and

important original papers.

Source: Data are from an exercise distributed at the Columbia University School of Public Health during the early 1970s.

51589_CH04_Printer.indd 234 09/02/13 12:34 PM

3chapte r

235

5chapte r

Sources of Data for Use in Epidemiology

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● identify bibliographic databases for locating epidemiologic research literature

●● note U.S. government sources of epidemiologic data (e.g., census, vital statistics, and others)

●● discuss criteria for assessing the quality and utility of epidemio- logic data

●● indicate privacy and confidentiality issues that pertain to epidemiologic data

●● discuss the uses, strengths, and weaknesses of various epidemio- logic data sources

●● locate a given source of data using resources available on the Internet

CHAPTER OUTLINE

I. Introduction II. Criteria for the Quality and Utility of Epidemiologic Data

III. Online Sources of Epidemiologic Data IV. Confidentiality, Sharing of Data, and Record Linkage V. Statistics Derived from the Vital Registration System

VI. Reportable Disease Statistics VII. Screening Surveys

51589_CH05_Printer.indd 235 09/02/13 12:39 PM

236 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Introduction

Why are data, especially large data sets, crucial for epidemiology and pub- lic health practice? As the basic methodology of public health, epidemiologic research usually requires use of large data sets in order to characterize whole pop- ulations. Data collected on entire populations (or representative samples thereof ) improve one’s ability to generalize observations or findings beyond the group studied. Second, data on populations are needed to provide adequate numbers for statistical inference, that is, for estimating parameters (characteristics of pop- ulation) (e.g., incidence, prevalence, and similar measures of morbidity and mor- tality). For diseases that occur at low frequencies, data must be accumulated on sufficient numbers of at-risk individuals to obtain reliable estimates.

This chapter provides information on where and how to obtain the data used in assessments of morbidity and mortality, for surveillance of disease outbreaks, and for program evaluation. Whether one is talking about incidence, prevalence, secular trends, descriptive epidemiology, or analytic studies of disease etiology, the findings are only as good as the data upon which they are based. As an exam- ple of the vital importance of morbidity and mortality counts, an editorial in The Los Angeles Times stated that accurate mortality figures are required in order to provide appropriate aid to disaster victims such as those from the Haitian earth- quake in January 2010.

The science of measuring mortality and morbidity is controversial. There are bit- ter disputes among groups of researchers who study death tolls in the world’s hot spots. Many governments would also prefer to discretely avoid any discussion

VIII. Disease Registries IX. Morbidity Surveys of the General Population X. Insurance Data

XI. Clinical Data Sources XII. Absenteeism Data

XIII. School Health Programs XIV. Morbidity in the Armed Forces: Data on Active Personnel and

Veterans XV. Other Sources: Census Data

XVI. Conclusion XVII. Study Questions and Exercises

51589_CH05_Printer.indd 236 09/02/13 12:39 PM

i n t r o D u C t i o n 237

of the civilian costs of war. Yet the numbers matter. They can influence politi- cal responses to unarmed conflicts, famines, and natural disasters. Statistics are routinely used to draw attention to evidence of systematic human rights violations and even genocide.*

Another use of health-related data is seeking linkages between exposures and disease. The following passage from MMWR illustrates this point and discusses the crucial function of statisticians in the process. Figure 5–1 shows a statistician at work in the field.

Have you ever wondered how an association between exposure and disease is evaluated? For example, how does the severity of salmonellosis depend on ingested dose of egg products? Or how is the relation between blood lead levels and gasoline lead levels determined? Each of these studies involves statistical analysis.

Since CDC’s inception, an important function of the agency has been the compilation, analysis, and interpretation of statistical information to guide actions and policies to improve health. Sources of data include vital statistics records, medical records, personal interviews, telephone and mail surveys, physical exami- nations, and laboratory testing. Public health surveillance data have been used to characterize the magnitude and distribution of illness and injury; to track health trends; and to develop standard curves, such as growth charts. Beyond the develop- ment of appropriate program study designs and analytic methodologies, statisti- cians have played roles in the development of public health data-collection systems and software to analyze collected data. [Centers for Disease Control and Preven- tion] CDC/[Agency for Toxic Substances and Disease Registry] ATSDR employs approximately 330 mathematical and health statisticians. They work in each of the four coordinating centers, two coordinating offices, and the National Institute for Occupational Safety and Health.24(p 22)

To aid public health practitioners, policy makers, and researchers, U.S. fed- eral agencies compile a great deal of relevant data and are a useful source of health-related information.4 This chapter informs the reader about some of the varied data sources that the government makes available to the general public. These data can be a valuable resource for generating indices of morbidity and mortality for descriptive epidemiologic studies. In addition, the authors describe a number of data sets that can be used for analytic studies that seek to under- stand the etiology of disease. Finally, because data collected by others may not always be perfectly suitable for all situations, this chapter covers some general issues related to primary data collection (i.e., collection of specialized data for a particular research goal).

*Muggah R, Kolbe A. Accurate death tolls matter. Los Angeles Times, July 12, 2011.

51589_CH05_Printer.indd 237 09/02/13 12:39 PM

238 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Throughout the chapter, we will discuss the nature of specific types of data, their strengths versus limitations, and their population coverage. Specific types of data include those gathered during research programs, as well as ongoing surveil- lance of communicable diseases and injuries. Government agencies and health insurance programs collect administrative data related to their activities. The Environmental Protection Agency and local governmental jurisdictions compile a wealth of environmental data. Finally, hospitals, clinics, and physicians gather data on patients in order to track progress and for purposes of accountability.

Nationwide online prescription records and standardized national medical records are likely to be adopted as a result of changes in healthcare funding. Both of these developments will have profound implications for public health practice, especially in the realm of evidence-based public health. The Internet and the growing processing capacity of computers are fostering innovations in data gathering and collection; both have accelerated the processes of capturing data and improving data quality. In the next section, we turn to a discussion of criteria for evaluating epidemiologic data with respect to quality and usefulness.

Figure 5–1 Statistician in the field, collecting data for a study on asphalt milling in Rapid City, South Dakota. Source: Reproduced from Centers for Disease Control and Prevention, Statistics and public health at CDC. MMWR. Vol 56, supplement, p. 22, 2006.

51589_CH05_Printer.indd 238 09/02/13 12:39 PM

C r i t e r i a f o r t h e Q u a l i t y a n D u t i l i t y 239

Criteria for the Quali ty and Uti l i ty of Epidemiologic Data

Figure 5–2 summarizes criteria that epidemiologists can apply to evaluate the quality of data for proposed empirical research studies or program planning and evaluation. These criteria include the nature of the data, availability of the data, completeness of the data, and strengths versus limitations of the data. The quality and specific types of data used for epidemiologic research affect permissible study designs (i.e., descriptive or inferential), possible statistical analyses (e.g., ranging from simple descriptive analyses to complex multivariate modeling), appropriate inferences that can be made, and one’s overall confidence about the soundness of a study’s findings.

The first criterion shown in the figure, nature of the data, includes whether the sources of data are vital statistics, case registries, records from medical practice, surveys of the general population, or cases from hospitals and clinics. The nature of the data affects the types of statistical analyses and inferences that are possible. Further examples of data types are those collected by inter- national agencies (e.g., World Health Organization), environmental monitor- ing agencies (e.g., U.S. Environmental Protection Agency), ongoing disease surveillance programs (e.g., those operated by the Centers for Disease Con- trol and Prevention), and health-related administrative data (e.g., collected by health insurance plans). The growth of medical informatics and implemen- tation of standardized national medical and online prescription records will continue to have major impacts on the nature and uses of epidemiologic data.

The criterion availability of the data relates to the investigator’s access. Medi- cal records and any associated data with personal identifiers usually are not

Figure 5–2 Criteria for evaluating the quality of epidemiologic data.

Epidemiologic data

Nature of the data

Availability of the data

Completeness

Strengths vs. limitations

Permissible study designs Possible statistical analyses Appropriate inferences Quality assurance

51589_CH05_Printer.indd 239 09/02/13 12:39 PM

240 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

available without a release from the individual patient and explicit approval by an institutional review board. Some data from population surveys, stripped of indi- vidually identifying characteristics, are available to the public from government and research organizations on electronic media or online. In order to protect the confidentiality of research subjects, some organizations perturb data before they are released. Data perturbation refers to the process of modifying identifying characteristics of data in order to protect the privacy of individual respondents.

The criterion completeness of population coverage encompasses two sub-criteria. The first, representativeness, refers to the degree to which a sample resembles a parent population. For a given data set, one should assess the degree to which data are representative of the parent population from which the data have been sampled. Is there evidence for omission of major subdivisions of the population, such as individuals from low income or minority groups? Is the population base from which the data have been taken clearly defined, or do the data encompass an unspecified mixture of different populations? A related concept is generaliz- ability (also called external validity), which denotes ability to apply the findings of a study to the population that did not participate in the study. If the data have been sampled from a narrow population subgroup (e.g., medical students), then is would be inappropriate to make generalizations of the results to other groups such as older adults. An aspect of completeness of population coverage is the sub-criterion thoroughness. This sub-criterion is related to the care that has been taken to identify all cases of a given disease including subclinical cases. Do the data represent only the severe cases (the tip of the iceberg) that have come to the attention of health authorities? Are there likely to be substantial numbers of unreported cases?

The criterion strengths versus limitations pertains to the data’s usefulness for various types of epidemiologic research, such as investigations of mortal- ity, detection of outbreaks of infectious disease, and studies of the incidence of chronic diseases. For example, data from death certificates usually are helpful is defining causes of death (and calculating mortality rates), but not as reliable for studying the etiology of diseases This criterion also includes whether there are limiting factors inherent in the data, for example, incomplete diagnostic information and case duplication. Thus, the criterion of strengths versus limita- tions tends to overlap with the criterion of completeness of population coverage. Because the criteria are general principles, one may apply them to data sources not specifically mentioned here, and for the evaluation of published epidemio- logic research.

51589_CH05_Printer.indd 240 09/02/13 12:39 PM

o n l i n e S o u r C e S o f e p i D e m i o l o g i C D a t a 241

Online Sources of Epidemiologic Data

Information from online sources provides a helpful starting point for baseline data for both descriptive and analytic epidemiologic research. A systematic retrieval and review of the existing published literature yield some of the basic facts regarding the occurrence and distribution of diseases. The types of online information include bibliographic databases, data from health-related organizations, and datasets that can be downloaded from federal government and other government websites.

Examples of online bibliographic databases are MEDLINE, TOXLINE, and commercial databases. For example, a premier source of health-related literature is the National Library of Medicine’s PubMed®, a bibliographic search engine available on the Internet at http://www.ncbi.nlm.nih.gov/entrez/ (accessed July 27, 2012). MEDLINE, the main part of PubMed®, focuses on biomedicine and contains a repository of over 21 million journal articles (as of 2012) from the life sciences. Articles contained in MEDLINE are indexed according to Medical Subject Headings (MESH® headings).6 Figure 5–3 shows the opening webpage for PubMed®.

In order to access citations in PubMed®, one enters a combination of search topics or names of authors into a search window and receives a listing of relevant article titles, authors, complete reference citations, and, depending on the specific journal and the year it was published, an abstract. In addition, each article is fol- lowed by a prompt to request more publications similar to the one highlighted. Several other methods for creating searchable terms are possible. Searches may be accomplished through online services from a library, office, or other location, or from a smartphone. In addition to providing access to the published work, it is becoming increasingly common for articles to provide supplementary data tables that can be used for secondary analyses.

The U.S. National Library of Medicine operates TOXLINE, Web address: http://toxnet.nlm.nih.gov/cgi-bin/sis/htmlgen?TOXLINE (accessed July 27, 2012). Highly relevant to epidemiology, the TOXLINE database is keyed to toxicology and contains bibliographical information on the effects of drugs and other chemicals. In 2012 TOXLINE contained over 4 million citations.

The Internet and World Wide Web, particularly their commercial search engines such as Google and ProQuest Dialog®, are resources for bibliographic citations and retrieval of entire articles. ProQuest provides a link to sociological abstracts. Psy- cINFO® (http://www.apa.org/pubs/databases/psycinfo/index.aspx) is a database for the literature on behavioral sciences and mental health (accessed July 27, 2012).

51589_CH05_Printer.indd 241 09/02/13 12:39 PM

242 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

The ERIC™ database (Educational Resources Information Collection) provides bibliographic records and complete articles related to education. The Web address is http://www.eric.ed.gov/ (accessed July 27, 2012).

Many university libraries offer helpful bibliographic resources. For example, the Health Sciences Library System at the University of Pittsburgh offers an excellent home page to link browsers with hundreds of sources of health statistics that are available on the World Wide Web (available at http://www.hsls.pitt. edu/, accessed July 27, 2012).

A final category of data sources encompasses data from health-related organi- zations and datasets that can be downloaded from federal government and other government websites. Information regarding these data sources is provided later in the chapter. The scope of available data available for download is quite exten- sive. See Exhibit 5–1 for a partial list of websites that are relevant to epidemiol-

ogy; some of these sites permit data downloads.

Figure 5–3 The opening webpage for PubMed®. Source: Reproduced from the US National Library of Medicine, National Institutes of Health. Available at http://www.ncbi.nlm.nih.gov/pubmed. Accessed July 26, 2012.

51589_CH05_Printer.indd 242 09/02/13 12:39 PM

o n l i n e S o u r C e S o f e p i D e m i o l o g i C D a t a 243

Selected Internet addresses of Interest to epidemiologists

Administration on Aging, Department of Health and Human Services (DHHS) http://www.aoa.gov/aoaroot/aging_statistics/index.aspx Agency for Healthcare Research and Quality http://www.ahcpr.gov/data

American Cancer Society http://www.cancer.org American Heart Association http://www.heart.org/HEARTORG/ American Lung Association http://www.lungusa.org American Public Health Association http://www.apha.org British Medical Journal http://bmj.com/ Centers for Disease Control and Prevention http://www.cdc.gov Directory of Health and Human Services Data Resources http://www.aspe.hhs.gov/datacncl/datadir Emory University Woodruff Health Sciences Center Library http://health.library.emory.edu/ Ethics in Science http://www.chem.vt.edu/ethics/ethics.html IPRC Indiana Prevention Resource Center, Indiana University, Bloomington http://www.drugs.indiana.edu JAMA Network, American Medical Association http://jamanetwork.com/ MMWR Morbidity and Mortality Weekly Report, Centers for Disease Control and Prevention data based on weekly reports from U.S. state health departments http://www.cdc.gov/mmwr/

e x

h Ib

It 5

–1

continues

51589_CH05_Printer.indd 243 09/02/13 12:39 PM

244 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

National Library of Medicine (NLM), National Institutes of Health (NIH) http://www.nlm.nih.gov/ National Center for Health Statistics (NCHS) http://www.cdc.gov/nchs/ National Heart, Lung, and Blood Institute http://www.nhlbi.nih.gov/ New England Journal of Medicine http://www.nejm.org Pan American Health Organization http://www.paho.org Pub Med®, National Center for Biotechnology Information (NCBI), National Institutes of Health (NIH) http://www.ncbi.nlm.nih.gov/sites/entrez The Food and Drug Administration http://www.fda.gov The WWW Virtual Library: Epidemiology (Biosciences and Medicine), University of California, San Francisco http://www.epibiostat.ucsf.edu/epidem/epidem.html The White House Issues http://www.whitehouse.gov/issues U.S. Census Bureau http://www.census.gov University of Michigan’s Monitoring the Future Study http://monitoringthefuture.org/index.html World Health Organization Statistical Information System (WHOSIS) http://www.who.int/whosis/en/ All sites accessed July 28, 2012. n

exhibit 5–1 continued

Confidential i ty, Sharing of Data, and Record Linkage

The investigator who conducts primary or secondary analyses of epidemiologic data is required to maintain adequate safeguards for privacy and confidentiality of this information; such privacy is legally mandated. Information that must be kept confidential is that which pertains to any personally identifiable features about a living individual; this includes information for which the research subject has not

51589_CH05_Printer.indd 244 09/02/13 12:39 PM

C o n f i D e n t i a l i t y , S h a r i n g o f D a t a 245

given permission for public release. Release of information regarding whether the subject participated in a study also is proscribed.4 Information that would permit identification of a deceased person also should be kept confidential, although the guidelines for release of information about the deceased are not well established.

The Privacy Act of 1974 introduced certain reforms necessary for the protec- tion of confidential records of individuals that are maintained by federal agencies in the United States. Specifically, one of the major provisions of the Privacy Act proscribes the release of confidential data by a federal government agency or its contractors, under most circumstances, without the permission of the client whose records are to be released. On the other hand, the Freedom of Information Act is directed toward the disclosure of government information to the public. It exempts personal and medical files, however, because release of such information would constitute an invasion of privacy. The Public Health Service Act protects the confidentiality of information collected by some federal agencies, such as the National Center for Health Statistics (NCHS). On August 21, 1996, the federal government enacted the Health Insurance Portability and Accountability Act (HIPAA). This law protects individually identifiable health information. For more information regarding the provision of the act, refer to Exhibit 5–2.

Another aspect of release of research information is called data sharing, a process that holds promise for both enhancing data quality and increasing knowledge from research. Noted epidemiologist Jonathan Samet writes, “[e]pide- miologists, academic entities, professional organizations, and funders need to become engaged proactively in data sharing . . . The community of epidemiolo- gists needs to develop its own culture of data sharing, to address the sweeping implications of data sharing, and to engage with researchers in other fields on this issue.”5(p 174)

The definition of data sharing is the voluntary release of information by one investigator or institution to another for purposes of scientific research.6 Illus- trations of data sharing include linkage of large data sets and the pooling of multiple studies in meta-analyses: statistical analyses that combine results from several research projects. One of the key scientific issues in data sharing is the pri- mary investigator’s potential loss of control over intellectual property. However, because of the value of research data to society, many investigators are willing to make their nonconfidential data available to the research community.

The term record linkage refers to joining data about a single entity from two or more sources, for example, employment records and mortality data. Record linkage has been facilitated by the growing number-crunching ability of today’s computer; this processing power helps researchers connect data fields on a single individual after sifting through large data sets that contain common identifying features

51589_CH05_Printer.indd 245 09/02/13 12:39 PM

246 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

the hIpaa privacy rule

The Health Insurance Portability and Accountability Act of 1996 (HIPAA), Public Law 104–191, was enacted on August 21, 1996. Sections 261 through 264 of HIPAA require the Secretary of HHS to publicize standards for the electronic exchange, privacy, and secu- rity of health information . . . [The date for most covered entities to comply with the Privacy Rule was April 14, 2003.]

Protected Health Information. The Privacy Rule protects all “individu- ally identifiable health information” held or transmitted by a covered entity or its business associate, in any form or media, whether electronic, paper, or oral. The Privacy Rule calls this information “protected health information (PHI).” . . . “Individually identifiable health information” is information, including demographic data, that relates to:

●● the individual’s past, present, or future physical or mental health or condition,

●● the provision of health care to the individual, or ●● the past, present, or future payment for the provision of health care to the individual,

and that identifies the individual or for which there is a reasonable basis to believe can be used to identify the individual . . . Individually identifi- able health information includes many common identifiers (e.g., name, address, birth date, Social Security number).

The Privacy Rule excludes from protected health informa- tion employment records that a covered entity maintains in its capac- ity as an employer and education and certain other records subject to, or defined in, the Family Educational Rights and Privacy Act, 20 U.S.C. §1232g.

De-Identified Health Information. There are no restrictions on the use or disclosure of de-identified health information . . . De-identified health information neither identifies nor provides a reasonable basis to identify an individual. There are two ways to de-identify information; either: 1) a formal determination by a qualified statistician; or 2) the removal of speci- fied identifiers of the individual and of the individual’s relatives, household members, and employers is required, and is adequate only if the covered

e x

h Ib

It 5

–2

continues

51589_CH05_Printer.indd 246 09/02/13 12:39 PM

S t a t i S t i C S 247

(e.g., Social Security numbers). Many of the European countries, particularly those in Scandinavia, which have developed extensive, nearly complete social and health records on the resident population, have used linked data in major epidemiologic research projects. Other applications of linked records include the study of clini- cal outcomes associated with the use of anti-inflammatory medications, genetic research, and planning of healthcare services.7 Record linkage systems offer a poten- tially rich source of information that could facilitate research on maternal and child concerns, chronic disease tracking, and the natural history of specific diseases.8

Table 5–1 demonstrates numerous sources of epidemiologic data. Examples of such data range from vital statistics to reports of absenteeism from work or school. Also, the table summarizes the nature of each type of data, their availabil- ity and completeness, population coverage, and strengths versus limitations. The following sections discuss some of the data sources in more detail.

Stat ist ics Derived from the Vital Registrat ion System

Mortality Statistics Data are collected routinely on all deaths that occur in the United States. As is true of other developed countries, U.S. mortality data have the advantage of being almost totally complete because deaths are unlikely to go unrecorded in the United States. Death certificate data in the United States include demographic information about the decedent and information about the cause of death, including the immediate cause and contributing factors. The death certificate is partially completed by the funeral director. The attending physician then com- pletes the section on date and cause of death. If the death occurred as the result of accident, suicide, or homicide, or if the attending physician is unavailable, then the medical examiner or coroner completes and signs the death certificate. Once this is done, the local registrar checks the certificate for completeness and

entity has no actual knowledge that the remaining information could be used to identify the individual . . . n

Source: Adapted from USDHHS. OCR Privacy Brief: Summary of the HIPAA Privacy Rule. Available at http://www.hhs.gov/ocr/privacy/hipaa/understanding/summary/ privacysummary.pdf. Accessed April 17, 2012.

exhibit 5–2 continued

51589_CH05_Printer.indd 247 09/02/13 12:39 PM

248 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

T ab

le 5

–1 O

ve rv

ie w

o f

Ep id

em io

lo gi

c D

at a

So ur

ce s

D at

a So

ur ce

N at

ur e

of D

at a

A va

ila bi

lit y

P op

ul at

io n

C ov

er ag

e St

re ng

th s

ve rs

us L

im it

at io

M or

ta lit

y st

at is

tic s

D at

a fr

om r

eg is

tr at

io n

vi ta

l e ve

nt s

A nn

ua lly

, f ro

m v

it al

gi st

ra tio

n sy

st em

s an

d po

lit ic

al s

ub di

vi si

on s

C om

pl et

e U

se fu

l f or

s tu

dy in

g m

or ta

lit y

ov er

al l

an d

by s

pe ci

fi c

ca us

M ed

ic al

d at

a fr

om b

ir th

co rd

s D

at a

on c

on ge

ni ta

l an

om al

ie s,

c om

pl ic

at io

of p

re gn

an cy

a nd

ild bi

rt h,

b ir

th w

ei gh

et c.

A nn

ua lly

, f ro

m v

it al

gi st

ra tio

n sy

st em

s an

d po

lit ic

al s

ub di

vi si

on s

C om

pl et

e R

ou tin

el y

av ai

la bl

e; s

om e

as pe

ct s

of m

or bi

di ty

m ay

b e

in co

m pl

et el

y re

po rt

ed (

e. g.

, e tio

lo gi

c fa

ct or

R ep

or ta

bl e

di se

as e

st at

is tic

s St

at is

tic s

ba se

d on

ys ic

ia n

re po

rt s,

n ew

se s

of n

ot if

ia bl

e co

m m

un ic

ab le

d is

ea se

W ee

kl y

re po

rt s

fo r

th e

U ni

te d

St at

es C

om pl

et e

U se

fu l f

or d

et ec

tio n

of o

ut br

ea ks

o f

in fe

ct io

us d

is ea

se s;

s om

e co

nd iti

on s

no t

co m

pl et

el y

re po

rt ed

M as

s di

ag no

st ic

a nd

re en

in g

su rv

ey s

D at

a th

at r

es ul

t fr

di ag

no st

ic a

nd s

cr ee

ni ng

st s

fo r

sp ec

if ie

d di

se as

O n

an a

d ho

c ba

si s

Va ria

bl e

U nk

no w

n co

m pl

et en

es s

D is

ea se

r eg

is tr

ie s

St at

is tic

s ba

se d

on e

xi st

in g

ca se

r eg

is tr

ie s

of c

an ce

st ro

ke , e

tc .

C on

tin uo

us , f

ro m

tio na

l, st

at e,

a nd

lo ca

l ju

ris di

ct io

Pr es

um ab

ly c

om pl

et e

fo r

se le

ct ed

d ia

gn os

di se

as es

U se

fu l f

or s

tu dy

in g

in ci

de nc

of d

is ea

se s

su ch

a s

ca nc

er ;

us ed

t o

se le

ct c

as es

in a

se -c

on tr

ol s

tu dy

M or

bi di

ty s

ur ve

ys o

f g en

er al

pu la

tio n

Fr om

t he

U .S

. H ea

lth

Ex am

in at

io n

Su rv

(H ES

) an

d th

e H

ou se

ho ld

te rv

ie w

S ur

ve y

(H IS

)

C on

tin uo

us , d

at a

re le

as ed

r do

w nl

oa d

C om

pl et

e fo

r pr

ob ab

ili ty

m pl

e U

se fu

l f or

e pi

de m

io lo

gi c

re se

ar ch

; s om

e H

IS d

at a

m ay

n ot

b e

ac cu

ra te

fo r

se lf-

re po

rt s;

H ES

d at

a co

nt ai

n m

or e

pr ec

is e

di ag

no se

s H

ea lth

in su

ra nc

e st

at is

tic s

C as

es g

iv en

m ed

ic al

c ar

e un

de r

pr ep

ai d

in su

ra nc

e co

ve ra

N ot

g en

er al

ly a

va ila

bl e,

th ou

gh s

om e

ag en

ci es

nd uc

t re

se ar

ch o

n th

ei r

ow n

in su

re d

C ov

er ed

p op

ul at

io n

U se

fu l f

or h

ea lth

ca re

u til

iz at

io n

st ud

ie s;

se ar

ch o

n m

or bi

di ty

a nd

m or

ta lit

y in

le ct

ed p

op ul

at io

51589_CH05_Printer.indd 248 09/02/13 12:39 PM

S t a t i S t i C S 249

Li fe

in su

ra nc

e st

at is

tic s

M or

ta lit

y da

ta a

nd r

es ul

of p

hy si

ca l e

xa m

in at

io ns

t ho

se a

pp ly

in g

fo r

co ve

ra ge

N ot

g en

er al

ly a

va ila

bl e

In su

ra nc

e po

lic yh

ol de

rs D

at a

fo r

se le

ct ed

p op

ul at

io n

H os

pi ta

l i np

at ie

nt s

ta tis

tic s

C as

es t

re at

ed in

h os

pi ta

l; de

pe nd

en t

on t

yp e

ho sp

it al

G en

er al

ly n

ot a

va ila

bl e

w ith

ou t

sp ec

ia l a

pp ro

va l

N ot

d et

er m

in ab

le D

ia gn

os tic

in fo

rm at

io n

m ay

b e

hi gh

er q

ua lit

y th

an t

ha t

fr om

o th

so ur

ce s;

d iff

ic ul

t to

r el

at e

ca se

s to

po pu

la tio

n de

no m

in at

or H

os pi

ta l o

ut pa

tie nt

at is

tic s

Pa tie

nt s

in c

lin ic

s an

d ou

tp at

ie nt

d iv

is io

(O PD

) of

h os

pi ta

G en

er al

ly n

ot a

va ila

bl e

w ith

ou t

sp ec

ia l a

pp ro

va l

N ot

d et

er m

in ab

le O

PD s

pr ov

id e

a la

rg e

vo lu

m e

of c

ar e;

PD h

os pi

ta l r

ec or

ds a

re s

om et

im es

t w

el l d

ev el

op ed

; d ia

gn os

tic d

at a

m ay

b e

in co

m pl

et e

D at

a on

d is

ea se

s tr

ea te

d in

ec ia

l c lin

ic s

an d

ho sp

it al

s D

ep en

de nt

o n

na tu

of c

lin ic

o r

ho sp

it al

; es

se nt

ia lly

m ed

ic al

c ar

e da

G en

er al

ly n

ot a

va ila

bl e

w ith

ou t

sp ec

ia l a

pp ro

va l

N ot

d et

er m

in ab

le C

ou nt

s of

p at

ie nt

s tr

ea te

d; d

iff ic

ul t

de te

rm in

e pr

ev al

en ce

r at

es ; p

op ul

at io

n de

no m

in at

or u

nk no

w n

D at

a fr

om p

ub lic

h ea

lth

cl in

ic s

D at

a fr

om p

hy si

ca l

ex am

in at

io ns

o f c

lie nt

s G

en er

al ly

n ot

a va

ila bl

e w

ith ou

t sp

ec ia

l a pp

ro va

l N

ot d

et er

m in

ab le

Po ss

ib le

u se

fo r

id en

tif ic

at io

n of

c as

of d

is ea

se a

nd fo

r st

ud y

of h

ea lth

rv ic

es ; p

op ul

at io

n de

no m

in at

un kn

ow n

D at

a fr

om r

ec or

ds o

f ph

ys ic

ia ns

’ p ra

ct ic

es M

ed ic

al c

ar e

pr ov

id ed

ph ys

ic ia

ns ’ o

ff ic

es G

en er

al ly

n ot

a va

ila bl

e w

ith ou

t sp

ec ia

l a pp

ro va

l N

ot d

et er

m in

ab le

M ay

b e

us ef

ul fo

r id

en tif

ic at

io n

ca se

s; r

ec or

ds m

ay v

ar y

in c

om pl

et en

es s

an d

qu al

ity ; d

up lic

at io

n of

c as

es fo

r pa

tie nt

s w

ho s

ee m

ul tip

le p

ro vi

de rs

A bs

en te

ei sm

d at

a Fr

eq ue

nc y

of a

bs en

te ei

fr om

w or

k or

s ch

oo l

B y

sp ec

ia l a

rr an

ge m

en t

w ith

s ch

oo l s

ys te

m o

r in

du st

Pr ob

ab ly

c om

pl et

e fo

r se

le ct

ed p

op ul

at io

n gr

ou ps

N on

sp ec

if ic

in di

ca to

r of

d is

ab ili

in s

el ec

te d

po pu

la tio

n; u

se fu

l f or

se ss

in g

ac ut

e di

se as

e ou

tb re

ak s

D at

a fr

om s

ch oo

l h ea

lth

pr og

ra m

s Fi

nd in

gs o

f p hy

si ca

l ex

am in

at io

ns o

f sc

ho ol

ch ild

re n

G en

er al

ly n

ot a

va ila

bl e

El em

en ta

ry a

nd s

ec on

da ry

ho ol

p op

ul at

io n

U ne

ve n

qu al

ity a

nd c

om pl

et en

es s

da ta

co nt

in ue

51589_CH05_Printer.indd 249 09/02/13 12:39 PM

250 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

D at

a So

ur ce

N at

ur e

of D

at a

A va

ila bi

lit y

P op

ul at

io n

C ov

er ag

e St

re ng

th s

ve rs

us L

im it

at io

St at

is tic

s on

m or

bi di

ty in

m ed

fo rc

es A

rm ed

fo rc

es m

or bi

di ty

d ho

sp it

al iz

at io

n ex

pe rie

nc e;

r es

ul ts

s el

ec tiv

e se

rv ic

e ex

am in

at io

G en

er al

ly a

va ila

bl e

D ra

ft ee

s an

d ca

re er

ili ta

ry p

er so

nn el

C om

pr eh

en si

ve m

or bi

di ty

d at

a on

se le

ct ed

p op

ul at

io n;

im po

rt an

t so

ur ce

r fo

llo w

-u p

st ud

ie s

St at

is tic

s on

v et

er an

s Ve

te ra

ns ’ h

os pi

ta liz

at io

n ex

pe rie

nc e

an d

de at

hs G

en er

al ly

a va

ila bl

e H

os pi

ta liz

at io

n of

t ho

us in

g Ve

te ra

ns A

ff ai

ho sp

it al

s; m

ay b

e in

co m

pl et

e fo

r ve

te ra

w ho

u se

o th

er fa

ci lit

ie s

U se

fu l f

or s

tu dy

in g

ca se

m ix

, de

m og

ra ph

y, a

nd h

os pi

ta liz

at io

n ex

pe rie

nc e

So ci

al S

ec ur

ity s

ta tis

tic s

D is

ab ili

ty b

en ef

it da

ta a

M ed

ic ar

e st

at is

tic s

D at

a ar

e re

le as

ed o

n co

m pu

te r-

re ad

ab le

m ed

ia N

at io

nw id

e U

se fu

l f or

s tu

dy in

g di

sa bi

lit y

La bo

r st

at is

tic s

In ju

rie s

an d

ill ne

ss es

in du

st ry

R ou

tin el

y re

po rt

ed b

y th

e U

.S . B

ur ea

u of

L ab

St at

is tic

W or

ke rs

in v

ar io

oc cu

pa tio

ns a

nd in

du st

rie s

in t

he U

ni te

d St

at es

U se

fu l d

at a

fo r

st ud

yi ng

a cc

id en

ts ,

in ju

rie s,

a nd

o cc

up at

io na

l d is

ea se

C en

su s

da ta

C ou

nt s,

e nu

m er

at io

ns ,

an d

ch ar

ac te

ris tic

s of

pu la

tio ns

b y

ge og

ra ph

lo ca

tio n

in t

he U

ni te

d St

at es

, i nc

lu di

ng a

ge ,

se x,

r ac

D ec

en ni

al c

en su

s an

d an

nu al

e st

im at

es C

om pl

et e

Ex tr

em el

y us

ef ul

fo r

en um

er at

in g

th e

po pu

la tio

n; s

om e

se gm

en ts

o f t

po pu

la tio

n m

ay b

e un

de rc

ou nt

ed (

e. g.

, th

e ho

m el

es s)

T ab

le 5

–1

co nt

in ue

51589_CH05_Printer.indd 250 09/02/13 12:39 PM

S t a t i S t i C S 251

accuracy and sends a copy to the state registrar. The state registrar also checks for completeness and accuracy and sends a copy to the NCHS, which compiles and publishes national mortality rates (e.g., in Vital Statistics of the United States).

Although mortality data are readily available and commonly used for indices of public health, they are hampered also by some limitations with which one should be familiar. The first is certification of the cause of death. When an older person with a chronic illness dies, the primary cause of death may be unclear. Death certificates list multiple causes of mortality as well as the underlying cause. However, assignment of the cause of death sometimes may be arbitrary. In illustration, diabetes may not be given as the immediate cause of death; rather, the certificate may list the cause of death as heart failure or pneumonia, which could be complications of diabetes. Another factor that detracts from the value of death certificates is lack of standardization of diagnostic criteria employed by various physicians in different hospitals and settings. Yet another problem is the stigma associated with certain diseases. For example, if the dece- dent died as a result of acquired immune deficiency syndrome (AIDS) or alco- holism and was a long-time friend of the attending physician, the physician may be reluctant to specify this information on a document that is available to the general public.

Regardless of what the true cause of death might be, a nosologist (person who classifies diseases) must review the death certificate and code the information for compilation; errors in coding are possible, but these can be minimized through standardized training and routine audits. Furthermore, the codes that are used for the causes of death change over time. Since 1900, an international classifi- cation scheme for coding mortality has aided in standardizing causes of death, although periodic changes in classifications have occurred. When the United Nations was formed after World War II, the World Health Organization took charge of this classification system. In 1948, the sixth revision of the Interna- tional Classification of Diseases (ICD) was published.9 The 10th revision is now entitled International Statistical Classification of Diseases and Related Health Problems (ICD-10).10 Epidemiologists and health managers rely on the ICD as the international standard diagnostic classification tool for morbidity and mor- tality.11 If the mortality data one accesses spans more than one version of the ICD, one must be especially careful in interpreting the results, because the codes and groupings of disease may have changed from one edition of the ICD to another. Therefore, sudden increases or decreases in a particular cause of death may not be real, but rather a reflection of a change in coding systems. An exam-

ple of a death certificate and the type of data collected are shown in Exhibit 5–3.

51589_CH05_Printer.indd 251 09/02/13 12:39 PM

252 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Sa m

pl e

D ea

th c

er ti

fic at

1. N

A M

E O

F D

E C

E D

E N

T •

• •

F IR

S T

( G

iv en

)

S TA

T E

O F

C A

LI F

O R

N IA

U S

E B

LA C

K IN

K O

N LY

( N

O E

R A

S U

R E

S , W

H IT

E O

U T

S O

R A

LT E

R AT

IO N

S )

V S

-1 1

(R E

V 1

/0 3)

S TA

T E

F IL

E N

U M

B E

R LO

C A

L R

E G

IS T

R AT

IO N

N U

M B

E R

2. M

ID D

4. D

AT E

O F

B IR

T H

m m

/d d/

cc yy

5. A

G E

/Y rs

7. D

AT E

O F

D E

AT H

m m

/d d/

cc yy

8. H

O U

R (

24 H

ou rs

)

6. S

E X

IF U

N D

E R

O N

E Y

E A

R M

on th

s D

ay s

H ou

rs M

in ut

es IF

U N

D E

R 2

4 H

O U

R S

3. L

A S

T (

FA M

IL Y

)

A K

A A

LS O

K N

O W

N A

S •

• •

In cl

ud e

fu ll

A K

A (

F IR

S T,

M ID

D LE

, L A

S T

)

9. B

IR T

H S

TA T

E /F

O R

E IG

N C

O U

N T

R Y

10 . S

O C

IA L

S E

C U

R IT

Y N

U M

B E

14 ./1

5. W

A S

D E

C E

D E

N T

S PA

N IS

H /H

IS PA

N IC

/L AT

IN O

? (I

f y es

, s ee

w or

ks he

et o

n ba

ck )

11 . E

V E

R IN

U .S

. A R

M E

D F

O R

C E

S ?

12 . M

A R

IT A

L S

TA T

U S

( at

T im

e of

D ea

th )

16 .

D E

C E

D E

N T

’S R

A C

E •

• •

U p

to 3

r ac

es m

y be

li st

ed (

se e

w or

ks he

et o

n ba

ck )

13 . E

D U

C AT

IO N

• •

• H

ig he

st L

ev el

/D eg

re e

(s ee

w or

ks he

et o

n ba

ck )

17 . U

S U

A L

O C

C U

PA T

IO N

• •

• T

yp e

of w

or k

fo r

m os

t o f l

ife . D

O N

O T

U S

E R

E T

IR E

D .

20 . D

E C

E D

E N

T ’S

R E

S ID

E N

C E

( S

tr ee

t a nd

n um

be r

or lo

ca tio

21 . C

IT Y

26 . I

N F

O R

M A

N T

’S N

A M

E , R

E LA

T IO

N S

H IP

28 . N

A M

E O

F S

U R

V IV

IN G

S P

O U

S E

• •

• F

IR S

31 . N

A M

E O

F F

AT H

E R

• •

• F

IR S

35 . N

A M

E O

F M

O T

H E

R •

• •

F IR

S T

)e ma

N nedia M(

T S

AL .03 EL

D DI

M .92 32 . M

ID D

38 . M

ID D

33 . L

A S

T 34

. B IR

T H

S TA

T E

38 . B

IR T

H S

TA T

47 . D

AT E

m m

/d d/

cc yy

43 . L

IC E

N S

E N

U M

B E

R 42

. S IG

N AT

U R

E O

F E

M B

A LM

E R

45 . L

IC E

N S

E N

U M

B E

40 . P

LA C

E O

F F

IN A

L D

IS P

O S

IT IO

N 39

. D IS

P O

S IT

IO N

D AT

E m

m /d

d/ cc

41 . T

Y P

E O

F D

IS P

O S

IT IO

N (S

)

44 . N

A M

E O

F F

U N

E R

A L

E S

TA B

LI S

H M

E N

T 46

. S IG

N AT

U R

E O

F L

O C

A L

R E

G IS

T R

A R

37 . L

A S

T (

M ai

de n)

27 . I

N F

O R

M A

N T

’S M

A IL

IN G

A D

D R

E S

S (

S tr

ee t a

nd n

um be

r or

r ur

al r

ou te

n um

be r,

ci ty

o r

to w

n, s

ta te

, Z IP

)

22 . C

O U

N T

Y /P

R O

V IN

C E

23 . Z

IP C

O D

E 24

. Y E

A R

S IN

C O

U N

T Y

25 . S

TA T

E /F

O R

E IG

N C

O U

N T

R Y

C E

R T

IF IC

A T

E O

F D

E A

T H

18 . K

IN D

O F

B U

S IN

E S

S O

R IN

D U

S T

R Y

( e.

g. , g

ro ce

ry s

to re

, r oa

d co

ns tr

uc tio

n, e

m pl

oy m

en t a

ge nc

y, e

tc .)

19 . Y

E A

R S

IN O

C C

U PA

T IO

V B

P K

V B

P K

R K

FUNERAL DIRECTOR/ LOCAL REGISTRAR

SPOUSE AND PARENT INFORMATION

INFOR- MANT

USUAL RESIDENCEDECEDENT’S PERSONAL DATA

SA MP

exhIbIt 5–3

51589_CH05_Printer.indd 252 09/02/13 12:39 PM

S t a t i S t i C S 253

PLACE OF DEATH CAUSE OF DEATH

PHYSICIAN’S CERTIFICATION CORONER’S USE ONLY

SA MP

51589_CH05_Printer.indd 253 09/02/13 12:39 PM

254 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Birth Statistics: Certificates of Birth and of Fetal Death Presumably, birth and fetal death statistics are nearly complete in their coverage of the general population. Although birth certificate data are needed to calcu- late birth rates, information also is collected about a range of conditions that may affect the neonate, including conditions present during pregnancy, congeni- tal malformations, obstetric procedures, birth weight, length of gestation, and demographic background of the mother. Some of the data may be unreliable, reflecting possible inconsistencies and gaps in the mother’s recall of events during pregnancy. It is also possible that certain malformations and illnesses that affect the neonate may not be detected at the time of birth. Many of the foregoing deficiencies of birth certificates also apply to the data contained in certificates of fetal death. In addition, variations from state to state in requirements for fetal death certificates further reduce their utility for epidemiologic studies. Birth and fetal death certificate data have been employed in studies of environmental influ- ences upon congenital malformations. For example, these data have been used in studies that search for clusters of birth defects in geographic areas where mothers may have been exposed to teratogens, such as pesticides or industrial pollution.

Reportable Disease Stat ist ics

By legal statute, physicians and other healthcare providers must report cases of certain diseases, known as reportable and notifiable diseases, to health authorities. The diseases are usually infectious and communicable ones that might endanger a population; examples are the sexually transmitted diseases, rubella, tetanus, measles, plague, and food-borne disease. Individual states may elect to maintain reports of communicable and noncommunicable diseases of local concern also.

The process of reporting diseases is known as public health surveillance, which denotes “. . . the ongoing systematic collection, analysis, interpretation, and dis- semination of health data.”12(p 290) Healthcare providers and related workers send reports of diseases to local health departments, which in turn forward them to state health departments and then to the Centers for Disease Control and Prevention (CDC). The CDC reports the occurrence of internationally quar- antinable diseases (e.g., plague, cholera, and yellow fever) to the World Health Organization. The method of reporting diseases in the United States (the infor- mation cycle) is illustrated in Figure 5–4.

Supplementing the notifiable disease surveillance system, the CDC oper- ates a surveillance system for a few diseases of interest, such as salmonellosis,

51589_CH05_Printer.indd 254 09/02/13 12:39 PM

r e p o r t a b l e D i S e a S e S t a t i S t i C S 255

Figure 5–4 The information cycle. Source: Reproduced from Centers for Disease Control and Prevention. Principles of Epidemiology. 2nd ed. Atlanta, GA: CDC; 1998, p 305.

Individual case reports

StateLocalReporter

CENTERS FOR DISEASE CONTROL AND PREVENTION

Aggregate reports and interpretations, Recommendations

World Health Organization

shigellosis, and influenza. For example, reports of influenza are tracked from October through May. The CDC collects information from four sources (as shown in Figure 5–5): laboratories across the United States, influenza mortality reports from 121 U.S. cities, sentinel physicians (a network of 150 family prac- tice physicians), and state epidemiologists. The surveillance system (shown in the figure) aids in monitoring influenza outbreaks (e.g., the 2009 H1N1 outbreak) and seasonal influenza outbreaks. For more information about surveillance sys- tems for reporting notifiable disease statistics, refer to Centers for Disease Con- trol and Prevention, Principles of Epidemiology.12

See Exhibit 5–4 for a detailed list of reportable (notifiable) infectious diseases. Some of the diseases and conditions are reportable in some states only; others are reportable in all states. For information regarding U.S. and state require- ments, refer to “Mandatory Reporting of Infectious Diseases by Clinicians, and Mandatory Reporting of Occupational Diseases by Clinicians,” a publication of the Centers for Disease Control and Prevention.13

The major deficiency of this category of data for epidemiologic research purposes is the possible incompleteness of population coverage. First, not every person who develops a disease that is on this list of notifiable conditions may seek medical attention; in particular, persons who are afflicted with asymptomatic and subclinical illnesses are unlikely to go to a physician. For example, an active case of typhoid fever will go unreported if the affected individual is unaware that he or she has the disease. Another factor associated with lack of complete popula- tion coverage is the occasional failure of physicians and other providers to fill out

51589_CH05_Printer.indd 255 09/02/13 12:39 PM

256 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Figure 5–5 Four different surveillance systems for influenza. Source: Reproduced from Centers for Disease Control and Prevention. Principles of Epidemiology. 2nd ed. Atlanta, GA: CDC; 1998, p 309.

Weekly influenza morbidity

NO REPORT NO ACTIVITY SPORADIC REGIONAL WIDESPREAD

Pneumonia and influenza mortality for 121 U.S. cities

the required reporting forms. This shortcoming can occur if responsible indi- viduals do not keep current with respect to the frequently changing requirements for disease reporting in a local area. Also, as discussed earlier, a physician may be unwilling to risk compromising the confidentiality of the physician–patient relationship, especially as a result of concern and controversy about reporting cases of diseases that carry social stigma. For example, incompleteness of AIDS reporting may stem from the potential sensitivity of the diagnosis.4 Robert Friis, who was previously associated with a local health department, observed that widespread and less dramatic conditions such as streptococcal pharyngitis (sore throat) sometimes are unreported. More severe and unusual diseases, such as diphtheria, are usually reported.

51589_CH05_Printer.indd 256 09/02/13 12:39 PM

r e p o r t a b l e D i S e a S e S t a t i S t i C S 257

Infectious Diseases Designated as Notifiable at the National Level during 2009*

Anthrax Arboviral diseases, neuroinvasive and non-neuroinvasive

California serogroup virus Eastern equine encephalitis virus Powassan virus St. Louis encephalitis virus West Nile virus Western equine encephalitis virus

Botulism foodborne infant other (wound and unspecified)

Brucellosis Chancroid Chlamydia trachomatis infections Cholera Coccidioidomycosis Cryptosporidiosis† Cyclosporiasis Diphtheria Ehrlichiosis/Anaplasmosis

Ehrlichia chaffeensis Ehrlichia ewingii Anaplasma phagocytophilum Undetermined

Giardiasis Gonorrhea Haemophilus influenzae, invasive disease Hansen’s disease (Leprosy) Hantavirus pulmonary syndrome Hemolytic uremic syndrome, post-diarrheal

e x

h Ib

It 5

–4

continues

51589_CH05_Printer.indd 257 09/02/13 12:39 PM

258 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Hepatitis, viral, acute Hepatitis A, acute Hepatitis B, acute Hepatitis B virus, perinatal infection Hepatitis C, acute

Hepatitis, viral, chronic Chronic hepatitis B Hepatitis C virus infection (past or present)

Human immunodeficiency virus (HIV) diagnosis§ Influenza-associated pediatric mortality Legionellosis Listeriosis Lyme disease Malaria Measles† Meningococcal disease Mumps Novel influenza A virus infections Pertussis Plague Poliomyelitis, paralytic Poliovirus infection, nonparalytic Psittacosis Q fever†

Acute Chronic

Rabies Animal Human

Rocky Mountain spotted fever Rubella† Rubella, congenital syndrome Salmonellosis Severe acute respiratory syndrome-associated coronavirus ( SARS-CoV)

disease

exhibit 5–4 continued

continues

51589_CH05_Printer.indd 258 09/02/13 12:39 PM

S C r e e n i n g S u r v e y S 259

Shiga toxin-producing Escherichia coli (STEC) Shigellosis Smallpox Streptococcal disease, invasive, Group A Streptococcal toxic-shock syndrome Streptococcus pneumoniae, drug resistant, all ages, invasive disease Streptococcus pneumoniae, invasive disease non-drug resistant, in children

aged <5 years Syphilis Syphilis, congenital Tetanus Toxic-shock syndrome (other than streptococcal) Trichinellosis Tuberculosis† Tularemia Typhoid fever Vancomycin-intermediate Staphylococcus aureus (VISA) infection Vancomycin-resistant Staphylococcus aureus (VRSA) infection Varicella (morbidity) Varicella (mortality) Vibriosis Yellow fever n

*Position statements the Council of State and Territorial Epidemiologists approved in 2008 for national surveillance were implemented beginning in January 2009. No new conditions were added to the notifiable disease list in 2009. †In a 2009 position statement the Council of State and Territorial Epidemiologists approved the modified national TB surveillance case definition. §AIDS has been reclassified as HIV stage III.

Source: Centers for Disease Control and Prevention. Summary of notifiable diseases–United States, 2009, MMWR. 2011;58(53):3.

exhibit 5–4 continued

Screening Surveys

The term screening refers to the application of various tests and procedures in order to identify suspected cases of disease. A screening survey is defined as an investigation of a particular group of persons in order to identify individuals

51589_CH05_Printer.indd 259 09/02/13 12:39 PM

260 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

who have unrecognized health conditions (e.g., infectious or chronic diseases). Because the individual may be unaware of these conditions, they need to be brought to the attention of the person’s healthcare provider for follow-up. For example, some community health agencies organize neighborhood screening clinics for hypertension or breast cancer. Another example is a health fair that may be organized by civic groups. The clientele for the screening programs are highly selected because they consist primarily of individuals who are sufficiently concerned about the disease to participate in screening. Epidemiologic studies might utilize the data yielded from screening programs of this type for research purposes; nevertheless, it would be difficult to generalize the results obtained to any other setting because of the nonrepresentative nature of the sample.

Many large corporations and other employers have set up multiphasic screen- ing programs for their employees. Multiphasic screening is defined as the admin- istration of two or more screening tests during a single screening program. In this type of screening, the employees of an entire large plant may be surveyed. Through the possible early detection of health problems, complications from chronic diseases may be reduced and the life of the employee extended. Because of the ongoing nature of multiphasic screening programs for employees as well as the possible coverage of a total working population, it may be fruitful to utilize data that have been collected for epidemiologic research. These data could be utilized in incidence studies and for research on occupational health problems. One negative feature of the data would be biases resulting from worker attri- tion and turnover. High loss to follow-up would compromise the validity of the study. A second difficulty is that such data may not contain etiologic informa- tion required for a specific analysis.

Disease Registr ies

A registry is a centralized database for the collection of information about a dis- ease. Registries are widely used for the compilation of statistical data on can- cer, prominent examples being the Connecticut Tumor Registry, the California Tumor Registry, and the New York State Cancer Registry. There are many other types of registries devoted to conditions as divergent as mental disabilities, strokes, unintentional injuries, and diabetes. The completeness of population coverage depends upon the ability of the registry staff to secure the cooperation of agencies and medical facilities that would submit data about diseases. If agencies that come into contact with new cases of disease do not report them to the registry, popu- lation coverage will be incomplete. The success of a registry is often dependent

51589_CH05_Printer.indd 260 09/02/13 12:39 PM

D i S e a S e r e g i S t r i e S 261

upon the conscientiousness of the staff and adequate funding. Nonreporting biases also are likely to occur as public concern grows about the confidentiality of medical data; patients may not want their personal records to be released to an outside agency by the service provider. Personal identifiers need to be attached to a medical record when it is released to a registry to permit record linkage or follow-up investigations. Coding algorithms that create a unique identifier for each medical record aid in maintaining the confidentiality of the data.

Several noteworthy applications of registries include patient tracking, develop- ment of information about trends in rates of disease, and the conduct of case-con- trol studies. For example, registries have been used to facilitate regular follow-up of patients with cancer and to study the natural history of infectious and chronic diseases. Population-based cancer registries, such as those incorporated in the Sur- veillance, Epidemiology, and End Results (SEER) program, have provided unique and valuable data on cancer survival, incidence, and treatment14 (Exhibit 5–5).

the Seer program

Surveillance, Epidemiology, and End Results (SEER) is a coordi- nated system of cancer registries strategically located across the United States. These registries routinely collect data on:

●● Patient demographics ●● Primary tumor site ●● Specific cancer markers (e.g., estrogen receptor status; proges- terone receptor status; human epidermal growth factor recep- tor 2, also known as proto-oncogene Neu (HER2/neu breast cancer))

●● Cancer stage at diagnosis ●● First course of treatment ●● Patient survival

The SEER Program, funded by NCI since 1973 as a result of the National Cancer Act of 1971, collects these data on every case of cancer reported from 20 U.S. geographic areas. These areas (shown in Figure 5–6) cover about 28% of the U.S. population and are representative of the demo- graphics of the entire U.S. population. n

e x

h Ib

It 5

–5

continues

51589_CH05_Printer.indd 261 09/02/13 12:39 PM

262 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Morbidity Surveys of the General Populat ion

Morbidity surveys collect data on the health status of a population group. Typi- cally, these surveys use a scientifically designed representative sample of a pop- ulation. The purposes of morbidity surveys are to determine the frequency of chronic and acute diseases and disability, collect measurements of bodily char- acteristics, conduct physical examinations and laboratory tests, and probe other health-related characteristics of specific concern to those who sponsor the survey. Morbidity surveys such as those authorized by the U.S. government strive to gather more comprehensive information than would be available from routinely collected data.14

The National Health Survey Act of 1956 authorized the establishment of a National Health Survey in order to obtain information about the health status of the U.S. population.15 (Note that the National Health Survey named in the Act refers generically to a group of related surveys and not a single survey.) The National Health Survey Act authorized three separate and distinct programs that

Figure 5–6 Geographical areas in the U.S. covered by the SEER Program. Source: Adapted and reprinted from National Cancer Institute, SEER as a Research Resource, NIH Publication No 10-7519, February 2010 and SEER, NIH Publication No 12-4772, March 2012.

Seattle-Puget Sound

Detrolt

Rural Georgla

Atlanta

Cherokee Nation

Los Angeles

San Jose- Monterey

San Francisco- Oakland

SEER Area Funded by NCI

SEER/NPCR Area Funded by NCI and CDC

NPCR-National Program of Cancer Registries CDC-Centers for Disease Control and Prevention

CA IA

AK (Alaska Natives)

exhibit 5–5 continued

51589_CH05_Printer.indd 262 09/02/13 12:39 PM

m o r b i D i t y S u r v e y S o f t h e g e n e r a l p o p u l a t i o n 263

are conducted by the NCHS: the National Health Interview Survey (NHIS, a household health interview survey), the Health Examination Survey (HES), and a family of surveys of health resources. The various components of the survey probe the amount, distribution, and effects of illness and disability in the United States and the services received for or because of such conditions. Information from the programs also is used for the development and improvement of surveys and other methods for obtaining health-related data.

National Health Interview Survey Here are some facts about the NHIS, an ongoing survey, according to the NCHS.

[T]he National Health Interview Survey (NHIS) has monitored the health of the nation since 1957. NHIS data on a broad range of health topics are collected through personal household interviews. For over 50 years, the U.S. Census Bureau has been the data collection agent for the National Health Interview Survey. Sur- vey results have been instrumental in providing data to track health status, health- care access, and progress toward achieving national health objectives. . . .

The NHIS is a large-scale household interview survey of a statistically representa- tive sample of the U.S. civilian noninstitutionalized population. Interviewers visit 35,000–40,000 households across the country and collect data about 75,000– 100,000 individuals. To ensure accuracy of the results, interviewers need to reach as many homes selected for the sample as possible. Once selected, respondents cannot be replaced with anyone else. . . .

On average, an NHIS interview takes about an hour to conduct. The questionnaire consists of two main parts: A core set of questions that remain basically unchanged from year to year and supplemental questions that change from year to year to col- lect additional data pertaining to current issues of national importance. . . .*

The range of conditions studied is comprehensive and includes diseases, injuries, disability, and impairments. Because the survey relies on reports of medical condi- tions by a principal respondent reporting for everyone in the household, the results should be interpreted with caution. These responses may be even less accurate than self-reports, which are known to reflect inadequately certain chronic illnesses.14

Health Examination Survey The HES is the second of the three different programs operated by NCHS as part of the National Health Survey.15 Data are collected from a sample of the civilian, noninstitutionalized population of the United States. Although the Household

*Reprinted and adapted from 12

51589_CH05_Printer.indd 263 09/02/13 12:39 PM

264 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Interview Survey provides data on self-reports of morbidity, the HES provides more direct information about morbidity through examinations, measurements, and clinical tests to yield data on unrecognized and untreated diseases. Many health researchers believe that direct assessment of the respondent yields opti- mal, standardized information about clinical, physiologic, and physical charac- teristics. The HES is conducted in a series of cycles that are limited to a specific segment of the U.S. population. Through the various tests and measurements that are taken, information is obtained about known conditions that the per- son might fail to disclose in an interview only or about previously undiagnosed conditions. One of the uses of HES data is to provide baseline measurements on physical, physiologic, and psychological characteristics not previously available for a defined population.

The HES has been amalgamated with the National Nutrition Surveillance Survey and renamed as the Health and Nutrition Examination Survey (HANES). HANES surveys include the first National Health and Nutrition Examination Survey (NHANES I). The purpose of NHANES I was to conduct a:

survey of the U.S. civilian noninstitutionalized population ages 1–74 years, using a multistage, clustered probability sample stratified by geographic region and pop- ulation size. Interviews and examinations with about 21,000 persons were con- ducted from 1971 through 1974. This sample was augmented with approximately 3,000 persons ages 25–74 in 1974 and 1975. . . Data on all examined persons include household and demographic information; nutrition information; medi- cal, dental, dermatological, and ophthalmological examinations; anthropometric measurements; hand-wrist X-rays (ages 1–17 only); and a variety of laboratory tests.17(p 13)

Related surveys and data collected by the NCHS include NHANES II and the Hispanic Health and Nutrition Examination Survey (HHANES) (Exhibit 5–6). Figure 5–7 shows portable survey laboratories that are moved to field locations for data collection as part of the NHANES.

Several healthcare surveys also are conducted as part of the NHS:

●● National Hospital Discharge Survey ●● National Ambulatory Medical Care Survey ●● National Nursing Home Survey

In addition, there are several vital statistics surveys:

●● National Natality Survey ●● National Fetal Mortality Survey ●● National Mortality Followback Survey

51589_CH05_Printer.indd 264 09/02/13 12:39 PM

m o r b i D i t y S u r v e y S o f t h e g e n e r a l p o p u l a t i o n 265

hispanic health and Nutrition examination Survey (hhaNeS), 1982–1984

From 1982 to 1984, the HHANES was conducted by NCHS to obtain data on the health and nutritional status of three Hispanic groups:

1. Mexican Americans from Texas, Colorado, New Mexico, and California

2. Cuban Americans from Dade County, Florida 3. Puerto Ricans from the New York City area

In the Mexican American portion, 9,894 persons were sampled, of whom 8,554 were interviewed and 7,462 were examined. In the Cuban- American portion, 2,244 persons were sampled, of whom 1,766 were interviewed and 1,357 were examined. In the Puerto Rican portion, 3,786 persons were sampled, of whom 3,369 were interviewed and 2,834 were examined. Respondents, whose ages ranged from 6 months to 74 years, were selected by using a multistage, clustered probability sample. Approxi- mately 76% of the Hispanic origin population of the United States resides in the sampled areas. Interviews, conducted in the household or in a mobile examination clinic, included basic demographic, health history, and health practices information. A variety of tests and medical examinations were performed in the mobile clinics. The survey focused on two major aspects of health: certain important chronic conditions (e.g., heart disease, diabe- tes, hypertension, and depression) and nutrition status. An extensive data- base (n = 12,000) was collected from English or Spanish interviews.

The data are available for download from http://dataarchives.ss.ucla .edu/da_catalog/da_catalog_titleRecord.php?studynumber=H2200V1 (accessed July 28, 2012).

e x

h Ib

It 5

–6

For more information regarding the data from these sources, refer to the articles by Rice18 and Gable.2 The CDC releases data from surveys and other sources via the CDC WONDER online databases on its website: http://wonder.cdc.gov/ datasets.html (active as of July 27, 2012). Users may download public use data sets by following links from the CDC WONDER website. An example of public use data can be found at the link to “AIDS Public Information Data on WONDER.”

51589_CH05_Printer.indd 265 09/02/13 12:39 PM

266 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Behavioral Risk Factor Surveillance System The Behavioral Risk Factor Surveillance System (BRFSS) gathers data primar- ily applicable to chronic diseases on behaviorally related phenomena (e.g., risk behaviors, preventive activities, and healthcare utilization). BRFSS is the largest telephone survey in the world. Established in 1984 by the CDC, BRFSS col- lects information monthly from the entire 50 states of the United States plus the District of Columbia, Puerto Rico, the U.S. Virgin Islands, and Guam. Data from the survey are applied to the evaluation of public health policies and activi- ties; BRFSS data also are used for providing quantitative support for state-based legislative initiatives. The BRFSS web address is http://www.cdc.gov/brfss/about .htm (accessed October 22, 2012).

California Health Interview Survey The California Health Interview Survey (CHIS) provides information on health and demographic characteristics of California residents. The CHIS uses tele- phone survey methods to study the population corresponding to age groups for adults, adolescents, and children. The CHIS collects information on physical and mental health conditions, health behaviors, health insurance coverage, and access to and use of healthcare services, including primary care and preventive services. CHIS has surveyed 42,000 to 56,000 randomly selected households during each wave of data collection beginning in 2001. A full data collection cycle takes 2 years to complete; the survey gathers data on a continuing basis in order to provide annual estimates. A unique feature of the website is AskCHIS, a user-friendly online query program that produces data tables and charts within a few minutes. The CHIS is housed within the UCLA Center for Health Policy

Figure 5–7 Portable units used by the National Health and Nutrition Examination Survey for data collection.

51589_CH05_Printer.indd 266 09/02/13 12:39 PM

C l i n i C a l D a t a S o u r C e S 267

Research at the University of California, Los Angeles. The Web address is http:// www.chis.ucla.edu/about.html (active as of July 27, 2012).

Insurance Data

Social Security, health insurance, and life insurance statistics are all examples of insurance data that have been used widely in epidemiologic studies. One of the major problems inherent in the data is that they lack information about people who are not insured and thus may not accurately represent all segments of society.

Some of the types of information provided by insurance statistics are as fol- lows. Social Security statistics yield data on recipients of disability benefits and participants in Medicare programs. Both types of data may be used in studies of the frequency and severity of disabling conditions. Health insurance statistics contain information about individuals who receive medical care through a pre- paid medical program. Some notable examples of these programs are the Health Insurance Plan of New York and the Kaiser Medical Plan. These plans as well as health maintenance organizations are proliferating as the emphasis on cost containment and primary prevention of disease grows. A number of investigators have employed data from these programs for epidemiologic studies. Data from prepaid plans may be valuable for long-term studies of chronic disease, especially incidence studies, because data collection is often continued over a number of years for each patient; special questionnaires may be added as needed. In contrast to health insurance statistics, life insurance statistics provide data on causes of mortality among insured groups and also on the results of physical examina- tions for those applying for insurance policies. Health and life insurance statistics may contain an overrepresentation of healthier individuals because unhealthy individuals may not be allowed to hold life insurance policies and because an ongoing health insurance program may result in a healthier population among insured individuals than among the noninsured. Findings derived from those populations may not necessarily be representative of the overall population of the United States.

Clinical Data Sources

Clinical data sources include hospital data, information gathered at special clinics and hospitals, data from physicians’ records, and results from clinical laboratories affiliated with clinical sites (not discussed further). See Figure 5–8 for examples

51589_CH05_Printer.indd 267 09/02/13 12:39 PM

268 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

of clinical data sources. As discussed previously, hospital cancer registries forward information to the SEER database. Data from clinical sources are valuable but may present challenges to epidemiologic researchers. Among these are lack of access because of confidentiality of medical records, lack of standardized record keeping and diagnostic procedures, and uncertain representativeness of the data.

Hospital Data Hospital statistics consist of both inpatient and outpatient data that are collected routinely when a patient is admitted to a hospital or enrolls in an outpatient treatment program. Often, these categories of data are deficient for epidemio- logic research applications because the individuals included do not represent any specific population (i.e., the population denominator is undefined). Patients who are treated in the hospital setting, especially those in large metropolitan hospitals, may be drawn from all over the metropolitan area or even from other countries. Furthermore, the types and completeness of information collected on each patient and the diagnostic procedures used by healthcare providers often lack standardization and can be highly variable.

Also, hospital data derived from inpatient and outpatient divisions and spe- cialized clinics are limited with respect to the socioeconomic composition of patients being treated. The majority of patients at certain specialized clinics and

Figure 5–8 Examples of clinical data sources. Source: Adapted from Data Flow in NCI’s SEER Registries. Available at: http://seer.cancer.gov/about/ factsheets/SEER_Data_Flow_.pdf. Accessed August 27, 2012.

Clinical Data Sources Case Identification

Patient Physicians’ Practices Hospitals

Special clinics and HospitalsLaboratory

Clinical Data Sources Case Identification

Patient Physicians’ Practices Hospitals

Special Clinics and HospitalsLaboratory

51589_CH05_Printer.indd 268 09/02/13 12:39 PM

C l i n i C a l D a t a S o u r C e S 269

the renowned hospitals in urban areas may represent the upper socioeconomic strata of our society. At the other extreme, hospital emergency departments and outpatient clinics may contain large proportions of lower socioeconomic individuals in some urban areas because some indigent patients need to use the hospital emergency department and related clinics of public hospitals as their primary source of medical care.

Increasingly in the United States and other countries, hospitals are imple- menting electronic health records that subsequently can be incorporated into electronic databases. An electronic health record is defined as “. . . an elec- tronic repository of patient-centric data that are identifiable, longitudinal and preferably life-long, cross-provider, cross-provider site, and cross the spectrum of health care, including primary care, acute hospital care, long-term care, and home care.”24(p 3) Electronic health records contribute to population health research by facilitating sharing of information among providers and enhancing data standardization.

Diseases Treated in Special Clinics and Hospitals As is true of data from hospitals in general, these data cannot be generalized readily to a reference population because the patients of a special clinic by defini- tion are a highly selected group. Case-control studies might be conducted with patients who present with rare and unusual diseases, but usually one would be unable to determine incidence rates and prevalence of disease without making assumptions about the size of the denominator.

An exception to the general rule about special clinics and hospitals is work done by investigators at the Mayo Clinic in Rochester, Minnesota. The Rochester Epidemiology Project provides a capability for population-based studies of dis- ease causes and outcomes that is unique in the United States, if not the world.20 This ability is due to a medical records-linkage system that has afforded access to details of the medical care provided to residents of Rochester and Olmsted County, Minnesota, since the early 1900s. This project exists, in part, because the Mayo Clinic is geographically isolated from other urban centers. Although best known as a tertiary referral center, the Mayo Clinic has always provided pri- mary and secondary care as well as tertiary care to local residents. Because Mayo offers care in every medical and surgical specialty and subspecialty, local residents are not obliged to seek providers throughout a large region but are able to obtain most of their medical care within the community. Indeed, in a comprehensive survey of community residents, 90% of those who sought medical assessment received care at the Mayo Clinic, the Olmsted Medical Group, or one of their

51589_CH05_Printer.indd 269 09/02/13 12:39 PM

270 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

affiliated hospitals, and 96% selected one of these providers when they had a major medical problem. This unusually close correspondence between a circum- scribed geographic population and its healthcare providers comprises a natural laboratory for population-based studies.

The Rochester Epidemiology Project studies are facilitated by the Mayo uni- fied medical record system, wherein all data about a specific patient are con- tained in a single file linked to a unique Mayo identification number. Today, the dossiers for each of the more than 5 million patients who have ever been seen at the Mayo Clinic in Rochester are maintained in a central repository and tracked by computerized bar codes. Fewer than 500 histories have been lost since the records were initiated about one century ago. In 1966, indices were created for the records of the other providers of medical care to Rochester and Olmsted County residents. The result is linkage of medical data from almost all sources of medical care available to, and used by, the local population in and near Rochester. These sources include the Mayo Clinic and its affiliated hospitals (Saint Mary’s and Rochester Methodist), the Olmsted Medical Group and its affiliated Olmsted Community Hospital, the University of Minnesota hospitals and the Department of Veterans Affairs Medical Center, located in Minneapolis, as well as other medical institutions in the region. During the past several decades, the Rochester Epidemiology Project has successfully provided the data and facilities to complete over 1,000 reports on the epidemiology of acute and chronic diseases. Several of these investigations have been descriptive studies of disease incidence or prevalence, especially long-term trends. Each year, more than half of the Olmsted County population is examined at one of the Mayo facilities, and most local residents have at least one Mayo contact dur- ing any specific three-year period. Thus, the Rochester Epidemiology Project records-linkage system provides what is essentially an enumeration of the popu- lation. Therefore, samples from this system should approximate samples of the general population. This assertion was validated in a survey in which every sub- ject contacted through a random-digit dialing telephone sample or by residence in a local nursing home or senior citizens’ complex was found to have a medical record in the community.

Data from Physicians’ Practices The records from private physicians’ medical practices would seem to be a logical source of information about health and illness. In reality, however, physicians’ medical records usually have limited applications in epidemiologic research. Because of professional codes of confidentiality and privacy, physicians may

51589_CH05_Printer.indd 270 09/02/13 12:39 PM

a b S e n t e e i S m D a t a 271

be either reluctant or forbidden to release any information about their patients without written informed consent. Also, patients of private physicians are a highly select group that can afford the higher cost of private medical care, mak- ing them unrepresentative of the total population. Finally, in the past, little effort was made to standardize the kinds of information collected about each indi- vidual patient, making it difficult to carry out a prevalence study or an incidence study without a considerable amount of missing or incomplete data. The records of physicians in private practice are likely to be highly idiosyncratic documents that cannot be linked readily to other data sources.

Although data exclusively from physicians’ records may be insufficient to gen- erate reliable measures of disease frequency, they nonetheless represent a valu- able supplement to analytic epidemiologic studies. For example, suppose you are interested in assembling a population of women at risk for breast cancer. The plan is to measure usual dietary intake, follow the women for the development of disease, and determine whether dietary exposures measured at baseline were associated with incidence. In this situation it would be important to exclude women who had developed cancer previously because cancers that developed among this subset would not be first cancers. Although such cancer history data could be collected by self-report, a more precise characterization of the cohort at risk could be obtained by verifying self-reported cancers with medical records. It also would be reasonable to take a sample of women who reported themselves free from cancer and verify self-reports against medical records.

Physicians’ records may be an important source of exposure data also. Let us continue with the hypothetical breast cancer study and suppose one decided to collect detailed information about oral contraceptive use. Again, although this information could be obtained by self-report from the subjects themselves, because the use of oral contraceptives requires a physician’s prescription, more detailed information about age at first use, duration of use, and formulation might be obtained from the medical record.

Absenteeism Data

Another kind of data that may be used for epidemiologic research are the records of absenteeism from work or school. This type of data, unfortunately, is sub- ject to a host of possible deficiencies. First of all, these data omit populations that neither work nor attend school. Second, not all people who are absent have an illness. Third, not all people who are ill take time off from work or school. Despite these deficiencies, the data are probably useful for the study of respiratory

51589_CH05_Printer.indd 271 09/02/13 12:39 PM

272 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

disease outbreaks and other rapidly spreading conditions, such as epidemics of influenza, which may be reflected in massive school absenteeism.

School Health Programs

The administration office and school nurse maintain records on the immuniza- tion history of pupils in school, findings of required physical examinations, and self-reports of previous illness. Detailed information may be retained about cog- nitive and other tests. Health-related data from this source are probably sporadic and incomplete, although there are exceptions to this general caveat. School health data have been used in studies of intelligence and mental retardation. For studies of disease etiology, there are some well-known examples where data rou- tinely collected on students have proven to be extremely valuable. Paffenbarger et al.21 used historical records on college students to identify causes of common chronic diseases. They studied nearly 45,000 men and women who attended the University of Pennsylvania from 1931 through 1940 or Harvard Univer- sity from 1921 through 1950. Standardized case taking by physicians in student health services provided data on medical, social, and psychological histories and extensive physical examinations. Although the methods of data collection varied slightly between universities and complete data were not obtained on all stu- dents, valuable measures were taken on such factors as vital capacity, pulse rise after exercise, urinalysis, and electrocardiogram. Information was obtained about mortality through the college alumni office, and causes of death were determined from official state or federal sources. The detailed medical, physiological, and lifestyle information plus the data on mortality afford an efficient analysis of precursive and causative factors.

Morbidity in the Armed Forces: Data on Active Personnel and Veterans

The types of information collected under this heading include reported morbid- ity among active armed forces personnel and veterans, results of routine physi- cal examinations, military hospitalization records, and results of selective service examinations. The last of these were, at one time, universally required of all qual- ified men upon reaching 18 years of age. With the abolition of the draft, physi- cal examinations are given selectively to volunteers for military service. Thus, this source of epidemiologic data, which tends to be representative primarily of

51589_CH05_Printer.indd 272 09/02/13 12:39 PM

o t h e r S o u r C e S : C e n S u S D a t a 273

volunteer groups, is not particularly useful for estimates of disease frequency for the general population.

Nonetheless, records on military personnel may be quite useful for studies of disease etiology. For example, the National Academy of Sciences-National Research Council assembled a large panel of twins to help sort out the influ- ences of “nature and nurture” on the pathogenesis of disease.22 The twin panel comprised approximately 16,000 white male twin pairs born between 1917 and 1927. Both members of each twin pair served in the U.S. military dur- ing the Korean War or World War II. Zygosity was determined by blood typ- ing for 806 pairs, by fingerprinting for 1,947 pairs, and by questionnaire for 10,732 pairs. Several investigators have used these data to examine the role of genetic factors in human obesity.23,24 Weight and height, measured dur- ing the induction physical examination, were available for 5,884 monozygotic (identical) and 7,492 dizygotic (fraternal) pairs. A comparison of the degree of similarity in various measures of obesity suggested that the identical twins were more alike than the nonidentical twins, an expectation consistent with genetic influences.23

Other Sources: Census Data

The U.S. Bureau of the Census provides much information of value to epidemi- ologic research, for example, general, social, and economic characteristics of the U.S. population. The U.S. Census is administered every 10 years to the entire population of the nation. The decennial census attempts to account for every person and his or her residence and to characterize the population according to sex, age, family relationships, and other demographic variables.25 Beginning with the 1940 census, a more detailed questionnaire also has been administered to representative samples of the population. The Census Bureau also makes annual estimates of the number of persons in the population. Some of the publications developed by the Bureau of the Census include the following:

●● Statistical Abstract of the United States ●● County and City Data Book ●● Decennial Censuses of Population and Housing ●● Historical Statistics of the United States, Colonial Time to 1970

Refer to the Census Bureau website (http://www.census.gov/) for more infor- mation about available publications and other information.

51589_CH05_Printer.indd 273 09/02/13 12:39 PM

274 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

Conclusion

This chapter covered a variety of types and sources of data used in epidemio- logic research. Epidemiologists need to find the best quality of data in order to describe the distribution of morbidity and mortality in a population or to conduct studies of disease etiology. To assess the potential utility of data, one needs to consider the nature, availability, representativeness, and completeness of the data. The criterion nature of the data includes whether the data are from vital statistics, case registries, physicians’ records, surveys of the general popula- tion, or hospital and clinic cases. The criterion availability of the data relates to the investigator’s ability to gain access to the data. The criterion representative- ness or external validity refers to generalizability of findings to populations other than the one from which the data have been obtained. Related to the extent of population coverage is the criterion completeness of the data, which refers to the thoroughness of identification of all cases with a particular health phenomenon, including subclinical cases. The criterion strengths versus limitations denotes the utility of the data for various types of epidemiologic research.

Some of the diverse sources of epidemiologic data include statistics compiled by government, industry, or organizations such as the United Nations. Much progress has been made in the development of computerized databases and the Internet; a helpful starting point for epidemiologic research studies is a system- atic retrieval of information from computerized bibliographic sources. Examples of epidemiologic data are those derived from the vital registration system, reports of absenteeism from work or school, disease registries, morbidity surveys of the general population, hospital statistics, and census tracts. Epidemiologic data from these sources have many valuable applications, including development of descriptive studies of trends in disease and analytic studies of disease etiology.

Study Questions and Exercises

1. Are you able to define the following? a. disease registry b. National Health Survey c. NHANES I and HHANES

2. What is likely to be the best routinely available data source for each of the following kinds of studies? a. incidence of influenza in the United States b. cancer morbidity

51589_CH05_Printer.indd 274 09/02/13 12:39 PM

S t u D y Q u e S t i o n S a n D e x e r C i S e S 275

c. congenital malformations d. prevalence of selected disabling conditions e. work-related accidents f. precursive factors for heart disease among college graduates g. ethnic differences in mortality

3. Death certificates are an important source of information for epidemio- logic studies. In the United States, death certificates have which of the following advantages (circle all that apply): a. There is a uniform national system of collection and coding. b. The cause of death is usually confirmed by autopsy. c. The international coding system for cause of death has remained con-

stant since 1900. d. Data collection is comprehensive; virtually no deaths go unrecorded. e. The decedent’s personal physician always completes the form and can

add his or her own knowledge of past illnesses. 4. Which of the following data sources are best able to provide numerator

data for the calculation of incidence of death by gunshot? a. hospital discharge survey b. autopsy or coroners’ records c. National Health Survey d. disease registries e. prepaid group practice insurance programs

5. An abrupt drop in mortality due to a specific cause is observed from one year to the next. Identify at least three possible reasons for such a change.

6. Pick up the local newspaper and search for an article on a recent medical finding or public health issue. Conduct a Medline search to find relevant published articles on the same topic.

7. Access the University of Pittsburgh’s “Guide to Locating Health Sta- tistics” on the Internet. Determine five vital statistics on your city or county: income, education, health care, land, and mortality rates.

8. State funding for a childhood injury prevention program has just become available. To gather baseline data on childhood injuries, the staff is discussing whether to conduct a survey or establish a surveil- lance system. Discuss the advantages and disadvantages of these two approaches.

9. During the previous six years, one to three cases per year of Kawasaki syndrome had been reported by a state health department. During the past 3 months, 17 cases have been reported. All but two of these cases have been reported from one county. The local newspaper carried an

51589_CH05_Printer.indd 275 09/02/13 12:39 PM

276 C h a p t e r 5 S o u r C e S o f D ata f o r u S e i n e p i D e m i o l o g y

article about one of the first reported cases, a young girl. Describe the possible causes of the increase in reported cases.

10. You have recently been hired by a state health department to run surveil- lance activities, among other tasks. All surveillance data are entered into a personal computer and transmitted to the Centers for Disease Control and Prevention each week. The state, however, has never generated its own set of tables for analysis. What three tables might you want to gener- ate by computer each week?

11. Suppose that last week, the state public health laboratory diagnosed rabies in four raccoons that had been captured in a wooded residential neighborhood. This information will be duly reported in the tables of the monthly state health department newsletter. Is this sufficient? Who needs to know this information?

Source: (Questions 8–11): Centers for Disease Control and Prevention. Principles of Epidemiology, 2nd ed. Atlanta, GA: CDC; 1998:332–334.

References

1. Sieber WK, Green T, Williamson GD. Statistics and public health at CDC. MMWR. 2006;55(suppl):22–24.

2. Gable CB. A compendium of public health data sources. Am J Epidemiol. 1990;131:381–394.

3. National Library of Medicine, National Institutes of Health, National Center for Biotechnology Information. PubMed. http://www.ncbi.nlm.nih.gov/entrez/. Accessed July 26, 2012.

4. Feinleib M. The epidemiologist’s responsibilities to study participants. J Clin Epidemiol. 1991;44(suppl I):73S–79S.

5. Samet JM. Data: to share or not to share? Epidemiology. 2009;20(2):172–174. 6. Hogue CJR. Ethical issues in sharing epidemiologic data. J Clin Epidemiol.

1991;44(suppl I):103S–107S. 7. Lilienfeld DE, Stolley PD. Foundations of Epidemiology. 3rd ed. New York: Oxford

University Press; 1994. 8. Merrill RM. Introduction to Epidemiology. 5th ed. Sudbury, MA: Jones and Bartlett

Learning; 2013. 9. World Health Organization (WHO). International Classification of Diseases. 6th rev.

Geneva, Switzerland: WHO; 1948. 10. World Health Organization (WHO). International Statistical Classification of

Diseases and Related Health Problems. 2nd ed. 10th revision. Geneva, Switzerland: WHO; 2004.

11. World Health Organization (WHO). Classifications: International Classification of Diseases (ICD). http://www.who.int/classifications/icd/en/. Accessed March 15, 2012.

51589_CH05_Printer.indd 276 09/02/13 12:39 PM

r e f e r e n C e S 277

12. Centers for Disease Control and Prevention. Principles of Epidemiology. 2nd ed. Atlanta, GA: 1998.

13. Chorba TL, Berkelman RL, Safford SK, Gibbs NP, Hull HF. Mandatory reporting of infectious diseases by clinicians, and mandatory reporting of occupational diseases by clinicians. MMWR Recomm Rep. 1990;39(RR-9):1–17.

14. Mausner JS, Kramer S. Epidemiology: An Introductory Text. 2nd ed. Philadelphia, PA: Saunders; 1985.

15. Miller HW. Plan and Operation of the Health and Nutrition Examination Survey: United States—1971–1973 (Vital and Health Statistics ser 1, no 10a). Department of Health, Education and Welfare publication (PHS)79–1310. Hyattsville, MD: National Center for Health Statistics; 1978.

16. Centers for Disease Control and Prevention, National Center for Health Statistics. National Health Interview Survey. http://www.cdc.gov/nchs/data/nhis/ brochure2010January.pdf. Accessed July 27, 2012.

17. Turczyn KM, Drury TF. An Inventory of Pain Data from the National Center for Health Statistics. Department of Health and Human Services publication (PHS)92– 1308. Hyattsville, MD: National Center for Health Statistics; 1992.

18. Rice DP. Data needs for health policy in an aging population (including a survey of data available in the United States of America). World Health Stat Q. 1992;45:61–67.

19. Friedman DJ. Assessing the potential of national strategies for electronic health records for population health monitoring and research. National Center for Health Statistics. Vital Health Stat. 2006;2(143).

20. Melton LJ III. History of the Rochester Epidemiology Project. Mayo Clin Proc. 1996;71:266–274.

21. Paffenbarger RS Jr, Notkin J, Kreuger DE, et al. Chronic disease in former college students, II. methods of study and observations on mortality from coronary heart disease. Am J Public Health. 1966;56:962–971.

22. Jablon S, Neel JV, Gershowitz H, et al. The NAS-NRC twin panel: methods of construction of the panel, zygosity diagnosis, and proposed use. Am J Hum Genet. 1967;19:133–161.

23. Stunkard AJ, Foch TT, Hrubec Z. A twin study of human obesity. JAMA. 1986;256:51–54.

24. Selby JV, Newman B, Quesenberry CP Jr, et al. Evidence of genetic influence on central body fat in middle-aged twins. Hum Biol. 1989;61:179–193.

25. U.S. Bureau of the Census. Statistical Abstract of the United States: 2008. 127th ed. Washington, DC: U.S. Bureau of the Census; 2007.

51589_CH05_Printer.indd 277 09/02/13 12:39 PM

51589_CH05_Printer.indd 278 09/02/13 12:39 PM

3chapte r

279

6chapte r

Study Designs: Ecologic, Cross-Sectional,

Case-Control

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● define the basic differences between observational and experimental epidemiology

●● identify an epidemiologic study design by its description ●● list the main characteristics, advantages, and disadvantages of

ecologic, cross-sectional, and case-control studies ●● describe sample designs used in epidemiologic research ●● calculate and interpret an odds ratio

CHAPTER OUTLINE

I. Introduction II. Observational Versus Experimental Approaches

in Epidemiology III. Overview of Study Designs Used in Epidemiology IV. Ecologic Studies V. Cross-Sectional Studies

VI. Case-Control Studies VII. Conclusion

VIII. Study Questions and Exercises

51589_CH06_Printer.indd 279 18/02/13 8:18 PM

280 C h a p t e r 6 S t u d y d e S i g n S

Introduction

An arsenal of study design options is available to the epidemiologist. The selection of a particular technique from this arsenal and its application to the study of health issues is a central theme of this chapter. The discussion dem- onstrates that the choice of a study design is, to a certain extent, dependent on the amount of information that is already known about a particular health issue proposed for investigation. When relatively little is known, the investiga- tor should not commence a costly and lengthy study. Rather, a more prudent approach would be to employ, if possible, a study design that uses existing data, is quick and easy to conduct, and is economical. As knowledge increases, and the complexity of the research questions increases, then more rigorous study designs may be merited.

The preceding paragraph, although an oversimplification, previews some of the factors involved in the full decision process of selecting a particular study design. This chapter will provide a more complete picture of the factors involved by presenting the various study designs in sequence from simpler, faster, and less expensive to more complex, time consuming, and expensive. This chapter explores three varieties of observational studies (ecologic, cross-sectional, and case-control). An attempt will be made to justify the added expense (in time, resources, and money) of each new design over its predecessors.

This chapter demonstrates that the major study designs differ from one another in several respects:

●● Number of observations made: In some cases, observations on subjects may be made at only a single point in time, whereas in others observations or examinations are made at two or more points in time.

●● Directionality of exposure: This measurement relative to disease varies. The investigator may elect to start with subjects who already have a disease and ask them retrospectively about previous exposures that may have led to the outcome under study, or he or she may start with a disease-free group for which exposures are determined first. The lat- ter group would then be followed prospectively for development of disease.

●● Data collection methods: Some methods require almost exclusive use of exist- ing, previously collected data, whereas others require collection of new data.

●● Timing of data collection: If long periods of time have elapsed between measurement of exposure and disease, questions might be raised about the quality and applicability of the data.

51589_CH06_Printer.indd 280 18/02/13 8:18 PM

O b S e r v a t i O n a l v e r S u S e x p e r i m e n t a l a p p r O a C h e S 281

●● Unit of observation: For some studies the unit of observation is an entire group, whereas for others the unit of observation is the individual.

●● Availability of subjects: Certain classes of subjects may not be available for epidemiologic research as a result of a number of considerations, including ethical issues.

In defining and characterizing the more common study designs used in epi- demiology, this chapter places particular emphasis on the following key points: how study subjects are selected, how each design fits into the spectrum of design options, and how each design has inherent strengths and weaknesses. The discus- sion will not belabor the points about each type of design but rather will provide a sense of how they differ from one another and how they are applied.

Observational Versus Experimental Approaches in Epidemiology

A basic typology of epidemiologic research will help put the various study designs in proper perspective.1 Consider two basic facets of research designs:

1. Manipulation of the study factor (M) means that the exposure of interest is controlled by the investigator, a government agency, or even nature, and not by the study subjects. For example, local water treatment plant person- nel may have chlorinated the water supply. Water consumers are exposed to chlorine and byproducts of the chlorination process because of the water treatment regulations, and not necessarily because of their own free choice.

2. Randomization of study subjects (R) refers to a process in which chance determines the likelihood of subjects’ assignment to exposure conditions. Thus, by a random process such as the flip of a coin, for example, an individual may be designated to receive either an intensive, experimental smoking cessation program or the current standard of care.

The various permutations of these two factors, M and R, produce three differ- ent study types: experimental, quasi-experimental, and observational (Table 6–1).

Table 6–1 Typology of Epidemiologic Research

M R Study Type

Yes Yes No

Yes No No

Experimental Quasi-experimental Observational

51589_CH06_Printer.indd 281 18/02/13 8:18 PM

282 C h a p t e r 6 S t u d y d e S i g n S

We shall learn later that an experimental study involves both M and R, and that a quasi-experimental study involves M but not R. An observational study involves neither M nor R.

Overview of Study Designs Used in Epidemiology

To compare and contrast the main types of study designs used in epidemiology, we provide a brief summary of each of the three types. Figure 6–1 depicts the inter- relationships of the various study designs, which are discussed in the next sections.

Experimental Studies In comparison with quasi-experimental and observational studies, experimental studies maintain the greatest control over the research setting; the investigator both manipulates the study factor and randomly assigns subjects to the exposed and nonexposed groups. As shown in Figure 6–1, researchers are able to control

Figure 6–1 Overview of epidemiologic study designs. Source: Adapted and reprinted from Lilienfeld AM. Advances in quantitative methods in epidemiology. Public Health Rep. 1980; 95(5:464).

THE EPIDEMIOLOGIC STUDY

Experimental Studies

Community Trials

Clinical Trials

Observational Studies

“Time-span Studies”

Cross-sectional and/or Retrospective Studies

Retrospective Studies

Prospective Studies

Cross-sectional Studies

Uncontrolled Assignment (Not Randomized)

Sampling with Regard to Disease or Effect

Sampling with Regard to Exposure, Characteristic, or Cause

History of Exposure or Characteristic (Prior to Time of Study)

Exposure or Characteristic at Time of Study

Controlled Assignment*

Randomized Assignment

*Assignment of subjects to study conditions

Non-randomized Assignment (Quasi-Experimental)

51589_CH06_Printer.indd 282 18/02/13 8:18 PM

O v e r v i e w O f S t u d y d e S i g n S u S e d i n e p i d e m i O l O g y 283

assignment of the study subjects to the conditions of the study; the investigator manipulates the study factor (e.g., drug or placebo).

From the perspective of epidemiology, one common experimental design is a clinical trial, used primarily in research and teaching hospitals for several pur- poses: to test the efficacy of new therapies, surgical procedures, or chemopreven- tive agents; to test etiologic hypotheses and estimate long-term effects; and to study the effects of interventions to modify health status. For example, dietary modification of fat intake may be tested within the context of a controlled clini- cal trial to determine acceptability, potential problems, and sources of dissatis- faction or confusion. Clinical trials thus may help demonstrate the feasibility of a large-scale population intervention. The number of subjects who are included in a clinical trial may limit its conclusions and generalizability. Large trials are indeed conducted, but their considerable expense, resource, and time constraints limit their use. In a clinical trial, usually participants are assigned randomly to the conditions of the study.

Quasi-Experimental Studies Quasi-experimental studies include community trials (community interven- tions), which are types of experimental designs that greatly enhance the poten- tial to make a widespread impact on a population’s health. Typical community interventions are oriented toward education and behavior change at the pop- ulation level. Examples of issues addressed are smoking cessation, control of alcohol use, weight loss, establishment of healthy eating behaviors, and encour- agement of increased physical activity. Community interventions also focus upon persons at high risk of disease within a particular population. Finally, successful community interventions may suggest public health policies, such as mandatory seat belt use or proscription of alcohol consumption by pregnant women.

Table 6–1 shows that quasi-experimental studies involve manipulation of the study factor but not randomization of study subjects; thus, in some respects they may be thought of as natural experiments. Before federal law mandated seat belt use in the United States, individual states varied in seat belt legislation; some states had seat belt laws, and others did not. Residents in the various states did not determine their own “exposure” to seat belts; rather, state politicians who enacted the seat belt laws were responsible for assignment of the “exposure.” A comparison of traffic fatalities in states with and without seat belt laws repre- sents a quasi-experimental design.

51589_CH06_Printer.indd 283 18/02/13 8:18 PM

284 C h a p t e r 6 S t u d y d e S i g n S

By contrasting appropriate indices before and after public health programs are implemented, the quasi-experimental design can be used to evaluate the extent to which the programs meet public health goals. For example, to evaluate the effectiveness of safety devices intended to prevent percutaneous injuries in a hospital setting, a quasi-experimental design was used. Results indicated that a 3-hour course on occupationally acquired blood-borne infections and a 2-hour hands-on training session with the devices decreased percutaneous injuries by 93%.2 Other applications for the quasi-experimental approach are to compare programs to determine reasons for success or failure of an intervention, to com- pare costs and benefits, and to suggest changes in current health policies or pro- grams. For example, the 1987 Omnibus Budget Reconciliation Act included regulation of antipsychotic drug use in nursing homes. An analysis of drug use in all Medicare- and Medicaid-certified nursing homes in Minnesota revealed that the rates of antipsychotic drug use declined by more than a third in apparent anticipation of, and as a result of, the legislation.3 Thus, the legislation appeared to achieve its intended effect.

Observational Studies In some instances an experiment would be impractical and in others, unethical. Accordingly, much of epidemiologic research is relegated to observational studies, which, as shown in Table 6–1, entail neither manipulation of the study factor nor randomization of study subjects. Rather, observational studies make use of careful measurement of patterns of exposure and disease in populations to draw inferences about etiology. There are two main subtypes of observational studies:

1. Descriptive studies include case reports, case series, and cross-sectional surveys. They are used to depict individuals’ health characteristics (e.g., morbidity from specific diseases or mortality) with respect to person, place, and time, and to estimate disease frequency and time trends. Although descriptive studies may be used for health planning purposes and allocation of resources, they are used also to generate etiologic hypotheses.

2. Analytic studies include ecologic studies, case-control studies, and cohort studies. These designs are employed to test specific etiologic hypotheses, to generate new etiologic hypotheses, and to suggest mechanisms of cau- sation. As a body of knowledge builds regarding likely etiologic factors for a disease, it becomes possible to generate preventive hypotheses and to suggest and identify potential methods for disease prevention.

51589_CH06_Printer.indd 284 18/02/13 8:18 PM

O v e r v i e w O f S t u d y d e S i g n S u S e d i n e p i d e m i O l O g y 285

Note that a cohort study differs from other observational study designs with regard to sampling for exposure. Sampled study participants are those who have had a specific exposure, which is the factor that determines assignment to a cohort. One type of cohort studies is called prospective cohort studies because the assessment of the study outcome (e.g., a disease) occurs prospectively after exposure has already occurred. (See Figure 6–1.)

The 2 by 2 Table The foregoing concludes the brief overview of three types of epidemiologic research designs: experimental, quasi-experimental, and observational. The remainder of this chapter defines and illustrates one group of observational stud- ies more fully. When thinking about study designs, it is helpful for the reader to visualize how the study groups are assembled in the context of the 2 by 2 table. The reader should bear in mind an important caveat, however: The model tends to underestimate the complexity of the potential linkage between exposure and disease. Exposures do not always fall neatly into the categories of exposed and nonexposed. Hence, the notion of one exposure–one disease is admittedly naive. Nevertheless, a comprehension of this rather simplistic model leads one to an understanding of more complex issues, such as a single exposure with multiple levels or more than one exposure.

Table 6–2 depicts the 2 by 2 table, an important tool in evaluating the association between exposure and disease. Note that the columns represent dis- ease status or outcome (yes or no) and that the rows represent exposure status (yes or no). To avoid confusion, remember that the first column always should refer to those with the disease and the first row should refer to those with the

Table 6–2 The 2 by 2 Table Represents the Association Between Exposure and Disease Status

Disease Status

Yes (people with disease)

No (people without disease) Total

Exposure Status

Yes (exposure present)

A (exposure & disease present)

B (exposure present, but no disease)

A + B (total number exposed)

No (no exposure)

C (no exposure, disease present)

D (no disease, no exposure)

C + D (total number with no exposure)

A + C (total number with disease)

B + D (total number without disease)

N (sample total)

51589_CH06_Printer.indd 285 18/02/13 8:18 PM

286 C h a p t e r 6 S t u d y d e S i g n S

exposure of interest. Although this recommended standard is not critical to the representation of data, consistency in usage will establish a common frame of reference for comparison of study designs and will reduce the likelihood of errors when one is calculating measures of effect.

The table cross-classifies exposure status and disease status. Thus, the total number of individuals with disease is A + C, and the total number free from disease is B + D. The total number exposed is A + B, and the total number not exposed is C + D. These four totals are referred to as the marginal totals. The entries (or cells) within the table represent the cross-classification of exposure and disease. Thus, the entry labeled A reflects the number of subjects who had both exposure and disease, B reflects the number of subjects with exposure but no disease, C represents subjects with disease but without exposure, and D is the number of individuals who have neither disease nor exposure. For most study design options, the researcher is aware of the joint classification (or distribution) of exposure and disease for each subject.

One means for keeping track of the different observational study designs is to think of each in terms of the point of reference for selection of the study groups. Inspection of Figure 6–2, a simplified version of Table 6–2, reveals several options. For example, one could start by selecting a sample number (N) and then determining each subject’s exposure and disease status. The results would be tabulated and entered into the four cells of the table (A, B, C, and D). The marginal totals would be determined afterward. This approach is a cross-sectional study. Alternatively, one could start with the marginal totals of exposed (A + B) and nonexposed (C + D) subjects and follow them for the

Figure 6–2 The 2 by 2 table.

Yes

Yes No Total

A + B

C + D

A + C B + D

Exposure Status

Disease Status

51589_CH06_Printer.indd 286 18/02/13 8:18 PM

e C O l O g i C S t u d i e S 287

development of the disease. The interior cells of the table would be filled at the conclusion of the period of follow-up. This approach represents a cohort study design. The third option would be to start with the column totals A + C (disease) and B + D (no disease) and determine exposures to complete the interior cell totals. This approach is called a case-control study. Note that for each of these study designs, information is known about each subject’s expo- sure and disease status. That is, it is possible to cross-classify each subject with respect to exposure and disease and thereby fill in each of the interior cells of the 2 by 2 table. In the fourth category of observational studies, the ecologic study, the interior cell counts are not known, as will be discussed in the fol- lowing section.

An ecologic study is one that examines a group as a unit of analysis. The ecologic approach differs from most study designs, which use an indi- vidual as the unit of analysis. Example: A study of mortality from lung disease in different cities that are known to have differing levels of air pollution would comprise an ecologic study. The unit of analysis is a city. Uses of ecologic studies: They can be used for generating hypotheses and also in analytic studies. Limitations: The ecologic fallacy Refer to the text for more details. n

What is an Ecologic Study?

Ecologic Studies

As mentioned earlier, for cross-sectional, case-control, and cohort studies, data on exposure and disease are known at the level of the individual. In ecologic studies, the unit of analysis is the group. Here is an example: In the southern California basin, a geographic area that spans more than 200 miles (330 km) from the U.S. California border with Mexico to the city of Santa Barbara, con- centrations of air pollutants vary greatly. The highest concentrations are in urban centers, such as central Los Angeles and sections of Long Beach near oil refiner- ies and ports; conversely, air pollution levels are lowest in the coastal areas that

51589_CH06_Printer.indd 287 18/02/13 8:18 PM

288 C h a p t e r 6 S t u d y d e S i g n S

are farthest from heavy industry. Suppose data were available on the average mortality and average particulate levels (one of the components of air pollution) during the year for each census tract in the basin. We could then assess the associ- ation between particulate pollution and mortality by plotting the mortality levels within each census tract. This hypothetical example illustrates one of the typical schemas for an ecologic study, which in this case uses a census tract, rather than individuals, as the unit of analysis. Table 6–3 provides additional examples of questions asked by ecologic studies.

Refer to Figure 6–3. In an ecologic study, the number of exposed persons [row total, A + B] (preferably termed the rate of exposure) and the number of cases [column total, A + C] (preferably termed the rate of disease) are known.

Table 6–3 Examples of Questions Investigated by Ecologic Studies

●● Is the ranking of cities by air pollution levels associated with the ranking of cities by mortality from cardiovascular disease, adjusting for differences in average age, percent of the population below poverty level, and occupational structure?

●● Have seat belt laws made a difference in motor vehicle fatality rates? This question could be addressed by comparing the motor vehicle fatality rates from years before and years after seat belt laws were passed.

●● Are daily variations in mortality in Boston related to daily variations in particle air pollution, adjusting for season of year and temperature?

●● What are the long-term time trends (1950–1995) for mortality from the major cancers in the United States, Canada, and Mexico?

Source: Adapted from ERIC Notebook, April 2000, Issue 12, pp. 1–2. Department of Veterans Affairs, Epidemiologic Research and Information Center at Durham, NC.

Figure 6–3 Illustration of sample selection for an ecologic study.

Yes

Yes No Total

A + B

C + D

A + C B + D

Exposure Status

Disease Status

51589_CH06_Printer.indd 288 18/02/13 8:18 PM

e C O l O g i C S t u d i e S 289

The number of nonexposed persons [row total, C + D] and noncases [column total, B + D] may be inferred. In summary, the marginal totals (row totals and column totals) are known. Neither the number of exposed cases [cell A] nor the numbers of persons in the other interior cells [B, C, and D] are known. As shown in Figure 6–3, the known information is surrounded by boxes.

This section covers two major types of ecologic studies: ecologic compari- son studies and ecologic trend studies. Ecologic comparison studies (sometimes called cross-sectional ecologic studies) involve an assessment of the correlation between exposure rates and disease rates among different groups or populations over the same time period; usually there are more than 10 groups or populations. (Note that the term cross-sectional is defined in the next section.) The data on a disease may include incidence rates, prevalence, or mortality rates for multiple defined populations. Data on rates of exposure also must be available on the same defined populations. Examples of exposure data include:

●● measures of economic development (e.g., per capita income and literacy rate) ●● environmental measures (e.g., mean ambient temperature, levels of humid-

ity, annual rainfall, and levels of mercury or microbial contamination in water supplies)

●● measures of lifestyle (e.g., smoking prevalence, mean per capita intake of cal- ories, annual sales of alcohol, and number of memberships in health clubs)

The important characteristic of ecologic studies is that the level of exposure for each individual in the unit being studied is unknown. Although one may have to do considerable work to amass the data needed for such studies, ecologic studies generally make use of secondary data that have been collected by the gov- ernment, some other agency, or other investigators. Thus, in terms of cost and duration, ecologic studies are clearly advantageous.

A second type of ecologic study, the ecologic trend (time series) study, involves correlation of changes in exposure with changes in disease over time within the same community, country, or other aggregate unit in order to ascertain trends. For example, within the United States there has been a consistent downward trend in the incidence of and mortality from coronary heart disease. The exact reasons for the decline are unknown. A cynic, however, might assert that some organizations that have worked hard to achieve such results may find it desir- able to claim responsibility (to ensure continued funding). Ecologic correlation data could be generated to support the claim that the downward trends reflect increased prescription of antihypertension medications or the number of coro- nary bypass surgeries performed.

51589_CH06_Printer.indd 289 18/02/13 8:18 PM

290 C h a p t e r 6 S t u d y d e S i g n S

A classic example of an ecologic correlation is the association between breast cancer and dietary fat.4 Rates of breast cancer mortality and estimates of per cap- ita dietary fat intake were collected for 39 countries. When presented graphically, the data led to a striking observation: Countries with high per capita intakes of dietary fat tended to be the same countries with high rates of breast cancer mor- tality (Figure 6–4).

A second example is a study of childhood lead poisoning in Massachusetts.5 More than 200,000 children from birth through 4 years of age were screened at physicians’ offices, hospitals, and state-funded screening sites, through nutri- tional supplementation programs, and by door-to-door screening in high-risk areas. Blood samples were drawn and analyzed for lead levels using standard pro- cedures. Communities were the unit of analysis, the value for each being the pro- portion of screened children with high blood levels of lead. The severity of blood lead poisoning was correlated with indices from U.S. census data. Community rates of lead poisoning were positively associated with a poverty index, the per- centage of houses built before lead-based paints were banned, and the percentage of the community that was African American. Median per capita income was inversely associated with lead poisoning. This example was an ecologic study

Figure 6–4 Ecologic correlation of breast cancer mortality and dietary fat intake. Source: Reprinted with permission from KK Carroll, Experimental Evidence of Dietary Factors and Hormone-Dependent Cancers. Cancer Research, Vol 35, p 3379, © 1975, American Association for Cancer Research.

CANADA

NETHERLANDS

DENMARK NEW ZEALAND

SWITZERLAND IRELAND US

BELGIUM AUSTRALIA

SWEDEN GERMANY

FRANCE AUSTRIA

NORWAY

FINLAND CZECH

ITALY

HUNGARYPORTUGAL

POLANDHONG KONG

SPAINBULGARIAROMANIA YUGOSLAVIA GREECE

PUERTO RICO

CHILE VENEZUELA

PANAMA COLOMBIA

PHILIPPINES MEXICO JAPAN TAIWAN

CEYLON EL SALVADORTHAILAND

Female

0 20 40 60 80 100 120 140 160 180

Total dietary fat intake (g/day)

A ge

-a dj

us te

d de

at h

ra te

(1 00

,0 00

p op

ul at

io n)

51589_CH06_Printer.indd 290 18/02/13 8:18 PM

e C O l O g i C S t u d i e S 291

because the outcomes were measured on groups (summary rates of lead poi- soning by community) and the exposures were measured on groups (based on census data). By identifying factors in communities associated with high rates of lead poisoning, this study provided data that may be relevant to the identifica- tion of high-risk communities where funding for screening may not be available.

Our third example of an ecologic study explores the unexpected inverse rela- tionship found in the Seven Countries Study between mean systolic blood pres- sure levels and stroke mortality rates.6 (An inverse relationship is the converse of a direct relationship: In a direct relationship, when a risk factor increases the out- come increases also; in an inverse relationship, the outcome decreases when the risk factor increases.) Thus, in the Seven Countries Study, investigators expected stroke mortality rates to increase as mean blood pressures levels increased, but the opposite relationship was found. Investigators obtained stroke mortality rates in a 25-year follow-up of 16 cohorts of men aged 40–59 in the United States, several European countries, and Japan. All age-eligible men in the targeted coun- tries were followed from 1958 to 1964. At the population (ecologic) level, the association between mean entry-level blood pressure levels and mortality due to strokes over 25 years was inverse and strong. When the analyses were repeated at the individual level within cohorts, the association was strongly positive among most of the cohorts examined. Although the findings for the group level seemed to contradict those for the individual level analysis, this study provided an excel- lent example of the ecologic approach in which countries are the unit of analysis.

A fourth example is an ecologic time series analysis that examined the relation- ship among mortality from homicides in São Paulo, Brazil, with three categories of factors: sociostructural indicators, investment in social policies, and public security. The dependent variable of the study was the homicide mortality rate per 100,000 inhabitants. The unit of analysis was the city of São Paulo, which was assessed annually between 1996 and 2008. The three categories of analytic factors included:

●● sociostructural indicators—the percentage of the population comprised of adolescents and the percentage of unemployed persons

●● investment in social policies—municipal investment in education and culture and state investment in health and sanitation

●● public security—municipal investment in public security and the number of firearms seized per 100,000 inhabitants

The researchers concluded that, in particular, a decline in unemployment, investment in social policies, and changes in public security policies were associ- ated synergistically with reductions in the homicide mortality rate.7

51589_CH06_Printer.indd 291 18/02/13 8:18 PM

292 C h a p t e r 6 S t u d y d e S i g n S

In summary, the ecologic approach has applications in a wide variety of situations. Some additional examples include the effect of fluoridation of the water supply on hip fractures,8 the association of naturally occurring fluoride levels and cancer incidence rates,9 and the relationship between neighborhood or local area social characteristics and health outcomes.10

The preceding examples reveal some of the applications and merits of ecologic studies; the following discussion presents some of their disadvantages. One of these is known as the ecologic fallacy, a term that will be defined subsequently. Let us begin with a hypothetical illustration. For the sake of argument, sup- pose that we found the mortality rates for emphysema to be lower in central Los Angeles, California (a highly industrialized area), than they were in the desert resort of Palm Springs, California (a less industrialized area). We might con- clude erroneously that areas with lower air pollution levels have higher emphy- sema mortality rates than do areas with higher air pollution levels. How could we explain this apparently contradictory finding? The answer lies in differentiating between lifelong residents and in-migrants (Figure 6–5).

After spending their working years in central Los Angeles, which is a major employment setting, some people migrate to Palm Springs during their retire- ment years, especially if they are afflicted with pulmonary difficulties. This retire- ment haven is also a magnet for in-migrants from northern industrialized cities throughout the United States. As Figure 6–5 shows, when we examine the com- posite emphysema mortality data (which do not disclose length of residence in Palm Springs), we may be misled into reaching an erroneous conclusion about the association between exposure and disease (refer to the bottom panel of the fig- ure). This incorrect observation results from the ecologic fallacy: Individuals’

Figure 6–5 Example of ecologic fallacy. Source: Adapted from ERIC Notebook, April 2000, Issue 12, pp. 1–2. Department of Veteran Affairs, Epidemiologic Research and Information Center at Durham, NC.

Lifelong Residents

In-Migrants

High Mortality

Low Mortality

High Mortality

Low Mortality

Lifelong Residents

In-Migrants

Reality

Observed Mortality

51589_CH06_Printer.indd 292 18/02/13 8:18 PM

e C O l O g i C S t u d i e S 293

levels of exposure to air pollution have not been specified. When degree of exposure is differentiated at the individual level by comparing in-migrants and lifelong residents of Palm Springs, the correct exposure and disease relationships become apparent (top panel of the figure). When those who have had long-term exposure to air pollution (in-migrants) are compared with lifelong residents with respect to emphysema mortality, the former have higher mortality.

We have seen that ecologic studies, which by definition use observations made at the group level, may not represent the exposure–disease relationship at the individual level. Hence, the term ecologic fallacy is defined as, “The bias that may occur because an association observed between variables on an aggregate level does not necessarily represent the association that exists at an individual level.”11

To give a second illustration of an ecologic fallacy, consider the association between wearing hats and protection against sunburn, an example that can be viewed from the perspective of the individual and the group. Suppose that among a sample of 10 individuals there are 7 (70%) who have sunburned foreheads and 6 (60%) who wear hats when they go outside. At this point, we do not specify at the individual level who wore hats and who was sunburned. We specify only the overall percent of who wore hats and the percent of those who were sunburned, just as we would in an ecologic study. The similar proportion of hats and sun- burns suggests that there is an association between the exposure (wearing hats) and disease (sunburn). The conclusion is illusory, however (Table 6–4).

To verify the error in this conclusion, note that among the six persons who wore hats outside, three were sunburned (50%). Among the remaining four per- sons who did not wear hats, however, all four (100%) had sunburned foreheads. Thus, the conclusion based on the association between hats and sunburns at the group level was incorrect.

Table 6–4 Hypothetical Ecologic Relationship Between Hats and Sunburn

Person Hat Wearer Sunburned Head

1 Yes Yes 2 Yes Yes 3 Yes Yes 4 No Yes 5 Yes No 6 No Yes 7 Yes No 8 No Yes 9 Yes No

10 No Yes

51589_CH06_Printer.indd 293 18/02/13 8:18 PM

294 C h a p t e r 6 S t u d y d e S i g n S

The reader may infer from the hat–sunburned head example that aggregate data on populations may not apply to individuals. To cite still another example, although rates of breast cancer tend to be higher in countries in which fat con- sumption is high than in those in which fat consumption is low, one cannot be certain that the breast cancer cases had high fat intakes. One could draw a similar correlation between the number of cars in a country and breast cancer rates, yet few would be willing to provide it as evidence of a cause–effect relationship. Thus, more rigorous study designs in which data on exposure and disease are collected on individuals are desirable as further support.

Other limitations of the ecologic study also must be acknowledged. Impreci- sion in the measurement of exposure and disease makes accurate quantification of the exposure–disease associations difficult. The ability to adjust for the influ- ence of extraneous variables is limited by the availability of such data and the analytic approaches for incorporating them.

In summary, despite the problems of the ecologic fallacy and other limitations, ecologic studies have earned a well-deserved place in epidemiologic research. They are quick, simple to conduct, and inexpensive. Thus, when little is known about the association between an exposure and disease, an ecologic study is a reasonable place to start if suitable data are available. If the investigator’s hypoth- esis is not supported, then few resources have been invested. In addition, when a disease is of unknown etiology, ecologic analyses represent a good approach for generating hypotheses.

Cross-Sectional Studies

The cross-sectional study (also termed prevalence study) is the first design to be covered in this chapter in which exposure and disease measures are obtained at the level of the individual. One starts by selecting a sample of subjects (N) and then determining the distribution of exposure and disease. However, cross- sectional studies are not required to include assessment of both exposure and disease. Some studies may be designed to provide only a measure of the burden of disease in a population, whereas others may focus exclusively on the distribu- tion of certain exposures.

The features of this type of study design include a single period of observation (Figure 6–6). Exposure and disease histories are collected simultaneously but may include assessment of history of disease or exposures, as shown in Figure 6–1. The unit of observation and analysis in cross-sectional studies is the individual. The majority of data are collected for the first time, primarily for the purpose

51589_CH06_Printer.indd 294 18/02/13 8:18 PM

C r O S S - S e C t i O n a l S t u d i e S 295

of the study, although they may be supplemented with secondary data, such as school records and medical records. Data from national surveys by the U.S. gov- ernment are frequently used.

As mentioned earlier in this chapter, cross-sectional studies are typically descriptive in nature. Primarily, they yield quantitative estimates of the mag- nitude of a problem but do not measure the temporal ordering of cause and effect.11 One could take two basic approaches to provide a measure of the mag- nitude of a public health problem: Collect data on each member of the popula- tion (i.e., a census, used frequently in prevalence studies) or take a sample of the population and from the sample make inferences about the remainder of the population. The more common method, however, is the second because in a shorter period of time, and much less expensively, one could derive reasonable estimates of the extent of a health problem through a survey on a subset of the population.

Sample Designs Sampling schemes for cross-sectional studies comprise two main types: prob- ability samples and nonprobability samples. They are defined as follows:12(p 15)

Figure 6–6 Illustration of subject selection in a cross-sectional study.

Yes

Yes No Total

A + B

C + D

A + C B + D

Exposure Status

Disease Status

A probability sample has the characteristic that every element in the popula- tion has a nonzero probability of being included in the sample. A nonproba- bility sample is one based on a sampling plan that does not have that feature.

51589_CH06_Printer.indd 295 18/02/13 8:18 PM

296 C h a p t e r 6 S t u d y d e S i g n S

Examples of probability samples include simple random samples, systematic samples, and stratified samples. Simple random samples refer to a type of sample in which each individual in the population (or other group) has an equal probability of being selected. Simple random samples require enumeration of all potential sub- jects before sampling, an expensive process that may not be feasible to implement. A systematic sample is, “The procedure of selecting according to some simple, systematic rule, such as all persons whose names begin with specified alphabetic letters, born on certain dates, or located at specified points on a master list.”11 Con- struction of a systematic sample does not necessarily require prior knowledge of the total number of sampling units. A sampling unit refers to “that element or set of elements considered for selection in some stage of sampling.”13(p 198) In epidemio- logic research, a sampling unit is usually a specific person selected for the study. It is possible to perform systematic sampling at the same time as the sampling frame is being constructed, a feature that makes systematic sampling the most widely used of all sampling procedures.12 (“A sampling frame is the actual list of sampling units from which the sample or some stage of the sample is selected.”13(p 198))

Suppose one wants to derive estimates of the magnitude of a health problem in a relatively small subset of the population. Simple random samples and sys- tematic samples will not ensure that sufficient numbers of this subgroup will be represented for meaningful estimates to be derived. A stratified sampling approach requires that the population be divided into mutually exclusive and exhaustive strata; sampling is then performed within each stratum (strata are “distinct subgroups according to some important characteristic, such as age or socioeconomic status . . .”11).

Nonprobability samples include quota samples and judgmental samples. An example of a quota sample design is one that requires interviewers to obtain infor- mation from a fixed number of subjects with particular characteristics regardless of their distribution in the population. A judgmental sample selects subjects on the basis of the investigator’s perception that the sampled persons will be repre- sentative of the population as a whole. Nonrandom samples are not appropriate for cross-sectional studies because the reliability of the estimates derived from such samples cannot be evaluated.12 Details on how to determine sample sizes and parameter estimates are beyond the scope of this book. The reader is referred to any of several excellent textbooks on the subject.12,14

Examples of Cross-Sectional Studies Some topics of cross-sectional studies include the extent of smokeless tobacco use, occurrence of neurodevelopmental disorders, health needs of minority pop- ulations, and trends in risk factors for disease. Murray et al.15 surveyed smokeless

51589_CH06_Printer.indd 296 18/02/13 8:18 PM

C r O S S - S e C t i O n a l S t u d i e S 297

tobacco use among ninth graders in four school districts representative of the Minneapolis–St. Paul metropolitan area of Minnesota. A questionnaire was administered in the classroom during the fall of 1985 to estimate the preva- lence of usage and to identify correlates of usage that might predict those teen- agers who were at greater risk than their peers of using smokeless tobacco. The study revealed that nearly 63% of boys and 24% of girls had ever used smoke- less tobacco, but only 18.5% of boys and 2.4% of girls had used it in the past week. Ethnicity also was associated with prevalence of usage, especially among boys. Self-reported prevalence ranged from a low among Asians and African Americans (21% and 22%, respectively), medium (45.5%) among whites, and high (60.8%) among Native Americans. The prevalence among boys from a two- parent household was 42.1%, 41.3% if only the mother lived at home, and 54.5% if the boy came from a father-only household. The smokeless tobacco use was much more common among cigarette smokers than among nonsmok- ers. Only 6% of nonsmoking girls reported using smokeless tobacco, in contrast to 16.4% of girls who smoked. A similar magnitude of difference was observed among the boys: 34.5% versus 56.1% for nonsmokers and smokers, respectively. Additional analyses revealed a clustering of unhealthy behaviors among teenag- ers who consumed three types of substances (alcohol, cigarettes, and marijuana). The prevalence of smokeless tobacco use among them was more than seven times higher than among non consumers—50.8% and 6.9%, respectively.

The preceding example was based on a survey of schoolchildren from a single major metropolitan area. Prevalence studies may also be performed with a much broader sampling frame. For example, although surgical sterilization has many advantages over other forms of contraception, data on prevalence of sterilization has been based primarily on surveys of women. Thus, to estimate the preva- lence of vasectomy and identify factors associated with sterilization, a nationally representative survey of male U.S. residents aged 15–44 years was conducted in 2002.16 The survey revealed that 13.3% of married men reported having had a vasectomy and 13.8% reported tubal sterilization in their partners. Vasectomy increased with older age, greater number of biological children, non-Hispanic white ethnicity, and ever having gone to a family planning clinic. Use of tubal sterilization was more likely among women with a history of live births; many of the women’s male partners had not attended college and were older. Thus, one in eight married men reported having vasectomies; men who relied on vasec- tomies had a somewhat different profile than those whose partners had tubal sterilizations.

Autism spectrum disorder (ASD) is a set of complex neurodevelop- ment disorders characterized by mild to severe problems in social interaction

51589_CH06_Printer.indd 297 18/02/13 8:18 PM

298 C h a p t e r 6 S t u d y d e S i g n S

and communication along with restricted repetitive behavior patterns. ASD symptoms begin before age 3 years and last into adulthood, although symptoms may improve over time. Although there is no one best treatment for ASD, the American Academy of Pediatrics recommends early behavioral intervention once a child is diagnosed.17 Nearly all children (94%) with ASD require health or related services beyond those generally needed by children.

The Survey of Pathways to Diagnosis and Services18 is a nationally represen- tative survey about children with special healthcare needs (NS-CSHCN) aged 6–17 years ever diagnosed with autism spectrum disorder (ASD), intellectual dis- ability, or developmental delay. Parents or guardians who previously participated in the 2009–2010 National Survey-CSHCN (sponsored by the Maternal and Child Health Bureau) and who reported that their child had ever been diagnosed with at least one of the three selected developmental conditions were recontacted via landline or cell phone to participate in the Pathways survey. To be eligible, the CSHCN had to be aged 6–17 years at the time of the Pathways interview and still living in the same household as the recontacted parent or guardian. Of the parents or guardians with eligible CSHCN, 71% were successfully recontacted and 87% agreed to participate. A total of 4,032 interviews were completed from February to May 2011, an average of 9 months after the initial interview.

The survey revealed that the median age when school-aged children with spe- cial healthcare needs (CSHCN) and autism spectrum disorder (ASD) were first identified was 5 years. (Refer to Figure 6–7.) Generalists and psychologists were more likely to identify children under the age of 5, while psychologists and psy- chiatrists more commonly identified children aged 5 years and over. Nine out of 10 school-aged CSHCN with ASD use one or more services to meet their developmental needs, with social skills training and speech or language therapy each used by almost three-fifths of these children. Finally, more than one-half of school-aged CSHCN with ASD use psychotropic medication.

Prevalence surveys are helpful for identifying resource needs for health interventions. For example, Friis et al. conducted a cross-sectional prevalence survey of cigarette smoking among Cambodian Americans in Long Beach, Cali- fornia.19 A stratified random sample of 1,414 adult respondents was selected from 15 census tracts with high concentrations of Cambodian Americans. The sex-specific prevalence of smoking was 24.4% among men and 5.4% among women. Significant covariates of current smoking were gender, age, education, marital status, and health status. The prevalence of smoking among Cambodian American men was higher than among other males in California. The research

51589_CH06_Printer.indd 298 18/02/13 8:18 PM

C r O S S - S e C t i O n a l S t u d i e S 299

suggested that culturally tailored interventions are needed to address the high prevalence of smoking among Cambodian Americans. Various aspects of the Cambodian culture such as the high level of respect given to religion, elders, and peers seem to be implicated in the high prevalence of smoking.

Another example of using cross-sectional prevalence data is for interventions for sensory impairments in vision, hearing, postural balance, or loss of feeling in the feet, all of which are known to increase with age. As the U.S. life expectancy increases, the prevalence of sensory impairments can be expected to increase and have negative effects on public health goals for older adults to maintain inde- pendent living, health, and quality of life.20 Minimizing the impact of sensory impairments is therefore important.

Prevalence estimates of sensory impairments21 were generated by using the most recent U.S. National Health and Nutrition Examination Survey (NHANES) data on vision (1999–2006), hearing (2005–2006), balance (2001– 2004), and peripheral neuropathy (1999–2004). NHANES is a cross-sectional survey that includes health interviews, health examinations, and laboratory tests

Figure 6–7 Percent distribution of child’s age when parent or guardian was first told that child had autism spectrum disorder among children aged 6–17 years with special health care needs and autism spectrum disorder: United States, 2011. Source: Reproduced from Pringle BA, Colpe LJ, Blumberg SJ, et al., Diagnostic history and treatment of school-aged children with autism spectrum disorder and special health care needs. NCHS Data Brief, No 97, May 2012.

Age 6 years and over 39.9%

Age 2 years and under

18.7%

Age 3 years 17.0%

Age 4 years 13.0%

Age 5 years 11.5%

51589_CH06_Printer.indd 299 18/02/13 8:18 PM

300 C h a p t e r 6 S t u d y d e S i g n S

on a complex sample designed to be nationally representative. In 1999–2006, non-Hispanic blacks, Mexican Americans, persons with low-income, and indi- viduals aged 60 years and over were oversampled for reliable subgroup preva- lence estimation. Analyses used sample weights that accounted for nonresponse and noncoverage as well as the complex survey design.

The data demonstrated that sensory impairments are a substantial problem for older Americans: One out of six has impaired vision; one out of four has impaired hearing; one out of four has loss of feeling in the feet; and three out of four have abnormal postural balance testing. Sensory impairments increase with age: Both vision and hearing impairments double and loss of feeling in the feet increases by 40% in persons aged 80 years and over compared with persons aged 70–79 years. One in five Americans below the poverty threshold has impaired vision—50% higher than other Americans. Balance problems are also common among the poor. Over one-half of those with impaired vision could improve their eyesight by using glasses or by getting a corrected prescription. For those with hearing problems, 72% might benefit from a hearing aid but do not use one.

Repeated cross-sectional surveys can be used to examine trends in disease or risk factors that can vary over time. The Centers for Disease Control and Preven- tion used data on persons 12 through 19 years of age from the National House- hold Surveys on Drug Abuse, High School Seniors Surveys, and National Health Interview Surveys to examine the prevalence of cigarette smoking among U.S. adolescents.22 Data were available from 1974 (1976 for the High School Seniors Surveys), 1980, 1985, and 1991. The results (Figure 6–8, part A) suggest that overall smoking levels declined at all survey periods, but that there was notable variation by race, sex, and time period. In general, the decline of smoking was most rapid between 1974 and 1980; the decline was faster for females than males and greater for African Americans than whites. Nelson et al.22 concluded that the slowing of the trend toward lower smoking prevalence was evidence of the success of increased tobacco advertising and promotional activities aimed at ado- lescents or inadequate antitobacco educational efforts.

With respect to the years 1991–2010, Figure 6–8, part B shows the annual percentage of smokers in the 8th, 10th, and 12th grades. After peaking in the mid-1990s, the percentage of smokers has shown a declining trend. In 2009 and 2010, non-Hispanic white students had a higher percentage of smokers than non-Hispanic blacks or Hispanic.23 (Figure 6–8, part C).

These examples illustrate a number of important applications of the cross- sectional study design. Perhaps the greatest utility of such studies is for collect- ing data to describe the magnitude and distribution of a health problem, data

51589_CH06_Printer.indd 300 18/02/13 8:18 PM

C r o s s - s e C t i o n a l s t u d i e s 301

1991

P er

ce nt

o f S

tu de

nt s

1993 1995 1997 1999 20012001 2003 2005 2007 2009 2010

19.2

13.6

7.1

10th Grade

8th Grade

12th Grade

essential for planning health services and administering medical care facilities. The survey of smokeless tobacco use among ninth graders provided evidence that usage was quite common even at this early age, representing a significant public health problem. The foregoing example also demonstrated that cross-sectional studies permit assessment of a population’s various characteristics; such assess- ment may help target appropriate interventions and educational materials. An example of such a characteristic would be the prevalence of single-parent house- holds. The survey of vasectomies in the United States revealed that 1 in 8 mar- ried men have had the procedure. Sometimes, repeated cross-sectional surveys

Figure 6–8 (continues)

Part B

Content removed due to copyright restrictions

51589_CH06_Printer.indd 301 20/02/13 12:52 AM

302 C h a p t e r 6 s t u d y d e s i g n s

examine quantitative factors that vary over time. Finally, prevalence studies may generate new etiologic hypotheses that can be tested in future studies.

The principal weaknesses of cross-sectional designs arise from their limited usefulness for inferring disease etiology. First, prevalent cases represent survivors: The study of prevalent cases makes it difficult to sort out factors associated with risk of disease from factors associated with survival, such as treatment and sever- ity. For diseases in which onset is difficult to determine (e.g., mental disorders), however, prevalence is an acceptable substitute for incidence. In other situations, the study of existing cases is the only feasible and affordable strategy to test etio- logic hypotheses. The second major limitation applies to the ability to study diseases of low frequency. Recall that the prevalence of a disease in a population is proportional to the incidence of the disease times its duration. Therefore, even a large survey may yield few cases of rare diseases or diseases that have short dura- tions. Third, because exposure and disease histories are taken at the same time

10th Grade

8th Grade

12th Grade

Non-Hispanic White Non-Hispanic Black Hispanic

7.6

14.7

22.9

4.3 6.7

10.1

6.7

12.2

15.0

P er

ce nt

o f S

tu de

nt s

Figure 6–8

(part B) Cigarette use among students in the past 30 days, by grade level, 1991–2010; (part C) Cigarette use among students in the past 30 days, by grade level and race/ethnicity, 2009–2011.* Sources:

(Parts B and C) U.S. Department of Health and Human Services, Health Resources and Services Administration, Maternal and Child Health Bureau. Child Health USA 2010. Rockville, Maryland: US Department of Health and Human Services, 2011.

Part C

* To derive percentages for each racial subgroup, data for 2009 and 2010 have been combined toincrease subgroup sample sizes and thus provide more stable estimates.

51589_CH06_Printer.indd 302 20/02/13 12:52 AM

C a S e - C O n t r O l S t u d i e S 303

in a cross-sectional study, one must be careful about the temporality issue (i.e., whether exposure or disease came first), making assertions about any apparent cause-effect relationship tenuous.

Case-Control Studies

A case-control study is a type of analytic study “of persons with the disease (or another outcome variable) of interest and a suitable control group of persons with- out the disease (comparison group, reference group).”11 As indicated in Figure 6–1, study subjects in the case and control groups are sampled with regard to the disease or effect under investigation. Consider two groups, one in which every- one has the disease of interest (cases) and a comparable one in which everyone is free from the disease (controls). The case-control study seeks to identify possible causes of the disease by finding out how the two groups differ with respect to an exposure (or suspected risk factor). That is, because disease does not occur randomly, the case group must have been exposed to some risk factor, either vol- untarily (e.g., through diet, exercise, or smoking) or involuntarily (e.g., through such factors as cosmic radiation, air pollution, occupational hazards, or genetic constitution). Therefore, a comparison of the frequency of exposure among cases and controls may permit inferences regarding their difference in disease status. Case-control studies are a mainstay of epidemiologic research.

From the standpoint of selection of study groups for a case-control design, one is going from effect to cause; usually case-control studies are retrospec- tive studies, because one collects causal (exposure) information retrospectively. (See Figure 6–1.) This term means that the researcher delves into the study of subjects’ past exposures after the disease has already occurred. In Figure 6–9, the column totals, denoted by boxes, represent the presence (i.e., cases) or absence (i.e., controls) of disease. In recent years, the case-control design has proven to be useful and efficient for evaluation of vaccine effectiveness,24 treatment efficacy,25 screening programs,26 and outbreak investigations.27

The number of observation points for a case-control study is only one. Cases and controls are selected, and data are collected about past exposures that may have contributed to disease. As is true also of cross-sectional studies, the unit of observation and the unit of analysis are the individual. The method of data col- lection typically involves a combination of both primary and secondary sources. Usually the data on exposure are collected by the investigators, although one can easily imagine situations where valuable information might be obtained from medical, school, and employee records. Data on disease are often collected by

51589_CH06_Printer.indd 303 18/02/13 8:18 PM

304 C h a p t e r 6 S t u d y d e S i g n S

someone other than the investigator, especially if one is making use of special registries or surveillance systems for case identification. In some situations, how- ever, the investigator might conduct a population screening survey to identify suitable cases. Notice that Figure 6–9 does not include the marginal totals A + B, C + D, or a total N. Because one is taking only a subset of the total popula- tion, namely the cases and some number of controls, these marginal totals are meaningless.

Selection of Cases Two tasks are involved in case selection: defining a case conceptually and iden- tifying a case operationally.28 The definition of a case is influenced by a number of factors, including whether there are standard diagnostic criteria, the severity of the disease, and whether the criteria to diagnose the disease are subjective or objective. At issue is misclassification. If the criteria are broad, the case group is more likely to include individuals who truly do not have the disease. Conversely, overly restrictive criteria may limit the number of subjects available for study. Although the selection criteria need to be weighed for each individual study, there is some evidence that the benefits of a more restrictive definition of a case outweigh the benefits of being overly inclusive.29

Sources of Cases Once a case has been defined conceptually, one can then proceed to develop a strategy for case identification. The researcher’s goal “is to ensure that all true cases have an equal probability of entering the study and that no false cases

Figure 6–9 Illustration of selection in a case-control study.

Yes

Yes No

A + C B + D

Exposure Status

Disease Status

51589_CH06_Printer.indd 304 18/02/13 8:18 PM

C a S e - C O n t r O l S t u d i e S 305

enter.”28(p 8) The ideal situation is to identify and enroll all incident cases in a defined population in a specified time period. For example, a tumor or disease registry or Vital Statistics Bureau may provide a complete listing of all available cases. The advantage of using incident cases is that, when all cases in a popula- tion are identified, there can be little question of their representativeness. In the real world, however, logistics or the lack of a suitable disease registry may restrict case selection to one or a few medical facilities. The main caveat in the selection of cases is the nonrepresentativeness of the cases derived from special care facilities such as hospitals and tertiary facilities. These institutions may receive only the most severe cases, ones in which the distribution of risk factors is atypical.

Per the earlier discussion regarding cross-sectional designs, the use of preva- lent cases creates challenges in separating causal exposures from consequential exposures; these are changes in exposures (e.g., deciding to smoke less) as a result of having diseases. Additional benefits of studying incident (as opposed to preva- lent) cases include subjects’ better recall of past exposures and a reduced likeli- hood that exposure has changed as a consequence of the disease.

Selection of Controls Suppose you have a strong and justified hypothesis about a new risk factor for a disease and want to conduct a case-control study. How should the controls be selected? To determine whether this risk factor is truly associated with the disease—not indirectly or incorrectly associated because of some third (con- founding) factor—the ideal controls should have the same characteristics as the cases (except for the exposure of interest). That is, if the controls were equal to the cases in all respects other than disease and the hypothesized risk factor, one would be in a stronger position to ascribe differences in disease status to the exposure of interest. Taking this example one step further, imagine a study of childhood vaccination and an adult chronic disease in which the cases and con- trols were all the same age, the same sex, and the same race, worked at the same job, ate the same foods, were educated at the same schools, and had the same leisure activities. Speculate that the cases and controls were identical in every respect except for vaccinations in childhood; all of the cases were exposed and none of the controls was exposed. This situation would provide the most clear- cut evidence that an exposure was indeed a risk factor for the disease. Clearly, such an ideal selection of cases and controls is extremely unlikely to occur, but the point should be obvious: Cases are presumed to have a given disease because of an excess (or deficiency) of an exposure. To identify whether the exposure

51589_CH06_Printer.indd 305 18/02/13 8:18 PM

306 C h a p t e r 6 S t u d y d e S i g n S

patterns of a group of cases are excessively high or excessively low, the investigator needs to know what exposure pattern should normally be expected. Selection of a group of disease-free individuals (controls) supposedly will reveal what a nor- mal or expected level of exposure should be in the absence of disease.

Note, however, that controls are neither always nor necessarily free from dis- ease. To examine the specificity of an exposure–disease association, one might identify controls with a different disease, assuming that the illness has a different etiologic basis from the disease of interest. For example, a review of 106 cancer case-control studies identified nine that used controls who had a type of cancer that was different from that of the cases.30

The epidemiologist needs to be able to isolate the effect of the risk factor on the outcome from influences that emanate from the controls. For exam- ple, suppose that the cases and controls differ in demographic characteristics, such as age or socioeconomic status. These demographic factors could operate as rival explanations to account for observed outcomes. That is, the researcher would be unable to distinguish the effects of the risk factor from the influence of demographic factors. To overcome the potential impact of systematic case- control differences in demographic and other characteristics upon the outcome, the investigator can match cases and controls on important characteristics. Two methods of matching are the use of matched pairs (individual matching) and frequency matching (group matching). An example of individual matching would be to match each case with one or more controls who are the same age and gender. Frequency matching means that approximately equal distributions of demographic variables such as age and gender are maintained among the cases and controls.

In addition to the problem of how to select the controls, a second issue is the number of controls to select. Estimating the number of controls is an aspect of statistical power (ability to identify a significant difference), a topic that is beyond the scope of this text. (For more information, refer to Cher- nick and Friis.31) However, here is a rule of thumb: Epidemiologic researchers sometimes use an equal number of cases and controls (one-to-one ratio), a perfectly acceptable procedure. Preferably, researchers may select more con- trols than cases (up to a three-to-one or four-to-one ratio). When the ratio of controls to cases is at about a four-to-one ratio, statistical power of the design is maximal. A single case-control study also may contain more than one con- trol group, as in the use of hospital controls for hospitalized cases as well as community controls who represent a less selected group than do the controls from a hospital.

51589_CH06_Printer.indd 306 18/02/13 8:18 PM

C a S e - C O n t r O l S t u d i e S 307

Sources of Controls The general concept guiding the selection of controls is that they should come from the same population at risk for the disease or condition as the cases being studied. That is, one should ask, if a “control” had developed the outcome under investiga- tion, would he or she have been ascertained as a case? Moreover, suitable controls should have the potential to become a case. For example, if you are conducting a study of ovarian cancer, women who have had bilateral oophorectomy (removal of their ovaries) would not be eligible as controls, since they could not develop ovarian cancer. Table 6–5 provides a guide for selection of controls who are comparable to the cases. Several options are available to the investigator regarding sources of con- trols. Each option has inherent strengths and weaknesses, advantages and disadvan- tages. Three common sources of controls are population-based controls, patients from the same hospital as the controls, and relatives of associates of cases.

Population-based controls Perhaps the best way to ensure that the distribution of exposure among the controls is representative of the exposure levels in the population is to select population-based controls. A method to identify such controls is to obtain a list that contains names and addresses of most residents in the same geographic area as the cases. For example, a driver’s license list would include most people between the ages of 16 and 65; a roster from the Centers for Medicare and Med- icaid Services (the former Health Care Financing Administration) would be a good source for subjects over the age of 65. Tax lists, voting lists, and telephone directories may be useful, provided that their coverage of the population is com- plete or nearly complete. One then could randomly select controls from the total

Table 6–5 Guide for Selection of Comparable Cases and Controls

Example Cases Controls

5 6

All cases diagnosed in the community

All cases diagnosed in a sampled population

All cases diagnosed in all hospitals in the community

All cases from one or more hospitals

All cases from a single hospital Any of the above

Sample of the general population in a community

Noncases, in a sample of the general population, or a specified subgroup

Sample of persons who reside in the same neighborhoods as cases

Sample of patients in one or more hospitals in the community who do not have the same or related diseases being studied

Sample of noncases from the same hospital Spouses, relatives, or associates of cases

51589_CH06_Printer.indd 307 18/02/13 8:18 PM

308 C h a p t e r 6 S t u d y d e S i g n S

list, making sure to verify that selected persons met all exclusion criteria. Another method would be to select controls matched to the cases on variables such as age or sex. An approach that historically worked well in countries where most households have a telephone is random-digit dialing (RDD).32 This technique uses a computer to generate randomly the last two to four digits of a telephone number for potential controls who have the same three-digit telephone prefix as the cases. The procedure is repeated until a suitable control is found. There is some evidence, however, that controls selected by RDD are better educated and more likely to be employed than survey controls.33 In addition, increasing use of caller identifications on home telephones, inundation by telemarketers, and replacement of land telephones with cellular telephones have decreased the yield of this approach. As a response to the excessive number of calls from tele- marketers, the Federal Trade Commission implemented the National Do Not Call (DNC) Registry on October 1, 2000. The DNC Registry does not apply to survey researchers and, consequently, does not limit their activities.

In comparison with other methods for selection of controls, the use of population-based controls is most likely to result in a control group that is rep- resentative of the exposure rate in the general or target population. Controls selected from the general population may have little incentive to participate in a research study, producing low participation rates and the need to contact more individuals to find an eligible control willing to participate. Consequently, the study becomes more expensive. Another consequence of low participation is that individuals who do ultimately agree to participate may be systematically differ- ent in the frequency of exposure than the target population they are intended to represent.

Patients from the same hospital as the cases For the reasons noted, a preferred approach is to conduct case-control studies in which both study groups (i.e., cases and controls) are population based. When selection of population-based study groups is not feasible, however, cases may need to be derived from one or more major hospitals. Although hospital-based studies are inherently subject to greater potential for errors than population- based studies, their use is certainly justified when little information has been reported about a particular exposure–disease association or when a population- based case registry is not available. After the decision has been made to select cases from hospitals, it is perfectly appropriate to select hospital controls.

There are several practical advantages to using hospital controls. The study personnel who are already in the hospital to interview cases may achieve time

51589_CH06_Printer.indd 308 18/02/13 8:18 PM

C a S e - C O n t r O l S t u d i e S 309

efficiency by also interviewing controls. This time saving, plus the fact that hospital controls may be more likely than population controls to participate, ultimately equates to cost savings.

Perhaps the main difficulty of using hospital controls is trying to decide the diagnostic categories from which to select the controls. A second major lim- itation of hospital controls is that they may not be representative of the true exposure rates in the target population; after all, they were ill enough to require medical attention and motivated enough to seek it. Taken to the extreme, a pos- sible category of controls is deceased persons. If cases were defined as individuals who had died from a particular cause, dead controls would permit testing of hypotheses regarding exposures and each cause of death.28

Relatives or associates of cases Earlier in this chapter, we stated that the ideal control is a person free of the outcome of interest and similar in every respect to the case except for the expo- sure of interest. This objective is not only difficult to achieve but also difficult to evaluate. One approach to control indirectly for factors that will not be mea- sured directly as part of the study is to select relatives, associates, or neighbors of cases as the control group. This strategy tends to be a good method to con- trol for possible differences in socioeconomic status, education, or other char- acteristics assumed to be determinants of friendship or neighborhood.28 At the same time, use of this category of controls is not without some disadvantages. Compared with the use of hospital controls, the method is more expensive and time consuming. Although one might intuitively expect “friend” controls to be highly cooperative, several investigators have noted that cases may be unwilling to provide the name of a friend to fill this role.34,35 A greater problem is that one may end up controlling for an important (unidentified) risk factor that could no longer be evaluated.

Measure of Association Used in Case-Control Studies The objective of case-control studies is to identify differences between groups of cases and controls in frequencies of exposures; these differences might be associ- ated with one group having and the other group not having a disease (or other condition) of interest. Although several measures of association (known as mea- sures of effect) between exposure and disease can be calculated, we introduce the most frequently calculated measure. The guiding principle is to determine how much more (or less) likely the cases are to be exposed than the controls.

51589_CH06_Printer.indd 309 18/02/13 8:18 PM

310 C h a p t e r 6 S t u d y d e S i g n S

In Figure 6–10, examine the cells that refer only to the cases. First, we calculate the odds of exposure among the cases. Odds signify the “ratio of the probability of occurrence of an event to that of nonoccurrence.”11 For example, if 30 persons with lung cancer (cell A) smoke and 15 persons with lung cancer do not smoke (cell C), the odds of lung cancer cases having smoked are A/C or 2 to 1.

Returning to Figure 6–10, you observe that the proportion of the cases exposed is A/(A + C). The proportion of cases not exposed is C/(A + C). The odds of exposure, given that an individual is a member of the case group, are simply the ratio of these two proportions: [A/(A + C)] ÷ [C/(A + C)]. To sim- plify this expression, one inverts the second term and multiplies it by the first: [A/(A + C)][(A + C)/C]. The terms (A + C) cancel out, and one is left with A/C; this term represents the odds of exposure among the case group.

Second, repeating the same calculations, one determines that the odds of exposure among the control group are B/D. To evaluate whether the odds of exposure for the case group are different from the odds of exposure for the con- trol group, we create a ratio of these two odds, or an odds ratio (OR): (A/C) ÷ (B/D). Note that this can be more conveniently expressed as (AD)/(BC), which is the cross-product of the cells from our 2 by 2 table. A calculation example based on data from Lopez-Carrillo36 is shown in Exhibit 6–1.

The OR literally measures the odds of exposure to a given disease. An OR of 1.0 (called the null value) implies that the odds of exposure are equal among

Sample calculation of an Odds ratio

Those of us who have a predilection for spicy foods have won- dered about the health hazards associated with consumption of chili peppers. López-Carrillo et al.36 conducted a population-based case-control study in Mexico City of the relationship between chili pepper consumption and gastric cancer risk. They reported that consumption of chili peppers was significantly associated with high

risk for gastric cancer (age- and sex-adjusted OR = 5.49). Subjects for the study consisted of 220 incident cases and 752 controls randomly selected from the general population. Interviews produced information regarding chili consumption. In the present example, the data from this study are abstracted to illustrate how to calculate the OR.

e x

h ib

it 6

–1

continues

51589_CH06_Printer.indd 310 18/02/13 8:18 PM

C a S e - C O n t r O l S t u d i e S 311

Figure 6–10 Distribution of exposures in a case-control study.

Yes

Yes No

A + C B + D

Exposure Status

Disease Status

Chili Pepper Consumption Cases of Gastric Cancer Controls

Yes

A = 204

C = 9

B = 552

D = 145

The OR (unadjusted for age and sex) is:

AD BC

(204)(145) (552)(9)

5.95= = n

Source: Data from López-Carrillo L, Avila MH, Dubrow R. Chili Pepper Consumption and Gastric Cancer in Mexico: A Case-Control Study, American Journal of Epidemiology, Vol 139, pp. 263–271, Johns Hopkins University, School of Hygiene and Public Health, © 1994.

exhibit 6–1 continued

the cases and controls and suggests that a particular exposure is not a risk factor for the disease in this study. An OR of 2.0 indicates that the cases were twice as likely as the controls to be exposed. The implication of this OR, given proper consideration to the issues of causality, is that this particular exposure is a risk factor for the disease. More specifically, an OR of 2.0 implies that this particular exposure is associated with twice the risk of disease. Not all risk factors increase risk; a factor that is associated with lower risk of disease (i.e., a protective factor) would manifest as an OR of less than 1.0. Note that it is customary for epide- miologists to express an OR as a point value plus a confidence interval (CI).

51589_CH06_Printer.indd 311 18/02/13 8:18 PM

312 C h a p t e r 6 S t u d y d e S i g n S

A confidence interval is a range of values that contains the OR with a certain degree of probability (e.g. the 95% level).

The reader is advised to interpret the OR with caution. The case-control study is retrospective in nature with only one period of observation. Therefore, rates (and consequently risk, which can be determined only from prospective studies) cannot be directly determined. The reason for this inability to determine rates and risks centers upon the way the study groups are assembled. Referring back to Figure 6–9, you can see that groups (A + B) and (C + D) do not represent the total populations exposed and not exposed to the factor. That is, there are no appro- priate denominators for the population at risk, and therefore no way to directly determine disease rates. Under certain conditions, however, as noted below, the OR provides a good approximation of the risk associated with a given exposure.

●● The controls are representative of the target population. The key issue is that the controls are representative of the target population in the fre- quency of the exposure of interest.

●● The cases are representative of all cases. Cases should be typical with respect to severity and diagnostic criteria.

●● The frequency of the disease in the population is small. Algebraically, the formula for relative risk approximates that of the OR when the number of cases is small relative to the population at risk. There is some debate over whether this assumption is necessary, however.

Table 6–6, which summarizes examples of case-control studies presented in this chapter, highlights the many uses of the case-control approach in such

Table 6–6 Examples of Research Conducted with Case-Control Studies

Cancer research Young women’s cancers resulting from in utero exposure to diethylstilbestrol Smoking and invasive cervical cancer Chili pepper consumption and gastric cancer Green tea consumption and lung cancer Parental smoking and childhood cancer Efficacy of colonoscopic screening Cigarette tar yield and risk of upper digestive tract cancers

Birth defects research Maternal anesthesia and fetal development of birth defects

Heart disease research Passive smoking at home and risk of acute myocardial infarction

Infectious disease research Household antibiotic use and antibiotic resistant pneumococcal infection

51589_CH06_Printer.indd 312 18/02/13 8:18 PM

C a S e - C O n t r O l S t u d i e S 313

diverse areas as research on cancer, birth defects, heart disease, and infectious disease. (This list is by no means exhaustive.) Note that several of the research studies cited in Table 6–6 deal with cancer etiology. Many forms of cancer have unknown causation and low prevalence in the general population; for this rea- son, case-control studies are used frequently to investigate cancer etiology. For example, the association between green tea consumption and lung cancer was the focus of a case-control study conducted in Shanghai, People’s Republic of China.37 A total of 649 incident cases of primary lung cancer diagnosed among women between early 1992 and early 1994 were selected from the Shanghai Cancer Registry. Investigators matched the cases to 675 control women selected at random and reported a significantly reduced risk of lung cancer (OR = 0.94) among women who consumed green tea and also did not smoke. The association was not significant among women who smoked.

To give another example from cancer etiology, we note that the case-control method is well suited to explore in greater detail unusual clinical observations based on a small number of cases. A classic historical example is exposure of female fetuses in utero to diethylstilbestrol (DES) and the development of vagi- nal adenocarcinoma as young women.38 A sudden increase in the number of vaginal cancers of a rare histologic type at an atypical age led to a small case- control study that successfully identified maternal exposure to DES as the cause.

The preceding example may give the erroneous impression that case-control studies examine only a single exposure or a series of related exposures. In fact, especially when one is exploring a disease for which relatively little is known about the etiology, the exposure data being collected can cover a broad range of known and suspected factors. Consider the case-control study conducted by Brinton et al.39 on cervical cancer. The cases included 480 patients with invasive cervical cancer diagnosed at 24 hospitals in 5 U.S. cities: Birmingham, Chicago, Denver, Miami, and Philadelphia. All patients were between the ages of 20 and 74 years. A total of 797 population controls were identified through RDD. Two controls were matched to each case by using the variables of telephone exchange, race, and five-year age group. Women who had a previous hysterectomy were excluded. Data on cases and controls were collected through extensive home interviews that included questions on a variety of known and hypothesized risk factors, including smoking, sexual behavior, pregnancy history, menstrual his- tory and hygiene practices, oral contraceptive use, medical history, diet, marital status, and family history.

A history of ever having smoked cigarettes was reported by 256 cases (61.4%) and 383 controls (48.1%). The corresponding OR was 1.7; when examined by

51589_CH06_Printer.indd 313 18/02/13 8:18 PM

314 C h a p t e r 6 S t u d y d e S i g n S

currency of smoking, the ORs were 1.4 for ever-smokers and 1.9 for current smokers. There was evidence for a dose-response relationship: Compared with never-smokers, the ORs for those who smoked fewer than 10, 10–19, 20–29, and 30 or more cigarettes per day were 1.2, 1.6, 1.7, and 3.2, respectively.

Case-control studies have explored the relationship between tobacco use and cancers at various sites. Smoking is a risk factor for oral and throat (upper diges- tive tract) cancers; risk of these forms of cancer is thought to be associated with the concentration of tars in cigarettes. A case-control study conducted in Italy and Switzerland (with 749 cases and 1,770 controls) confirmed an association between cigarette consumption and upper digestive tract neoplasms.40 Investiga- tors reported a direct relationship between the tar yield of cigarettes and such cancers. The ORs between current smokers and never-smokers were 6.1 for cigarettes with lower tar concentrations and 9.8 for cigarettes with higher tar concentrations.

The United Kingdom Childhood Cancer Study examined parental smoking behavior as a risk factor for childhood cancers.41 The study included 3,838 chil- dren with cancer and 7,629 control children. Cancers were classified into four major groups: leukemia, lymphomas, central nervous system tumors, and other solid tumors. Information was collected on self-reported smoking habits of par- ents. Statistical analyses adjusted for factors such as parental age. In general, the study did not demonstrate that parental smoking was a significant risk factor for childhood cancers, although risk of one form of cancer (hepatoblastoma) was significant when both parents smoked (OR = 4.74).

Case-control studies have evaluated the efficacy of cancer screening pro- grams.42 The efficacy of colonoscopic screening and polypectomy for prevent- ing colorectal cancer (CRC) was assessed in a small-scale case-control study. The cases consisted of 40 asymptomatic persons diagnosed with CRC and 160 nor- mal controls. Researchers selected subjects from a high-risk population of first- degree relatives of CRC patients. It was found that cases and controls varied in frequency of screening, with procedures such as screening colonoscopy occurring less frequently among the cases than among the controls.

The remainder of this section illustrates the use of case-control studies to investigate conditions other than cancer (e.g., heart disease, birth defects, infec- tious disease). For example, exposure to side-stream cigarette smoke has been hypothesized to be a risk factor for heart disease. A study conducted in Argentina examined the risk associated with passive smoking and acute myocardial infarc- tion (AMI).43 Both cases (n = 336) and controls (n = 446) consisted of never- smokers admitted to the same network of hospitals. Cases were patients with

51589_CH06_Printer.indd 314 18/02/13 8:18 PM

C a S e - C O n t r O l S t u d i e S 315

AMI; controls were patients admitted with acute disorders unrelated to smoking or known to be risk factors for AMI. Trained interviewers administered a struc- tured questionnaire that contained items on smoking by close relatives, such as children and spouses. The risk of AMI was significantly associated with passive smoking; the ORs ranged upward of 1.68 depending on the contrast category (e.g., numbers of relatives who smoked and the number of cigarettes smoked).

Regarding the use of case-control studies to explore risk of birth defects, an example is the report based on a record linkage of Swedish healthcare registries. This research identified 6 infants with neural tube defects (NTDs) born to mothers who had undergone surgery in the first trimester of pregnancy when only 2.5 were expected.44 The observation was evaluated in more detail by inves- tigators at the Centers for Disease Control and Prevention.45 Cases were infants with major central nervous system defects, including NTDs (anencephaly, spina bifida, and encephalocele), microcephaly, and hydrocephaly (hydrocephalus), ascertained through the population-based surveillance system known as the Metropolitan Atlanta Congenital Defects Program. Controls were a 1% random sample of infants born in the same geographic region over the same time period. Trained interviewers, blinded to the case-control status of the infant, collected data from the mothers by telephone. Maternal anesthesia exposure immedi- ately before becoming pregnant or during the first 3 months of gestation was ascertained. Data were collected from the mothers of 694 case infants and the mothers of 2,984 control infants. There were no differences between mothers of the cases and controls with respect to mean age, parity, smoking status, use of alcohol-containing beverages, weight gain during pregnancy, or education. Maternal exposure to general anesthesia during the first trimester, however, was reported by 1.7% of case mothers versus only 1.1% of control mothers. This equates to an OR of 1.7, with a 95% CI of 0.8 to 3.3. Because the 95% CI includes the null value of 1.0, the results are consistent with the hypothesis of no association. Further analysis of the data revealed that a stronger association was evident for one subtype of defect, hydrocephalus (OR, 3.8; 95% CI, 1.6–9.1). Thus, the mothers of infants with hydrocephalus were nearly four times as likely to have reported early exposure to general anesthesia as mothers of infants with- out congenital malformations, an association that does not appear to be due to chance alone. When multiple defects were considered, stronger associations were identified. For example, there were eight infants with both hydrocephalus and eye defects. Three of the mothers of these eight infants (37.5%) reported general anesthesia exposure, an odds of exposure 39.6 times greater (95% CI, 7.5–209.2) than among control mothers. These intriguing data warrant additional research

51589_CH06_Printer.indd 315 18/02/13 8:18 PM

316 C h a p t e r 6 S t u d y d e S i g n S

to determine why the surgeries were performed as well as the types of surgeries, premedications for surgery, and use of general anesthesia and whether there were complications during or after the surgery.

Case-control studies are often the method of choice in infectious disease research. Outbreaks of new and unusual diseases, such as toxic shock syndrome, anthrax, and even occurrence of antibiotic-resistant organisms, may be explored fruitfully by applying case-control methods. A case-control study examined the issue of whether a person’s risk of antibiotic-resistant infection is increased if a member of that person’s family is exposed to antibiotics.46 These data were derived from patients enrolled in health maintenance organizations in western Washington State and northern California. The study outcome was affliction with penicillin nonsusceptible pneumococcal disease (143 cases) versus penicillin susceptible disease (79 controls). A significant association between antibiotic use within 2 months prior to diagnosis and antibiotic resistance was found. Use of antibiotics by other family members within four months prior to diagnosis was not related to antibiotic resistance.

Summary of Case-Control Studies In this section we have indicated the numerous attractive features of case- control studies. Compared with large-scale surveys or prospective studies, case-control studies tend to be smaller in size: hundreds to thousands of subjects versus thousands to tens of thousands. As such, they are relatively quick and easy to complete as well as cost effective. The smaller sample size increases the likeli- hood that a case-control study will be repeated. In fact, because consistency is critical to epidemiologic research, it is highly desirable that several investiga- tors repeat studies of a particular outcome in different populations. Although progression from case-control studies to a cohort study may be a logical pur- suit, prospective studies are not feasible for some exposures and outcomes; a meaningful cohort study of a rare disease would require a large study group, a long period of follow-up, or both. In comparison with cohort studies, case- control studies are particularly useful for investigations into the etiology of rare diseases.

Three major limitations of case-control studies are 1) unclear temporal rela- tionships between exposures and diseases; 2) use of indirect estimates of risk; and 3) indeterminate representativeness of the cases and controls in some situations. Furthermore, not only can errors be introduced in selection of subjects but also in measurement of exposures. If the exposure is rare in the population, then case-control studies may be inefficient. That is, despite a large number of cases,

51589_CH06_Printer.indd 316 18/02/13 8:18 PM

S t u d y Q u e S t i O n S a n d e x e r C i S e S 317

one may still end up with few exposed cases. For a more detailed coverage of case-control studies refer to Schlesselman’s classic textbook47 or a more contem- porary work by Armenian.48

Conclusion

This chapter provided an overview of major study designs that are used in epi- demiology. The authors noted that study designs differ in a number of key respects, including the unit of observation, the unit of analysis, the timing of exposure data in relation to occurrence of disease end point, complexity, rigor, and amount of resources required. Some studies are designed to infer etiology; others are designed to affect parameters of health. Study designs differ with respect to control over exposure and the ability to assign subjects to study condi- tions. Two major categories of study design exist: observational and experimen- tal. This chapter presented three major types of observational designs: ecologic, cross-sectional, and case-control. The measure of effect (association) used in case-control studies, OR, was illustrated. Each of the three types of observational designs covered in this chapter are key components of the arsenal of epidemio- logic study designs.

Study Questions and Exercises

1. Define in your own words the following terms: a. ecologic study b. ecologic comparison study c. ecologic trend study d. ecologic fallacy e. cross-sectional study f. case-control study

2. Compare ecologic, cross-sectional, and case-control studies with respect to their strengths and weaknesses, and advantages and disadvantages.

3. The following question lists examples of observational studies. Indicate the type of study design that is being described. a. A study examined the effect of hormone replacement therapy on

cancer; cancer cases were identified by using a cancer registry in northern California. Controls were selected from a random sample of Bay Area cities.

51589_CH06_Printer.indd 317 18/02/13 8:18 PM

318 C h a p t e r 6 S t u d y d e S i g n S

b. Data from the Behavioral Risk Factor Survey were used in a secondary data analysis to examine the effect of income inequality and race on preventive health practices.

c. The level of unemployment was used as a measure of economic dis- tress in Germany. Researchers examined the association between distress and general anxiety syndrome across states (e.g., Saxony) in Germany.

4. What are the differences between probability and nonprobability sam- ples used in cross-sectional designs?

5. You are interested in conducting a case-control study of childhood leu- kemia and exposure to environmental toxins in utero. Describe how you would select cases and controls for this study and how you would define exposure and outcome factors. How could the same problem be investi- gated using an ecologic study design?

6. Describe the advantages and disadvantages of each of the following types of controls in a case-control study: a. Population-based b. Hospital cases c. Relatives

7. Calculate the OR for the following 2 by 2 table:

8. An investigator wanted to determine whether vitamin deficiency was associated with birth defects. By reviewing the birth certificates during a single year in a large U.S. county, the researcher located 189 infants born with NTDs. A total of 600 other births were selected at random from the certificates. Mothers were given a dietary questionnaire. Among mothers who gave birth to an infant with an NTD, 84 reported no use of supple- mentary vitamins; a total of 137 control mothers did not use a vitamin supplement. Construct the appropriate 2 by 2 table and calculate the OR between vitamin use and NTDs.

9. The association between job-related exposure to welding fumes and chronic obstructive pulmonary disease (COPD) was explored in a case- control study. The following data were reported for 399 COPD patients: 37 currently employed as welders; the remainder had no occupational

Outcome Yes No

Factor Yes 37 68 No 24 121

51589_CH06_Printer.indd 318 18/02/13 8:18 PM

r e f e r e n C e S 319

exposure. Among 800 controls, 48 were employed as welders. Set up a 2 by 2 table and calculate the OR.

10. Hypothesize that cell phone use by drivers is related to fatal automobile accidents. Design a hypothetical case-control study, giving attention to the following points: definition of the outcome, selection of controls, and difficulties in conducting such a study.

11. Why would you exclude as controls in a case-control study of gyneco- logic cancer women who cannot develop the disease?

12. The following ORs are reported for several hypothetical examples. Give your interpretation of the results, assuming all results are statistically significant unless otherwise specified. a. OR (low-fat diet and colon cancer) + 0.6 b. OR (aerobic exercise and dental caries) + 1 (not significant) c. OR (exposure to side-stream cigarette smoke and lung cancer) + 1.3 d. OR (infectious disease of the pelvis and ectopic [tubal] pregnancy) + 3.0

13. A random-digit dialed survey conducted in the City of Long Beach, California, reported that a greater proportion of nonsmokers endorsed a ban on smoking in alcohol-serving establishments than did smokers. What type of study design was this?

14. Case-control studies allow the investigator to examine only one outcome at a time, but they permit examination of several different exposures at a time. Select a disease or other health outcome with which you are famil- iar and see how many potential exposures you can identify.

15. How would you design an ecologic study to investigate the following prob- lems? How might the ecologic fallacy come into play in each situation? a. lung disease and air pollution b. birth defects and hazardous waste c. cancer and radiation leakage from a power plant

References 1. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic Research: Principles and

Quantitative Methods. Belmont, CA: Lifetime Learning; 1982. 2. Valls V, Lozano MS, Yanez R, et al. Use of safety devices and the prevention of

percutaneous injuries among healthcare workers. Infect Control Hosp Epidemiol. 2007;28:1352–1360.

3. Garrard J, Chen V, Dowd B. The impact of the 1987 federal regulations on the use of psychotropic drugs in Minnesota nursing homes. Am J Public Health. 1995;85: 771–776.

51589_CH06_Printer.indd 319 18/02/13 8:18 PM

320 C h a p t e r 6 S t u d y d e S i g n S

4. Carroll KK. Experimental evidence of dietary factors and hormone-dependent cancers. Cancer Res. 1975;35:3374–3383.

5. Sargent JD, Brown MJ, Freeman JL, Bailey A, Goodman D, Freeman DH. Childhood lead poisoning in Massachusetts communities: its association with sociodemographic and housing characteristics. Am J Public Health. 1995;85:528–534.

6. Menotti A, Blackburn H, Kromhout D, et al. The inverse relation of average popula- tion blood pressure and stroke mortality rates in the seven countries study: a para- dox. Eur J Epidemiol. 1997;13:379–386.

7. Peres AFT, de Almeida JF, Vicentin D, et al. Fall in homicides in the city of São Paulo: an exploratory analysis of possible determinants. Rev Bras Epidemiol. 2011;14(4):709–721.

8. Jacobsen SJ, O’Fallon WM, Melton LJ 3rd. Hip fracture incidence before and after the fluoridation of the public water supply, Rochester, Minnesota. Am J Public Health. 1993;83:743–745.

9. Steiner GG. Cancer incidence rates and environmental factors: an ecological study. J Environ Pathol Toxicol Oncol. 2002;21:205–212.

10. Pickett KE, Pearl M. Multilevel analyses of neighbourhood socioeconomic context and health outcomes: a critical review. J Epidemiol Community Health. 2001;55: 111–122.

11. Porta M, ed. A Dictionary of Epidemiology. 5th ed. New York, NY: Oxford University Press; 2008.

12. Levy PS, Lemeshow S. Sampling for Health Professionals. Belmont, CA: Lifetime Learning; 1980.

13. Babbie E. The Practice of Social Research. 6th ed. Belmont, CA: Wadsworth; 1992. 14. Kish L. Survey Sampling. New York, NY: Wiley; 1965. 15. Murray DM, Roche LM, Goldman AI, Whitbeck J. Smokeless tobacco use among

ninth graders in a North-Central metropolitan population: cross-sectional and pro- spective associations with age, gender, race, family structure, and other drug use. Prev Med. 1988;17:449–460.

16. Anderson JE, Warner L, Jamieson DJ, et al. Contraceptive sterilization use among married men in the United States: results from the male sample of the National Survey of Family Growth. Contraception. 2010;82(3):230–235.

17. Myers SM, Johnson CP. Management of children with autism spectrum disorders. Pediatr. 2007;120(5):1162–1182.

18. Pringle BA, Colpe LJ, Blumberg SJ, et al. Diagnostic history and treatment of school- aged children with autism spectrum disorder and special health care needs. NCHS data brief, no 97. Hyattsville, MD: National Center for Health Statistics. 2012.

19. Friis RH, Garrido-Ortega C, Safer AM, et al. Socioepidemiology of cigarette smok- ing among Cambodian Americans in Long Beach, California. J Immigr Minor Health. 2012; 14(2):272–280.

20. Diwan S, Moriarty D. A conceptual framework for identifying unmet health care needs of community dwelling elderly. J Appl Gerontol. 1995;14:47–63.

21. Dillon CF, Gu Q, Hoffman H, Ko CW. Vision, hearing, balance, and sensory impairment in Americans aged 70 years and over: United States, 1999–2006. NCHS data brief, no 31. Hyattsville, MD: National Center for Health Statistics. 2010.

22. Nelson DE, Giovino GA, Shopland DR, Mowery PD, Mills SL, Eriksen MP. Trends in cigarette smoking among U.S. adolescents, 1974 through 1991. Am J Public Health. 1995;85:34–40.

51589_CH06_Printer.indd 320 18/02/13 8:18 PM

r e f e r e n C e S 321

23. U.S. Department of Health and Human Services, Health Resources and Services Administration, Maternal and Child Health Bureau. Child Health USA 2011. Rockville, Maryland: U.S. Department of Health and Human Services, 2011.

24. Comstock GW. Evaluating vaccination effectiveness and vaccine efficacy by means of case-control studies. Epidemiol Rev. 1994;16:77–89.

25. Selby JV. Case-control evaluations of treatment and program efficacy. Epidemiol Rev. 1994;16:90–101.

26. Weiss NS. Application of the case-control method in the evaluation of screening. Epidemiol Rev. 1994;16:102–108.

27. Dwyer DM, Strickler H, Goodman RA, Armenian HK. Use of case-control studies in outbreak investigations. Epidemiol Rev. 1994;16:109–123.

28. Lasky T, Stolley PD. Selection of cases and controls. Epidemiol Rev. 1994;16:6–17. 29. Brenner H, Savitz DA. The effects of sensitivity and specificity of case selection on

validity, sample size, precision, and power in hospital-based case-control studies. Am J Epidemiol. 1990;132:181–192.

30. Linet MS, Brookmeyer R. Use of cancer controls in case-control studies. Am J Epidemiol. 1987;125:1–11.

31. Chernick MR, Friis RH. Introductory Biostatistics for the Health Sciences: Modern Applications Including Bootstrap. Hoboken, NJ: John Wiley & Sons, Inc.; 2003.

32. Hartge P, Brinton LA, Rosenthal JF, et al. Random digit dialing in selecting a population-based control group. Am J Epidemiol. 1984;120:825–833.

33. Olson SH, Kelsey JL, Pearson TA, et al. Evaluation of random digit dialing as a method of control selection in case-control studies. Am J Epidemiol. 1992;135:210–222.

34. Shaw GL, Tucker MA, Kase RG, et al. Problems ascertaining friend controls in a case-control study of lung cancer. Am J Epidemiol. 1991;133:63–66.

35. Jones S, Silman AJ. Re: “Problems ascertaining friend controls in a case-control study of lung cancer.” Am J Epidemiol. 1991;134:673–674.

36. López-Carrillo L, Avila MH, Dubrow R. Chili pepper consumption and gastric can- cer in Mexico: a case-control study. Am J Epidemiol. 1994;139:263–271.

37. Zhong L, Goldberg MS, Gao Y-T, et al. A population-based case-control study of lung cancer and green tea consumption among women living in Shanghai, China. Epidemiology. 2001;12:695–700.

38. Herbst AL, Scully RE. Adenocarcinoma of the vagina in adolescence. Cancer. 1970;25:745–757.

39. Brinton LA, Schairer C, Haenszel W, et al. Cigarette smoking and invasive cervical cancer. JAMA. 1986;255:3265–3269.

40. Gallus S, Altieri A, Bosetti C, et al. Cigarette tar yield and upper digestive tract can- cers: case-control studies from Italy and Switzerland. Ann Oncol. 2003;14:209–213.

41. Pang D, McNally R, Birch JM. Parental smoking and childhood cancer: results from the United Kingdom Childhood Cancer Study. Br J Cancer. 2003;88:373–381.

42. Niv Y, Dickman R, Figer A, et al. Case-control study of screening colonoscopy in relatives of patients with colorectal cancer. Am J Gastroenterol. 2003;98:486–489.

43. Ciruzzi M, Pramparo P, Esteban O, et al. Case-control study of passive smoking at home and risk of acute myocardial infarction. J Am Coll Cardiol. 1998;31:797–803.

44. Kallen B, Mazze RI. Neural tube defects and first trimester operations. Teratology. 1990;41:717–720.

51589_CH06_Printer.indd 321 18/02/13 8:18 PM

322 C h a p t e r 6 S t u d y d e S i g n S

45. Sylvester GC, Khoury MJ, Lu X, Erickson JD. First-trimester anesthesia exposure and the risk of central nervous system defects: a population-based case-control study. Am J Public Health. 1994;84:1757–1760.

46. Kwan-Gett TS, Davis RL, Shay DK, et al. Is household antibiotic use a risk factor for antibiotic-resistant pneumococcal infection? Epidemiol Infect. 2002;129:499–505.

47. Schlesselman JJ. Case-Control Studies: Design, Conduct, Analysis. New York, NY: Oxford University Press; 1982.

48. Armenian HK, ed. The Case-Control Method: Design and Applications. Oxford: Oxford University Press; 2009.

51589_CH06_Printer.indd 322 18/02/13 8:18 PM

3chapte r

323

7chapte r

Study Designs: Cohort Studies

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

● differentiate cohort studies from other epidemiologic study designs ● list the main characteristics, advantages, and disadvantages of

cohort studies ● describe at least three research questions that lend themselves to

cohort studies ● calculate and interpret a relative risk ● give three examples of published studies discussed in this chapter

CHAPTER OUTLINE

I. Introduction II. Cohort Studies Defined

III. Sampling and Cohort Formation Options IV. Temporal Differences in Cohort Designs V. Practical Considerations

VI. Measures of Effect: Their Interpretation and Examples VII. Summary of Cohort Studies

VIII. Conclusion IX. Study Questions and Exercises

51589_CH07_Printer.indd 323 11/02/13 2:12 PM

324 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

Introduction

This chapter explores one of the most powerful observational epidemiologic designs—the cohort study—which overcomes many of the problems associated with temporality of data collection and obtaining information about exposures that are uncommon in the population. A distinguishing feature of each type of study design—whether observational or experimental—is the temporality of data collection with respect to exposure and disease. The term temporality refers to the timing of information gathering, that is, whether the information about cause and effect was assembled at the same time point or whether information about the cause was garnered before or after the information about the effect. When information about exposures is collected before an outcome occurs, and there is an association between exposures and outcomes, one is more confident about a possible cause-and-effect relationship. We will learn that cohort studies preserve the temporality of cause (exposure) happening before effect (disease).

Cross-sectional and case-control study designs (and many types of ecologic study designs) are premised upon exposure information and disease informa- tion that are collected at the same time. In a cross-sectional study, one might administer a survey that contains questions about current or past exposures (e.g., exposure to secondhand cigarette smoke) and various outcomes (e.g., respira- tory symptoms). Data for a case-control study might be collected from patient interviews and reviews of medical records. Even though exposure and outcome information are obtained simultaneously, the frame of reference for exposure assessment in case-control studies is retrospective, meaning that respondents are interviewed about exposures that occurred in the past. All in all, in case-control and cross-sectional studies, researchers obtain information about health out- comes and exposures after they have occurred.

Although the strategy of collecting exposure and outcome information at the same time, and after they have occurred, is efficient for generating and testing hypotheses, the strategy does lead to almost unavoidable challenges regarding interpretation of results. In particular, cross-sectional studies present difficulties in distinguishing the causes (e.g., certain exposures) from the consequences (e.g., certain outcomes) of the disease, especially if the outcome marker is a biological or physiological parameter. Similarly, case-control studies may raise concerns that recall of past exposures differs between the cases (i.e., those study participants who have the disease or outcome of interest) and the controls (i.e., those study participants who do not). In addition, although investigators may query subjects

51589_CH07_Printer.indd 324 11/02/13 2:12 PM

C o h o r t S t u d i e S d e f i n e d 325

about exposures that took place many years in the past, there has been no actual lapse of time between measurement of exposure and disease. Finally, neither cross-sectional nor case-control designs are especially well suited for exposures that are uncommon in the population. However, cohort studies overcome many of the challenges presented by temporality and uncommon exposures.

Cohort Studies Defined

Cohorts and Cohort Effects A cohort is defined as a population group, or subset thereof (distinguished by a common characteristic), that is followed over a period of time. The term cohort is said to originate from the Latin cohors, which is one of 10 divisions of an ancient Roman military legion. The common characteristic may be either that the group members experience an exposure associated with a specific setting (e.g., an occupational cohort or a school cohort) or that they share a nonspecific expo- sure associated with a general classification (e.g., a birth cohort, defined as being born in the same year or era). For example, people who belong to the same birth cohort may be exposed to similar environmental and societal changes, whereas those who belong to different birth cohorts may grow up exposed to dissimilar environmental conditions that are reflected in differences in health outcomes. The influence of membership in a particular cohort is known as a cohort effect.

The term cohort analysis refers to “the tabulation and analysis of morbidity or mortality rates in relationship to the ages of a specific group of people (cohort) identified at a particular period of time and followed as they pass through differ- ent ages during part or all of their life span.”1 Wade Hampton Frost helped to draw attention to the method of cohort analysis, even though he did not origi- nate this methodology.2 Table 7–1 reproduces Frost’s data. “To illustrate cohort analysis, Frost first arranged tuberculosis mortality rates from Massachusetts . . . in a table with age on one axis and year of death on the other . . . Arranged in this way, one could quickly see the age-specific mortality for each of the avail- able years on one axis, and the time trend for each age group on the other. What proved to be most interesting in this instance were the rates in the cells of the table that lay on the diagonals, starting with the youngest ages and earliest years. These ‘diagonal rates’ were analogous to tuberculosis mortality rates . . . experi- enced by each cohort of persons as they simultaneously aged and passed through time.”2(pp 9–10)

51589_CH07_Printer.indd 325 11/02/13 2:12 PM

326 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

Table 7–1 shows that the 1880 mortality rates from tuberculosis were high among males aged 0–4 and 5–9 years—the highest rates occurring among the cohort aged 0–4 years—when compared with mortality for the cohort as it aged. You can confirm this observation by tracing along the diagonal shown in the table. With respect to 1890 and subsequent decades, the groups that were 0–4 and 5–9 years of age in 1880 are combined as one moves along the diagonal. In comparison to 1880, mortality rates for that same cohort were lower during sub- sequent points of observation; however, during 1900 and 1910 they were higher than in 1890 and then declined again. The mortality pattern was similar for the female cohort that was aged 0–4 and 5–9 years in 1880.

Another example of a cohort effect is the use of tobacco products in the United States (Figure 7–1). A low proportion (< 5%) of the population smoked cigarettes around the early 1900s. As a result of widespread distribution of free cigarettes to the troops during World War I, however, the prevalence of smok- ing in the population began to increase gradually, reaching a peak in the 1960s.3

Table 7–1 Death Rates per 100,000 from Tuberculosis, All Forms, for Massachusetts, 1880 to 1930, by Age and Sex, with Rates for Cohort of 1880 Indicated

Age 1880 1890 1900 1910 1920 1930

Male 0–4 760 578 309 209 108 41 5–9 43 49 31 21 24 11 10–19 126 115 90 63 49 21 20–29 444 361 288 207 149 81 30–39 378 368 296 253 164 115 40–49 364 336 253 253 175 118 50–59 366 325 267 252 171 127 60–69 475 346 304 246 172 95 70+ 672 396 343 163 127 95

Female 0–4 658 595 354 162 101 27 5–9 71 82 49 45 24 13 10–19 265 213 145 92 78 37 20–29 537 393 290 207 167 92 30–39 422 372 260 189 135 73 40–49 307 307 211 153 108 53 50–59 334 234 173 130 83 47 60–69 434 295 172 118 83 56 70+ 584 375 296 126 68 40

Source: Reproduced with permission. Frost WH. The Age Selection of Mortality from Tuberculosis in Successive Decades. Am J Epidemiol, 1995, Vol 141, p. 95. © The Johns Hopkins University School of Hygiene and Public Health.

51589_CH07_Printer.indd 326 11/02/13 2:12 PM

C o h o r t S t u d i e S d e f i n e d 327

When smoking first became popular, the age at which the habit was initiated varied greatly; that is, there were considerable differences according to age, sex, and education levels.

Although some people began smoking as young adults, a large number of people adopted the habit much later in life. Over the years, more and more people took up smoking and commenced smoking earlier in life. This trend is depicted graphically in Figure 7–2. One of the net effects was a shift in the distribution of the age of onset of lung cancer.4 Consider, for example, the birth cohort of 1850. If smoking prevalence in this cohort was similar to that of the general population, then most individuals did not begin smoking until around 1915, when the average cohort member was in his or her 60s. Because there is a delay between the onset of smoking and the development of cancer, these individuals would not develop cancer for 10 years or more, perhaps around the age of 70. In contrast, individuals born in 1890 were only 20 when smoking became popular. As a result, a greater proportion of this age cohort would have started smoking at an earlier age, so the distribution of the entire age at onset curve for lung cancer would be shifted toward earlier ages. The more traditional approach of examining trends through repeated cross-sectional surveys leads to a distorted impression of the smoking–cancer association. In particular, it leads

Figure 7–1 Annual adult per capita cigarette consumption and major smoking-and-health events–United States, 1900–2011. Source: Reproduced from U.S. Department of Health and Human Services. Ending the Tobacco Epidemic: Progress Toward a Healthier Nation. Washington: U.S. Department of Health and Human Services, Office of the Assistant Secretary for Health, August 2012.

1900 0

1,000

2,000

N um

be r

of C

ig ar

et te

3,000

Great Depression

End of WW II

1st Smoking- Cancer Concern

Fairness Doctrine Messages on TV

and Radio

1st Surgeon General’s Report

Master Settlement Agreement

Federal Cigarette Tax More Than Doubles

4,000

5,000

1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010

51589_CH07_Printer.indd 327 11/02/13 2:12 PM

328 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

to an underestimation of the past smoking behavior of the older segments of the current population. That is because smoking behavior (age at initiation and total duration) is greatly influenced by the calendar year of birth.

Table 7–2 demonstrates a final example of a cohort effect, this time for lung cancer death rates in the United Kingdom and the United States.5 The table presents the midpoint of birth cohorts on the y-axis (left side) and the midpoint of 5-year age groups on the x-axis (across the top). From the table, one is able

Figure 7–2 Changes in prevalence of cigarette smoking among successive birth cohorts of U.S. men, 1900–1987. Source: Reproduced from Strategies to Control Tobacco Use in the United States: A Blueprint for Public Health Action in the 1990s. Washington, DC: National Cancer Institute, National Institutes of Health; Publication No. 92-3316. 1992:82.

19901910

1961–1970 1901–1910

1911–1920

1921–1930

1931–1940

1941–1950

1901–1910

1911–1920

1921–1930

1931–1940

1941–1950

1951–1960

1961–1970

P er

ce nt

ag e

Year

1920 1930 1940 1950 1960 1970 1980

1951–1960

51589_CH07_Printer.indd 328 11/02/13 2:12 PM

C o h o r t S t u d i e S d e f i n e d 329

Table 7–2 Lung Cancer Death Rate per 100,000 for the United Kingdom and United States

Age (Midpoint of 5-year Age Group) United Kingdom

Midpoint of Birth Cohort

32.5 37.5 42.5 47.5 52.5 57.5 62.5 67.5 72.5 77.5 82.5

1873 167.30 1878 243.01 259.83 1883 305.25 377.48 391.85 1888 329.25 431.62 506.16 509.23 1893 289.47 428.10 538.81 650.73 679.95 1898 219.13 353.38 512.62 662.33 764.63 812.65 1903 126.17 232.18 374.27 528.11 682.74 796.69 832.76 1908 59.72 125.16 228.23 369.44 514.73 655.69 756.35 767.28 1913 24.87 57.76 120.75 215.69 344.15 479.32 616.24 678.27 669.39 1918 9.78 25.00 55.08 111.60 202.32 316.22 437.96 553.20 592.54 1923 3.76 9.47 22.38 53.96 106.37 184.94 294.74 402.30 475.24 1928 3.53 9.06 21.01 46.80 92.11 158.66 245.40 326.52 1933 2.80 6.29 16.30 36.15 69.26 122.78 184.53 1938 2.49 5.90 12.96 29.62 59.51 102.18 1943 2.24 4.97 11.24 26.26 49.74 1948 1.56 4.07 9.72 20.74 1953 1.09 3.13 8.02 1958 0.77 2.13 1963 0.65

United States 32.5 37.5 42.5 47.5 52.5 57.5 62.5 67.5 72.5 77.5 82.5

1873 116.80 1878 138.60 176.30 1883 148.90 199.60 222.60 1888 157.80 232.20 268.30 325.40 1893 135.50 219.70 302.60 380.60 431.60 1898 95.90 180.70 277.30 371.00 464.00 477.70 1903 58.20 114.92 199.88 306.95 418.93 502.80 543.33 1908 30.40 68.71 127.55 228.01 329.42 458.80 546.20 584.96 1913 11.60 31.79 76.79 152.13 244.30 359.11 470.70 565.40 580.60 1918 4.90 13.99 38.61 84.32 150.01 255.74 367.06 485.89 529.90 1923 1.70 5.73 17.26 44.03 90.91 162.93 262.46 374.07 470.90 1928 1.97 7.05 21.54 47.86 95.28 167.41 268.18 359.60 1933 2.00 7.34 19.30 45.44 86.59 159.35 233.60 1938 2.03 6.15 17.43 40.26 80.52 132.70 1943 1.80 5.29 15.19 34.64 66.10 1948 1.12 4.32 11.63 26.20 1953 0.98 3.85 9.50 1958 1.16 3.30 1963 1.20

Source: Adapted from National Cancer Institute. Risks Associated with Smoking Cigarettes with Low Machine-Measured Yields of Tar and Nicotine, p. 128. Smoking and Tobacco Control Monograph No. 13, NIH Pub. No. 02-5074. Bethesda, MD: U.S. Department of Health and Human Services, National Institutes of Health, National Cancer Institute; October 2001.

51589_CH07_Printer.indd 329 11/02/13 2:12 PM

330 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

to infer cohort effects as a particular cohort ages and also compare mortality between the two continents. For example, the respective lung cancer death rates per 100,000 of the 1923 cohort at age 32.5 years are 3.76 and 1.70. The cohort effect can be traced along a diagonal: The respective rates at the intersection of year 1928 and age 37.5 years are 9.06 and 7.05. The overall trend shown in the table is for age-specific lung cancer death rates to start at lower levels in the United States but then show more rapid increases than in the United Kingdom, perhaps due to differences in age of initiation of smoking, particularly among males.

Life Table Methods According to Chernick and Friis, “Life tables give estimates for survival during time intervals and present the cumulative survival probability at the end of the interval.”6(p 339) The uses of life tables include making projections of life expec- tancy, portraying the survival times of patients who have undergone a medical procedure, and demonstrating the survival of patients who have been diagnosed with a chronic disease such as cancer. The National Center for Health Statistics produces life tables for the U.S. population by sex, race, and Hispanic origin.7 Two major types of life tables are created: cohort (or generation) life tables and period (or current) life tables. A cohort life table shows the mortality experience of all persons born during a particular year, such as 1900. “Based on age-specific death rates observed through consecutive calendar years, the cohort life table reflects the mortality of an actual cohort from birth until no lives remain in the group.”7(p 1) For more information, see Arias.7

A period life table gives an overview of the present mortality experience of a population and shows projections of future mortality experience. The term life expectancy refers to the number of years that a person is expected to live, at any particular year. With respect to a year of interest (e.g., 2007), a period life table enables us to project the future life expectancy of persons born during the year as well as the remaining life expectancy of persons who have attained a certain age. Table 7–3 and Exhibit 7–1 show an abridged life table for the total U.S. popula- tion in 2007. From the table, you can see (column ex) that the life expectancy at birth (0–1 years) was 77.9 years and at age 99–100 years was 2.4 years. For more information about life table methods, refer to Chernick and Friis.6

There are additional ways to describe the mortality experience of the population. One measure, which takes into account the effect of prema- ture death caused by diseases, is known as years of potential life lost (YPLL).8 For example, we might assume that the average person lives until age 65. If an

51589_CH07_Printer.indd 330 11/02/13 2:12 PM

C o h o r t S t u d i e S d e f i n e d 331

Table 7–3 Life Table for the Total Population: United States, 2007 (Abridged)

Probablity of dying

between ages x to x + 1

Number surviving to

age x

Number dying

between ages x to

x + 1

Person- years lived between ages x to

x + 1

Total number of

person-years lived above

age x

Expectation of life at

age x

Age qx lx dx Lx Tx ex

0–1 0.006761 100,000 676 99,406 7,793,398 77.9 1–2 0.000460 99,324 46 99,301 7,693,992 77.5 2–3 0.000286 99,278 28 99,264 7,594,691 76.5 3–4 0.000218 99,250 22 99,239 7,495,427 75.5 4–5 0.000176 99,228 17 99,219 7,396,188 74.5 5–6 0.000164 99,211 16 99,203 7,296,969 73.6 6–7 0.000151 99,194 15 99,187 7,197,766 72.6 7–8 0.000140 99,179 14 99,173 7,098,579 71.6 8–9 0.000124 99,166 12 99,159 6,999,407 70.6 9–10 0.000105 99,153 10 99,148 6,900,247 69.6 10–11 0.000091 99,143 9 99,138 6,801,099 68.6 11–12 0.000094 99,134 9 99,129 6,701,961 67.6 12–13 0.000132 99,125 13 99,118 6,602,831 66.6 13–14 0.000209 99,112 21 99,101 6,503,713 65.6 14–15 0.000314 99,091 31 99,075 6,404,612 64.6 15–16 0.000426 99,060 42 99,039 6,305,537 63.7 16–17 0.000529 99,018 52 98,991 6,206,498 62.7 17–18 0.000627 98,965 62 98,934 6,107,506 61.7 18–19 0.000715 98,903 71 98,868 6,008,572 60.8 19–20 0.000796 98,832 79 98,793 5,909,705 59.8 20–21 0.000881 98,754 87 98,710 5,810,911 58.8

67–68 0.015959 81,237 1,296 80,589 1,393,772 17.2 68–69 0.017288 79,940 1,382 79,249 1,313,183 16.4 69–70 0.018755 78,558 1,473 77,822 1,233,934 15.7 70–71 0.020424 77,085 1,574 76,298 1,156,112 15.0 71–72 0.022385 75,511 1,690 74,666 1,079,814 14.3 72–73 0.024679 73,820 1,822 72,909 1,005,149 13.6 73–74 0.027320 71,999 1,967 71,015 932,239 12.9

96–97 0.239389 6,499 1,556 5,721 19,597 3.0 97–98 0.258999 4,943 1,280 4,303 13,876 2.8 98–99 0.279625 3,663 1,024 3,151 9,573 2.6 99–100 0.301225 2,638 795 2,241 6,422 2.4 100+ 1.000000 1,844 1,844 4,181 4,181 2.3

individual succumbs at age 60, that person has lost 5 years of life. YPLL is com- puted by summing years of life lost for each individual in a population such as the United States for a specific cause of mortality (Figure 7–3). Another measure is disability-adjusted life years (DALYs), which adds the time a person has a dis- ability to the time lost to early death.9 Thus, one DALY indicates one year of life lost to the combination of disability and early mortality.

51589_CH07_Printer.indd 331 11/02/13 2:12 PM

332 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

explanation of the columns of the Life table

Column 1—Age (x to x + 1)—Shows the age interval between the two exact ages indicated. For instance, ‘‘20–21’’ means the 1-year inter- val between the 20th and 21st birthdays.

Column 2—Probability of dying (qx)—Shows the probability of dying between ages x to x + 1. For example, in the age interval 20–21 years, the probability of dying is 0.000881. The ‘‘probability of dying’’ col-

umn forms the basis of the life table; all subsequent columns are derived from it.

Column 3—Number surviving (lx)—Shows the number of persons from the original hypothetical cohort of 100,000 live births who survive to the beginning of each age interval. The lx values are computed from the qx values, which are successively applied to the remainder of the original 100,000 persons still alive at the beginning of each age interval. Thus, out of 100,000 babies born alive, 99,324 will complete the first year of life and enter the second; 99,143 will reach age 10; 98,754 will reach age 20; and 6,499 will live to age 96.

Column 4—Number dying (dx)—Shows the number dying in each succes- sive age interval out of the original 100,000 live births. For example, out of 100,000 persons born alive, 676 will die in the first year of life; 87 between ages 20 and 21; and 1,844 will die after reaching age 100. Each figure in column 4 is the difference between two successive figures in column 3.

Column 5—Person-years lived (Lx)—Shows the number of person-years lived by the hypothetical life table cohort within an age interval x to x + 1. Each figure in column 5 represents the total time (in years) lived between two indicated birthdays by all those reaching the earlier birthday. Thus, the figure 98,710 for males in the age interval 20–21 is the total number of years lived between the 20th and 21st birthdays by the 98,754 (column 3) persons who reached their 20th birthday out of 100,000 males born alive.

Column 6—Total number of person-years lived (Tx)—Shows the total number of person-years that would be lived after the beginning of the age inter- val x to x + 1 by the synthetic life table cohort. For example, the figure 5,810,911 is the total number of years lived after attaining age 20 by the 98,754 persons reaching that age.

e X

h IB

It 7

–1

continues

51589_CH07_Printer.indd 332 11/02/13 2:12 PM

C o h o r t S t u d i e S d e f i n e d 333

Survival Curves In addition to life tables, one of the methods for portraying survival times is to use survival curves. In order to construct a survival curve, the following informa- tion would be required about each subject: time of entry into the study, time of death (or other outcome), and status of the patient at that time (either dead or censored, which may mean that the patient is lost to follow-up).

Figure 7–3 Years of potential life lost (YPLL) before age 65, 2009, United States, all races, both sexes, all deaths. Source: Reproduced from Centers for Disease Control and Prevention, National Center for Injury Prevention and Control. Years of potential life lost (YPLL) before age 65. Available at http://webappa.cdc.gov/sasweb/ncipc/ypll10.html. Accessed April 17, 2012.

Cause of Death

All Causes

Unintentional Injury Malignant Neoplasms

Heart Disease

Perinatal Period

Suicide

Homicide

Liver Disease

Congenital Anomalies

Cerebrovascular

Diabetes Mellitus

All Others 2,784,958

218,223

236,008

256,021

457,042

540,536

735,420

851,141

1,367,026

1,865,472

2,090,684

11,402,531

Percent

100.0%

18.3%

16.4%

12.0%

7.5%

6.4%

4.7%

4.0%

2.2%

2.1%

1.9%

24.4%

YPLL

exhibit 7–1 continued

Column 7—Expectation of life (ex)—Shows, at any given age, the average number of years remaining to be lived by those surviving to that age on the basis of a given set of age-specific rates of dying. Thus, the average remain- ing lifetime for persons who reach age 20 is 58.8 years.

For more information regarding how the numbers shown in the table are calculated, refer to the original source. n

Source: Adapted from Arias E. United States life tables, 2007. National vital statistics reports, Vol 59, No 3, pp. 2–3, 8–9, Hyattsville, MD: National Center for Health Statistics. 2011.

51589_CH07_Printer.indd 333 11/02/13 2:12 PM

334 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

Take the case of a clinical study of survival of patients diagnosed with malig- nant pleural mesothelioma (MPM) classified according tumor histology (epitheli- oid mesothelioma and nonepithelioid mesothelioma).10 A group of 128 incident cases of MPM were followed over a period of months. The definition of survival time was the time between diagnosis with MPM and a patient’s death or “last known follow-up.” (Refer to Figure 7–4.) The dots in the figure denote censored observations. The x-axis portrays survival time in months; the y-axis indicates the fraction surviving. Each step of the curve represents the death(s) of one or more patients. The survival curve for patients with the nonepithelioid tumor histology (n = 37) drops much more rapidly than the curve for patients with epi- thelioid histology (n = 91); statistical analyses indicated a significant difference in survival between the two tumor histologies.

Statistical procedures (e.g., the method of Kaplan and Meier) are available to evaluate the statistical significance of differences among survival curves. Survival

Figure 7–4 Kaplan-Meyer survival probability plots of malignant plural mesothelioma patients (n = 128). Survival of patients with an epithelioid tumor (n = 91) and those with a mixed sarcomatoid tumor [nonepithelioid tumor] (n = 37); patients with a nonepithelioid tumor had significantly reduced survival compared to those with an epithelioid tumor. Source: Reproduced from Christensen BC, Godleski JJ, Roelofs CR, et al. Asbestos burden predicts survival in plur mesothelioma. Environmental Health Perspectives. 2008, Vol 116, p. 725.

0 10 20 30 40 50 60

Time (months)

Epithelioid Nonepithelioid

Censored

S ur

vi va

l F ra

ct io

0.0

0.2

0.4

0.8

1.0

0.6

51589_CH07_Printer.indd 334 11/02/13 2:12 PM

S a m p l i n g a n d C o h o r t f o r m a t i o n o p t i o n S 335

curves have numerous applications in research on infectious diseases, clinical trials, occurrence of psychiatric disorders, and many other situations. For more detailed information, consult Chernick and Friis.6

Cohort Studies The bases for forming a cohort are almost limitless with regard to the unifying feature, but the rationale for studying a particular cohort should be guided by the scientific question of interest, rather than mere availability of a group for study. Also known as a prospective or longitudinal study, a cohort study is distinguished from other observational research designs by the fact that it starts with a group of subjects who lack a positive history of the outcome of interest and are at risk for the outcome. From the standpoint of selecting study groups, cohort stud- ies can be thought of as going from cause to effect. That is, the exposure(s) of interest is (are) determined for each member of the cohort at baseline or time of study; then the group is followed through time to document the incidence of an outcome among the exposed and nonexposed members. Possible outcome mea- sures include the incidence of disease (cohort studies measure disease incidence directly), mortality, health status, and certain biological parameters, which are examined for changes that occur as a result of exposure to a risk factor.

In contrast to other observational designs, an additional characteristic of cohort studies is that they include at least two observation points: one to deter- mine exposure status and eligibility and a second (or more) to determine the number of incident cases that developed during follow-up. This feature (i.e., two or more observation points) permits the calculation of disease rates, which can- not be obtained with only a single time point of observation. In cohort studies, the individual forms the unit of observation and the unit of analysis, as is also true of cross-sectional studies and case-control studies. Cohort studies almost always involve the collection of primary data, although secondary data sources are used sometimes for both exposure and disease assessment.

Sampling and Cohort Formation Options

Although all cohort studies share certain common features, such as measurement of exposure before disease onset and at least two periods of observation, some types of cohort studies differ from one another. One of the differences is the sampling strategy used to define the cohort; these two strategies are population- based samples and exposure-based samples. We make note of these differences now because they have implications for the measures of association that we’ll learn to calculate with the data generated in a later chapter. These measures of

51589_CH07_Printer.indd 335 11/02/13 2:12 PM

336 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

association are called measures of effect (effect measures). An effect measure is “[a] quantity that measures the effect of a factor on the frequency or risk of a health outcome.”1 The point we wish to make here is that the population-based cohort studies enable estimation of more measures of association than is possible with some cohorts that are assembled based on exposures.

Population-Based Cohort Studies In population-based cohort studies, the cohort includes either an entire popula- tion or a representative sample of the population. As illustrated in Figure 7–5, a population-based cohort study starts with N; the total of exposed (A + B) and nonexposed (C + D) subjects is determined as part of the research process. Population-based cohorts have been used in studies of coronary heart disease (CHD). Perhaps the best known cohort study of CHD was initiated in 1948 in Framingham, Massachusetts.11 When the study commenced, the town had a pop- ulation of 28,000; the study design called for a random sample of 6,500 from the targeted age range of 30 to 59 years. This sample, representative of the population, was followed subsequently for changes in risk factors and incidence of disease.

An example of a cohort based on an entire population comes from the city of Tecumseh, Michigan, which was selected to examine the contribution of envi- ronmental and constitutional factors to the maintenance of health and the ori- gins of illness. Begun in 1959–1960, the Tecumseh study successfully enrolled 8,641 persons, 88% of the community residents.12 For some applications, cohort

Figure 7–5 Illustration of sample selection in a one-sample cohort study.

Yes

Yes No

A + B

Exposure Status

Disease Status

C + D

Total

51589_CH07_Printer.indd 336 11/02/13 2:12 PM

S a m p l i n g a n d C o h o r t f o r m a t i o n o p t i o n S 337

studies may be larger than a given community. The Iowa Women’s Health Study included a random sample of 41,83713 women between the ages of 55 and 69 in 1986 in the entire state.

The important point to remember is that these cohorts—Framingham, Tecumseh, and Iowa—were not selected because of a particular exposure (or risk factor) for disease. In addition, the frequency of the exposures within the cohort was expected to be representative of the parent population from which the cohort was selected. Thus, in a population-based cohort study the proportion of the population exposed can be determined either directly (when the entire popu- lation has been selected) or indirectly (when a known fraction of the population has been selected). Let us explain how exposure could be determined.

With a population-based cohort, exposures are unknown until the first period of observation when exposure information is collected. For example, after admin- istration of questionnaires, collection of biologic samples, clinical examinations, or physiologic testing, the cohort can be divided into two or more exposure cate- gories as a result of what is learned from the subjects. For a simple dichotomy (i.e., exposed versus nonexposed), the nonexposed subjects become an internal com- parison group. As an example, one could take a sample of male college students and test blood samples for antibodies to human papilloma virus (HPV); some are exposed (test positive for the antibodies) and the rest are not exposed (no detect- able antibodies). Sometimes, exposure may be categorized along a continuum, called a continuous variable, which is a type of variable that has an infinite set of possible values within a specified range. (Blood pressure measurements represent a continuous variable, whereas the designation of exposed/ nonexposed forms a dichotomy that signifies a discrete variable.) For a continuous variable, such as certain dietary intake measures or blood pressure, one typically constructs mul- tiple levels of exposure. A statistical procedure is used to subdivide the exposure variable into quantiles, which are divisions of a distribution into equal, ordered subgroups1 (e.g., quartiles or quintiles). These subdivisions are then used to define the levels of exposure. Subjects in one of the extreme categories, such as the upper or lower quintile, serve as the comparison (or referent) category. The incidence rate among this referent category becomes the expected rate of disease occurrence.

One of the benefits of population-based cohort studies is that results are gen- eralizable. By definition, this type of cohort study uses either an entire population or a random sample of a population. Thus it is possible to define the demo- graphic and other characteristics of the parent population used in the research. Consequently researchers are aided in their ability to generalize the findings of the study to populations or groups that did not participate in the study.

51589_CH07_Printer.indd 337 11/02/13 2:12 PM

338 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

Exposure-Based Cohort Studies Exposure-based cohort studies permit investigations of exposures that are uncommon. They are a suitable alternative to population-based cohort studies (or case-control studies, for that matter), which are not efficient for rare expo- sures. This assertion can be illustrated best with an example. Suppose that experi- ments with animals provided evidence that exposure to lead causes long-term neurologic toxicity. The amounts of lead used were much greater than those found in most human exposures. Certain occupational groups, however, such as those involved in battery production, might have sufficient occupational expo- sure to incur significant health risks. Although one could consider using a case- control study, because the proportion of the population employed in the battery manufacturing industry is low, there would likely be few cases or controls with exposure in the study sample. An alternative approach might be to assemble an exposure-based cohort of employees in battery production factories, quantify levels of exposure using job titles and assignments, and determine incidence rates of neurotoxicity.

Except in certain circumstances, the use of special exposure groups to form study groups precludes determination of the proportion of the target population exposed; as a result this design limits the public health inferences (generaliza- tion to a larger population or to other groups). As shown in Figure 7–6, such designs typically involve select subgroups with known exposures, for example, an

Figure 7–6 Illustration of sample selection in a multisample cohort study.

Yes

Yes No Total

A + B

C + D

Exposure Status

Disease Status

51589_CH07_Printer.indd 338 11/02/13 2:12 PM

S a m p l i n g a n d C o h o r t f o r m a t i o n o p t i o n S 339

exposed cohort (A + B) and a nonexposed, comparison cohort (C + D), perhaps from a different industry. Thus, in most situations the cohort members are homogeneous with respect to exposure, and because of the selection procedure, the frequency of exposure in the population cannot be determined.

Sources of exposure-based cohorts Examples of exposure-based cohorts include:

● Prepaid medical care plans, such as Kaiser Permanente or Group Health of Puget Sound, keep detailed medical information about a potentially large number of readily accessible subjects and maintain regular contact for follow-up information. In this case, the cohort is defined by insurance membership and the exposure of interest may be some medical condition or test recorded in the medical record.

● Physicians, nurses, and other health professionals have been the focus of several cohort studies. Because these individuals typically belong to national organizations (e.g., the American Medical Association), they are often easier to follow over the long term than other occupational groups because it is important for them to retain their association with the orga- nization and they can be contacted through it. Their knowledge of disease makes them good respondents for surveys; they can be expected to report previous medical conditions and recent diagnoses reliably.

● Childhood cancer survivors are becoming increasingly common, due mainly to remarkable improvements in therapy. The Childhood Cancer Survivor Study14 represents the largest (>14,000 five-year survivors ini- tially diagnosed between 1970 and 1986 from 25 centers) and most exten- sively characterized cohort of childhood and adolescent cancer survivors in North America. It serves as a resource for addressing important issues, such as risk of second malignancies, endocrine and reproductive outcomes, car- diopulmonary complications, and psychosocial implications, among this unique and ever-growing population.

● Veterans, because of the benefits they receive from the U.S. government, usually remain in contact with the relevant agency, making long-term follow-up feasible.

● College graduates are a final example of special resource groups that have been investigated in several noteworthy epidemiologic studies. (Later in this chapter we will cover in detail findings from a longitudinal study of Harvard alumni.15)

51589_CH07_Printer.indd 339 11/02/13 2:12 PM

340 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

Comparison (nonexposed) groups for use in exposure-based cohort studies Exposure-based cohort studies entail the comparison of disease rates between exposed and nonexposed groups. Continuing with our earlier example of neuro- toxicity and lead exposure among workers in the battery manufacturing indus- try, there are essentially three options in defining the comparison group. First, it may be that certain workers within the industry have no exposure to lead. For example, sales staff, secretaries, and management may have significantly less (or more) exposure to lead than factory workers. Thus, an internal compari- son would be possible (exposed versus nonexposed) within the factory or fac- tories under investigation. However, consider the situation in which everyone in the factory or industry is exposed to lead, and a nonexposed group could not be identified. The second option would therefore be the construction of a nonexposed group from a separate industrial cohort, similar in demographics and geography to the battery factory, but without lead exposure. A third option would be to compare disease rates among workers in the battery manufacturing industry with available population rates. Because population rates are summary rates, however, perhaps only specific to age-, sex-, and race-defined subgroups, they may have limited utility in some situations.

Take the example of a cohort study of lung cancer among uranium ore miners. Careful attention is paid to collection of data on other exposures that contribute to lung cancer risk, including use of tobacco products and diet. In fact, in this population a high percentage of the cohort consists of current smokers. Popula- tion rates of lung cancer, however, are based on the entire population. They are not adjusted for smoking, nor are smoking status-specific rates available. In this situation, because of the smoking levels alone, comparison of lung cancer rates among the miners with population lung cancer rates would not be informative.

Outcome Measures in Cohort Studies In a previous section, we mentioned that although many cohort studies gather information on the incidence of disease as the principal outcome measure, sev- eral other types of outcomes may be assessed. Table 7–4 illustrates types of out- comes used for cohort studies and lists three categories of outcomes: discrete events, levels of disease markers, and changes in disease markers.16 Discrete events cover single events and multiple occurrences, an example of the former being death and the first occurrence of a disease such as cancer, and the latter referring to repeated occurrence of disease such as recurrent heart disease (heart

51589_CH07_Printer.indd 340 11/02/13 2:12 PM

t e m p o r a l d i f f e r e n C e S i n C o h o r t d e S i g n S 341

Table 7–4 Types of Outcomes for Cohort Studies

Discrete events

● Single events

● Mortality

● First occurrence of a disease or health-related outcome

● Incidence (density) ● Cumulative incidence (risk) ● Ratios (incidence density and cumulative incidence)

● Multiple occurrences:

● Of disease outcome ● Of transitions between states of health/disease ● Of transitions between functional states

Level of a marker for disease or state of health Change in a functional/physiologic/biochemical/anatomical marker for disease or health

● Rate of change

● Patterns of growth and/or decline ● “Tracking” of markers of disease/health

● Change in level with time (age)

Source: Adapted with permission from Tager IB. Outcomes in cohort studies. Epidemiologic Reviews. 1998, Vol 20, p. 16.

attacks) or strokes. Examples of discrete measures include the age-standardized annual death rate, annual age-specific death rates, and the cumulative incidence of disease of specific time intervals (e.g., 5 years).

Cohort studies that include multiple occurrences as outcomes involve the repeated assessment of these outcomes over time. The multiple occurrences can involve discrete events, as in the case of repeated heart attacks or changes in sta- tus of outcome markers. Research questions may address the association between changes in risk markers over time and health status, studies of the effects of aging on the natural history of disease, and transitions of health status, such as the shift to functional disability among some elderly persons.

Temporal Dif ferences in Cohort Designs

Although the basic feature of all cohort studies is measurement of exposure and follow-up for disease, there are several variations in cohort designs that depend on the timing of data collection on exposure and outcomes. These variations are prospective and retrospective cohort studies.

51589_CH07_Printer.indd 341 11/02/13 2:12 PM

342 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

Prospective Cohort Studies A prospective cohort study is purely prospective in nature and is characterized by determination of exposure levels at baseline (the present) and follow-up for occurrence of disease at some time in the future (Figure 7–7). The sampling strategy may be population-based or defined by a special exposure of interest. There are numerous advantages to prospective studies:

● Prospective cohort studies enable the investigator to collect data on exposures. The collection of exposure information at baseline may result in the most direct and specific test of the study hypothesis. Examples include assessment of diet, physical activity, alcohol use, occupation, coping skills, and quality of life, each of which can be assessed with a few specific items or a comprehensive battery of items.

● The size of the cohort to be recruited is under greater control by the inves- tigators than is the size of a retrospective cohort (see next section). Cohort studies that rely on historical records are sometimes fixed in size.

● Biological and physiological assays can be performed with decreased con- cern that the outcome will be affected by the underlying disease process. Examples include measures of serum factors or nutrient levels and medical examinations (e.g., specific functional tests, antibody titers, or cholesterol levels).

● Direct measures of the environment (e.g., indoor radon levels, electro- magnetic field radiation, cigarette smoke concentration, or chlorination byproducts in the water supply) can be made in order to define exposures precisely.

Figure 7–7 Cohort design options on timing of data collection. E, exposure; D, disease.

Design Past Present Future

Prospective

Retrospective

Historical prospective D

51589_CH07_Printer.indd 342 11/02/13 2:12 PM

t e m p o r a l d i f f e r e n C e S i n C o h o r t d e S i g n S 343

Retrospective Cohort Studies Despite the substantial benefits of prospective cohort studies, investigators must wait for cases to accrue while conducting such a study. Depending upon the size of the cohort and the prevalence of a disease in the population, several years could elapse before meaningful analyses are feasible. An alternative is a retrospec- tive cohort study that makes use of historical data to determine exposure level at some baseline in the past; “follow-up” for subsequent occurrences of disease between baseline and the present is performed. A design that makes use of both retrospective (to determine baseline exposure) and prospective (to determine dis- ease incidence in the future) features is the historical prospective cohort study (also known as an ambispective cohort study).

There are several advantages to retrospective cohort studies:

● In a relatively short period of time, a significant amount of follow-up data may be accrued. For example, Sellers et al.17 performed a follow-up begin- ning in 1991 of a cohort of 426 families originally ascertained between 1944 and 1952 at the Dight Institute of Genetics at the University of Minnesota. Three-generation pedigrees were constructed at baseline with data collected from mothers, aunts, sisters, and daughters of breast can- cer patients. Records of breastfeeding, reproductive history, and validated occurrences of cancer were stored. Thus, when the family members were recontacted and interviewed regarding subsequent occurrences of cancer, almost 50 years of follow-up was completed during a 5-year period of funding.

● The amount of exposure data collected can be quite extensive and can be available to the investigator at minimal cost. For example, Hartmann and colleagues18 at the Mayo Clinic, using an index of surgical procedures, were able to construct a cohort of women who received prophylactic mas- tectomy between 1963 and 1986. Details were available in the medical record on type of surgery, age at surgery, family history of cancer, and complications following surgery. Follow-up through 1997 was performed to identify subsequent occurrences of breast cancer. Analyses based on a median 14 years of follow-up were possible, even though the actual study took less than 5 years to complete.

Many beginning epidemiology students find the distinction between case-control studies and retrospective cohort studies difficult to grasp. The nuances may be subtle but are noteworthy. Recall that case-control studies begin with ascertainment of study subjects on the basis of disease status. Data are then

51589_CH07_Printer.indd 343 11/02/13 2:12 PM

344 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

collected regarding exposures that occurred prior to disease onset. Because there is only one time point of observation, there is no longitudinal component and disease rates cannot be computed. Retrospective cohort studies begin with expo- sure, although these measurements occurred some time in the past. The sub- sequent occurrence of disease, perhaps supplemented with additional exposure assessments, is the primary focus of research activity. A retrospective cohort study incorporates an entire cohort of study subjects, whereas a case-control study involves identified cases and controls only.

Practical Considerat ions

Given the considerable advantages of cohort studies over other observational study designs, some readers may wonder why cohort studies are not the only designs used. In this next section, we provide a brief answer under the unify- ing theme of practical considerations. Such considerations include availability of exposure data, size and cost of the cohort used, data collection and data manage- ment, follow-up issues, and sufficiency of scientific justification.

Availability of Exposure Data Although development of prospective cohort studies may leverage data collected for other reasons besides the cohort study itself, the quality and extent of his- torical exposure data are absolutely crucial for retrospective cohort studies. In most situations, investigators will find themselves trying to weigh the trade-offs between a retrospective study design—with its associated benefits of more imme- diate follow-up time—versus the value of collecting the primary exposure data in the most ideal manner in a prospective cohort design. There are no simple rules to guide this decision, which must be carefully evaluated for each particular research question.

Size and Cost of the Cohort From a scientific standpoint there is little question that the larger the size of the cohort, the greater the opportunity to obtain answers in a timely manner. For a fixed rate of disease or outcome, only by increasing the denominator can the number of cases be increased during an interval. For example, suppose a form of cancer has an incidence rate of 10 cases per 100,000 per year. A total of only 50 cases would be expected among a cohort of 50,000 persons during a 10-year observation period. By doubling the size of the cohort to 100,000, one

51589_CH07_Printer.indd 344 11/02/13 2:12 PM

p r a C t i C a l C o n S i d e r a t i o n S 345

would expect 100 cases to occur. This larger number might be more appropriate for some types of statistical analyses. Of course, there is a direct relationship between size and cost, and resource constraints typically influence design deci- sions. One approach to design cohorts with the greatest future value is to focus initial development on the collection of risk factor data and biological samples, with subsequent (future) or parallel (nested studies) grants to obtain funding for analyses of the samples. For example, Zheng and colleagues19 obtained funding to establish the Shanghai Women’s Cohort Study of 75,000 women by limiting activities during the first 5 years to collection of diet and risk factor data, blood samples, tissue blocks, and urine samples. To keep costs within funding lim- its, initial aims did not focus on assays of the biological specimens. Subsequent renewal of the funding for that study was obtained to begin testing hypotheses related to hormone metabolism (urine), growth factor levels (serum), and genetic polymorphisms (DNA from lymphocytes in the blood).

Data Collection and Data Management An axiom of epidemiologic research design is that larger studies necessarily are more demanding than smaller ones with regard to challenges in data collection and data management. Coordination of activities in the field is especially com- plex when multiple sites are necessary for recruitment. Additional challenges may arise from data entry, especially if individual sites enter their own data for transmittal to the coordinating center. In these situations, explicit protocols for quality control (e.g., double entry of data and scannable forms) should be con- sidered in the design and implementation stage. The organizational and admin- istrative burdens are increased even further when there are multiple levels of data collection (such as telephone interviews, mail-out questionnaires, consent forms to access medical records, and collection of biological samples) at multiple time periods (especially when active follow-up is needed). Cohort study research pro- tocols may require elaborate data-management systems to monitor the status of the various components of data collection. Such management systems support individuals in the field who are charged with the multifaceted components as well as study managers and investigators who monitor overall progress. Manage- ment of data from cohort studies can be incredibly challenging and should be considered when staffing needs are being defined. Challenges arise from data collected from multiple sources, merging of files, and “cleaning” of data files. Cleaning is necessary in the case of missing values, out-of-range entries, and inconsistencies (e.g., someone responds that they never smoke but subsequently

51589_CH07_Printer.indd 345 11/02/13 2:12 PM

346 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

report a habit of 2 packs per day). These issues are not peculiar to cohort studies, because cross-sectional surveys and case-control studies can be quite large, too. However, the relative inefficiency of cohort studies for investigation of rare dis- eases means that they are typically larger than other designs.

Follow-up Issues The value of cohort studies can be realized only if an effective system can be implemented to follow the cohort for subsequent occurrence of disease or other outcomes. It may be helpful at this time to distinguish between active versus pas- sive follow-up.

Active follow-up denotes the situation in which the investigator, through direct contact with the cohort, must obtain data on subsequent incidence of the outcome (disease, change in risk factor, change in biological marker). Such contact may be accomplished through follow-up mailings, phone calls, or writ- ten invitations to return to study sites/centers for subsequent medical evalua- tion and, in some cases, biospecimen collection. Follow-up requires a substantial amount of effort, especially for large cohorts. For instance, the Minnesota Breast Cancer Family Study20 followed up on study participants with a mailed survey, a reminder postcard 30 days later, a second survey, and a telephone call to non- responders. In the present era of telephone technology—answering machines, caller identification, automated telemarketing solicitations, and widespread use of cellular telephones—active follow-up is becoming increasingly labor inten- sive and oftentimes frustrating, especially in our mobile society where addresses change often. For some cohort studies, however, a persistent and labor-intensive effort is the only option for follow-up.

Contrast the foregoing scenario with passive follow-up, which does not require direct contact with cohort members. Passive follow-up is possible when databases containing the outcomes of interest are collected and maintained by organizations outside the investigative team. Epidemiologists sometimes are able to achieve record linkage between databases and the study cohort. An excellent example amenable to passive follow-up is cancer, for which the federal and many state governments mandate reporting. The Iowa Women’s Health Study13 is able to conduct follow-up for cancer incidence within the state in this manner.

Passive follow-up is clearly not an option for many diseases. The only end- point for which there is universal coverage is death; the National Center for Health Statistics collects mortality data, which are made available to the scien- tific community through the National Death Index (NDI). The success of link- age to the NDI depends upon the extent of demographic information collected

51589_CH07_Printer.indd 346 11/02/13 2:12 PM

m e a S u r e S o f e f f e C t 347

on participants. Social Security numbers are the most effective data element for linkage. If Social Security numbers are unavailable, record linkage sometimes is possible for persons who have uncommon last names.

Sufficiency of Scientific Justification The preceding sections emphasized that the establishment of a cohort study requires major investments in resources (time, money, and energy). Thus, there should be considerable scientific rationale for a cohort study. This rationale should be grounded on prior research from various perspectives: study designs other than cohort studies; several different investigators; and several different study populations. Additional justification for cohort studies may come from laboratory experiments or animal studies. Furthermore, the situation in which investigators would like to explore more than one outcome from a particular exposure provides one of the greatest justifications for a cohort study. Because of the nature of case-control studies, only a single outcome can be investigated at a time.

Cohort studies are the only observational study design that permits examina- tion of multiple outcomes at the level of the individual within a single study. Consider as an example the complex issue of hormone replacement therapy (HRT). Numerous epidemiologic studies have noted that HRT is associated with a slight, but detectable, increased risk of breast cancer. In fact, recommen- dations have been published in leading medical journals for the avoidance of HRT among women at elevated risk of breast cancer because of a family his- tory.19 Complicating such a recommendation is the observation from other epi- demiologic studies that have associated HRT with a number of health benefits, including lower risks of CHD,21,22 osteoporosis,23,24 and Alzheimer’s disease (AD).25,26 The pros and cons of HRT use can be examined within an appropri- ately designed cohort study27 with measures of HRT collected at baseline and a follow-up protocol designed to collect data on the multiple outcomes of interest.

Measures of Effect : Their Interpretat ion and Examples

In the simplest case of two levels of exposure (yes/no), two incidence rates are calculated. The relative risk is defined as the ratio of the risk of disease or death among the exposed to the risk among the unexposed.1 Recall that risk is esti- mated in epidemiologic studies only by the cumulative incidence. When the

51589_CH07_Printer.indd 347 11/02/13 2:12 PM

348 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

relative risk is calculated with incidence rates or incidence density, then the term rate ratio is more precise.

Relative risk Incidence rate in the exposed

In =

ccidence rate in the nonexposed

Using the notation from the 2 by 2 table (Figure 7–8), the relative risk can be expressed as [A/(A + B)] ÷ [C/(C + D)]. A sample calculation is shown in Exhibit 7–2.

Some comments regarding interpretation of relative risk are in order. A rela- tive risk of 1.0 implies that the risk (rate) of disease among the exposed is no different from the risk of disease among the nonexposed. A relative risk of 2.0 implies that risk is twice as high, whereas a relative risk of 0.5 indicates that the exposure of interest is associated with half the risk of disease.

Examples of Cohort Studies Table 7–5 presents examples of major cohort studies, including when they were initiated, their main focus, study population, measurement of exposure at base- line and follow-up, and frequency of exposure follow-up. More detailed coverage of some of these studies is provided in this chapter and throughout the text. Exhibit 7–3 showcases four cohort studies that concern women’s health.

The remainder of this section provides examples of cohort studies built upon a common research theme: physical activity and CHD. The idea that physical activity is beneficial to humans is certainly not a new one. In The Dialogues,

Figure 7–8 The 2 by 2 table.

Yes

Yes No Total

A + B

C + D

A + C B + D

Exposure Status

Disease Status

51589_CH07_Printer.indd 348 11/02/13 2:12 PM

m e a S u r e S o f e f f e C t 349

Sample problem: relative risk

The following is an example of how to use a fourfold table for cal- culation of relative risk. Deykin and Buka28 studied suicidal ideation and attempts in a population of chemically dependent adoles- cents believed by the researchers to be a group at high risk for self- destructive behavior. Boys who had been exposed to physical or sexual abuse and who stated that a report of abuse or neglect had

been filed with authorities were more likely than boys who were absent such exposures to have made a suicide attempt. The data for history of sexual abuse among the boys are charted below:

History of Sexual Abuse Suicide Attempt No Suicide Attempt Totals

Yes No

A = 14 C = 49

B = 9 D = 149

A + B = 23 C + D = 198

Relative risk (14/23)/(49/198) = 0.609/0.247 = 2.46 n

Source: Data from Deykin EY, Buka SL. 1994. Suicidal ideation and attempts among chemically dependent adolescents. American Journal of Public Health. Vol 84, pp. 634–639; American Public Health Association, © 1994. The fourfold table was constructed by the authors from these data and from the percentage of suicide attempts reported for boys with a positive history and boys with a negative history of sexual abuse. Deykin and Buka reported 60.9% and 24.8% attempting suicide, respectively, and a relative risk of 2.4.

e X

h IB

It 7

–2

Timeus tells Socrates that, “Moderate exercise reduces to order, according to their affinities, the particles and affections which are wandering about the body.”29(p 405) Two thousand years later, experts still are unable to agree on how much or how often exercise is needed or whether habitual physical activ- ity (as opposed to exercise) is sufficient to maintain one’s health. The rediscov- ery of the potential importance of physical activity was spurred by a landmark study of British transportation workers by Morris and colleagues in 1953.30 Hoping to uncover “social factors which may be favourable or unfavourable to its occurrence,”30(p 1053) the investigators conducted a study of roughly 31,000 men aged 35 to 64. Rates of angina pectoris, coronary thrombosis, and sudden death were obtained for drivers, conductors, and underground railway work- ers. Conductors had jobs that were physically more demanding than those of the other two groups and, consequently, experienced significantly lower rates of CHD. The proposition that physical activity might be protective generated

51589_CH07_Printer.indd 349 11/02/13 2:12 PM

350 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

T ab

le 7

–5

Ex am

pl es

o f M

aj or

C oh

or t

St ud

ie s

St ud

y [Y

ea r

B eg

un ]

M ai

n Fo

cu s

St ud

y P

op ul

at io

[S am

pl e

Si ze

] B

as el

in e

Ex po

su re

st ru

m en

ts *

Fo llo

w -u

p Ex

po su

In st

ru m

en ts

* Fr

eq ue

nc y

of E

xp os

ur e

Fo llo

w -u

Th e

Fr am

in gh

am H

ea rt

ud y

[1 94

8] R

is k

fa ct

or s

fo r

ca rd

io va

sc ul

ar d

is ea

se R

es id

en ts

o f

Fr am

in gh

am , M

A , a

ge s

28 –6

2 ye

ar s

[n =

5 ,2

09 ]

In te

rv ie

w , c

lin ic

am in

at io

n (P

E, la

te st

In te

rv ie

w , c

lin ic

am in

at io

n (P

E, la

te st

Ev er

y 2

ye ar

C ol

or ad

o Pl

at ea

u U

ra ni

um M

in er

s St

ud y

[ 19

50 ]

O cc

up at

io na

l r is

k fa

ct or

s fo

r ca

nc er

W hi

te , m

al e

un de

rg ro

un d

ur an

iu m

in er

s, C

ol or

ad o

Pl at

ea u

= 3

,4 15

]

C lin

ic e

xa m

in at

io n

(P E)

, q ue

st io

nn ai

re ,

en vi

ro nm

en ta

l m ea

su re

s (a

irb or

ne r

ad ia

tio n)

Q ue

st io

nn ai

(i n-

pe rs

on o

r m

ai le

d) ,

en vi

ro nm

en ta

l m

ea su

re s

(a irb

or ne

di at

io n)

Tr ie

nn ia

A m

er ic

an C

an ce

r So

ci et

C an

ce r

Pr ev

en tio

n St

ud y

1 [1

95 9]

C ig

ar et

te s

m ok

in g

an d

ca nc

er m

or ta

lit y

U .S

. m en

a nd

w om

ag ed

3 0

ye ar

s an

d ol

de r

[n =

1 ,0

45 ,0

87 ]

Se lf-

ad m

in is

te re

d qu

es tio

nn ai

re d

el iv

er ed

v ol

un te

er s

Se lf-

ad m

in is

te re

d qu

es tio

nn ai

re Ev

er y

2 ye

ar s

Th e

A la

m ed

a C

ou nt

y St

ud y

[1 96

5] Fa

ct or

s as

so ci

at ed

w ith

al th

a nd

m or

ta lit

y R

es id

en ts

o f A

la m

ed a

C ou

nt y,

C A

, a ge

s 16

–9 4

ye ar

s [n

= 6

,9 28

]

M ai

le d

qu es

tio nn

ai re

, te

le ph

on e

in te

rv ie

or h

om e

in te

rv ie

w o

f no

nr es

po nd

en ts

M ai

le d

qu es

tio nn

ai re

; te

le ph

on e

in te

rv ie

or h

om e

in te

rv ie

w o

f no

nr es

po nd

en ts

A t

ye ar

s 9,

1 8,

a nd

2 9

H on

ol ul

u H

ea rt

P ro

gr am

[ 19

65 ]

C or

on ar

y he

ar t

di se

as e

an d

st ro

ke in

m en

o f

Ja pa

ne se

a nc

es tr

M en

o f J

ap an

es e

an ce

st ry

li vi

ng o

n O

ah u,

I, ag

es 4

5– 65

y ea

[n =

8 ,0

06 ]

M ai

le d

qu es

tio nn

ai re

, in

te rv

ie w

, c lin

ex am

in at

io n

(P E,

la b,

st s)

Th e

O ra

l C on

tr ac

ep tio

n St

ud y

of t

he R

oy al

ol le

ge o

f G en

er al

ac tit

io ne

rs [

19 68

]

O ra

l c on

tr ac

ep tiv

e us

e an

d ca

nc er

M ar

rie d

pr em

en op

au sa

l B

rit is

h w

om en

= 4

7, 00

M ed

ic al

fo rm

c om

pl et

by p

hy si

ci an

b as

ed o

n pa

tie nt

in te

rv ie

w o

r m

ed ic

al r

ec or

Sa m

e Ev

er y

6 m

on th

N ur

se s’

H ea

lth S

tu dy

97 6]

O rig

in al

ly o

ra l

co nt

ra ce

pt iv

e us

e an

d ca

nc er

, e xp

an de

d to

om en

’s h

ea lth

M ar

rie d

fe m

al e

U .S

. re

gi st

er ed

n ur

se s,

a ge

s 30

–5 5

ye ar

[n =

1 21

,7 00

]

M ai

le d

qu es

tio nn

ai re

M ai

le d

qu es

tio nn

ai re

b (t

oe na

il sa

m pl

La b

(h om

e bl

oo d

dr aw

)

Ev er

y 2

ye ar

A t

ye ar

A t

ye ar

1 3

51589_CH07_Printer.indd 350 11/02/13 2:12 PM

m e a S u r e S o f e f f e C t 351

Po rt

P iri

e C

oh or

t St

ud y

[1 97

9] Le

ad e

xp os

ur e

an d

ch ild

ve lo

pm en

t In

fa nt

s bo

rn in

P or

t Pi

rie , S

ou th

A us

tr al

ia ,

19 79

–1 98

2 [n

= 7

23 ]

La b

(b lo

od s

am pl

fr om

p re

gn an

t m

ot he

r an

d um

bi lic

al c

or d

bi rt

La b

(b lo

od s

am pl

es )

A t

6, 15

, a nd

m on

th s,

a nn

ua lly

t o

7 ye

ar s,

a nd

a t

11 –1

3 ye

ar s

M ul

tic en

te r

A ID

S C

oh or

t St

ud y

[1 98

4] R

is k

fa ct

or s

fo r

H IV

ga y

m en

U .S

. h om

os ex

ua l m

en ,

ag es

1 8–

70 y

ea rs

= 4

,9 54

]

C lin

ic e

xa m

in at

io n

(P E,

b) , s

el f-

ad m

in is

te re

d qu

es tio

nn ai

re , i

nt er

vi ew

C lin

ic e

xa m

in at

io n

(P E,

la b)

, s el

f- ad

m in

- is

te re

d qu

es tio

nn ai

re ,

in te

rv ie

Ev er

y 6

m on

th s

C or

on ar

y A

rt er

y R

is k

D ev

el op

m en

t in

Y ou

A du

lts (

C A

R D

IA )

[1 98

R is

k fa

ct or

s fo

r co

ro na

he ar

t di

se as

e in

y ou

ad ul

B la

ck a

nd w

hi te

U .S

. m

en a

nd w

om en

, a ge

s 18

–3 0

ye ar

s [n

= 5

,1 15

]

Te le

ph on

e in

te rv

ie w

, cl

in ic

e xa

m in

at io

n (P

la b,

t es

ts ),

s el

f- ad

m in

- is

te re

d qu

es tio

nn ai

re ,

in te

rv ie

C lin

ic e

xa m

in at

io n

(P E,

la b,

t es

ts ),

s el

f- ad

m in

is te

re d

qu es

tio n-

na ir

e, in

te rv

ie w

A t

ye ar

s 2,

5 , a

nd 7

N ew

Y or

k U

ni ve

rs ity

om en

’s H

ea lth

S tu

[1 98

En do

ge no

us h

or m

on es

d ris

k of

b re

as t

ca nc

er N

ew Y

or k

w om

en , a

ge s

35 –6

5 ye

ar s

= 1

4, 29

La b,

s el

f- ad

m in

is te

re d

qu es

tio nn

ai re

La b

A nn

ua l

Io w

a W

om en

’s [

H ea

lth ]

St ud

y [1

98 6]

C an

ce r

in w

om en

Io w

a w

om en

, a ge

s 55

–6 9

ye ar

s [n

= 4

1, 83

7] M

ai le

d qu

es tio

nn ai

w ith

s el

f b od

y gi

rt h

m ea

su re

N on

St ud

y of

O st

eo po

ro tic

ac tu

re s

[1 98

6] R

is k

fa ct

or s

fo r

fr ac

tu re

s N

on bl

ac k

U .S

. w om

en ,

ag es

6 5

ye ar

s an

d ol

de r

[n =

9 ,7

04 ]

Q ue

st io

nn ai

re , c

lin ic

am in

at io

n (P

E, t

es ts

ab st

ra ct

io n

(m ed

ic at

io n

la be

ls )

Q ue

st io

nn ai

C lin

ic e

xa m

in at

io n

(P E,

t es

ts )

Ye ar

C ar

di ov

as cu

la r

H ea

lth

St ud

y (C

H S)

[ 19

89 ]

R is

k fa

ct or

s fo

r ca

rd io

- va

sc ul

ar d

is ea

se in

o ld

ad ul

U .S

. m en

a nd

w om

en ,

ag es

6 5

ye ar

s an

d ol

de r

[n =

5 ,2

01 ]

H om

e in

te rv

ie w

, c lin

ex am

in at

io n

(P E,

la b,

st s)

, q ue

st io

nn ai

re ,

ab st

ra ct

io n

(m ed

ic at

io n

la be

ls )

C lin

ic e

xa m

in at

io n

(P E,

t es

ts ),

es tio

nn ai

r e ,

ab st

r a ct

io n

(m ed

ic at

io n

la be

ls )

C lin

ic e

xa m

in at

io n

(l ab

) an

d al

l o f a

bo ve

A nn

ua l

Tr ie

nn ia

co nt

in ue

51589_CH07_Printer.indd 351 11/02/13 2:12 PM

352 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

St ud

y [Y

ea r

B eg

un ]

M ai

n Fo

cu s

St ud

y P

op ul

at io

[S am

pl e

Si ze

] B

as el

in e

Ex po

su re

st ru

m en

ts *

Fo llo

w -u

p Ex

po su

In st

ru m

en ts

* Fr

eq ue

nc y

of E

xp os

ur e

Fo llo

w -u

W om

en ’s

H ea

lth In

iti at

iv e

(W H

I) O

bs er

va tio

na l

St ud

y [1

99 4]

W om

en ’s

h ea

lth U

.S . w

om en

, ag

es 5

0– 79

y ea

[n =

1 00

,0 00

]

Te le

ph on

e in

te rv

ie w

, cl

in ic

e xa

m in

at io

n (P

la b,

t es

ts ),

in te

rv ie

w ,

se lf-

ad m

in is

te re

d qu

es tio

nn ai

re ,

ab st

ra ct

io n

(m ed

ic at

io n

la be

ls )

M ai

le d

qu es

tio nn

ai re

lin ic

e xa

m in

at io

n (P

E, la

b, t

es ts

ab st

ra ct

io n

(m ed

ic at

io n

la be

ls ),

lf- ad

m in

is te

re d

qu es

tio nn

ai re

A nn

ua l a

t ye

ar 3

* E xp

os ur

e in

st ru

m en

ts d

ef in

iti on

s: P

E, p

hy si

ca l e

xa m

in at

io n

or p

hy si

ca l m

ea su

re s,

s uc

h as

a nt

hr op

om et

ric s,

s tr

en gt

h, e

tc .;

la b,

m ea

su re

s in

b lo

od a

nd o

th er

ec im

en s;

t es

ts , m

ed ic

al t

es ts

, s uc

h as

e le

ct ro

ca rd

io gr

ap hy

, t re

ad m

ill , b

on e

m in

er al

d en

si ty

, e tc

.; ab

st ra

ct io

n, a

bs tr

ac tio

n of

m ed

ic at

io n

in fo

rm at

io n

fr om

ed ic

at io

n la

be ls

, m ed

ic al

r ec

or d

ab st

ra ct

io n;

e nv

ir on

m en

ta l m

ea su

re s,

m ea

su re

s in

t he

e nv

ir on

m en

t (a

ir or

w at

er );

q ue

st io

nn ai

re , e

ith er

s el

f- ad

m in

is te

re d

in te

rv ie

w er

-a dm

in is

te re

d (n

ot e

xp lic

itl y

st at

ed );

a nd

in te

rv ie

w , i

nt er

vi ew

er -a

dm in

is te

re d

qu es

tio nn

ai re

So ur

ce :

A da

pt ed

w it

h pe

rm is

si on

f ro

m W

hi te

E ,

H un

t JR

, C

as so

D .

Ex po

su re

m ea

su re

m en

t in

c oh

or t

st ud

ie s:

t he

c ha

lle ng

es o

f pr

os pe

ct iv

e da

ta c

ol le

ct io

Ep id

em io

lo gi

c R ev

ie ws

, V ol

2 0,

N o

1, p

p. 4

4– 45

, 1 99

T ab

le 7

–5

co nt

in ue

51589_CH07_Printer.indd 352 11/02/13 2:12 PM

m e a S u r e S o f e f f e C t 353

Four cohort Studies that Investigate Women’s health

Nurses’ Health Study: Affiliated investigators, housed at Brigham and Women’s Hospital in Boston, mailed questionnaires beginning in 1976 to 170,000 nurses who resided in the 11 most populous U.S. states. Nurses (approximately n = 122,000 responding) were selected because they were knowledgeable about the technically

worded questionnaire items and motivated to remain over the long term in the cohort study. The aim of Frank Speizer, the study’s originator, was to examine the long-term potential consequences of use of oral contra- ceptives. Mailed every 2 years, the original questionnaire searched for the occurrence of various diseases and also probed health-related topics, such as smoking, hormone use, and menopausal issues. A later phase of the project initiated in 1980 was expanded to include diet and quality of life topics. A subset of the cohort submitted toenail samples (used for min- eral analyses) and blood samples, employed in studies of biomarkers. In 1989, Dr. Walter Willett started the Nurses’ Health Study II, which focused on oral contraceptive use, diet, and lifestyle risk factors in a population younger that the original Nurses’ Health Study cohort. The Nurses’ Health Study has generated an impressive list of scientific publications and many fundamental contributions to the store of health knowledge.1

Women’s Health Initiative (WHI): The WHI, begun in 1991, concen- trates on the major causes of death, disability, and frailty among post- menopausal women: coronary heart disease, breast and colorectal cancer, and osteoporotic fractures. Two major goals of the WHI are to provide estimates of the extent to which known risk factors predict heart disease, cancers, and fractures, and to identify new risk factors for these and other diseases in women. One of the largest preventive studies of its kind in the United States, the WHI lasted 15 years. The WHI encompassed three major components: a randomized clinical trial for disease prevention, a study of community approaches to developing healthful behaviors, and an observational study (OS). The OS followed more than 93,000 postmeno- pausal women between the ages of 50 to 79 over an average of 9 years. The respondents completed periodic health forms and visited a clinic 3 years

e X

h IB

It 7

–3

continues

51589_CH07_Printer.indd 353 11/02/13 2:12 PM

354 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

after enrollment.2 The data and stored biological specimens collected from study participants are expected to serve as a resource for continued analyses. The WHI Extension Study currently is funded through 2010.3

Study of Osteoporotic Fractures: From September 1986 to October 1987, this prospective cohort study enrolled 9,704 women aged 65 years and older. Subjects were selected from the rosters of four clinical cen- ters: the Kaiser-Permanente Center for Health Research, Portland, OR; the University of Minnesota, Minneapolis; the University of Maryland, Baltimore; and the University of Pittsburgh. The University of California, San Francisco, acted as the research-coordinating center. Investigators administered questionnaires, interviews, and examinations to obtain information on anthropometric characteristics, estrogen use, and medical history. Follow-up measures included the incidence of fractures, validated by radiographic reports.4

Iowa Women’s Health Study: The Iowa Women’s Health Study (IWHS), started in 1986, is a cohort of 41,836 postmenopausal women aged 55–69 at baseline. The primary aims of the study were to:

1. Determine if the distribution of body fat (waist/hip) predicts incidence of chronic diseases, with the primary endpoints being total mortality and incident cancers of the breast, endometrium, and ovaries, and

2. Determine to what degree diet and other lifestyle factors influence risk of chronic disease.

Questionnaires at baseline and for five follow-up surveys (1987, 1989, 1992, 1997, and 2004) provided self-reported information on demograph- ics, reproductive history, medical history, hormone replacement therapy, dietary intake (FFQ), physical activity, and other factors.5 n

Sources: Adapted and reprinted from:

1. http://www.channing.Harvard.edu/nhs/. Accessed August 7, 2012. 2. http://www.nhlbi.nih.gov/whi/. Accessed August 7, 2012. 3. The Women’s Health Initiative, WHI Matters. 2006: 11;1. 4. Sellmeyer DE, Stone KL, Sebastian A, et al. A high ratio of dietary animal to vegetable

protein increases the rate of bone loss and the risk of fracture in postmenopausal women. American Journal of Clinical Nutrition. 2001;73:118–122.

5. Prevention and Etiology Research Program, Iowa Women’s Health Study. http://www .cancer.umn.edu/research/programs/peiowa.html. Accessed August 7, 2012.

exhibit 7–3 continued

51589_CH07_Printer.indd 354 11/02/13 2:12 PM

m e a S u r e S o f e f f e C t 355

tremendous interest and spurred numerous investigations to confirm and refine the hypothesis. Some examples of relevant studies follow.

The first example is a retrospective cohort study of railroad workers.31 The cohort included 191,609 railroad industry employees in the United States between the ages of 40 and 64.31 The “exposure” was work-related physical activ- ity. Based on job descriptions, three groups were formed: clerks, switchmen, and section men, with activity levels of low, moderate, and heavy, respectively. Any cohort study that entails long periods of follow-up raises a legitimate concern whether exposures at the baseline period change over time, resulting in exposure misclassification. One advantage of using the railroad industry for this study was that the labor contracts between management and the unions contained senior- ity provisions that prevented a man from carrying his seniority from one job to another job controlled by a separate union. Because seniority brings benefits in terms of privileges and income, job changes associated with a switch to a different labor contract were uncommon. Follow-up for CHD end points was accomplished through the Railroad Retirement Board, which maintained an account for each man employed by any interstate railroad in the United States. Because the retirement and disability benefits to members were greater than those received from Social Security, follow-up rates were high. Members’ deaths that were not detected by the board occurred almost exclusively among men who left the industry completely; the investigators estimated that this number was only 11 to 12 per 1,000 workers.

Average annual age-adjusted mortality rates of CHD per 1,000 men were calculated among men who had accumulated 10 years of service by the end of 1951 and were employed in 1954. Mortality rates per 1,000 were 5.7, 3.9, and 2.8 for clerks, switchmen, and section men, respectively. If one considers the rates among the sedentary clerks as the reference, then the relative risk of CHD death for the moderately active switchmen was 0.68, and 0.49 for the very active section men.

Although the study findings supported the hypothesis that physical activity is protective for CHD, there were some limitations inherent in the historical expo- sure data. In this particular situation, few data were available on other risk fac- tors that might underlie the observed association. For example, no information was available about smoking, body mass index, blood pressure, family history of CHD, and hypertension. Therefore, additional studies were warranted.

The second example is of an ambispective (or historical prospective) cohort study. Reported by Paffenbarger and colleagues in 1984,15 the study examined a history of athleticism and CHD in a cohort of male Harvard alumni from

51589_CH07_Printer.indd 355 11/02/13 2:12 PM

356 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

1916 to 1950 (N = 16,936). Exposure assessments occurred at two time periods: a historical measure of physical activity based on college archives of student health and athletics, and a questionnaire mailed in 1962 or 1966 for which the response rate was 70%. The alumni questionnaire assessed post-college physical exercise, other elements of lifestyle, health status, and histories of parental disease. The assessment of physical activity included the number of stairs climbed per day, the number of city blocks or equivalent walked each day, and sports actively played (in hours per week). Responses were used to estimate kilocalories of energy expen- diture per week. Follow-up of the cohort was achieved through questionnaires mailed in 1972 by the alumni office. The first questionnaire ascertained self- reports of physician-diagnosed CHD events. The second questionnaire mailed in the same year to the survivors of the deceased cohort members ascertained dates of death. Weekly updates of death lists by the alumni office provided the means to obtain death certificates for causes. To account for the varying amount of follow-up, person-years of observation were calculated. Heart attack rates were computed according to activity level as a student and as an alumnus. Men who participated in fewer than 5 hours per week of intramural sports as a student and fewer than 500 kilocalories per week of leisure time activity as an alumnus were designated as the reference group (85.5 per 10,000 person-years). Men who were college varsity athletes but became sedentary as adults had the same rate of first CHD attacks (relative risk, 1.0) as the reference group. In contrast, men who were active as adults (2,000+ kilocalories per week) had a much-reduced risk of CHD, regardless of whether they had been sedentary or active as students.

The third and final example study is purely prospective in nature. Peters et al.32 designed a study to address one of the major concerns of the previous body of literature on the subject of physical activity (occupational or leisure). Most occupational physical activity is not of sufficient intensity or duration to affect cardiorespiratory fitness. Exercise physiologists argued that physical fit- ness, not physical activity, was the appropriate and relevant exposure. A cohort of 2,779 firefighters and police officers in Los Angeles County between the ages of 35 and 55 was established. In contrast to the two previous studies, which based exposure levels on job title or questionnaires, level of physical work capacity was determined by use of a bicycle ergometer. The cohort was divided into two expo- sure categories (low and high physical work capacity) based on a median split (the 50% point in a distribution). Measurements were taken also on a number of other risk factors, including blood pressure, relative weight, family history, cholesterol, and skinfold measurements. Follow-up for heart attacks was accom- plished using county workers’ compensation files, death certificates, and medical

51589_CH07_Printer.indd 356 11/02/13 2:12 PM

m e a S u r e S o f e f f e C t 357

records; person-years of observation were tabulated. Because heart attacks and expensive hospital stays were fully reimbursable by insurance, coverage was thought to be complete. Analyses were performed to control for smoking, obesity, blood pressure, cholesterol, family history, relative weight, and physical activity. Results suggested that the least physically fit had more than a twofold greater risk of heart attacks than the most physically fit. Risk was especially prominent (relative risk, 6.6) for those who had at least two additional risk factors (smoking, high serum cholesterol, or high blood pressure).

Nested Case-Control Studies Although this section would seem to be more logically placed within a chapter on case-control studies, an understanding of nested case-control studies requires an understanding of cohort studies. A nested case-control study is defined as a type of case-control study “. . . in which cases and controls are drawn from the popu- lation in a cohort study.”1 For example, suppose we have data from an ongoing cohort study of the relationship between use of birth control pills and breast cancer. The population of the cohort study would comprise both exposed and nonexposed persons; the former and latter would consist of women who do and do not take birth control pills, respectively. To perform a nested case-control study, the investigator would select a subset of the population from the cohort study; this subset would comprise the controls. The cases of breast cancer identi- fied from the cohort study would comprise the cases in the case-control study.

What are the advantages of a nested case-control study? This design provides a degree of control over confounding factors, because relevant exposure informa- tion and other data have been collected during the course of the cohort study. Confounding factors mask an association between an exposure and an outcome because of the influence of a third variable that was not considered in the study design. In comparison with standard case-control studies, nested case-control designs enable the investigator to differentiate more clearly between a hypoth- esized study-related exposure and extraneous (confounding) exposures. Another advantage of nested case-control studies is the reduced cost of collecting detailed exposure information from a subset of the cohort only; this procedure obviously would be less costly than obtaining information from every single person in the cohort.

An example of a nested case-control study is an investigation of suicide among electric utility workers. The study examined the association between exposure to extremely low-frequency magnetic fields and suicide.33 Cases (536 deaths

51589_CH07_Printer.indd 357 11/02/13 2:12 PM

358 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

from suicide) and controls (n = 5,248) were selected from a cohort of 138,905 male utility workers. Findings supported an association between occupational exposure to electromagnetic fields and suicide.

A second example comes from Dearden et al.,34 who conducted a case-control study that aimed to identify factors associated with teen fatherhood. Data came from the National Child Development Study, a longitudinal investigation of all children born in Great Britain between March 3 and March 9, 1958. This type of design, in which a subset of members from a larger cohort study is selected for analysis, was a nested case-control study. Data were collected at birth and at 7, 11, 16, and 23 years of age35; information about fatherhood status was avail- able on 5,997 males. Cases were defined as teens who had fathered a child before their 20th birthday (n = 209). Controls, selected from the same follow-up group, were divided into two groups: 844 nonteen fathers (those who became fathers between the ages of 20 and 23), and 798 nonfathers (those who had no children by age 23). Teen fathers were found to be three times as likely as nonfathers to engage in lawbreaking behavior and to be absent from school, three to four times as likely to show signs of aggression, and eight times as likely to leave school at age 16.

Summary of Cohort Studies

Cohort studies have several clearly identified strengths. The cohort study is the first observational study design covered that permits direct determination of risk. Because one starts with disease-free subjects, this design provides stronger evidence of an exposure–disease association than the case-control scheme. In addition, cohort studies provide evidence about lag time between exposure and disease. In comparison with case-control studies, which have a greater potential for sampling errors, cohort studies (especially population-based studies) facilitate generalization of findings. A tremendous advantage of cohort studies is that, if properly designed and executed, they allow examination of multiple outcomes. While case-control studies may not be efficient for exposures that are rare in the population, cohort studies are able to increase the efficiency for rare exposures by selecting cohorts with known exposures (such as certain occupational groups).

The main limitation of cohort studies, at least for those that are purely pro- spective, is that they take considerable effort to conduct. Because they almost always use larger sample sizes than case-control studies, more time is required to collect the exposure information. Additional time passes while one waits for

51589_CH07_Printer.indd 358 11/02/13 2:12 PM

C o n C l u S i o n 359

the outcomes to occur. The amount of time required to accumulate sufficient end points for meaningful analysis can be reduced by increasing the size of the cohort, but this increase has to be balanced against the longer time to assemble and measure the cohort as well as the increased financial costs. Given the large size of cohort studies and the need for multiple observation points, they are more difficult to implement and carry out than other observational designs, especially for rare diseases. Loss to follow-up can be a significant problem, limiting the sample size for analysis and raising questions about the results if loss is too high. With long-term follow-up, some exposures may change over time. This misclas- sification of exposure would attenuate the estimates of the relative risk. It is even conceivable that participation in the study itself may lead to changes in exposure. For example, suppose the investigator recruits a cohort to study the association of dietary fat and disease. As a result of participation, subjects’ motivation to learn more about the hypothesis may subsequently lead to adoption of a low-fat diet. Ethical issues arise if good data already indicate that a particular exposure is harmful and one does nothing to intervene with at-risk subjects. Despite these limitations, the cohort study design is an important and valuable tool. For an in-depth coverage of cohort studies, refer to Samet and Munoz.36 In conclusion, the major types of observational study designs used in epidemiology are cross- sectional studies, case-control studies, and cohort studies. For the convenience of readers, Table 7–6 summarizes these designs by comparing their characteristics, advantages, and disadvantages.

Conclusion

This chapter has covered the cohort study, one of the most powerful epidemio- logic study designs. We noted that cohort studies overcome many of the prob- lems associated with temporality of data collection and rare exposures. Both the term cohort and the method of cohort analysis were defined. We provided sev- eral examples of cohort analyses, including Frost’s tabular data on tuberculosis mortality. Related to cohort analyses are life tables and survival curves, which, respectively, estimate and graphically portray survival (often of patients in clini- cal trials) over time.

The remainder of the chapter focused on methods associated with cohort studies. Population-based cohort studies were distinguished from exposure-based cohort studies. We also presented methods for the selection of comparison groups in cohort studies. Other issues included types of outcome measures, temporal

51589_CH07_Printer.indd 359 11/02/13 2:12 PM

360 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

T ab

le 7

–6

C om

pa ri

so n

of O

bs er

va ti

on al

S tu

dy D

es ig

St ud

y A

tt ri

bu te

N am

e of

S tu

dy D

es ig

Ec ol

og ic

C ro

ss -S

ec ti

on al

C as

e- C

on tr

ol C

oh or

Ty pe

o f S

tu dy

D es

cr ip

tiv e

an d

an al

yt ic

D es

cr ip

tiv e

A na

ly tic

A na

ly tic

N um

be r

of O

bs er

va ti

on s

Si ng

le p

oi nt

o f o

bs er

va tio

n (e

co lo

gi c

co m

pa ris

on s

tu dy

). Si

ng le

p oi

nt o

f o bs

er va

tio n.

C ol

le ct

e xp

os ur

e an

d di

se as

e hi

st or

ie s

si m

ul ta

ne ou

sl y.

Si ng

le p

oi nt

o f o

bs er

va tio

n. O

ut co

m e

de te

rm in

pr os

pe ct

iv el

y. Ex

po su

re d

et er

m in

ed a

t ba

se lin

In fo

rm at

io n

on D

ir ec

tio na

lit y

of E

xp os

ur e

U su

al ly

u nk

no w

n. U

nk no

w n

w he

th er

d is

ea se

o r

ex po

su re

c am

e fi

rs t.

Ex po

su re

in fo

rm at

io n

co lle

ct ed

r et

ro sp

ec tiv

el y.

Ex po

su re

in fo

rm at

io n

co lle

ct ed

a t

ba se

lin e.

O ut

co m

e in

fo rm

at io

n co

lle ct

ed p

ro sp

ec tiv

el y.

D at

a So

ur ce

s/ C

ol le

ct io

n M

et ho

ds Ex

is tin

g da

ta , s

uc h

m or

ta lit

y ra

te s.

D at

a fr

om s

am pl

e su

rv ey

s. M

ay a

ls o

us e

se co

nd ar

y da

ta (

e. g.

, s ch

oo l r

ec or

ds ,

na tio

na l s

ur ve

ys ).

Pr im

ar y

da ta

fr om

s ur

ve ys

p at

ie nt

s. M

ay a

ls o

us e

se co

nd ar

y da

ta (

e. g.

, f ro

di se

as e

re gi

st rie

s, m

ed ic

re co

rd s)

Pr im

ar y

da ta

fo r

ex po

su re

d ou

tc om

e. S

om e

co ho

de si

gn s

us e

se co

nd ar

y da

ta .

U ni

t of

O bs

er va

tio n

Th e

gr ou

p Th

e in

di vi

du al

Th e

in di

vi du

al Th

e in

di vi

du al

M ea

su re

s U

se d

C or

re la

tio n,

c hi

-s qu

ar e

Pr ev

al en

ce e

st im

at es

O dd

s ra

tio R

el at

iv e

ris k

A dv

an ta

ge s

Q ui

ck a

nd e

as y

to c

on du

ct .

In ex

pe ns

iv e.

C an

b e

co m

pl et

ed w

ith in

sh or

t tim

e fr

am e.

ex pe

ns iv

O ft

en s

m al

le r

in s

iz e

th an

os pe

ct iv

e st

ud ie

s. Q

ui ck

a nd

e as

y to

c om

pl et

e. C

os t

ef fe

ct iv

e. U

se fu

l f or

s tu

dy in

g ra

di se

as es

Pe rm

it di

re ct

d et

er m

in at

io n

of r

is k.

Pr ov

id e

st ro

ng er

e vi

de nc

e of

e xp

os ur

e– di

se as

e as

so ci

at io

ns t

ha n

ot he

r ob

se rv

at io

na l d

es ig

ns . U

se fu

l fo

r st

ud yi

ng r

ar e

ex po

su re

51589_CH07_Printer.indd 360 11/02/13 2:12 PM

C o n C l u S i o n 361

D is

ad va

nt ag

es Ec

ol og

ic fa

lla cy

. Im

pr ec

is e

m ea

su re

m en

t of

e xp

os ur

e– di

se as

e re

la tio

ns hi

ps .

Sh ou

ld n

ot b

e us

ed fo

r st

ud yi

ng d

is ea

se e

tio lo

gy .

N ot

u se

fu l f

or d

is ea

se s

of lo

fr eq

ue nc

y. Ex

po su

re –d

is ea

se t

em po

ra lit

y no

t as

ce rt

ai ne

U nc

er ta

in ty

o f t

ex po

su re

–d is

ea se

la tio

ns hi

p. In

ab ili

ty t

o pr

ov id

e di

re ct

tim at

e of

r is

k. N

ot u

se fu

l f or

r ar

e ex

po su

re s.

R ec

al l b

ia s.

Le ng

th y

to c

on du

ct .

A tt

rit io

n. D

iff ic

ul t

to c

ar ry

o ut

. Se

le ct

io n

bi as

. D

iff ic

ul t

to u

se w

ith r

ar e

di se

as es

A pp

lic at

io ns

Te st

s pe

ci fi

c et

io lo

gi c

hy po

th es

es .

D ev

el op

n ew

e tio

lo gi

c hy

po th

es es

. Su

gg es

t m

ec ha

ni sm

s of

us at

io n.

Id en

tif y

ne w

m et

ho ds

fo r

di se

as e

pr ev

en tio

D ep

ic t

m or

ta lit

y an

d m

or bi

di ty

fr om

s pe

ci fi

c di

se as

es .

G en

er at

e hy

po th

es es

. Pl

an in

te rv

en tio

ns .

D es

cr ib

e th

e m

ag ni

tu de

d di

st rib

ut io

n of

a h

ea lth

ob le

m .

A ss

is t

in p

la nn

in g

he al

se rv

ic es

In fe

ct io

us d

is ea

se o

ut br

ea k

in ve

st ig

at io

n. C

hr on

ic d

is ea

se in

ve st

ig at

io n

w he

n et

io lo

gy u

nk no

w n.

H yp

ot he

si s

te st

in g.

St ud

ie s

of c

hr on

ic d

is ea

et io

lo gy

. R

is k

es tim

at io

n. H

yp ot

he si

s te

st in

Ex am

pl es

A ss

oc ia

tio n

be tw

ee n

br ea

ca nc

er m

or ta

lit y

an d

di et

ar y

fa ts

. A

ir po

llu tio

n an

d m

or ta

lit y

by c

iti es

Pr ev

al en

ce o

f s m

ok el

es s

to ba

cc o

us e

am on

g sc

ho ol

c hi

ld re

n. N

um be

r of

v as

ec to

m ie

s. Pr

ev al

en ce

o f c

on ge

ni ta

l m

al fo

rm at

io ns

A ss

oc ia

tio n

be tw

ee n

ch ili

pp er

c on

su m

pt io

n an

d ga

st ric

c an

ce r

ris k.

A ne

st he

si a

an d

ne ur

al t

ub e

de fe

ct s.

Sm ok

in g

an d

ce rv

ic al

c an

ce r.

Ph ys

ic al

a ct

iv ity

a nd

ro na

ry h

ea rt

d is

ea se

or ta

lit y.

C hi

ld a

bu se

a nd

s ui

ci de

. Ex

po su

re t

o to

xi c

m at

er ia

an d

m or

ta lit

y am

on g

w or

ke rs

51589_CH07_Printer.indd 361 11/02/13 2:12 PM

362 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

differences in designs, and practical considerations in the operation of cohort studies. Relative risk, a measure of interpretation, was defined and illustrated. Finally, the chapter concluded with many examples of cohort studies, a review of the related topic of nested case-control studies, and a comparison of OS designs.

Study Questions and Exercises

1. Define in your own words the following terms: a. Cohort b. Cohort effect c. Population-based cohort d. Exposure-based cohort e. Comparison groups in cohort studies f. Prospective cohort study g. Retrospective cohort study h. Ambispective cohort study

2. What are secular trends and cohort effects? Explain the relationship between these two terms.

3. Explain what is meant by the term relative risk and explain how it is used in cohort studies.

4. Describe the essential differences between life tables and survival curves. 5. A cohort study was conducted to study the association of coffee drink-

ing and anxiety in a population-based sample of adults. Among 10,000 coffee drinkers, 500 developed anxiety. Among the 20,000 noncoffee drinkers, 200 cases of anxiety were observed. What is the relative risk of anxiety associated with coffee use?

6. How is a case-control study different from a retrospective cohort study? List the key criteria that, in general, would influence you to select one approach over the other.

7. Are relative risks of 2.0 and 0.5 the same or different in strength of association?

8. Cohort studies have some advantages over case-control studies in terms of the confidence with which their results are viewed. Suppose there have been four case-control studies of an exposure–disease association and that the range of the odds ratios is from 28.0 to 49.0. Would you advo- cate a cohort study? Justify your answer.

51589_CH07_Printer.indd 362 11/02/13 2:12 PM

r e f e r e n C e S 363

9. Cohort studies allow the investigator to examine multiple outcomes and multiple exposures. Consider the following three exposures: smoking, low vitamin D intake, and severe cold weather. How many different out- comes could you examine in a cohort study that measured all three expo- sures at baseline?

10. High rates of follow-up are essential to the validity of cohort studies. What are some approaches that can be employed to ensure compliance when linkage to a central disease registry is not an option?

11. Summarize the strengths and weaknesses and advantages and disadvan- tages of the various types of observational study designs: ecologic, cross- sectional, case-control, and cohort.

12. Describe how you would conduct a nested case-control study of low socioeconomic status as a risk factor for teenage pregnancy.

13. Explain what is meant by the statement that cohort studies overcome the problem of temporality, which is not addressed by other types of obser- vational study designs.

14. What are some of the practical issues that influence the design of a cohort study?

15. Discuss some of the possible outcomes for cohort studies, distinguishing between discrete events and disease markers.

References

1. Porta M. A Dictionary of Epidemiology. 5th ed. New York: Oxford University Press; 2008.

2. Comstock GW. Cohort analysis: W.H. Frost’s contributions to the epidemiology of tuberculosis and chronic disease. Soz Präventivmed. 2001;46:7–12.

3. Tolley HD, Crane L, Shipley N. Smoking prevalence and lung cancer death rates. In: Strategies to Control Tobacco Use in the United States: A Blueprint for Public Health Action in the 1990s. Bethesda, MD: National Institutes of Health; 1991.

4. National Center for Health Statistics. Mortality from diseases associated with smok- ing, United States, 1950–64. Mon Vital Stat Rep. 1966;20.

5. National Cancer Institute. Risks associated with smoking cigarettes with low machine- measured yields of tar and nicotine. Smoking and Tobacco Control Monograph No. 13. NIH Pub. No. 02-5074. Bethesda, MD: U.S. Department of Health and Human Services, National Institutes of Health, National Cancer Institute; October 2001.

6. Chernick MR, Friis RH. Introductory Biostatistics for the Health Sciences: Modern Applications Including Bootstrap. Hoboken, NJ: John Wiley & Sons, Inc.; 2003.

7. Arias E. United States Life Tables, 2007. National Vital Statistics Reports. Hyattsville, MD: National Center for Health Statistics; 2011:59, No. 3.

51589_CH07_Printer.indd 363 11/02/13 2:12 PM

364 C h a p t e r 7 S t u d y d e S i g n S : C o h o r t S t u d i e S

8. Gordis L. Epidemiology. 4th ed. Philadelphia, PA: Saunders Elsevier; 2008. 9. World Health Organization. Disability adjusted life years (DALY). http://www.who

.int/healthinfo/boddaly/en/index.html. Accessed October 30, 2007. 10. Christensen BC, Godleski JJ, Roelofs CR, et al. Asbestos burden predicts survival in

plural mesothelioma. Environ Health Perspect. 2008;116 (6):723–726. 11. Dawber TR, Meadors GF, Moore FE. Epidemiological approaches to heart disease:

the Framingham Study. Am J Public Health. 1951;41:279–286. 12. Francis T Jr., Epstein FH. Survey methods in general populations: Tecumseh,

Michigan. In: Acheson RM, ed. Comparability in International Epidemiology. Princeton, NJ: Milbank Memorial Fund; 1965:333–342.

13. Bisgard KM, Folsom AR, Hong CP, Sellers TA. Mortality and cancer rates in nonre- sponders to a prospective study in older women: 5-year follow-up. Am J Epidemiol. 1994;139:990–1000.

14. Robison LL, Mertens AC, Boice JD, et al. Study design and cohort characteristics of the Childhood Cancer Survivor Study: a multi-institutional collaborative project. Med Pediatr Oncol. 2002;38:229–239.

15. Paffenbarger RS, Hyde RT, Wing AL, Steinmetz CH. A natural history of athleticism and cardiovascular health. JAMA. 1984;252:491–495.

16. Tager IB. Outcomes in cohort studies. Epidemiol Rev. 1998;20:15–28. 17. Sellers TA, King RA, Cerhan JR, et al. Fifty-year follow-up of cancer incidence in a

historical cohort of Minnesota breast cancer families. Cancer Epidemiol Biomarkers Prev. 1999;12:1051–1057.

18. Hartmann LC, Schaid DJ, Woods JE, et al. Efficacy of bilateral prophylactic mas- tectomy in women with a family history of breast cancer. N Engl J Med. 1999;340: 77–84.

19. Zheng W, Chow WH, Yang G, et al. The Shanghai Women’s Health Study: rationale, study design, and baseline characteristics. Am J Epidemiol. 2005;162:1123–1131.

20. Hoskins KF, Stopler JE, Calzone KA, et al. Assessment and counseling for women with a family history of breast cancer. A guide for clinicians. JAMA. 1995;273: 577–585.

21. Hu FB, Grodstein F. Postmenopausal hormone therapy and the risk of cardiovascular disease: the epidemiologic evidence. Am J Cardiol. 2002;90(suppl):26F–29F.

22. Psaty BM, Heckbert SR, Atkins D, et al. A review of the association of estrogens and progestins with cardiovascular disease in postmenopausal women. Arch Intern Med. 1993;153:1421–1427.

23. Nelson HD, Humphrey LL, Nygren P, et al. Postmenopausal hormone replacement therapy: scientific review. JAMA. 2002;288:872–881.

24. Lobo RA. Benefits and risks of estrogen replacement therapy. Am J Obstet Gynecol. 1995;173:782–789.

25. Kawas C, Resnick S, Morrison A, et al. A prospective study of estrogen replacement therapy and the risk of developing Alzheimer’s disease: the Baltimore Longitudinal Study of Aging. Neurology. 1997;48:1517–1521.

26. Dye RV, Miller KJ, Singer DJ, Levine AJ. Hormone replacement therapy and risk for neurodegenerative diseases. Int J Alzheimers Dis. 2012; 2012: 258-454.

27. Sellers TA, Mink PJ, Anderson KE, et al. The role of hormone replacement therapy in the risk for breast cancer and total mortality in women with a family history of breast cancer. Ann Intern Med. 1997;127:973–980.

51589_CH07_Printer.indd 364 11/02/13 2:12 PM

r e f e r e n C e S 365

28. Deykin EY, Buka SL. Suicidal ideation and attempts among chemically dependent adolescents. Am J Public Health. 1994;84:634–639.

29. Fox SM III, Naughton JP, Haskell WL. Physical activity and the prevention of coro- nary heart disease. Ann Clin Res. 1971;3:404–432.

30. Morris JN, Heady JA, Raffle PAB, et al. Coronary heart disease and physical activity of work. Lancet. 1953;2:1053–1120.

31. Taylor HL, Klepetar E, Keys A, et al. Death rates among physically active and sed- entary employees of the railroad industry. Am J Public Health. 1962;10:1697–1707.

32. Peters RK, Cady LD Jr., Bischoff DP, et al. Physical fitness and subsequent myocar- dial infarction in healthy workers. JAMA. 1983;249:3052–3056.

33. van Wijngaarden E, Savitz DA, Kleckner RC, et al. Exposure to electromagnetic fields and suicide among electric utility workers: a nested case-control study. West wJ Med. 2000;173:94–100.

34. Dearden KA, Hale CB, Woolley T. The antecedents of teen fatherhood: a retrospec- tive case-control study of Great Britain youth. Am J Public Health. 1995;85:551–554.

35. Shepherd PM. The National Child Development Study: An Introduction to the Background to the Study and the Methods of Data Collection (Working Paper I, National Child Development Study User Support Group). London, UK: City University, Social Statistics Unit; 1985.

36. Samet JM, Munoz A, eds. Cohort studies. Epidemiol Rev. 1998;20(1).

51589_CH07_Printer.indd 365 11/02/13 2:12 PM

51589_CH07_Printer.indd 366 11/02/13 2:12 PM

3chapte r

367

8chapte r

Experimental Study Designs

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● state how study designs compare with respect to validity of causal inference

●● distinguish between a controlled experiment and a quasi-experiment ●● describe the scope of intervention studies ●● define the term controlled clinical trials and give examples ●● explain the phases in testing a new drug or vaccine ●● discuss blinding and crossover in clinical trials ●● define what is meant by community trials ●● discuss ethical aspects of experimentation with human subjects

CHAPTER OUTLINE

I. Introduction II. Hierarchy of Study Designs

III. Intervention Studies IV. Clinical Trials V. Community Trials

VI. Conclusion VII. Study Questions and Exercises

51589_CH08_Printer.indd 367 09/02/13 12:52 PM

368 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

Introduction

Although observational research can provide important clues and even compelling evidence regarding means to improve public health, it would be beneficial to know for certain whether an identified risk factor is truly associated with an out- come before embarking on a major public health intervention. The study design that is most convincing for conferring evidence of associations between risk fac- tors and outcomes is one that involves true experimentation. In a traditional laboratory experiment involving yeast, fruit flies, or mice, for example, an inves- tigator has complete control over the exposure (perhaps a chemical suspected of being a health risk) and can determine issues such as timing of exposure, intensity, and duration; these issues are aspects of manipulation of the study factor. Thus, when a true experimental design is used, assigning definitive relationships between cause and effect is fairly simple.

With humans, complete control over an exposure is clearly not possible, especially if the exposure is harmful. However, there are ways to test hypotheses about preventive interventions in human subjects. According to the National Institutes of Health, a clinical trial is “a prospective biomedical or behavioral research study of human subjects that is designed to answer specific questions about biomedical or behavioral interventions (drugs, treatments, devices, or new ways of using known drugs, treatments, or devices). Clinical trials are used to determine whether new biomedical or behavioral interventions are safe, effica- cious, and effective.”1

Although clinical trials are most commonly applied to test new therapies they are also appropriate to test preventive interventions. The key point is that clini- cal trials enroll individual subjects and enable randomization of subjects to either receive or not receive the intervention. Sometimes the questions being tested for public health interventions do not enable manipulation at the level of the individual (such as a public service announcement on television that reaches everyone) or randomization (since there is no way to randomly show the public service announcement on some televisions but not others). These particular chal- lenges provide a glimpse into the need for study designs that include groups or other defined populations rather than individuals. Thus, the principle of experi- ments has been extended to community-based trials to test certain preventive and public health trials.

To give an example of why experimental studies are important to verify epidemiologic findings, let us highlight the case of hormone replacement therapy (HRT). A large number of epidemiologic studies had shown that HRT

51589_CH08_Printer.indd 368 09/02/13 12:52 PM

i n t r o d u C t i o n 369

use had significant benefits against coronary heart disease. A series of clinical trials had failed to demonstrate any benefit.2 Conversely, an equally large body of epidemiologic research had observed that women who took HRT for control of menopausal symptoms had elevated risks of breast cancer. The only way to truly resolve the question of risks versus benefits of HRT was a very large clini- cal trial, which is part of the motivation for the Women’s Health Initiative (see Exhibit 8–1). This clinical trial demonstrated that the epidemiologic findings on cancer were generally accurate, but the benefits on cardiovascular disease had been overestimated.3 The results had a dramatic effect on public health, as use of HRT decreased 40–80% after the trial was stopped.

Questions and answers regarding the Women’s health Initiative

Q. What is the purpose of the WHI study on combination hormone therapy?

A. The long-term studies in the WHI were initiated because over the years a number of research studies presented a complicated picture of the risks and benefits of hormone therapy, and its continued use for prevention of cardiovascular diseases was con- troversial. This situation led the NIH to conduct a large clinical trial of the risks and benefits of hormone therapy. The WHI set out to examine the long-term effect of estrogen plus progestin on the prevention of heart disease and hip fractures, while moni- toring for possible increases in risk for breast and colon cancer. The estrogen-plus-progestin regimen was given to women who have a uterus since progestin is known to protect against endo- metrial cancer, a known effect of unopposed estrogen. A sepa- rate study of estrogen alone in women who had a hysterectomy was also begun.

Q. Why were the women in the WHI estrogen-plus-progestin study told to stop study pills in July 2002?

A. When it reviewed the study data in May 2002, the WHI Data and Safety Monitoring Board saw an increased risk of breast cancer in women

e x

h Ib

It 8

–1

continues

51589_CH08_Printer.indd 369 09/02/13 12:52 PM

370 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

taking estrogen plus progestin. The Board also saw that the previously identified risks for heart attacks, strokes, and blood clots in the lungs and legs had persisted. Therefore, in the judgment of the Board, the overall risks outweighed the benefits of taking estrogen plus progestin.

Q. What were the main findings in the WHI study on estrogen plus progestin?

A. The main findings show that compared to women taking placebo pills: ●● The number of women who developed breast cancer was higher in women taking estrogen plus progestin.

●● The numbers of women who developed heart attacks, strokes, or blood clots in the lungs and legs were higher in women taking estro- gen plus progestin.

●● The numbers of women who had hip and other fractures or colorec- tal cancer were lower in women taking estrogen plus progestin.

●● There were no differences in the number of women who had endo- metrial cancer (cancer of the lining of the uterus) or in the number of deaths.

Q. What are the increased risks for women taking estrogen plus progestin?

A. For every 10,000 women taking estrogen plus progestin pills: ●● 38 developed breast cancer each year, compared to 30 breast can- cers for every 10,000 women taking placebo pills each year.

●● 37 had a heart attack each year, compared to 30 out of every 10,000 women taking placebo pills.

●● 29 had a stroke each year, compared to 21 out of every 10,000 women taking placebo pills.

●● 34 had blood clots in the lungs or legs each year, compared to 16 out of every 10,000 women taking placebo pills.

Q. What are the reduced risks for women taking estrogen plus progestin? A. For every 10,000 women taking estrogen plus progestin pills:

●● 10 had a hip fracture each year, compared to 15 out of every 10,000 women taking placebo pills.

●● 10 developed colon cancer each year, compared to 16 out of every 10,000 women taking placebo pills.

Exhibit 8–1 continued

continues

51589_CH08_Printer.indd 370 09/02/13 12:52 PM

h i e r a r C h y o f S t u d y d e S i g n S 371

Experimental (a.k.a. intervention) studies are used in both clinical medicine and public health, and there are applications of epidemiologic principles in each. We begin with an overview of some general principles of experimentation in humans and then touch on the application of experimental procedures to test new therapies and devices. Although this is outside the realm of public health, epidemiologic principles have a role in some types of these studies. Thus epi- demiologic principles provide a foundation for further elaboration on the use of experimental study designs in public health interventions. The authors pro- vide numerous examples and discuss how these experimental designs differ from observational studies (i.e., cross-sectional, case-control, and cohort studies) and describe methods for evaluating the outcomes of community interventions. The unique advantages and disadvantages of experimental designs in compari- son with observational designs are presented. This discussion will facilitate an understanding of why experimental studies are generally accepted as having the greatest relative influence in conferring evidence.

Hierarchy of Study Designs

Exhibit 8–2 ranks study designs according to their validity for causal infer- ences. As the exhibit indicates, all of the observational study designs from case studies to prospective cohort studies may be considered less powerful for etio- logic inference than experimental studies. The latter are generally regarded as the most scientifically rigorous method of hypothesis testing available involving

Q. What are the conclusions from these findings? A. The main conclusions are:

●● The estrogen-plus-progestin combination studied in WHI does not prevent heart disease.

●● For women taking this estrogen-plus-progestin combination, the risks (increased breast cancer, heart attacks, strokes, and blood clots in the lungs and legs) outweigh the benefits (fewer hip frac- tures and colon cancers). n

Source: Adapted and reprinted from the Women’s Health Initiative study website. Available at http://www.nhlbi.nih.gov/health/women/q_a.htm. Accessed August 9, 2012.

Exhibit 8–1 continued

51589_CH08_Printer.indd 371 09/02/13 12:52 PM

372 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

human participants. The emphasis is on rigor and not feasibility, for as we shall learn subsequently, not all research situations permit the use of this design.

Experimental designs enable us to overcome some of the deficiencies inherent in observational designs. The use of experimentation to derive knowledge about the causes of disease has intuitive appeal. By exercising control over who will receive the exposure as well as the level of the exposure, the investigator more confidently may attribute cause and effect to associations than in nonexperimen- tal (observational) designs.

Ranked immediately below controlled experiments are quasi-experiments. The investigator is unable to randomly allocate subjects to the conditions (inter- vention or control) of a quasi-experimental study. As a result, there may be contamination across the conditions of the study. It becomes more difficult to differentiate between the effects of the intervention and the control conditions than in a controlled experiment. (Of course, it also is possible for contamination to occur in a clinical trial.)

Validity for etiologic Inference according to Study Designs

Validity Ranking Type of Study Design

Highest

Lowest

Experimental study Controlled experiment/randomized clinical trial Quasi-experiment/community trial Prospective cohort study Retrospective cohort study Nested case-control study Time-series analysis Cross-sectional study Ecologic study Case study Anecdote

Source: Adapted from Künzli N, Tager IB. The semi-individual study in air pollution epidemiology: a valid design as compared to ecologic studies, Environmental Health Perspectives. 1997:(105)10;1079. National Institute of Environmental Health Sciences, U.S. Department of Health and Human Services.

e x

h Ib

It 8

–2

51589_CH08_Printer.indd 372 09/02/13 12:52 PM

i n t e r v e n t i o n S t u d i e S 373

The reader should not conclude that experimental designs are always the most appropriate design for investigating the causes of disease. Experimental designs are not necessarily appropriate for testing all conceivable hypotheses, such as those in the fields of occupational and environmental health. To give an example, an epidemiologist might want to examine the relative contributions of smoking and radon exposure to lung cancer among uranium miners. For ethical reasons, this investigation could not be conducted as an experimental study because it would involve deliberate exposure of subjects to agents suspected of being harm- ful. An observational study design is the only realistic approach in this scenario. Alternatively, one could test the hypothesis that elimination of exposure to smok- ing and radon reduces lung cancer risk and this could be tested prospectively in an experimental study.

Intervention Studies

An intervention study is defined as “[a]n investigation involving intentional change in some aspect of the status of the subjects, e.g., introduction of a preven- tive or therapeutic regimen or in intervention designed to test a hypothesized relationship; usually an EXPERIMENT such as a RANDOMIZED CON- TROLLED TRIAL.”4 Intervention studies are employed to test the efficacy of a preventive or therapeutic measure. In comparison, the goal of observational studies is to generate enough knowledge about the etiology and natural history of a disease to formulate strategies for prevention.

Recall the two basic facets of research designs: manipulation of the study fac- tor and randomization of study subjects. Controlled experimental studies involve randomization of subjects to exposures under the control of the investigator, whereas quasi-experimental studies involve external control of exposure without randomization. Both strategies are employed in intervention studies.

●● Clinical trials (focus on the individual) ●● Community trial or community intervention (focus on the group or

community) Note: Controlled clinical trials may be conducted both at the individual and community levels (for narrowly defined outcomes). n

Intervention Designs Include:

51589_CH08_Printer.indd 373 09/02/13 12:52 PM

374 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

Broadly defined, there are two types of intervention designs: clinical trials and community trials. The key difference between the two types of intervention is that in clinical trials the focus is individuals, whereas the focus of community trials is groups or community outcomes. This difference in focus limits the types of interventions that are possible under each approach. Clinical trials are usu- ally tightly controlled in terms of eligibility, delivery of the intervention, and monitoring of outcomes. The duration of clinical trials ranges from days to years. Participation is generally restricted to a highly selected group of individuals: com- monly volunteers who have been diagnosed with a disease; volunteers who are screened subjects at high risk for disease; or other types of volunteers who may be interested in participating in the clinical trial and are deemed eligible for such participation. One example of a clinical trial is termed a randomized controlled trial (RCT), which is most appropriate for testing narrow hypotheses regarding vaccines, treatments, or individuals’ behavior change.5 An example of behavior change is elimination of tobacco use among smokers. Usually community trials cannot exert rigid control over members of the group or community, are typically delivered to all members rather than narrowly defined subsets, tend to be of lon- ger duration, and usually involve primary prevention efforts. Thus, RCTs at the community level are typically used only in special situations when there is a sim- ple intervention (e.g., a response to an education campaign using mass media) and there are a sufficient number of communities to involve in the experiment.5

Clinical Trials

As a concept, clinical trials have a venerable history that spans a time period from early biblical and Greek references to increasing activity during the 18th and 19th centuries to present methodologic sophistication.6 The first efforts were not formal clinical trials, as we know them today; however, the attempts at experimentation led the way to contemporary methods. For example, in 1537 Ambroise Paré applied an experimental treatment for battlefield wounds that used what he called a “digestive” made from turpentine, rose oil, and egg yolks. He observed that this concoction was more effective in treating wounds than the application of boiling oil, the standard treatment of the day. Later, in 1600, the East India Shipping Company found that lemon juice protected sailors from scurvy by comparing sailors on a ship supplied with lemon juice with sailors on ships that were not supplied. In a 1747 study of scurvy, James Lind designed one of the first experiments that used a concurrently treated control group.6 His experiment involved feeding 12 sailors who were suffering from scurvy

51589_CH08_Printer.indd 374 09/02/13 12:52 PM

C l i n i C a l t r i a l S 375

6 different types of diets; Lind noted that sailors who received citrus fruits had the best recovery from their malady. Figure 8–1 shows the image of a scurvy victim; the symptoms of scurvy included sunken face, red-rimmed eyes, and red spots on the skin. The lesions of scurvy are shown on the victim’s forearm.

Other pioneering landmarks in the development of clinical trials include Jenner’s efforts to develop a smallpox vaccine in the late 18th century and experi- ments with anesthetics, such as ether and chloroform, in the mid-19th century. While the earliest planned experiments were carried out without the benefit of a control or comparison, subsequent research contributed to the development of control treatments and randomization. More recent historical developments have included the use of multicenter trials in which recruitment of participants is extended across several to hundreds of accrual sites with data sent to a coor- dinating center for analysis. Multicenter trials have been instrumental in the

FigurE 8–1 Image of a scurvy victim. Source: Image from the History of Medicine (NLM). Scurvy victim. Krankenphysiogomik. 1842

51589_CH08_Printer.indd 375 09/02/13 12:52 PM

376 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

development of treatments for infectious diseases (e.g., polio) and recently in chronic diseases that are of noninfectious origin. (Refer to Exhibit 8–3, which explains the rationale for clinical trials.)

the Why, What, When, and Where of clinical trials

We’ve all taken medication. It may have been in the form of over-the-counter cough medicine or prescription pills to treat chronic conditions such as diabetes. But how did researchers dis- cover if the medicine is effective, if it’s safe, and if there are any potential side effects?

Testing and evaluating drugs is serious business, and clinical trials are right at the center of the process, according to Dorothy Cirelli, chief of the Patient Recruitment and Referral Center at the National Institutes of Health (NIH), U.S. Department of Health and Human Services (HHS). “Medical advances would not occur without clinical trials,” she said.

Well known as a world leader in medical research, NIH developed the first treatment for the human immunodeficiency virus (HIV) as well as innovative therapies for breast cancer, leukemia, and lymphoma. For over a century, the agency has conducted clinical studies that explore the nature of illnesses. Clinical studies are currently under way for nearly every kind of cancer, HIV, cardiovascular disease, diabetes, obesity, and many other conditions both common and rare.

There are several types of clinical trials. By far the largest number test new therapeutic drugs and fall in the domain of medical interventions rather than public health intervention. Note, however, that epidemiologic research is needed to monitor long term risks of drugs, devices, or other established treatments. However, there are other clinical trials that clearly relate to public health. Prevention studies test drugs, vaccines, or lifestyle changes that may help prevent disease. Diagnostic studies evaluate new ways of detecting or classifying disease.

The Food and Drug Administration (FDA), another HHS agency, is responsible for reviewing the scientific work of drug developers and

e x

h Ib

It 8

–3

continues

51589_CH08_Printer.indd 376 09/02/13 12:52 PM

C l i n i C a l t r i a l S 377

A clinical trial is defined as “[a] research activity that involves the administration of a test regimen to humans to evaluate its efficacy and safety. The term is subject to wide variation in usage from the first use in humans [refer to foregoing examples] without any control treatment to a rigorously designed and executed experiment involving RANDOM ALLOCATION of test and

implementing a rigorous drug approval process. The FDA, which some have called the world’s largest consumer protection agency, works to pro- tect the public by ensuring that products are safe, effective, and labeled for their intended use.

From animals to humans: Every year, hundreds of clinical trials are con- ducted at medical centers across the country. Drugs must be studied in properly controlled trials in order to determine if they work as intended to achieve benefits for patients. Drugs that have not been previously used in humans must undergo preclinical test-tube analysis and/or animal studies involving at least two mammals to determine toxicity.

The toxicity information is then used to make risk/benefit assessments, and determine if the drug is acceptable for testing in humans.

NIH or the company developing the drug must conduct studies to show any interaction between the body and the drug. In addition, researchers must provide the FDA with information on chemistry, manufacturing, and controls. This ensures the identity, purity, quality, and strength of both the active ingredient and the finished dosage form.

The sponsor of the proposed new drug then develops a plan for testing the drug in humans. The plan is submitted to the FDA with information on animal testing data, the composition of the drug, manufacturing data, qualifications of its study investigators, and safety of the people who will participate in the trial. This information forms what is known as the Inves- tigational New Drug Application (IND).

The entire drug development process is lengthy and expensive. On aver- age it takes about 10 years to complete. But it is an effective system that generally protects the public from dangerous and ineffective drugs. n

Source: Sections reprinted and adapted from Brooks J. Clinical trials: how they work, why we need them. In Closing the Gap. Washington, DC: Office of Minority Health, Public Health Service, U.S. Department of Health and Human Services; December 1997/January 1998:1–2.

Exhibit 8–3 continued

51589_CH08_Printer.indd 377 09/02/13 12:52 PM

378 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

control treatments.”4 One starts by determining eligibility of potential subjects. Eligibility rules must be carefully defined and rigidly enforced. Criteria for inclu- sion will vary by the type and nature of the intervention proposed. Once eligible subjects agree to participate, they are then randomly assigned to one of the study groups. Figure 8–2 illustrates a single intervention and single control (or pla- cebo) arm of a trial; yet more than one experimental intervention can be run in parallel. Table 8–1 provides a glossary of terms used in clinical trials.

Prophylactic and Therapeutic Trials A prophylactic trial is designed to evaluate the effectiveness of a substance (such as a vaccine against measles or polio) or a prevention program (such as vita- min supplementation or patient education) that is used to prevent a disease.

FigurE 8–2 Schematic diagram of a clinical trial.

Randomization to groups

Sample

Nonparticipants

Control group

Intervention group

Measure outcome

Lost to follow-up

51589_CH08_Printer.indd 378 09/02/13 12:52 PM

C l i n i C a l t r i a l S 379

A therapeutic trial involves the study of curative drugs or a new surgical procedure to evaluate how well they bring about an improvement in the patient’s health. The control arm subjects would receive the standard of care (for a surgery or drug trial), a placebo (for a vitamin trial), or no intervention (for the patient education trial). Some patient education programs might use individual or group counseling in the intervention group and provide control participants with an informational brochure—the usual method of education.

Outcomes of Clinical Trials Outcomes, or results of clinical trials, are referred to as clinical end points. To assess the results, investigators compare rates of disease, death, recovery, or other appropriate outcomes. The outcome of interest is measured in the intervention and control arms of the trial to evaluate efficacy. We have previously noted that the outcome of a clinical trial must be measured in a comparable manner in both the intervention and control conditions. The analysis plan must be specified in advance and typically includes an early stopping rule in case the trial shows early evidence of efficacy or insufficient evidence to justify continuation.

Table 8–1 Glossary of Terms Used for Clinical Trials

Arm Any of the treatment groups in a randomized trial. Most randomized trials have two arms, but some have three arms, or even more.

Drug–drug interaction A modification of the effect of the drug when administered with another drug.

Efficacy (of a drug or treatment)

The maximum ability of a drug or treatment to produce a result, regardless of dosage.

Informed consent The process of learning the key facts about a clinical trial before deciding whether or not to participate.

Placebo effect A physical or emotional change occurring after a substance is taken or administered that is not the result of any special property of the substance.

Protocol A study plan on which all clinical trials are based. The plan is carefully designed to safeguard the health of the participants as well as answer specific research questions. A protocol describes what types of people may participate in the trial; the schedule of tests, procedures, medications, and dosages; and the length of the study.

Recruiting The period during which a trial is attempting to identify and enroll participants. Recruitment activities can include advertising and other ways of soliciting interest from possible participants.

Risk-benefit ratio The risk to individual participants versus the potential benefits. Side effects Any undesired actions or effects of a drug or treatment. Negative

or adverse effects may include headache, nausea, hair loss, skin irritation, or other physical problems.

Toxicity An adverse effect produced by a drug that is detrimental to the participant’s health.

Source: Adapted from Glossary of Clinical Trials Terms. http://clinicaltrials.gov/ct2/info/ glossary#icd. Accessed March 6, 2012.

51589_CH08_Printer.indd 379 09/02/13 12:52 PM

380 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

For some drugs, such as antihypertensive agents, it may not be feasible to conduct randomized large clinical trials that evaluate major clinical end points, such as strokes or myocardial infarctions.7 As an alternative, surrogate end points may be used in small short-term trials. A surrogate end point for drug therapy in hypertension would be measures of subclinical disease or physical measures, including reduction of blood pressure.

Examples of Clinical Trials The first example of a clinical trial is the double-blind Medical Research Council Vitamin Study.8 Neural tube defects (NTDs; e.g., anencephaly, spina bifida, or encephalocele) are among the most severe congenital malformations. The possi- bility that folic acid (a B vitamin) might be involved was raised as early as 1964.9 Preliminary interventions had been promising but not conclusive. Therefore, a randomized trial was conducted at 33 centers in 7 countries. The trial examined whether vitamin supplementation around the time of conception could prevent NTDs. The supplements used were folic acid or a mixture of 7 other vitamins: A, D, B1, B2, B6, C, and nicotinamide. Eligible subjects included women who were planning a pregnancy, had had a previous child with an NTD, and/or were not already taking vitamin supplements. A total of 1,817 women were randomized into one of four groups: folic acid, other vitamins, both, or neither. Subjects did not know to which group they were assigned. To monitor possible toxicity associ- ated with the supplementation, recording forms were provided to all subjects. Of those randomized, 1,195 gave birth to a child with a known outcome. Whenever an NTD was reported, independent corroboration was sought. Classification was made without knowing to which group the mother had been randomized.

The rate of NTDs among women receiving folic acid (alone or in combina- tion with other vitamins) was 1% (5 out of 514). The rate among those allocated to the other groups (other vitamins or nothing) was 3.5%, yielding a relative risk of 0.28 (95% confidence interval [CI], 0.12–0.71). The rate among women in the other vitamin-only group was only slightly lower than the rate among the women receiving nothing (relative risk, 0.80; 95% CI, 0.32–1.72). The trial results were regularly monitored. By April 12, 1991, the difference between the folic acid-supplemented group and the others was firmly established by case 27. The data-monitoring committee recommended that the trial be stopped; the steering committee agreed. The authors of the report8 concluded that folic acid supplementation could now be recommended for all women who had a previously affected pregnancy. Furthermore, they suggested that public health measures should be taken to ensure that all women of childbearing age receive

51589_CH08_Printer.indd 380 09/02/13 12:52 PM

C l i n i C a l t r i a l S 381

adequate dietary folic acid and that consideration should be given to fortification of staple foods with it.

Another application of clinical trials is to evaluate the effectiveness of educa- tion efforts to prevent the spread of sexually transmitted diseases (STDs). The factors that contribute to an increased risk of STDs are multiple and complex and include social, behavioral, and environmental influences. Efforts to reduce the incidence of STDs among inner-city residents are especially difficult, given the fact that most education efforts are relegated to public clinics. Because the clinics often have limited resources, educational programs must be delivered by staff who may be inadequately trained in prevention and unprepared to pro- vide information and skills to culturally diverse populations. O’Donnell and col- leagues10 conducted a clinical trial of video-based educational interventions on condom acquisition among men and women seeking services at a large STD clinic in South Bronx, New York. A total of 3,348 African-American and His- panic male and female patients were assigned to one of three arms: video only, video plus interactive session, or control. The videos were 20 minutes in length, culturally appropriate, and designed to model appropriate strategies for over- coming barriers to consistent condom use. A proxy measure of condom use was employed. After clinic services and participation in the assigned treatment group, subjects were given coupons for free condoms at a nearby pharmacy. The rate of redemption of coupons for condoms was 21.2% among patients who received no intervention, 27.6% among patients who received video alone, and 36.9% among patients who received video plus interactive group sessions. O’Donnell et al. pointed out that because of the high prevalence of STDs in the population served by public clinics, the observed increases in the number of people practic- ing safer sex could have a significant impact on public health.

The foregoing examples of clinical trials show how they are important to public health practitioners with respect to a range of applications (e.g., use of vitamin supplements among pregnant women and evaluating the effectiveness of programs to prevent the spread of STDs). Sometimes clinical trials may not yield significant results, evoking consternation for the researchers as the cartoon in Figure 8–3 suggests.

Blinding (Masking) Objectivity of the data is a major concern regarding clinical trials. For exam- ple, some clinical trials make use of volunteers who are extremely grateful for the opportunity to participate. At the same time, the investigator may assess the trial’s outcome by determining the participants’ subjective impressions, for

51589_CH08_Printer.indd 381 09/02/13 12:52 PM

382 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

example, by means of a self-report questionnaire. This evaluation method may tend to overstate the intervention’s clinical responses and benefits. Furthermore, subjects who learn that they had been randomized to the placebo arm of the trial may not wish to continue participation. For these reasons, a commonly used approach is a single-blind design. (A synonym for the term blinding is masking.) In this design, the subject is unaware of group assignment. Informed consent is obtained before assignment, and the experimenter must treat and monitor all groups in a similar manner. Another concern and potential for problems lies in how the experimenter assesses the trial’s outcome. This concern is especially war- ranted if the trial evaluates a new drug, the drug’s manufacturer is financing the trial, and/or those performing the evaluations stand to benefit if the new drug is

FigurE 8–3 Researchers who are analyzing the results of a medical experiment. Source: Reproduced from Science Cartoons Plus.com. Available at: http://www.sciencecartoonsplus.com/gallery/medical/index.php#. Accessed on August 28, 2012.

51589_CH08_Printer.indd 382 09/02/13 12:52 PM

C l i n i C a l t r i a l S 383

shown to be more effective than existing medications. To reduce the likelihood of biased assessment, an approach called a double-blind design is used; neither the subject nor the experimenter is aware of group assignment. For example, in a clinical trial of a new drug, one way to implement such a procedure is to have the placebo or treatment agents come in a preassigned container, the contents of which are unknown to the investigator and subjects at the time of the trial. All subjects are then treated and monitored in a similar manner.

Phases of Clinical Trials Phases in a clinical trial refer to the stages that must occur in the development of a vaccine, drug, or treatment before it can be licensed for general use. A long and arduous process is required to bring a vaccine from the laboratory setting to eventual licensing,11 and the same may be said for licensing of new drugs. The process of bringing a new vaccine to market requires the balance between protecting the public from a potentially deleterious vaccine and satisfying urgent needs for new vaccines. To illustrate, while researchers have devoted a great deal of attention to producing a vaccine for HIV, the problems in perfecting such a vaccine have been daunting. Generally, licensing of a vaccine requires a phase I, phase II, and phase III evaluation. A description of these steps for the development of a new vaccine is in Exhibit 8–4.

Stages in the Development of a Vaccination program

Pre-licensing evaluation of vaccine ●● Phase I trials: Safety in adult volunteers ●● Phase II trials: Immunogenicity and reactogenicity in the target population

●● Phase III trials: Protective efficacy

Post-licensing evaluation ●● Safety and efficacy of vaccine ●● Disease surveillance ●● Serologic surveillance ●● Measurement of vaccine coverage

e x

h Ib

It 8

–4

51589_CH08_Printer.indd 383 09/02/13 12:52 PM

384 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

In the evaluation of vaccines, phase I trials involve testing the new vaccine in adult volunteers, typically fewer than 100. Following successful demonstra- tion of a response in a small-scale study among the volunteers (e.g., antibody formation in response to a vaccine), the testing proceeds to phase II. This phase expands the testing to a group of approximately 100 to 200 subjects who are selected from the target population for the vaccine. Antibody responses and clin- ical reactions to the vaccine are examined. There may be a double-blind design with random allocation to study or placebo conditions. The third phase, which is used to assess protective efficacy in the target population, is the main test of the vaccine. Vaccine efficacy refers to the reduction in the incidence rate of a disease in a vaccinated population compared with an unvaccinated population.11 After phase III testing has been completed, a license to manufacture the vaccine may be granted. Post-licensing evaluations of the vaccine need to continue to monitor its safety and efficacy. Figure 8–4 shows a billboard used to advertise the highly successful vaccination program for polio.

FigurE 8–4 Billboard used to promote public health awareness of polio vaccinations within the community, circa 1950s. Source: Reproduced from Centers for Disease Control and Prevention, Public Health Image Library, image number 8286. Available at http://phil.cdc.gov/phil/. Accessed May 3, 2012.

51589_CH08_Printer.indd 384 09/02/13 12:52 PM

C l i n i C a l t r i a l S 385

Another program is the development of the HPV vaccine. At least 15% of human malignant diseases are attributable to the consequences of persistent viral or bacterial infection. Chronic infection with oncogenic human papillomavirus (HPV) types is a necessary, but not sufficient, cause in the development of more cancers than any other virus. Genital human papillomavirus (HPV) is the most common sexually transmitted infection in the United States; an estimated 6.2 million persons are newly infected every year.12 Although the majority of infections cause no symptoms and are self-limited, persistent genital HPV infec- tion can cause cervical cancer in women and other types of anogenital cancers and genital warts in both men and women.

Approximately 100 HPV types have been identified, over 40 of which infect the genital area, and 16 of which have been shown in epidemiologic stud- ies to pose high risk. A series of clinical trials were performed to test a vaccine (GARDASIL™, manufactured by Merck and Co., Inc., Whitehouse Station, New Jersey) that targeted the four most common high-risk HPV subtypes (6, 11, 16, and 18). Thus, GARDASIL™ is a “quadrivalent vaccine” that immunizes against more than 70% of cervical cancers. Based on these clinical trial results, in June 2006 the HPV vaccine was licensed for use among females aged 9–26 years for prevention of HPV-type-related cervical cancer, cervical can- cer precursors, vaginal and vulvar cancer precursors, and anogenital warts.

Epidemiologic studies of the natural history of HPV infection in men13 have expanded the indications for the vaccine to boys and young men. Thus, in a period of only several decades, knowledge has accumulated about HPV as a cause of cervical (and other) cancers to a proven intervention that can prevent the disease. In a recent review the significance was aptly summarized: “The HPV- related cancers are dominated by cervical cancer in the developing world, where cervical cancer screening is limited. In this setting, widespread uptake of cur- rent HPV vaccines by adolescent girls could reduce this cancer’s incidence and mortality by approximately two-thirds, with cost-effective screening programs of adult women having the potential to reduce mortality more rapidly.”14(p 18)

The phases for testing an anticancer drug follow a similar pattern. (See Table 8–2.) Laboratory studies in vitro and in vivo among animals may have suggested a new agent that has promising antitumor action. Phase I consists of testing the agent among human subjects after animal studies have been con- ducted. Phase II is concerned with testing the efficacy of the drug with various tumor types. Note that phase I trials to determine maximum tolerated dose and phase II trials to provide evidence of potential efficacy do not involve random- ization. If those phases are successful, then one can proceed to a phase III trial,

51589_CH08_Printer.indd 385 09/02/13 12:52 PM

386 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

which uses randomization of subjects to intervention and control arms. Phase III checks the new therapy against available therapies. Successful phase III studies typically lead to approval of the drug by the Food and Drug Administration. At the conclusion of phase III trials, much information has been gathered about the agent.15 In practice, some clinical trials may have more than three phases (e.g., phase IV trials—post-marketing research to gather more data about risks and benefits of a drug). Since clinical trials are designed to provide evidence of short-term efficacy, long-term complications or side effects cannot be detected. Hence, phase IV studies, using epidemiologic surveillance methods, are used to monitor for these adverse effects.

The lengthy process that is required to bring a new drug, vaccine, or treatment to the healthcare marketplace not only helps to protect the consumer, but also delays bringing needed therapies to critically ill patients. Because of demands that advocacy groups have made on government agencies, proposals have been made to shorten the lead time to make new drugs more rapidly available. These legislative and policy modifications are intended to bring more speedy relief to patients who are afflicted with rapidly fatal and debilitating conditions.

Randomization The method of choice for assigning subjects to the treatment or control condi- tions of a clinical trial is randomization. Researchers must be very concerned

Table 8–2 A Description of Clinical Trials for a Cancer Drug

Phase Goals Objectives

I Initial testing in humans following animal studies. Organized as escalating dose trials in which subjects are entered into a series of progressively higher dosage levels until life-threatening, irreversible, or fatal toxicity is experienced.

Identify dose-limiting toxicities. Establish maximally tolerated dose; optimal dosage range. Describe pharmacology of agent (e.g., metabolism, distribution, excretion).

II Testing in selected tumor types ranging from highly chemosensitive to chemo-resistant.

Determine activity and therapeutic efficacy in a range of tumor types. Validate toxicity and dosage data.

III Randomized trial comparing new therapy with existing therapies in terms of duration and quality of survival. Trial may have multiple study arms and involve sample stratification.

Determine value of new therapy in relation to existing therapies. Generate and publish recommendations for the medical community.

IV Post-marketing surveillance using using epidemiologic methods.

To determine potential long-term adverse effects of the treatment.

51589_CH08_Printer.indd 386 09/02/13 12:52 PM

C l i n i C a l t r i a l S 387

about errors that might be introduced when some other method of subject assignment is used. In nonrandom assignment, any observed differences between study arms might simply reflect differences observed among participants of the trial. There are two general methods for randomization of subjects to the condi- tions of the trial, known as fixed and adaptive randomization.6 Fixed randomiza- tion is easier to perform than adaptive randomization. However, a full discussion of them is beyond the scope of this text. The general concept of fixed randomiza- tion is that once subjects have been selected, pass the eligibility determination, and agree to participate, they have an equal probability of being assigned to the intervention or control arm. The simplest form of randomization is the “flip of a coin,” but more elaborate protocols that employ random numbers tables or computer algorithms are typically used. As part of randomization in some tri- als, subjects are stratified in order to improve comparability among conditions. For example, in order to control for sex and age differences, equal proportions of gender and age groups are included in the treatment and control groups. For randomization to be fully effective, a reasonably large number of subjects must be enrolled in order to obtain equal distributions of demographic and other vari- ables in the study conditions. For the sake of illustration, a study with only two men and two women would have too few subjects, because random assignment could result in each study arm having only one gender (i.e., one group with all men and the other with all women).

Crossover Designs A treatment crossover refers to, “any change of treatment for a patient in a clinical trial involving a switch of study treatments.”6(p 306) These crossovers may be planned or unplanned. To illustrate, the protocol may specify in advance that a patient or group of subjects may be switched from one treatment condition to another treatment condition during the course of the trial and after a prede- termined period of follow-up. Sometimes, in such crossovers, the patient is said to serve as his or her own control. An unplanned crossover refers to a switch of patients to different treatment conditions for various reasons. For example, patients in a coronary bypass treatment may have misgivings about this invasive surgical procedure, or patients in a medical care group may require surgical treat- ment because of deterioration in their condition.16

Figure 8–5 illustrates a clinical trial for two experimental conditions (treat- ment and placebo) with a crossover design. As in a randomized controlled trial discussed previously, study subjects are assigned randomly to treatment and con- trol conditions at baseline. The investigator then measures the outcome at t1.

51589_CH08_Printer.indd 387 09/02/13 12:52 PM

388 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

Then participants wait during a time interval known as the washout period. Subsequently, study subjects switch experimental conditions as shown in the figure and the trial is repeated.

One of the disadvantages of a crossover design is the carryover effect, which is a type of bias that may affect the results during the second period of the trial (t2). For example, a drug that is administered to the intervention group might remain active and need to clear from the body during washout. The purpose of the washout period is to control biases that might be introduced by the first treatment. Crossover trials are not suitable for all medical conditions, but in general only those for which temporary relief is being investigated and not for ill- ness that can be cured by the treatment. Examples of conditions appropriate for crossover trials include tests of new drugs to control asthma or pain from chronic conditions such as arthritis.

Ethical Aspects of Experimentation with Human Subjects In their classic article, Miké and Schottenfeld17 identified some of the ethical issues surrounding experimentation with human subjects, particularly with respect to clinical trials. These issues include informed consent, withhold- ing treatment known to be effective, protecting the interests of the individual patient, monitoring for toxicity and side effects, and deciding when to withdraw a patient from the study.

The ethical aspects of experimentation with human subjects are capable of generating much heated debate and strong emotional responses. It is generally

FigurE 8–5 Clinical trial for two experimental conditions (treatment and placebo) with a crossover design.

Group 1 Intervention (Treatment)

Control (Placebo)

Washout Period

Time

Group 1Group 2

E xp

er im

en ta

l C on

di tio

Baseline (t0) t1 t2

Group 2

Measure Outcome

51589_CH08_Printer.indd 388 09/02/13 12:52 PM

C l i n i C a l t r i a l S 389

agreed that the benefits of participation in an experimental study must clearly outweigh any possible risks to the subject. One side of the issue is that human experimentation, especially with drugs, may bring about certain iatrogenic reac- tions (adverse effects caused by the medications) that could have been avoided had the subject not participated in the experiment. The other viewpoint is that an experimental design (i.e., being in a control condition) requires that medica- tion be withheld from people who might benefit from it.

Both of these issues speak to a characteristic known as clinical equipoise. Stated briefly, clinical equipoise demands that, at the time a clinical trial is being carried out, there is genuine scientific uncertainty in the medical community over which of the drugs or treatments being tested is more efficacious and safer. If there is sufficient evidence that the new drug, for example, is superior to the current drug, it would be unethical to randomize subjects in a clinical trial. One way to test for equipoise is to review the proportion of trials that favor the new treatment; if far better than 50%, one could conclude that equipoise had been violated. A recent review of cancer clinical trials suggests that the principle is being upheld by government-sponsored trials.18

In order to address both of these issues—the withholding of needed medica- tion and the possibility of adverse effects from the medication—experimental drug trials use what is known as a sequential design. Let us consider a clinical trial to evaluate a new drug. In contrast with a sequential design, many drug trials use a preestablished number of subjects who are assigned a priori to the study and control conditions. The results of the trial are evaluated after all sub- jects have been assigned to the conditions of the trial, usually after an extended period of time. In a sequential design, investigators continuously monitor results and add subjects (i.e., there is no pre-established number of subjects). The trial is interrupted as soon as the results are statistically significant and either confirm or reject a positive outcome for the drug being tested. If the drug produces improvement in the patients’ conditions, it then becomes avail- able for use by members of the control group as well (or the drug is discon- tinued if it is found to produce adverse effects). In addition, in some clinical trials, experimental therapies are evaluated only on patients for whom all other treatments have failed.

There has been increasing interest in developing rigorous evaluations of psy- chological therapy interventions.19 Selection of appropriate control subjects in such evaluations poses ethical concerns for subjects in placebo or no-treatment conditions. One concern relates to withholding benefits that might accrue from a psychosocial intervention. This concern is similar to the ethical issue of

51589_CH08_Printer.indd 389 09/02/13 12:52 PM

390 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

withholding an effective drug from a needy control patient. Another issue is finding a control group that is sufficiently comparable to the study group for which the intervention is being evaluated.

The NIH has instituted a policy “. . . that requires oversight and monitoring of all intervention studies to ensure the safety of participants and the validity and integrity of the data.”20 This goal is accomplished through the use of a data and safety monitoring board (DSMB) composed of an independent group of experts who advise investigators and funding agencies. For example, in the case of the Division of Microbiology and Infectious Diseases (DMID), “The primary responsibilities of the DSMB are to 1) periodically review and evaluate the accu- mulated study data for participant safety, study conduct and progress, and, when appropriate, efficacy, and 2) make recommendations to DMID concerning the continuation, modification, or termination of the trial.”21 Although phase III clinical trials are required to have a DSMB, phase I and phase II trials, in certain circumstances, also may require them.

As part of their participation in a clinical trial, study subjects are required to give informed consent by signing an informed consent document. This document describes the risks and benefits of participating in the study and dis- closes the purpose of the clinical trial, how long it will last, and the procedures involved.22 Upon receiving full information about the trial, a potential study subject is in a position to decide whether or not to participate.

Reporting of the Results of Clinical Trials The results of an RCT may impact patient care greatly.23 Readers of the results of an RCT need to comprehend key aspects of the trial—design, conduct, and analysis—as well as to determine the generalizability of its findings. To meet that need, a panel of experts developed the CONSORT statement.24 The acronym CONSORT stands for Consolidated Standards of Reporting Trials. The current revision guides the reporting of randomized trials by providing a 22-item check- list and a flowchart, to be used by those who review, write, and assess the findings from an RCT. One of the main aims of CONSORT is to help authors optimize the quality of their reports of simple RCTs. Figure 8–6 is a CONSORT flow- chart, which describes the flow of subjects through the phases of an RCT.

Summary of Clinical Trials This brief overview of clinical trials reveals a number of their strengths. As opposed to the several varieties of observational studies, clinical trials provide the greatest control over the study situation. The investigator has the ability to

51589_CH08_Printer.indd 390 09/02/13 12:52 PM

C l i n i C a l t r i a l S 391

control the amount of exposure (e.g., drug dosage), the timing and frequency of the exposure, and the period of observation for end points. For large trials, the ability to randomize subjects to study assignments reduces the likelihood that the groups will differ significantly with respect to the distribution of risk factors that might influence the outcome.

Clinical trials have several weaknesses including limited applicability of findings to the larger community and difficulties in enforcing the research pro- tocol. Although randomized controlled clinical trials play an important role in

FigurE 8–6 Flow diagram of the progress through the phases of a randomized trial. Source: Reprinted with the permission from Moher D, Schultz KF, Altman DG. The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials. Lancet. 2001;357:1193.

E n

ro llm

en t

A llo

ca ti

o n

F o

llo w

-u p

A n

al ys

Assessed for eligibility (n = ...)

Randomized (n = ...)

Allocated to intervention (n = ...) Received allocated intervention (n = ...) Did not receive allocated intervention (give reasons)(n = ...)

Lost to follow-up (n = ...) (give reasons) Discontinued intervention (n = ...)(give reasons)

Analyzed (n = ...) Excluded from analysis (give reasons)(n = ...)

Lost to follow-up (n = ...) (give reasons) Discontinued intervention (n = ...)(give reasons)

Allocated to intervention (n = ...) Received allocated intervention (n = ...) Did not receive allocated intervention (give reasons)(n = ...)

Excluded (n = ...) Not meeting inclusion criteria (n = ...) Refused to participate (n = ...) Other reasons (n = ...)

51589_CH08_Printer.indd 391 09/02/13 12:52 PM

392 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

efforts to improve health and the delivery of medical care, they are limited in terms of the scope of their potential impact. Because they require such great control over subjects, they are not typically employed to evaluate the poten- tial efficacy of large-scale public health interventions. The setting for delivery and evaluation of the treatment tends to be artificial, so that the experimenter may find it difficult to determine whether the treatment would work well in the larger community or for unselected patient populations. Obviously, there is necessarily less control over all factors in a community setting than in a clinical trial. This recognition has led some to describe clinical trial evidence of efficacy whereas effectiveness is the term used to describe impact in the “real world.” A second disadvantage is that adherence to protocols may be difficult to enforce, especially if the treatment produces undesirable side effects and presents a significant burden to the subjects. By the time a clinical trial is con- ducted, fairly good evidence usually exists in support of the exposure–disease association; withholding a potentially beneficial treatment from the control arm presents an ethical dilemma.

Community Trials

A community trial is an “[e]xperiment in which the unit of allocation to receive a preventive, therapeutic, or social intervention is an entire community or politi- cal subdivision. Examples include the trials of fluoridation of drinking water and of heart disease prevention in North Karelia (Finland) and California.”4 Community trials are intervention trials at the level of entire communities. Such trials help to determine the potential benefit of new policies and programs such as those for the prevention of obesity. According to Rossi and Freeman, an inter- vention is “Any program or other planned effort designed to produce changes in a target population.”25(p 15) Note that the word community as used here is not meant to be taken literally; a community may well be some other defined unit, such as a county, state, or school district.

Like clinical trials, community trials start by determining eligible communi- ties and their willingness to participate. Permission to enroll the community is typically given by someone capable of providing consent, such as a mayor, gov- ernor, or school board. To be able to evaluate the impact of a program properly, it is desirable to have some baseline measures of the problem to be addressed in the intervention and control communities. Such measures may include, for example, disease prevalence or incidence; knowledge, attitudes, and practice; or purchase of lean relative to fatty cuts of meat. After the relevant baseline

51589_CH08_Printer.indd 392 09/02/13 12:52 PM

C o m m u n i t y t r i a l S 393

measures have been taken, communities are randomized to receive or not receive the intervention. Both study assignments are followed for a period of time, and the outcomes of interest are measured (Figure 8–7). Another name for a trial that randomizes units such as communities to the conditions of an intervention is a cluster randomized trial. “Cluster trials randomize intact social units, such as households, primary care practices, hospital wards, classrooms, neighborhoods and entire communities, to differing intervention arms.”26

FigurE 8–7 Schematic diagram of a community trial.

Measure outcome

Randomly assign intervention

Observe occurrence of disease for a specified period of time

Population A Population B

Do nothing Initiate

program

51589_CH08_Printer.indd 393 09/02/13 12:52 PM

394 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

Years of etiologic research on atherosclerotic disease have contributed greatly to our understanding of risk factors for the disease. Many of these risk factors, such as elevated levels of serum cholesterol and low-density lipoprotein, can be reduced by lowering intake of dietary fat and cholesterol; cigarette smokers can quit; and hypertensive individuals can lower their body weight, exercise more, or use appropriate medication. By the 1970s, enough was known about these risk factors and their potential for modification that a number of community interventions were tested; a few of these interventions include the North Karelia Project, the Minnesota Heart Health Program, and the Stanford Five-City Project (Exhibit 8–5 and Figure 8–8).

As another example of a community trial, the design and methods of the Pawtucket Heart Health Program are described.27 Nine cities in Rhode Island that met certain size and population stability criteria were identified. Pawtucket was randomly selected for the intervention, and an unnamed city with similar

Stanford Five-city project Design, Methods, and results

The Stanford Five-City Project was a major community trial designed to lower risk of cardiovascular diseases. Two treatment cities (Monterey and Salinas) and two control cities (Modesto and San Luis Obispo), located in northern California and rang- ing in size from 35,000 to 145,000 residents, were selected for

study (the fifth city, Santa Maria, was included only for morbidity and mortality surveillance).

Random assignment of cities to treatment and control conditions was precluded by constraints on community selection, particularly concerning independent media markets (Figure 8–8).

The intervention was a six-year, integrated, comprehensive, communi- tywide multifactor risk reduction education program. The interventions had multiple target audiences and used multiple communication channels and settings. Before the intervention began, data from a baseline popula- tion survey were used to develop an overview of knowledge, attitudes, and behavior in treatment communities. The audience was segmented by age,

e x

h Ib

It 8

–5

continues

51589_CH08_Printer.indd 394 09/02/13 12:52 PM

C o m m u n i t y t r i a l S 395

ethnicity, socioeconomic status, overall cardiovascular risk, media use, organizational membership, and motivation to change behavior.

Formative evaluation with these audience subgroups was used to refine educational strategies, programs, and materials. [Formative evaluation is defined later in this chapter.] Social learning theory guided the develop- ment of educational materials.

To evaluate the effect of the intervention on cardiovascular risk fac- tors, four independent cross-sectional surveys of randomly selected house- holds and four repeated surveys of a cohort were conducted. All persons from households randomly chosen from commercial directories, age 12 to 74 years, were eligible for recruitment into the survey. Eligible persons were contacted by mail, telephone, and in person, and they were invited to attend survey centers located in the four cities. Trained health profession- als interviewed participants about their demographic background, health knowledge, cardiovascular risk-related attitudes and behavior, including a 24-hour diet recall, and medical history. Weight was measured using a bal- ance beam, and blood pressure was measured using a semiautomated cuff, after which venous blood samples were drawn for determination of total and high-density lipoprotein cholesterol concentrations. Prior fasting was not required. Cigarette smoking status was confirmed with biochemi- cal testing.

Morbidity and mortality rate data are still being collected and have yet to be analyzed. However, changes in risk factors have been reported. These risk factor results . . . [showed] . . . changes during the six-year educational intervention in the individuals surveyed in the treatment towns compared with the control towns. With the exception of obesity, as measured by body mass index, risk factor changes were in healthful directions in both treatment and control towns, the changes in treatment towns exceeded those in control towns, and there was general consistency of these treatment-control differences across risk factors and in both the cohort and serial cross-sectional surveys. n

Source: Reprinted with permission from Fortmann SP et al. Community intervention trials: reflections on the Stanford Five-City Project experience. American Journal of Epidemiology, Vol 142, No 6, pp. 579–580, © 1995, The Johns Hopkins University School of Hygiene and Public Health.

Exhibit 8–5 continued

51589_CH08_Printer.indd 395 09/02/13 12:52 PM

396 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

sociodemographic characteristics was chosen as the comparison city. The focus of the intervention was to help individuals adopt new, healthy behaviors and to create a supportive physical and behavioral environment. The program tar- geted three dimensions of activities: risk factors (e.g., elevated blood cholesterol, elevated blood pressure, and smoking), behavioral change (e.g., training, aid in development of social support, and maintenance strategies), and commu- nity activation (achieving goals while working through community groups and organizations). During a 7-year intervention period, “over 500 community orga- nizations were involved at some level. These included all 27 public and private schools, most religious and social organizations and larger work sites, all super- markets and many smaller grocery stores, 19 restaurants, and most departments of city government. In addition, a total of 3,664 individuals volunteered to assist in program delivery.”28(p 778) Efforts to create a supportive environment included identification of low-fat foods in grocery stores, installation of an exercise course, nutrition programs at the public library, and highlights of heart-healthy selections

FigurE 8–8 Design of the nine-year Stanford Five-City Project, encompassing the initial surveys and the intervention. The project was subsequently extended to 18 years, including four years of education maintenance activity (to 1990) and six additional years of surveillance (through 1992). C1–C4, cohorts 1–4; I1–I4, independent samples 1–4. Source: Reprinted with permission from SP Fortmann et al, Community intervention trials: reflections on the Stanford Five-City Project experience, American Journal of Epidemiology, vol 142, No 6, p 579, © 1995, The Johns Hopkins University School of Hygiene and Public Health.

Apr 1 2 3 4 5 6

7/87

May Mar

Sep Jun

Apr Feb

Jan

Feb

Apr

Project Year

Survey- Longitudinal Cohort

Survey- Independent Samples

Epidemiologic Surveillance

Community Education

7/78 7/79 7/80 7/81 7/82 7/83 7/84 7/85 7/86

1 78–79

2 79–80

3 80–81

4 81–82

5 82–83

6 83–84

7 84–85

8 85–86

9 86–87

51589_CH08_Printer.indd 396 09/02/13 12:52 PM

C o m m u n i t y t r i a l S 397

on restaurant menus. Efforts to permeate the community, its organizations, and social groupings placed particular emphasis on behavior change, low cost, ease of adoption, and visibility.

Evaluation of the community intervention included cross-sectional surveys in the intervention and control communities at baseline, four times over the course of the intervention, and after the intervention was completed. There was some evidence of lower mean cholesterol and blood pressure levels in Pawtucket, with significantly lower projected disease rates. According to the authors of the report, however, “The hypothesis that projected cardiovascular disease risk can be altered by community-based education gains limited support from these data. Achieving cardiovascular risk reduction at the community level was feasible, but maintaining statistically significant differences between cities was not.”28(p 777)

The Community Intervention Trial for Smoking Cessation was conducted as a multicenter project beginning in 1989. The project sought to reach cigarette smokers and bring about long-term cessation of the habit.29 This intervention trial involved 11 matched pairs of communities throughout the United States and Canada and was believed to affect more than 200,000 adult smokers.

The Community Clinical Oncology Program (CCOP) was established in 1983 in order to increase the input of research findings regarding cancer treat- ment into practice settings.30 This ongoing program seeks to engage commu- nity physicians in clinical trials sponsored by the National Cancer Institute. The vision of the CCOP is the creation of a network of settings for participation in clinical trials for cancer treatment and the development of innovative strategies for prevention and control of cancer. Cooperative groups and a clinical trials net- work have been involved in such cancer prevention and control research activi- ties as the implementation of the Tamoxifen and Finasteride Prevention Trials.

Many community trials employ community-based interventions for preven- tion of HIV. These interventions may be characterized as rigorous and theory based with multiarms, multisites, or using multiple outcome measures. An illus- tration of a multisite study was a randomized controlled community trial that evaluated the impact of an HIV/AIDS prevention program in Tanzania.31 Prevention programs were adopted in Arusha and Kilimanjaro, neighboring areas in Tanzania. The target study sites were all public primary schools, which were stratified for random assignment to intervention or comparison conditions. Investigators reported that intervention groups showed a significant increase in knowledge of AIDS information and had significantly more positive attitudes toward people with AIDS relative to the comparison groups. Another example of an AIDS-related community trial is shown in Exhibit 8–6.

51589_CH08_Printer.indd 397 09/02/13 12:52 PM

398 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

Summary of Community Trials Community trials are crucial because they represent the only way to estimate directly the realistic impact of a change in behavior or some other modifiable exposure on the incidence of disease. They are inferior to clinical trials with respect to the ability to control entrance into the study, delivery of the inter- vention, and monitoring of outcomes. With clinical trials one can potentially randomize a large enough number of subjects to ensure that the study groups are comparable with respect to both measured and unmeasured variables. This state- ment is less true of community trials; fewer study units (e.g., communities or subjects) are capable of being randomized. Accordingly, the likelihood remains that the intervention and control communities may differ with respect to racial composition, education level, age distribution, or some other unmeasured vari- able. In a dynamic population, residents who received the intervention may

project reSpect

Project RESPECT is an example of a randomized controlled trial for prevention of HIV and other STDs.32 The trial sought to establish the efficacy of several counseling methods for reducing high-risk sexual behaviors. This effort involved collaboration among a clinic in Long Beach, California, as well as four other inner-city clinics across the United States.

The project had four interventions. In a study that has several differ- ent interventions, each type is called an arm. The four arms of project RESPECT were enhanced counseling (arm 1), brief counseling (arm 2), and didactic messages (arms 3 and 4). Arms 1 through 3 were followed up actively after enrollment with periodic questionnaires and STD tests for one year. Outcome measures were use of condoms and new occurrence of STDs.

The study sample comprised a total of 5,758 HIV-negative heterosexual patients (14 years of age and older) who presented at the five STD clinics for examinations. Project investigators randomly assigned subjects to one of the four arms of the trial (with approximately equal numbers of subjects in each arm). The results of this study demonstrated that brief counseling sessions for risk reduction (as opposed to didactic sessions) are effective in increasing condom use and preventing the occurrence of new STDs. n

e x

h Ib

It 8

–6

51589_CH08_Printer.indd 398 09/02/13 12:52 PM

C o m m u n i t y t r i a l S 399

move away, and people may move into the community after the intervention has begun; such shifting in the composition of the study population may lead to loss of effect. If there are significant secular trends with respect to the preva- lence of the exposure being modified, then it may be extremely difficult to dem- onstrate an effect of the intervention. For example, the prevalence of cigarette smoking among U.S. adults is decreasing irrespective of specific interventions in a given community. A related phenomenon is that of nonintervention influ- ences. Continuing with the example of cigarette smoking, suppose the American Cancer Society decides to implement a national stop-smoking campaign char- acterized by television, radio, and newspaper messages. Its ability to succeed can affect both control and treatment communities involved in an intervention.

Evaluation of Community Interventions The benefit of an intervention should never be assumed. Rather, it is important for the investigator to quantify and properly evaluate whether a program has achieved its intended results. Evaluations may be undertaken for a variety of reasons, and the form of the evaluation must obviously follow the function. A thorough coverage of evaluation is beyond the scope of this book. The reader is referred to the text by Rossi et al.25 for greater detail. It is our intent to pro- vide an overview as well as some of the specifics germane to assessment. We note that evaluation is, ideally, a continuing activity that comprises four stages following in sequence: formative, process, impact, and outcome (Exhibit 8–7). The examples that we have provided in this chapter apply mostly to impact evaluation.

In most instances, community interventions make use of quasi-experimental designs because random selection of individual study subjects often is not pos- sible in such interventions. As discussed previously, quasi-experimental designs permit manipulation of the study factor but do not permit random assignment of study subjects to study conditions. Community interventions may involve the random selection of entire communities or other units, but this method of sample selection is not equivalent to random selection of individual subjects. The following section describes four major variations on quasi-experimental designs: posttest only; pretest/posttest; pretest/posttest/control; and Solomon four-group. Table 8–3 gives an overview of these four quasi-experimental designs.

Posttest One approach to evaluation is simply for the researcher to make observations only after the program has been delivered. The U.S. educational system typically

51589_CH08_Printer.indd 399 09/02/13 12:52 PM

400 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

Table 8–3 Overview of Quasi-Experimental Study Designs

Type of Study Design Group(s) Pretest Intervention Posttest

Posttest Only (has only one group) Intervention O X X Pretest/Posttest (has only

one group) Intervention X X X

Pretest/Posttest/Control (has two groups)

Intervention Control

X X

X O

X X

Solomon Four-Group (has four groups)

Intervention 1 Intervention 2

X O

X X

Control 1 X O X Control 2 O O X

Note. O = not used; X = used.

the Four Stages of evaluation

Formative Evaluation: “A way of making sure program, plans, procedures, activities, materials, and modifications will work as planned. Begin formative evaluation as soon as the idea for a pro- gram is conceived.”

Process Evaluation: Its purpose “is to learn whether the program is serving the target population as planned and whether the number

of people being served is more or less than expected. Begin process evalua- tion as soon as the program goes into operation.”

Impact Evaluation: Measures “whatever changes the program creates in the target population’s knowledge, attitudes, beliefs, or behaviors.” Requires collection of baseline information before the program starts and subsequent data after first encounter with the target group. Informs pro- gram planners whether they are making progress toward goals.

Outcome Evaluation: “For ongoing programs . . . conduct outcome evalu- ations at specified intervals . . . For one-time programs conduct outcome evaluation after the program is finished. The purpose is to learn how well the program has accomplished its ultimate goal.” n

Source: Adapted from Thompson NJ, McClintock HO. Demonstrating Your Program’s Worth: A Primer on Evaluation for Programs to Prevent Unintentional Injury. 1998, pp. 21–22. Atlanta, GA: National Center for Injury Prevention and Control, Centers for Disease Control and Prevention.

e x

h Ib

It 8

–7

51589_CH08_Printer.indd 400 09/02/13 12:52 PM

C o m m u n i t y t r i a l S 401

follows this pattern for evaluation of classroom instruction. Students enter a class, endure lectures and presentations from the instructor, and then complete examinations to demonstrate that they have acquired knowledge. The advan- tage of this evaluation method is that it makes explanation of results easy. The obvious limitation is that there is usually no measure for baseline comparison. Perhaps the students who score highest on examinations came into the class with greater prior knowledge of the subject. In a similar vein, an evaluation of the effectiveness of a community program to increase consumption of fresh fruits and vegetables (a posttest-only design) would be unable to determine whether eating habits had actually changed.

Pretest/posttest In view of the criticisms leveled at the posttest-only design, one way to improve the evaluation is simply to add a baseline period of observation. The intervention and posttest observations would still be done. By making observations before and after the program is put into effect, one can measure change relative to baseline.

Pretest/posttest/control Although the pretest/posttest design is better than the posttest-only design, the limitation of the former is that there is no measure of external influences that might induce changes in both study groups. One approach to address this deficiency is to add a control group that does not receive the intervention. Observations are made in both intervention and control groups before and after the program. Thus, the effect of any external influences can be estimated by any measured changes that occur in the nonintervention group. The “true” effect of the intervention therefore would be the “observed effect” in the intervention group minus the “nonintervention effect” from external factors as measured in the control group.

Solomon four-group assignment As alluded to earlier in this chapter, the mere fact that individuals are observed may result in behavior change. In a famous experiment designed to determine the effects of varying light intensity on the productivity of women assembling small electronic parts,33 it was observed that a change, whether positive or nega- tive, in the intensity of illumination produced an increase in worker produc- tivity. The investigators reasoned that the workers took the fact that they had been singled out as an experimental group and given a great deal of attention

51589_CH08_Printer.indd 401 09/02/13 12:52 PM

402 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

as evidence that the firm was interested in their personal welfare. Termed the Hawthorne effect (after the site where the experiment was conducted), the phe- nomenon is not specific to social experiments but applies to any circumstance that involves human subjects. This is the basis for the placebo control in clinical trials of pharmacologic agents.

One possible solution to the problem of the Hawthorne effect is to design a study that includes four equivalent groups: two intervention and two con- trol groups. Two are observed before and after the program, and the other two are observed only after the program. Thus, one has the pretest/posttest/control design with two additional arms, neither of which had a pretest observation. This design allows one to determine the effect of both the treatment and the observa- tions. Although the Solomon four-group assignment has been used extensively in social science and educational research, epidemiologists seldom employ it. An obvious reason for its limited use is the increased cost inherent in adding two additional study arms. Furthermore, a reanalysis of the Hawthorne experiment cast considerable doubt on whether the work actually demonstrated any observa- tion effect at all34; some researchers believe that such an effect is probably rare.25

Miscellaneous Issues Several other issues apply to the design of experiments: external validity, statis- tical power, and noncompliance. As defined previously, external validity refers to the generalizability of the results of an intervention study (either clinical or community trial) to a target population beyond the subjects of the study. The generalizability of findings from intervention studies is limited to populations that have demographic characteristics similar to those of the original participants in the intervention. Trickett et al. noted that concerns have been raised about the external validity of information obtained from highly controlled interventions when the results are applied to socioculturally diverse communities.35 External validity is very much connected with the manner in which the study subjects have been selected (e.g., white, middle-class men only, or a more diverse sample). Findings from a clinical trial of a new medication using a sample that has nar- row demographic composition, such as all white men, can be generalized only to a similarly narrow target population. As a result, the research community is required to increase the diversity of human subjects who participate in epidemio- logic studies.

The issue of statistical power is linked to the size of the samples selected for the various conditions of the intervention. It refers to the ability of the study

51589_CH08_Printer.indd 402 09/02/13 12:52 PM

C o m m u n i t y t r i a l S 403

design to detect the hypothesized outcomes of the study. Studies that have larger samples in comparison to smaller samples have greater statistical power and reduced measurement error; of course, larger samples are more costly to collect than smaller samples.

Finally, noncompliance is a factor in experimental study design that has the potential to vitiate or nullify the effects of the intervention. For example, if the subjects in a randomized controlled clinical trial do not take a prescribed medication, the outcome of the trial may be reduced or may not be detectable. Table 8–4 summarizes intervention studies (comparisons of clinical trials and community trials).

Table 8–4 Summary of Intervention Studies (Clinical Trials Versus Community Trials)

Clinical Trials Community Trials

Prospective study that often uses a true experimental design.

Prospective study that often uses a quasi- experimental design.

Used for testing new medications, procedures (therapeutic trials), and substances to prevent disease (prophylactic trials).

Designed to produce changes (especially health-related) in a target population.

Focused on the individual. Focused on the community, school district, county, or state.

Randomization of subjects to the exposure (treatment or control conditions), in order to promote comparability of subjects with respect to measured and unmeasured variables.

Fewer study units can be randomized to the study conditions. Intervention and control communities may differ with respect to race, education, age, and other unmeasured variables.

Manipulation (external control of exposure) is possible.

Duration ranges from days to years (in the case of some trials).

Generally longer time duration than clinical trials.

Strict human subject protocol to regulate eligibility of subjects; informed consent to participate required.

Informed consent of participating subjects required for collection of baseline, outcome, and other information.

Participation restricted to a highly selected group of individuals with disease, at high risk for disease, or volunteers.

Participation involves all members of a targeted community; intervention is applied more generally than in a clinical trial.

Tightly controlled in terms of eligibility, delivery of intervention (treatment), and monitoring of outcome.

Less rigid control over intervention in comparison to clinical trials.

Appropriate for testing narrow hypotheses, such as those related to vaccine treatment testing and evaluating new drugs.

Appropriate for determining the potential benefit of new policies and programs, such as those related to health.

Usually two groups (experimental and control) are compared in terms of an outcome.

Baseline measures taken in the intervention and control communities with follow-up at the end of the intervention.

continues

51589_CH08_Printer.indd 403 09/02/13 12:52 PM

404 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

Conclusion

This chapter has provided an overview and description of experimental study designs, which are called intervention studies in epidemiology. Intervention studies include both controlled experiments and quasi-experiments, examples being clinical trials and community trials (community interventions). Clinical trials focus upon individuals and community trials upon groups (i.e., outcomes that occur in the community). While both clinical trials and community trials can take the form of controlled experiments, community trials, because of their complexity, are often conducted as quasi-experimental designs. A limitation of many community trials is that they produce weak or inconsistent findings.35 In addition, challenges arise from the gap between obtaining knowledge from community interventions and applying that knowledge to the community.35 Related to the types of study designs is their hierarchy with respect to ability to generate causal inferences. Other issues covered in this chapter were the appli- cations and phasing of clinical trials as well as randomization and crossover designs.

Clinical Trials Community Trials

Outcomes of clinical trials (clinical end points): compare rates of disease, death, recovery; outcome of interest is measured in the intervention and control arms of the trial to evaluate efficacy.

Outcomes of community trials: programs may be evaluated according to the four stages of evaluation: formative, process, impact, and outcome. Pre- and posttest measures compared.

Advantages: provide general control over the study situation with respect to amount of exposure, timing, frequency, and period of observation.

Advantages: estimate realistically the impact of behavior change or other modifiable exposure in the incidence of disease.

Disadvantages: artificial setting of the treatment delivery and evaluation result in lack of generalizability (external validity). Difficulty of adherence to a protocol when trial produces side effects or preliminary results show it to be efficacious.

Disadvantages: loss of effect may occur due to the shifting in the study population composition. Secular trends with respect to prevalence of the exposure being modified make it difficult to determine the effect of the intervention.

Blinding and double blinding are used to promote objectivity.

Blinding and double blinding not generally used.

Variations: crossover design, a switch of treatments for the study patients.

Variations: posttest only, pretest/posttest, pretest/posttest/control, and the Solomon four-group assignment.

Examples: the folic acid randomized clinical trial; effectiveness of education efforts to reduce STDs. Refer to text.

Examples: the Pawtucket Heart Healthy Program; Project RESPECT. Refer to text.

Table 8–4 continued

51589_CH08_Printer.indd 404 09/02/13 12:52 PM

S t u d y Q u e S t i o n S a n d e x e r C i S e S 405

Study Questions and Exercises

1. Define the following terms: a. Experimental study b. Quasi-experimental study c. Intervention study d. Controlled clinical trial

2. Compare and contrast the two categories of intervention studies (con- trolled clinical trials versus community interventions) discussed in this chapter. What are some advantages and disadvantages of controlled experimental designs in comparison to quasi-experimental designs?

3. Explain the purposes of blinding and randomization in clinical trials. 4. Describe the phases of a clinical trial to license a vaccine or new medicine. 5. What is meant by a crossover design? Distinguish between a planned and

unplanned crossover. 6. Can you foresee ethical problems that might arise in the development of

new pharmaceutical agents for HIV? 7. What are some of the design strategies that are used to improve the valid-

ity of community trials? Questions 8 and 9 pertain to the chapters Study Designs: Cohort Studies; Study Designs: Ecologic, Cross-Sectional, Case-Control; and Experimental Study Designs.

8. Epidemiologic studies of the role of a suspected factor in the etiology of a disease may be observational or experimental. The essential difference between experimental and observational studies is that in experimental studies: (Choose one answer.) a. The study and control groups are equal in size. b. The study is prospective. c. The study and control groups are always comparable with respect to

all factors other than the exposure. d. The investigator determines who shall be exposed to the suspected

factor and who shall not. e. Controls are used.

9. From the descriptions provided, identify the type of study design that is being described: a. Smoking histories are obtained from all patients entering a hospi-

tal who have lip cancer and are compared with smoking histories of patients with cold sores who enter the same hospital.

51589_CH08_Printer.indd 405 09/02/13 12:52 PM

406 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

b. The entire population of a given community is examined, and all who are judged free of bowel cancer are questioned extensively about their diet. These people are then followed for several years to see whether their eating habits will predict their risk of developing bowel cancer.

c. To test the efficacy of vitamin C in preventing colds, army recruits are randomly assigned to two groups: one given 500 mg of vitamin C daily, and one given a placebo. Both groups are followed to determine the number and severity of subsequent colds.

d. The physical examination records of the incoming first-year class of 1935 at the University of Minnesota are examined in 1980 to see whether the freshmen’s recorded height and weight at the time of admission to the university were related to their chance of developing coronary heart disease by 1981.

e. Fifteen hundred adult men who worked for Lockheed Aircraft were initially examined in 1951 and were classified by diagnostic criteria for coronary artery disease. Every three years they have been reexam- ined for new cases of the disease; attack rates in different subgroups have been computed annually.

f. A random sample of middle-aged sedentary women was selected from four census tracts, and each subject was examined for evidence of osteoporosis. Those found to have the disease were excluded. All others were randomly assigned to either an exercise group, which fol- lowed a two-year program of systematic exercise, or a control group, which had no exercise program. Both groups were observed semian- nually for incidence of osteoporosis.

g. Questionnaires were mailed to every 10th person listed in the city tele- phone directory. Each person was asked to provide his or her age, sex, and smoking habits and to describe the presence of any respiratory symptoms during the preceding 7 days.

References

1. Glossary. National Institutes of Health. Office of Extramural Research (OER) Web site. http://grants.nih.gov/grants/glossary. Accessed August 8, 2012.

2. Petitti D. Commentary: hormone replacement therapy and coronary heart disease: four lessons. Int J Epidemiol. 2004;33:461-3.

3. Writing Group for the Women’s Health Initiative Investigators. Risks and benefits of estrogen plus progestin in healthy postmenopausal women: principal

51589_CH08_Printer.indd 406 09/02/13 12:52 PM

r e f e r e n C e S 407

results from the Women’s Health Initiative randomized controlled trial. JAMA. 2002;288(3):321-333.

4. Porta M, ed. A Dictionary of Epidemiology. 5th ed. New York: Oxford University Press; 2008.

5. Susser M. Some principles of study design for preventing HIV transmission: rigor or reality. Am J Public Health. 1996;86:1713–1716.

6. Meinert CL. Clinical Trials: Design, Conduct, and Analysis. New York: Oxford University Press; 1986.

7. Psaty BM, Siscovick DS, Weiss NS, et al. Hypertension and outcomes research: from clinical trials to clinical epidemiology. Am J Hypertens. 1995;9:178–183.

8. Medical Research Council Vitamin Study Research Group. Prevention of neu- ral tube defects: results of the Medical Research Council Vitamin Study. Lancet. 1991;338:131–137.

9. Hibbard BM. The role of folic acid in pregnancy with particular reference to anaemia, abruption, and abortion. J Obstet Gynaecol Br Commonw. 1964;71:529–542.

10. O’Donnell LN, San Doval A, Duran R, O’Donnell C. Video-based sexually transmit- ted disease patient education: its impact on condom acquisition. Am J Public Health. 1995;85:817–822.

11. Begg N, Miller E. Role of epidemiology in vaccine policy. Vaccine. 1990;8:180–189. 12. Weinstock H, Berman S, Cates W Jr. Sexually transmitted diseases among American

youth: incidence and prevalence estimates, 2000. Perspect Sex Reprod Health. 2004;36:6–10.

13. Giuliano AR, Lee JH, Fulp W, et al. Incidence and clearance of genital human papillomavirus infection in men (HIM): a cohort study. Lancet. 2011;377(9769): 932–40.

14. Lowy DR, Schiller JT. Reducing HPV-associated cancer globally. Cancer Prev Res. 2012;5(1):18–23.

15. Engelking C. Clinical trials: impact evaluation and implementation considerations. Semin Oncol Nurs. 1992;8:148–155.

16. Gordis L. Epidemiology. 3rd ed. Philadelphia, PA: Elsevier Saunders; 2004. 17. Miké V, Schottenfeld D. Observations on the clinical trial. Clin Bull. 1972;2:

130–135. 18. Djulbegovic B. The paradox of equipoise: the principle that drives and limits thera-

peutic discoveries in clinical research. Cancer Control. 2009;16(4):342–7. 19. Schwartz CE, Chesney MA, Irvine MJ, Keefe FJ. The control group dilemma in clin-

ical research: applications for psychosocial and behavioral medicine trials. Psychosom Med. 1997;59:362–371.

20. National Institutes of Health. Further guidance on a data and safety monitoring for phase I and phase II trials. Notice: OD-00-038, June 5, 2000. http://grants.nih.gov/ grants/guide/notice-files/NOT-OD-00-038.html. Accessed July 8, 2012.

21. National Institutes of Health, National Institute of Allergy and Infectious Diseases, Division of Microbiology and Infectious Diseases. Policy and Guidelines for Data and Safety Monitoring. http://www3.niaid.nih.gov/research/resources/DMIDClinRsrch/ PDF/dsm_pol_guide.pdf. Accessed July 8, 2012.

22. U.S. National Institutes of Health. An Introduction to Clinical Trials. http:// clinicaltrials.gov/ct/info/whatis. Accessed March 6, 2010.

23. Begg C, Cho M, Eastwood S, et al. Improving the quality of reporting of randomized controlled trials: the CONSORT statement. JAMA. 1996;276:637–639.

51589_CH08_Printer.indd 407 09/02/13 12:52 PM

408 C h a p t e r 8 e x p e r i m e n ta l S t u d y d e S i g n S

24. Moher D, Schultz KF, Altman DG. The CONSORT statement: revised recommendations for improving the quality of reports of parallel-group randomized trials. Lancet. 2001;357:1191–1194.

25. Rossi PH, Freeman HE. Evaluation: A Systematic Approach. 7th ed. Newbury Park, CA: Sage; 2003.

26. Weijer C, Grimshaw JM, Taljaard, et al. Ethical issues posed by cluster randomized trials and health research. Trials. 2011;12:100.

27. Carleton RA, Lasater TM, Assaf A, et al. The Pawtucket Heart Health Program, I: an experiment in population-based disease prevention. R I Med J. 1987;70:533–538.

28. Carleton RA, Lasater TM, Assaf AR, et al. The Pawtucket Heart Health Program: community changes in cardiovascular risk factors and projected disease risk. Am J Public Health. 1995;85:777–785.

29. COMMIT Research Group. Community Intervention Trial for Smoking Cessation (COMMIT): summary of design and intervention. J Natl Cancer Inst. 1991;83: 1620–1628.

30. National Cancer Institute. Community Clinical Oncology Program. History and vision. Available at: http://ccop.cancer.gov/about/history-vision. Accessed November 5, 2012.

31. Klepp K, Ndeki SS, Leshabari MT, et al. AIDS education in Tanzania: promoting risk reduction among primary school children. Am J Public Health. 1997;87:1931–1936.

32. Kamb ML, Fishbein M, Douglas JM, et al. Efficacy of a risk-reduction counseling to prevent human immunodeficiency virus and sexually transmitted diseases: a random- ized controlled trial. JAMA. 1998;280:1161–1167.

33. Roethlisberger FJ, Dickson W. Management and the Worker. Cambridge, MA: Harvard University Press; 1939.

34. Franke RH, Kaul JD. The Hawthorne experiments: first statistical interpretation. Am Sociol Rev. 1978;43:623–642.

35. Trickett EJ, Beehler S., Deutsch C, et al. Advancing the science of community-level interventions. Am J Public Health. 2011;101:1410–1419.

51589_CH08_Printer.indd 408 09/02/13 12:52 PM

3chapte r

409

9chapte r

Measures of Effect

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● explain the meaning of absolute and relative effects ●● calculate and interpret the following measures: risk difference,

population risk difference, etiologic fraction, and population etio- logic fraction

●● define the role of statistical tests in epidemiologic research ●● apply Hill’s criteria for evaluation of epidemiologic associations

CHAPTER OUTLINE

I. Introduction II. Absolute Effects

III. Relative Effects IV. Statistical Measures of Effect V. Evaluating Epidemiologic Associations

VI. Models of Causal Relationships VII. Conclusion

VIII. Study Questions and Exercises

51589_CH09_Printer.indd 409 09/02/13 12:53 PM

410 C h a p t e r 9 M e a s u r e s o f e f f e C t

Introduction

One of the major challenges to an epidemiologist is presentation of research findings in a meaningful and interpretable manner. Much of the basic vocabu- lary in this context is termed measures of effect in “epi-speak.” An effect measure is “[a] quantity that measures the effect of a factor on the frequency or risk of a health outcome. Three such measures are ATTRIBUTABLE FRACTIONS, which measure the fraction of cases due to a factor; risk and rate differences, which measure the amount a factor adds to the risk or rate of a disease; and risk and rate ratios, which measure the amount by which a factor multiplies the risk or rate of disease.”1

This chapter will extend the discussion of an odds ratio (OR) and a relative risk (RR). Classified as measures of relative effects, these two measures were defined and illustrated elsewhere in the text. Several additional measures of effect, useful when one is evaluating the potential implications of an exposure–disease associa- tion, are introduced. For the science of public health, correct extrapolation of the findings of individual studies to the larger population is critical. Armed with knowledge of the measures presented in this chapter, public health practitioners can be more effective in planning programs, delivering resources, and evaluating proposed interventions.

We will also demonstrate that the exposure–disease association can have quite different implications for risk to the individual and impact upon the population. A risk that is relatively modest for the individual can be very meaningful for the population. Thus, individuals may be less inclined to lower their risk factor status for a particular adverse health outcome when quality of life is reduced; from the population perspective, public health officials may be more inclined to advocate for reduction of that same risk factor. An example will be provided later in the chapter.

Absolute Effects

One of the simplest ways to compare the disease burden in two groups is to cal- culate the absolute difference in disease frequency. This type of comparison also is referred to as a difference measure of association, or attributable risk.2 An abso- lute effect may be based on differences in incidence rates, cumulative incidence, prevalence,3 or mortality.4 An attributable risk is also known as a rate difference or risk difference.3,5

51589_CH09_Printer.indd 410 09/02/13 12:53 PM

a b s o l u t e e f f e C t s 411

Measures of risk differences aid in assessing the impact of a component cause, which is one of a set of multiple causes linked to a particular effect. With respect to the issue of component causes in epidemiology, Rothman’s comments are par- ticularly relevant:

A cause is an act or event or a state of nature which initiates or permits, alone or in conjunction with other causes, a sequence of events resulting in an effect. A cause which inevitably produces the effect is sufficient. The inevitability of disease after a sufficient cause calls for qualification: disease usually requires time to become manifest, and during this gestation, while disease may no longer be preventable, it might be fortuitously cured, or death might intervene . . . Most causes that are of interest in the health field are components of sufficient causes but are not suffi- cient in themselves . . . Causal research focuses on components of sufficient causes, whether necessary or not.6(p 588)

Along the lines of Rothman’s statements, it is asserted elsewhere in this book that many chronic diseases, for example, coronary heart disease (CHD), result not from a single exposure but rather from the combined influences of several exposures, such as environmental and lifestyle factors, that operate over a long time period. Therefore, removal of only one of the exposures (e.g., high serum cholesterol) that leads to a chronic disease (i.e., CHD) would not result in com- plete elimination of the disease; other risk factors would still be operative and contribute to the rate of disease. One approach to estimate the realistic poten- tial impact of removing an exposure from the population is to calculate the risk difference in disease frequency (i.e., incidence rates) between the exposed and the nonexposed groups. According to Rothman, a risk difference “represents the incidence rate of disease with the exposure as a component cause.”3(p 35)

Risk Difference

Ie Incidence rate ofdisease in exposed gro= uup

Incidence rate ofdisease in nonexpo neI = ssed group

Risk difference: The difference between the incidence rate of disease in the exposed group (Ie) and the incidence rate of disease in the nonexposed group (Ine); risk difference = Ie - Ine.

51589_CH09_Printer.indd 411 09/02/13 12:53 PM

412 C h a p t e r 9 M e a s u r e s o f e f f e C t

As mentioned earlier, the measure of disease frequency used in the determination of absolute effects may be incidence density, cumulative inci- dence, prevalence, or mortality. Thus, to be perfectly accurate, when the measure of disease frequency is cumulative incidence, the term risk difference could be used. When incidence density measures are used as the measure of disease fre- quency, the term rate difference is most appropriate. For prevalence and mortality, the most precise terms would be prevalence difference and mortality difference, respectively. Regardless of the measure of disease frequency used, the basic con- cept of absolute effects is the same: The measure of disease frequency among the nonexposed group is subtracted from the measure of disease frequency among the exposed group.

As an example, hip fractures (often among persons with osteoporosis) pose a significant public health burden for the elderly population. (Figure 9–1 illus- trates osteoporosis.) In 2002, there were an estimated 44 million U.S. adults with osteoporosis or low bone mass.7 Investigators at the Mayo Clinic in Rochester, Minnesota, examined seasonal variations in fracture rates, comparing the rates during winter with those during summer.8 For women younger than age 75, the

Figure 9–1 Osteoporosis: A risk factor for fractures.

51589_CH09_Printer.indd 412 09/02/13 12:53 PM

a b s o l u t e e f f e C t s 413

incidence Ie of fractures per 100,000 person-days was highest in the winter (0.41), and the incidence Ine was lowest in the summer (0.29). The risk difference between the two seasons (Ie - Ine) was 0.41 - 0.29, or 0.12 per 100,000 person-days.

Population Risk Difference

Ip = Overall incidence rate of disease in a population Pe = Proportion of the population exposed Pne = Proportion of the population not exposed

Population risk difference is defined as a measure of the benefit to the population derived by modifying a risk factor. This measure addresses the question of how many cases in the whole population can be attributed to a particular exposure. To understand fully the concept of population risk difference, consider that the incidence rate (risk) of disease in the population (denoted by the symbol IP), for the simplest case of a dichotomous exposure (exposed or nonexposed), is made up of four components: the incidence rate (risk) of disease in the exposed group (Ie); the incidence rate (risk) of disease in the nonexposed group (Ine); the pro- portion of the population exposed (Pe); and the proportion of the population not exposed (Pne). The nonexposed group is sometimes called the reference group. The relationship among the four components may be expressed by the following formula:

I I P I PP e e ne ne= +( )( ) ( )( )

Ignore, for the moment, the proportion exposed (Pe) and the proportion not exposed (Pne). If one were to remove the effects of exposure associated with higher rates of disease, the overall rate of disease in the population then would be expected to decrease to the rate observed among the nonexposed, or reference, group. Thus, subtraction of the rate (risk) of disease among the nonexposed (Ine) from the rate of disease among the population (IP) provides an indication of the potential impact of a public health intervention designed to eliminate the harm- ful exposure.

Population risk difference: The difference between the rate (risk) of disease in the nonexposed segment of the population (Ine) and the overall rate (IP).

51589_CH09_Printer.indd 413 09/02/13 12:53 PM

414 C h a p t e r 9 M e a s u r e s o f e f f e C t

Just as for the risk difference, the measures of disease frequency used to calculate population risk differences may be generalized to include the cumu- lative incidence (risk), incidence density (rates), prevalence, or mortality. Remember: Risk difference is the risk in the exposed minus the risk in the non- exposed; population risk difference is the risk in the population minus the risk in the nonexposed subset of the population.

As another example, nonsteroidal anti-inflammatory drugs (NSAIDs) are the most frequently used drugs in the United States, with 111 million prescrip- tions filled annually and an estimated cost of about $2 billion annually for over-the-counter NSAIDs.9 The risks associated with use of NSAIDs include significant upper gastrointestinal bleeding; especially among older persons.10 To examine the association of NSAID usage and peptic ulcer disease among elderly persons, Smalley et al.11 determined the incidence rate of serious ulcer disease among users and nonusers of NSAIDs. The study was based on 103,954 elderly Tennessee Medicaid recipients followed from 1984 to 1986. A total of 1,371 patients were hospitalized with peptic ulcer disease after 209,068 person-years of follow-up. The incidence (density) rate of peptic ulcer disease in the study popu- lation (IP) was calculated to be 6.6 per 1,000 person-years [(1,371/209,068) × 1,000]. The rate (Ine) among nonusers of NSAIDs was only 4.2 per 1,000 person-years. The population risk difference (IP - Ine) was 6.6 - 4.2, or 2.4 per 1,000 person-years. The risk difference may be computed also: It was known that the observed incidence rate (Ie) of peptic ulcer disease among users of NSAIDs was 16.7 per 1,000 person-years. Therefore, the risk difference (Ie - Ine) was 16.7 - 4.2, or 12.5 per 1,000 person-years.

Relat ive Effects

Interpretation of the absolute measures of effect can sometimes be enhanced when expressed relative to a baseline rate. For example, an RR provides an esti- mate of the magnitude of an association between exposure and disease.5 Such a ratio also can be described as a relative effect. Note that all relative effects contain an absolute effect in the numerator.

Previously, we defined RR as the ratio of the cumulative incidence rate in the exposed (Ie) to the cumulative incidence rate in the nonexposed (Ine), or Ie/Ine. This is actually a simplification of the true formula for RR, in which the numera- tor is Ie - Ine (the risk difference). If one divides both terms in the numerator by Ine, one is left with the formula (Ie/Ine) - (Ine/Ine). The first term, Ie/Ine, was previously defined as the RR. Because any number (or variable) divided by itself

51589_CH09_Printer.indd 414 09/02/13 12:53 PM

r e l a t i v e e f f e C t s 415

is 1, the second term becomes 1, and the expression becomes RR = 1. Typically the = 1 is ignored. Occasionally, however, one may encounter statements such as “30% greater risk among the exposed”; this statement implies that the RR ratio of Ie/Ine is 1.3 but that the 1 has been subtracted. The interpretation is exactly the same. RRs between 1.0 and 2.0 sound bigger when stated as a percentage, however (e.g., 1.3 versus 30%).

Etiologic Fraction One of the conceptual difficulties with RR is that the rate of disease in the refer- ent (nonexposed) group is not necessarily 0. In fact, for a common disease that is theorized to have multiple contributing causes, the rate may still be quite high in the referent group as a result of other causes in addition to the exposure of interest. One implication of multiple contributing causes is that, even in the absence of exposure to the single factor of interest, a number of cases still would have developed among the nonexposed population. An approach to estimating the effects due to the single exposure factor is to compute the etiologic fraction. It is defined as the proportion of the rate in the exposed group that is due to the exposure. Also termed attributable proportion or attributable fraction, it can be estimated by two formulas. To estimate the number of cases among the exposed that are attributable to the exposure, one must subtract from the exposed group those cases that would have occurred irrespective of membership in the exposed population.

Etiologic fraction e ne

= −( )I I I

(Eq. 1)

Note that the difference between Equation 1 and RR is the rate in the denom- inator—Ie instead of Ine. The numerator represents an acknowledgment that not all the cases among the exposed group can be fairly ascribed to the exposure; some fraction would have occurred anyway, and this fraction is estimated by the rate in the nonexposed group. This formula can be applied to data from cohort or cross-sectional studies. The appropriate measures of disease frequency must be utilized: cumulative incidence, incidence density, or mortality from cohort stud- ies or prevalence of disease from cross-sectional studies.

With a little arithmetic, it is possible to express Equation 1, the formula for etiologic fraction, in another convenient form. If one considers Equation 1 as

51589_CH09_Printer.indd 415 09/02/13 12:53 PM

416 C h a p t e r 9 M e a s u r e s o f e f f e C t

two separate fractions, one obtains 1 − (Ine/Ie). Note that (Ine/Ie) is merely the reciprocal of the original definition of RR. Thus, one is left with 1 − −(1/RR). If one expresses the 1 as RR/RR, the formula requires only an estimate of RR, obviously beneficial for those situations in which the actual incidence rates are unknown (Equation 2). Thus, this formula may be applied when the data at hand, whether from a report or a published article, include only the summary measures. More important, because the OR provides an estimate of RR, this formula is applicable to data from case-control studies.

Etiologic fraction

RR RR

= −( )1

(Eq. 2)

For example, what fraction of peptic ulcer disease in elderly persons is attrib- utable to NSAIDs? Recall from the previous example that Ine was 4.2 and Ie was 16.7 per 1,000 person-years.

11 The risk difference was computed to be 16.7 - 4.2, or 12.5 per 1,000 person-years. The etiologic fraction from Equation 1 is 12.5 ÷ 16.7, or 74.9%. Thus, roughly three-fourths of the cases of peptic ulcer disease that occurred among NSAID users were attributed to that exposure.

To demonstrate that both formulas are equivalent, one may compute the etio- logic fraction using Equation 2. To do this, one must first compute the RR. In this example the answer is 16.7 ÷ 4.2, or 3.98. The etiologic fraction is therefore 2.98 divided by 3.98. Both formulas should yield the same answer, an outcome that the reader may wish to verify.

In general, low RRs equate to a low etiologic fraction, and high RRs equate to a high etiologic fraction. A reasonable question to ask at this point is: What does risk difference reveal beyond what one could already infer from the RR? Perhaps this question is best answered with an illustration. Take the case of two diseases, A and B, and two exposure factors, X and Y. The rate of disease A is 2 per 100,000 per year among individuals exposed to factor X and 1 per 100,000 per year among those not exposed to factor X. The rate of disease B is 400 per 100,000 per year among individuals exposed to factor Y and 200 per 100,000 per year among those not exposed to factor Y. Therefore, for either disease the RR associated with the relevant exposure is 2 (i.e., 2 ÷ 1 or 400 ÷ 200). Exposure factors X and Y both appear to pose a significant health hazard, a doubling of risk of disease. Consider what is obtained by examining the risk difference: For disease A the risk difference is 2 - 1 or 1 per 100,000 per year, and for disease

51589_CH09_Printer.indd 416 09/02/13 12:53 PM

r e l a t i v e e f f e C t s 417

B the risk difference is 400 - 200 or 200 per 100,000 per year. Although the RRs for factor X and factor Y are the same, the risk differences for the two factors are quite disparate. If one were to design an intervention to improve public health, the RRs for factors X and Y would not be terribly informative. The risk differ- ence calculations would suggest, however, that control of exposure Y might pay greater dividends than control of exposure X (ignoring, for the moment, critical issues such as cost and feasibility).

Population Etiologic Fraction As we have seen, from the perspective of those with a disease, the etiologic frac- tion gives an indication of the potential benefit of removing a particular exposure to a putative disease factor. That is, does a particular exposure account for 5% of the etiology of the disease or 95%? An alternative perspective to consider is that of the population. The population etiologic fraction provides an indication of the effect of removing a particular exposure on the burden of disease in the popula- tion. A possible scenario is one in which a dichotomous (present or absent) expo- sure factor is associated with risk of disease and 25% of the population is exposed to the factor. As was pointed out earlier, the total rate of disease in the population may be thought of as a weighted average of the rate of disease among the 25% of the population exposed and the rate of disease among the 75% of the population not exposed. (Note that the concept of a weighted average is applied to the direct method of age adjustment.) If the offending exposure is reduced, the lower limit of disease rate that can be achieved is the background rate observed among the nonexposed segment of the population. Again, two formulas for the population etiologic fraction will be presented.

The population etiologic fraction (also termed the attributable fraction in the population) represented by Equation 3 is the proportion of the rate of disease in the population that is due to the exposure. It is calculated as the population risk difference divided by the rate of disease in the population.

Population etiologic fraction p ne

= −( )I I I

(Eq. 3)

As an example, consider again the study of NSAIDs and peptic ulcer dis- ease among elderly persons.11 Ine was 4.2, IP was 6.6 per 1,000 person-years, and the population risk difference was computed to be 6.6 - 4.2, or 2.4 per

51589_CH09_Printer.indd 417 09/02/13 12:53 PM

418 C h a p t e r 9 M e a s u r e s o f e f f e C t

1,000 person-years. For this example, the population etiologic fraction is (2.4/6.6) × 100 = 36.4%. Therefore, if everyone in the population stopped tak- ing NSAIDs, the rate of peptic ulcer disease would decrease by more than one- third. Notice that compared with the etiologic fraction of those with the disease, this value of 36.4% is far less than the etiologic fraction of 74.9%.

When the incidence rate in the population is unknown, an alternative for- mula (Equation 4) may be applied. This formula requires information about two components: the RR of disease associated with the exposure of interest, and the prevalence of the exposure in the population (Pe).

Population etiologic fraction

R e

= −P

( )

(

RR − + ×

1 1 100

) (Eq. 4)

0 13 3 98 1 0 13 3 98 1 1

100 0 387 1 387

1 . ( . )

. ( . ) . .

− − +

× = × 000 27 9= . %

Case-control studies do not allow an estimation of disease rates in the total population or in the nonexposed population and, therefore, the Equation 3 population etiologic fraction cannot be used. Equation 4, however, lends itself to interpretation of data from case-control studies because the OR can be sub- stituted for RR. The missing piece of information is the prevalence of the expo- sure in the population. Recall from the chapter titled “Study Designs: Ecologic, Cross-Sectional, Case-Control” that the purpose of a control group is to provide an estimate of the expected frequency of the exposure of interest. With certain assumptions, the frequency of exposure among the control group can be used to approximate the overall frequency of exposure in the population.

Example 1: Given that the prevalence (Pe) of current NSAID use in the study by Smalley et al.11 was 0.13, compute the population etiologic fraction using Equation 4. The RR had been previously determined to be 3.98. Plugging the values for RR and Pe into Equation 4, one obtains:

This answer is slightly lower than the results obtained by using Equation 3 for the population etiologic fraction because the prevalence figure Pe did not include former or indeterminate users of NSAIDs. Mathematically the two formulas yield the same result, however.

51589_CH09_Printer.indd 418 09/02/13 12:53 PM

r e l a t i v e e f f e C t s 419

Example 2: Suppose you are dealing with an exposure that confers a high RR for disease (e.g., RR = 20), but the prevalence (Pe) of the exposure in the popula- tion is low (e.g., 1 per 100,000). Compare the etiologic fraction with the popu- lation etiologic fraction using these data. Compute the etiologic fraction using Equation 2. We obtain:

20 1 20

100 95

0 00001 20 1 0 00001 20 1 1

− × =

− − +

. ( ) . ( )

1100 0 019= . %

Thus, 95% of the cases that occurred among the exposed were attributable to the exposure. Because the exposure was rare in the population, however, it contributed little to the total disease rate.

These two examples illustrate that the impact of an exposure on a population depends upon:

●● the strength of the association between exposure and resulting disease.

●● the overall incidence rate of disease in the population. ●● the prevalence of the exposure in the population.

One may also infer that exposures of high prevalence and low RR can have a major impact on the public’s health. For example, an individual’s risk of car- diovascular disease mortality associated with an elevated serum cholesterol level may be low. That is, the etiologic fraction is low. However, because a substantial proportion of the population has high cholesterol (i.e., hypercholesterolemia has a high prevalence), the benefit to the population from reducing cholesterol could be substantial. In contrast, the foregoing example of a rare exposure with a high RR for disease demonstrates that a single exposure factor can account for the vast majority of cases of disease among the exposed but that removal of that particu- lar exposure from the population will have little impact on the overall incidence of disease.

From Equation 4, we obtain:

51589_CH09_Printer.indd 419 09/02/13 12:53 PM

420 C h a p t e r 9 M e a s u r e s o f e f f e C t

Stat ist ical Measures of Effect

In addition to the preceding methods of expressing epidemiologic study results (absolute and relative effects), epidemiologists frequently employ and rely on statistical tests to help interpret observed associations. An illustration of statisti- cal tests arises from a study of the effects of passive smoking (by parents) on the prevalence of wheezing respiratory illness among their children.12 The results indicate that mothers who smoked at the time of the survey were 1.4 times more likely to report wheezing respiratory illness among their children than mothers who did not smoke. The reasons for this outcome may be as follows:

1. Passive smoking by a parent does, in fact, increase children’s risk of wheez- ing respiratory illness.

2. Some additional exposure has not been properly allowed for in the analysis.

3. The results represent nothing more than a chance (random) finding.

Only after options 2 and 3 have been ruled out can one reasonably conclude that passive smoking increases children’s risk of wheezing respiratory illness.

Significance Tests Underlying all statistical tests is a null hypothesis, usually stated as, “There is no difference in population parameters among the groups being compared.” The parameters may consist of the prevalence or incidence of disease in the popula- tion. For example, the prevalence or incidence might represent an actual count of cases of disease identified by surveillance programs, or by other means such as positive serological evidence of infection from elevated antibody titers. A discus- sion of the particular statistical test to be employed, the choice of which is deter- mined by a number of considerations, is beyond the scope of this book. Suffice it to say that in deciding whether to fail to reject or to reject the null hypothesis, a test statistic is computed and compared with a critical value obtained from a set of statistical tables. The significance level is the chance of rejecting the null hypothesis when, in fact, it is true.

The P Value The P value indicates the probability that the findings observed could have occurred by chance alone. The converse is not true: A nonsignificant difference is not necessarily attributable to chance alone. For studies with a small sample size, the sampling error is likely to be large, which may lead to a nonsignificant test even when the observed difference is caused by a real effect.

51589_CH09_Printer.indd 420 09/02/13 12:53 PM

s t a t i s t i C a l M e a s u r e s o f e f f e C t 421

Confidence Interval A confidence interval (CI) is a statistical measure that is considered by many epidemiologists to be more meaningful than a point estimate; the latter is a sin- gle number—for example, a sample mean, an incidence rate, or an RR—that is used to estimate a population parameter. A CI is expressed as a computed interval of values that, with a given probability, contains the true value of the population parameter.1 The degree of confidence is usually stated as a percent- age; the 95% CI is commonly used. Although it is beyond the scope of this book to demonstrate how to construct CIs, it is important, nonetheless, to know how to interpret them. A CI can be interpreted as a measure of uncertainty about a parameter estimate (e.g., a mean, OR, RR, or incidence rate).

●● In terms of utility, a 95% CI contains the “true” population estimate 95% of the time.

●● Thus, if one samples a population 100 times, the 95% CI will contain the true estimate (i.e., the population parameter) 95 times. Alternatively, if one were to repeat the study 100 times, one would observe the same outcome 5 times just by chance.

●● CIs are influenced by the variability of the data and the sample size.

The hypothetical example presented in Table 9–1 reports the OR for a case- control study with three different sample sizes. The exposure, disease, study pop- ulation, and survey instrument are the same in all three cases. In fact, everything is identical except for the size of the study groups.

Perhaps the first sample size was obtained for a small-scale pilot study. Twenty cases and 20 controls are included. An OR of 2 is observed, but the 95% CI includes 1; the results are therefore consistent with no association. Suppose, alternatively, that one is able to study 50 per group instead of only 20. The same point estimate of association is observed, and the 95% CI also includes the null value of 1. The degree of precision of the magnitude of the OR is improved; the

Table 9–1 Odds Ratios, P Values, and 95% Confidence Intervals for a Case-Control Study with Three Different Sample Sizes

Sample Size

Parameter Computed 20 50 500

OR 2.00 2.00 2.000 P 0.50 0.20 0.001 95% CI 0.5, 7.7 0.9, 4.7 1.5, 2.6

51589_CH09_Printer.indd 421 09/02/13 12:53 PM

422 C h a p t e r 9 M e a s u r e s o f e f f e C t

interval is narrower, but the results are still not statistically significant. In the final scenario there are unlimited resources, and one is able to study 500 indi- viduals in each group. With this extra effort and expense, the same study results are obtained: an OR of 2. The larger sample size has allowed for a more precise estimate of the effect to be obtained (the 95% CI is narrower). The outcome is now statistically significant, as the null value of 1 is now excluded from the 95% CI for the OR. The point to be made is that the estimate of an effect from an epidemiologic study is not necessarily incorrect just because the sample size is small; a small sample size merely may not produce precise results (i.e., there is a wide CI around the estimate of effect).

Clinical Versus Statistical Significance The preceding discussion of statistical significance should suggest to the reader that P values are only a part of the evaluation of the validity of epidemiologic data.

One also should be aware of an important caveat of large sample sizes: Small differences in disease frequency or low magnitudes of RR may be statistically significant. Such minimal effects may have no clinical significance, however. For example, suppose an investigator conducted a survey among pregnant women in urban and suburban populations to assess folic acid levels. Furthermore, suppose that there were 2,000 women in each group, that the average folic acid levels differed by 1.3%, and that this difference was statistically significant. In this example, the sample size was large enough to detect subtle differences in expo- sure; biologically and clinically, such small differences may be quite insignificant.

The converse of the large sample size issue is that, with small samples, large differences or measures of effect may be clinically important and worthy of addi- tional study. Thus, mere inspection of statistical significance could cause over- sight. The lack of statistical significance may simply be a reflection of insufficient statistical power to detect a meaningful association. Statistical power is defined as, “. . . the ability of a study to demonstrate an association if one exists. The power of a study is determined by several factors, including the frequency of the condition under study, the magnitude of the effect, the study design, and sample size.”1 One example of the magnitude of the effect is how large a relative risk is found, that is, whether RR = 1, 5, or 10.

Another problem inherent in the use of statistical significance testing is that it may lead to mechanical thinking. In his Cassel Memorial Lecture to the Soci- ety for Epidemiologic Research Annual Meeting in June 1995, Rothman13 noted that John Graunt’s famous epidemiologic contributions were made in the absence of a knowledge of statistical significance testing.

51589_CH09_Printer.indd 422 09/02/13 12:53 PM

e v a l u a t i n g e p i d e M i o l o g i C a s s o C i a t i o n s 423

Evaluating Epidemiologic Associat ions

The ability to evaluate critically epidemiologic associations reported in the literature is a realistic and attainable goal for the public health practitioner. Although the basic skills to perform such an evaluation are covered in this book, there is no substitute for practice. As an aid to the reader, five key questions that should be asked are presented below.

Could the Association Have Been Observed by Chance? The major tools that are used to answer this question are statistical tests. Although any public health practitioner should have a basic understanding of biostatistics, he or she should not underestimate the value of a competent biostatistician as a source of help. A small P value (i.e., highly significant result) for an observed association should provide some assurance that the results were not obtained simply by chance, but one must always remember that a very small P value does not imply that the association is real.

Could the Association Be Due to Bias? The term bias refers to systematic errors, and is discussed in detail elsewhere in this text. At this point it is sufficient to say that one should critically evaluate how the study groups were selected, how the information about exposure and disease was collected, and how the data were analyzed. Errors at any of these stages may lead to results that are not valid.

Could Other Confounding Variables Have Accounted for the Observed Relationship? Confounding refers to the masking of an association between an exposure and an outcome because of the influence of a third variable that was not considered in the design or analysis. The issue of confounding and how to control it is covered elsewhere in this book. Based on one’s understanding of the natural history and epidemiology of a disease, one needs to consider whether important known con- founding factors have been omitted from the study.

To Whom Does This Association Apply? Although population-based samples are important in epidemiologic research, and although these sampling procedures enhance the likelihood of generalizabil- ity of results, they do not guarantee such an outcome. Furthermore, in some

51589_CH09_Printer.indd 423 09/02/13 12:53 PM

424 C h a p t e r 9 M e a s u r e s o f e f f e C t

situations a great deal can be learned from an unrepresentative study sample. If a study has been properly conducted among a certain stratum of the popula- tion, for example, white women between the ages of 55 and 69 who live in the state of Iowa, then one could certainly generalize to other white women who live in the Midwest. If the diets of the women in Iowa are indicative of the diets of American women of this age group, however, then any observed diet–disease associations may apply to a much broader population.

In addition to the representativeness of the sample, many investigators believe that participation rates are crucial to the validity of epidemiologic findings. Par- ticipation rates, the percentage of a sample that completes the data collection phase of a study, must be at a sufficiently high level. For example, some top-tier public health journals may not publish a report in which the participation rate was less than 70%. Ironically, high participation rates do not necessarily ensure generalizability, and in certain circumstances generalizability may be high even if participation rates are low. Consider a study of a potential precursor of colorectal cancer: the rate of proliferation of cells in the rectal mucosa. Measurement of the proliferation rate of the rectal epithelium requires a punch biopsy, obtainable as part of a sigmoidoscopy or colonoscopy procedure. Suppose one conducts a case-control study of patients with adenomatous polyps (a known precursor of colorectal cancer) and controls free from colon polyps or cancer. Cases are found to have significantly higher rates of rectal cell proliferation than the controls. Because of the invasive nature of the procedure, however, the participation rates are only 10% among the eligible cases and 5% among eligible controls. Does this necessarily mean that the findings cannot be generalized? The key issue is whether the exposure of interest influenced the decision process of the eligible cases and controls to participate. In this example, it is difficult to imagine how an unmeasured characteristic, such as the rate of rectal cell proliferation, could possibly influence participation. Therefore, despite participation rates that usu- ally would be regarded as unacceptable, one may still be able to generalize the findings, especially the underlying biology, to a broader population.

Does the Association Represent a Cause-and-Effect Relationship? The answer to this question is determined by careful consideration of each of Hill’s criteria of causality.14 These criteria are: strength of the association, tem- porality, dose-response, consistency, biologic plausibility, specificity, analogy, and coherence.

51589_CH09_Printer.indd 424 09/02/13 12:53 PM

M o d e l s o f C a u s a l r e l a t i o n s h i p s 425

Models of Causal Relat ionships

Drawing upon the concepts presented earlier in the chapter, this section introduces models of disease causation. Relationships between suspected disease- causing factors and outcomes fall into two general categories: not statistically associated and statistically associated.15 Among statistical associations are non- causal and causal associations. Possible types of associations are formatted in Figure 9–2.

We have already considered the role of statistical significance in evaluating an association and noted that evaluation of statistical significance is used to rule out the operation of chance in producing an observed association; a nonstatisti- cally associated (independent) relationship is shown in box A of the diagram (left side).

As shown in Figure 9–2, a statistical association may be either noncausal or causal. What is meant by a noncausal (secondary) association? Suppose factor C is related to disease outcome A. The association may be due to the operation of a third factor B that is related to both C and A. Thus, the association between C and A is secondary to the association of C with B and C with A. For example, periodontal disease (C) is associated with chronic obstructive pulmonary disease (A).16 One possible explanation for this association is the secondary association

Figure 9–2 Map of possible associations between disease-causing factors and outcomes. Source: Data from B MacMahon and TF Pugh, Epidemiology Principles and Methods. Boston, MA: Little, Brown and Company; 1970.

Relationships between factors and outcomes

B. Statistically associatedA. Not statistically associated (independent)

a. Indirectly associated

b. Directly associated

2. Causally associated 1. Noncausally (secondarily)

associated

51589_CH09_Printer.indd 425 09/02/13 12:53 PM

426 C h a p t e r 9 M e a s u r e s o f e f f e C t

of smoking (B) with both periodontal disease (C) and chronic obstructive pulmonary disease (A). This model suggests that the increased risk of chronic obstructive pulmonary disease associated with periodontal disease is related to the role that smoking may play as a cofactor in both conditions. Here is a map of a secondary association: C ← B → A.1

With respect to causal associations, the relationship between factor and out- come may be indirect or direct. An indirect causal association involves the oper- ation of an intervening variable, which is a variable that falls in the chain of association between C and A. An illustration of an indirect association is the postulated relationship between low education levels (C) and obesity (A) among men.17 Men who have lower education levels tend to be more obese than those who have higher education levels. It is plausible that the relationship between C and A operates through the intervening variable of lack of leisure time physi- cal activity (B). An indirect association involves an intervening variable in the association between C and A. This relationship may be formatted as follows: C → B → A.1 Note that the arrow between C and B has been reversed in con- trast with an indirect noncausal association.

Multiple Causality The foregoing section provided models of causality that employ more than one factor. As stated earlier in this chapter, the measure risk difference implies mul- tivariate causality by isolating the effects of a single exposure from the effects of other exposures. The example on NSAIDs examined the difference between risk of peptic ulcer among users and nonusers of NSAIDs, where the risk difference was 12.5 per 1,000 person-years. The risk of peptic ulcer caused by other expo- sures was 4.2 per 1,000 person-years.

The issue of disease causality is exceedingly complicated. To describe exposure– disease relationships, epidemiologists have developed complex models of disease causality. These models acknowledge the multifactor causality of diseases, even those that seem to have “simple” infectious agents. Often, these models involve an ecologic approach by relating disease to one or more environmental factors. “The requirement that more than one factor be present for disease to develop is referred to as multiple causation or multifactorial etiology.”18(p 27) Examples of several influential models are the:

●● epidemiologic triangle ●● web of causation ●● wheel model ●● pie model

51589_CH09_Printer.indd 426 09/02/13 12:53 PM

M o d e l s o f C a u s a l r e l a t i o n s h i p s 427

Web of causation The web of causation is “. . . a popular METAPHOR for the theory of sequential and linked multiple causes of diseases and other health states.”1 The web of cau- sation implicates broad classes of events and represents an incomplete portrayal of reality.15 Although the web of causation for most diseases is complex, one may not need to understand fully the causality of any specific disease in order to prevent it. An example of the web of causation of avian influenza is provided in Figure 9–3. Follow the infection of the human host from the virus reservoir in wild birds. As of 2007, the virus had not mutated into a form that could be spread readily from person to person.

Wheel model The wheel model is similar to the epidemiologic triangle and web of causation with respect to involving multiple causality (Figure 9–4). Observe that the model explains the etiology of disease by calling into play host and environment

Figure 9–3 The web of causation for avian influenza.

Virus reservoir (wild birds)

Wild birds contact

domestic flocks

Virus mutation in domestic birds

Human contact with poultry

Human contact with bird feces

Poultry raising at home

Sick birds sold Sick birds

eaten

Cross- contamination

of food with poultry juices

Poor hand washing

techniques

Coinfection in animals (e.g.,

swine)

Reassortment of genetic material in human host

Human host affected

Virus mutation

Human to human spread

Emergence of pandemic

Has not occurred at present

Speculative

51589_CH09_Printer.indd 427 09/02/13 12:53 PM

428 C h a p t e r 9 M e a s u r e s o f e f f e C t

interactions. Environmental components are biologic, social, and physical. The circle designated as “host” refers to human beings or other hosts affected by a disease. The circle called “genetic core” acknowledges the role that genetic fac- tors play in many diseases. The wheel model de-emphasizes specific agent factors and, instead, differentiates between host and environmental factors in disease causation. The biologic environment is relevant to infectious agents, by taking into account the environmental dimensions that permit survival of microbial agents of disease.

A wheel model may be used to account for the occurrence of childhood lead poisoning.18 In this example, preschool children are typical hosts. The physi- cal environment provides many opportunities for lead exposure from lead-based paint in older homes, playground equipment, candy wrappers, and other sources. Some children ingest paint chips from peeling surfaces as a result of pica, the predilection to eat nonfood substances. Because lead-based paints often are located in poorer neighborhoods that have substandard housing, the social environment is associated with childhood poisoning. Limited access to medi- cal care in such communities may restrict screening of preschool children for lead exposure. Elimination of childhood lead poisoning requires visionary public

Figure 9–4 The wheel model of man–environment interactions. Source: Modified with permission from JS Mausner and S Kramer, Mausner & Bahn Epidemiology: An Introductory Text, 2nd ed. Philadelphia, PA: W.B. Saunders Company;1974, p 36.

Host (e.g., Humans)

Genetic Core

Biologic Environment Social

Environment

Physical Environment

51589_CH09_Printer.indd 428 09/02/13 12:53 PM

M o d e l s o f C a u s a l r e l a t i o n s h i p s 429

health leadership to advocate for detection of lead-based paints and other sources of environmental lead exposure as well as the implementation of screening pro- grams. Such efforts will help to protect vulnerable children against the sequelae of lead poisoning.

Pie model Another model of multiple causality (multicausality) is the causal pie model.19 As Figure 9–5 shows, the model indicates that a disease may be caused by more than one causal mechanism (also called a sufficient cause), which is defined as “a set of minimal conditions and events that inevitably produce disease.”19(p S144)

Each causal mechanism is denoted in Figure 9–5 by the numerals I through III. An example of different causal mechanisms for a disease is provided by the etiol- ogy of lung cancer: lung cancer caused by smoking; lung cancer caused by expo- sure to ionizing radiation; and lung cancer caused by inhalation of carcinogenic solvents in the workplace.

Rothman and Greenland note that, “A given disease can be caused by more than one causal mechanism, and every causal mechanism involves the joint action of a multitude of component causes.”19(p S145) The component causes, or factors, are denoted by the letters shown within each pie slice. A single let- ter indicates a single component cause. A single component could be common to each causal mechanism (shown by the letter A that appears in each pie); in

Figure 9–5 Three sufficient causes of disease. Source: From KJ Rothman and S Greenland, Causation and causal inference in epidemiology, Am J Public Health, 2005; vol 95, p S145. Reprinted with permission from the American Public Health Association.

B E

C D

B H

F G

III

C J

F I

One Causal Mechanism Single Component Cause

51589_CH09_Printer.indd 429 09/02/13 12:53 PM

430 C h a p t e r 9 M e a s u r e s o f e f f e C t

other cases, the component causes for each causal mechanism could be different for each mechanism (shown by the letters that differ across the pies). Returning to the lung cancer example, a common factor that could apply to all causal mechanisms for lung cancer is a genetic predisposition for cancer. Several other component causes might be different for each causal mechanism involved in the etiology of lung cancer.

In models of multicausality, most of the identified component causes are neither necessary nor sufficient causes (defined in the section on absolute effects). Accordingly, it is possible to prevent disease when a specific component cause that is neither necessary nor sufficient is removed; nevertheless, when the effects of this component cause are removed, cases of the disease will continue to occur.

Conclusion

This chapter covered two new measures of effect—absolute and relative effects— that may be used as aids in the interpretation of epidemiologic studies. In addi- tion, the chapter presented guidelines that should be taken into account when one is interpreting an epidemiologic finding. Absolute effects, the first variety of which is called risk differences, are determined by finding the difference in measures of disease frequency between exposed and nonexposed individuals. A second type of absolute effect, called population risk difference, is found by computing the difference in measures of disease frequency between the exposed segment of the population and the total population. Relative effects are char- acterized by the inclusion of an absolute effect in the numerator and a refer- ence group in the denominator. One type of relative effect, the etiologic fraction, attempts to quantify the amount of a disease that is attributable to a given expo- sure. The second type of relative effect, the population etiologic fraction, pro- vides an estimate of the possible impact on the population rates of disease that can be anticipated by removal of the offending exposure. With respect to inter- pretation of epidemiologic findings, one should be cognizant of the influence of sample size upon the statistical significance of the results. Large sample sizes may lead to clinically unimportant, yet statistically significant, results; small sample sizes may yield statistically nonsignificant results that are clinically important. Therefore, we presented a series of five questions that should be asked when one attempts to interpret an epidemiologic observation. The chapter closed with a discourse on causal models, which may be particularly instructive when trying to interpret epidemiologic data.

51589_CH09_Printer.indd 430 09/02/13 12:53 PM

s t u d y Q u e s t i o n s a n d e x e r C i s e s 431

Study Questions and Exercises

1. Calculate the etiologic fraction when the RR for disease associated with a given exposure is 1.2, 1.8, 3, and 15.

2. The impact of an exposure on a population does not depend upon: a. the strength of the association between exposure and disease. b. the prevalence of the exposure. c. the case fatality rate. d. the overall incidence rate of disease in the population. The next seven questions (3–9) are based on the following data: The death rate per 100,000 for lung cancer is 7 among nonsmokers and 71 among smokers. The death rate per 100,000 for coronary thrombosis is 422 among nonsmokers and 599 among smokers. The prevalence of smoking in the population is 55%. (If necessary, refer to the chapter on cohort studies for formulas for RR.)

3. What is the RR of dying of lung cancer for smokers versus nonsmokers? 4. What is the RR of dying of coronary thrombosis for smokers versus

nonsmokers? 5. What is the etiologic fraction of disease due to smoking among individu-

als with lung cancer? 6. What is the etiologic fraction of disease due to smoking among individu-

als with coronary thrombosis? 7. What is the population etiologic fraction of lung cancer due to smoking? 8. What is the population etiologic fraction of coronary thrombosis due to

smoking? 9. On the basis of the RR and etiologic fractions associated with smoking

from lung cancer and coronary thrombosis, which one of the following statements is most likely to be correct? a. Smoking seems much more likely to be causally related to coronary

thrombosis than to lung cancer. b. Smoking seems much more likely to be causally related to lung cancer

than to coronary thrombosis. c. Smoking seems to be equally causally related to both lung cancer and

coronary thrombosis. d. Smoking does not seem to be causally related to either lung cancer or

coronary thrombosis. e. No comparative statement is possible between smoking and lung can-

cer or coronary thrombosis.

51589_CH09_Printer.indd 431 09/02/13 12:53 PM

432 C h a p t e r 9 M e a s u r e s o f e f f e C t

10. A cohort study was conducted to investigate the association between coffee consumption and anxiety in a population-based sample of adults. The data are presented in Appendix 9. a. What is the RR of anxiety associated with coffee use? b. Calculate the risk (rate) difference. c. What is the etiologic fraction? d. Determine the population etiologic fraction.

References

1. Porta M. A Dictionary of Epidemiology. 5th Ed. New York: Oxford University Press, 2008.

2. Kelsey JL, Thompson WD, Evans AS. Methods in Observational Epidemiology. New York: Oxford University Press; 1986.

3. Rothman KJ. Modern Epidemiology. Boston: Little, Brown; 1986. 4. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic Research: Principles and

Quantitative Methods. Belmont, CA: Lifetime Learning; 1982. 5. Hennekins CH, Buring JE. Epidemiology in Medicine. Boston: Little, Brown; 1987. 6. Rothman KJ. Causes. Am J Epidemiol. 1976;104:587–592. 7. National Osteoporosis Foundation. America’s Bone Health: The State of Osteoporosis

and Low Bone Mass in Our Nation. Washington, DC: National Osteoporosis Foundation; 2002.

8. Jacobsen SJ, Sargent DJ, Atkinson EJ, et al. Population-based study of the contribution of weather to hip fracture seasonality. Am J Epidemiol. 1995;141:79–83.

9. Laine I. GI risk and risk factors of NSAIDs. J Cardiovas Pharmacol. 2006;47 (Suppl 1):S60–66.

10. Risser A, Donovan D, Heintzman J, Page T. NSAID prescribing precautions. Am Fam Physician. 2009; 80: 1371-1378.

11. Smalley WE, Ray WA, Daugherty JR, Griffin MR. Nonsteroidal anti-inflammatory drugs and the incidence of hospitalizations for peptic ulcer disease in elderly persons. Am J Epidemiol. 1995;141:539–545.

12. Stoddard JJ, Miller T. Impact of parental smoking on the prevalence of wheezing respiratory illness in children. Am J Epidemiol. 1995;141:96–102.

13. Rothman gives Cassel Memorial Lecture at SER. Epidemiol Monit. 1995;16:1. 14. Hill AB. The environment and disease: association or causation? Proceedings of the

Royal Society of Medicine. 1965; 58:295–300. 15. MacMahon B, Pugh TF. Epidemiology Principles and Methods. Boston: Little, Brown;

1970. 16. Hyman JJ, Reid BC. Cigarette smoking, periodontal disease, and chronic obstructive

pulmonary disease. J Periodontol. 2004;75:9–15. 17. Ward H, Tarasuk V, Mendelson R. Socioeconomic patterns of obesity in Canada:

modeling the role of health behaviour. Appl Physiol Nutr Metab. 2007;32:206–216. 18. Mausner JS, Kramer S. Mausner & Bahn Epidemiology: An Introductory Text. 2nd

ed. Philadelphia: Saunders; 1985. 19. Rothman KJ, Greenland S. Causation and causal inference in epidemiology. Am

J Public Health. 2005;95:S144–S150.

51589_CH09_Printer.indd 432 09/02/13 12:53 PM

C o h o r t s t u d y d a t a f o r C o f f e e u s e a n d a n x i e t y 433

appendix

9 Cohort Study Data for Coffee Use and Anxiety

Table 9A–1 Cohort Study Data for Coffee Use and Anxiety

Anxiety

Coffee Use Yes No Total

Yes 500 9,500 10,000 No 200 19,800 20,000

51589_CH09_Printer.indd 433 09/02/13 12:53 PM

51589_CH09_Printer.indd 434 09/02/13 12:53 PM

3chapte r

435

10chapte r

Data Interpretation Issues

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● distinguish between random and systematic errors ●● state and describe three main sources of bias ●● identify techniques to reduce bias at the design and analysis phases

of a study ●● define what is meant by the term confounding and provide three

examples ●● describe the methods that can be used to control confounding

CHAPTER OUTLINE

I. Introduction II. Validity of Study Designs

III. Sources of Error in Epidemiologic Research IV. Techniques to Reduce Bias V. Methods to Control Confounding

VI. Bias in Analysis and Publication VII. Conclusion

VIII. Study Questions and Exercises

51589_CH10_Printer.indd 435 09/02/13 12:53 PM

436 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

Introduction

As Exhibit 10–1 suggests, findings from epidemiologic studies are often quite newsworthy. One of the real dangers of obtaining epidemiologic information solely from media reports, however, is the selective nature of the coverage; the media may focus on the one positive result among a larger quantity of negative data. Another more troubling issue is whether the study being reported was even scientifically valid. One must not only understand the study’s results and impli- cations, but also be able to evaluate critically the study’s design and methodol- ogy, a task that requires considerably greater knowledge and skills than merely assimilating the findings. Despite the peer review process adopted by scientific journals, methodologically flawed studies do appear in print and do attract media attention. One cannot assume that just because a study was published in a reputable journal the findings should not be questioned.

press coverage: Leaving Out the Big picture

In the past . . . years, thorough readers of the Los Angeles Times would have learned about an extraordinary range of potential can- cer causes. Among these putative hazards of modern life are hot dogs, breast implants, dioxin, stress, asbestos, allergy drugs, gas leaks, living in Orange County, tubal ligation, sunscreen, Asian food, pesticides, vasectomy, liquor, working in restaurants, Retin-

A, vegetables, dietary fat, delayed child-bearing, impurities in meat, and lesbianism. This litany of fear was accompanied by a similar, although shorter, series of reports on dietary habits and lifestyles that may reduce cancer risk. Parallel coverage appeared in other newspapers and magazines and on television. To many scientists, though, the media would do well to curb its appetite for such news. The problem, many researchers say, is that journalists often misunderstand the context of the research. Because of the limitations of risk factor epidemiology, most individual studies can- not produce authoritative findings. “Articles published in medical journals are often misconstrued by the lay press to be more definitive than they really are,” says Larry Freedman, a biostatistician at the National Cancer

e x

h iB

it 1

0 –1

continues

51589_CH10_Printer.indd 436 09/02/13 12:53 PM

V a l I D I t y o f s t u D y D e s I g n s 437

Where do possible deficiencies in media reports and published research leave the public health practitioner? To gain a more complete picture of any particular report, one really should retrieve and read the original article firsthand. More important, one should have the basic skills to evaluate critically the report as to selection of study subjects, measurement of exposure and outcome, analysis of data, and interpretation of results. This chapter provides a foundation for such skills.

Validity of Study Designs

The validity of a study is defined as “The degree to which the inference drawn from a study, [is] warranted when account is taken of the study methods, the rep- resentativeness of the study sample, and the nature of the population from which it is drawn.”1 Study validity embodies two components: internal and external validity.

Internal Validity A study is said to have internal validity when there have been proper selection of study groups and a lack of error in measurement. The goal is to be able to ascribe

Exhibit 10–1 continued

Institute. “Broccoli prevents cancer, garlic prevents cancer—all these things do appear in the literature. But epidemiologists understand very well that these studies are far from definitive. It’s only when a body of evidence exists over many, many studies that epidemiologists should really get seri- ous about giving the public advice.” Instead of presenting surveys of the big, evolving picture, he and others say the media tend to report each new study in isolation, as a new breakthrough. Such reporting, some scientists say, is encouraged by press releases put out by journals and researchers’ institutions. But whoever is to blame, says Noel Weiss, an epidemiologist at the University of Washington, Seattle, the result is “just too many false alarms. When we do have a serious message, I fear it won’t be heeded because of the large number of false messages.” n

51589_CH10_Printer.indd 437 09/02/13 12:53 PM

438 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

any observed effect to the exposure under investigation. Thus, the manner of selection of cases and controls, or exposed and nonexposed groups, must be critically reviewed. Maintenance of internal validity also necessitates appropriate measurement of the following:

●● Exposure. Was characterization of exposure based on a questionnaire? If so, was the questionnaire administered in person or as a telephone survey, or was it mailed to the study participant for self-administration? To illustrate, the investigator may collect some types of data with reasonable accuracy by using mailed questionnaires, but careful probing that can be achieved only through in-person interviews may be required to collect other types of data. Were the instruments validated? It is important to know whether the questionnaires used actually measured what they purported to measure. Is the reliability known? If the instrument were administered to the same individual on separate occasions, would it provide the same response? Were biologic samples collected to quantify exposure? If so, were the procedures to collect the samples standardized according to timing of collection? For example, suppose you were interested in urinary hormone levels of pre- menopausal women. It would be important to know whether the samples were all collected during the same phase of the menstrual cycle. Were the laboratory assays used appropriate? Although it is not possible to cover all variations on the theme, this brief overview indicates some of the types of questions that should be pondered with respect to exposure assessment.

●● Outcome. Whether the outcome of interest is a particular disease, behavior, or intermediate marker, the criteria used to define the outcome should be fully described. Was the outcome based solely on self-report, documen- tation in a medical record, or was an examination performed by trained health professionals according to a standard protocol? How were the sub- jects with and without the outcome of interest identified? Were all eligible subjects successfully located? Did a high proportion participate? If the study was prospective in nature, were all end points identified? Was there loss to follow-up? Did loss to follow-up differ between the exposed and nonexposed groups? Clearly, a number of important considerations pertain to assessment of the outcome and the formation of the study groups.

●● Association between exposure and disease. The two preceding categories reflect aspects of measurement of exposure and outcome; this category relates to assessment of the association between them by raising the fol- lowing questions: Were the data properly analyzed? Was adjustment made for extraneous factors that might influence the results? Some types of data

51589_CH10_Printer.indd 438 09/02/13 12:53 PM

V a l I D I t y o f s t u D y D e s I g n s 439

analysis require certain assumptions about the nature (distribution) of the data. Were the assumptions tested? Do the assumptions appear reasonable given the context of the study? Are there crucial analyses that appear to have been omitted?

External Validity The preceding section discussed internal validity, a requirement that must be satisfied for a study to have external validity. External validity is a more encompassing process than the ability to extrapolate from a sample population to a target population. External validity implies the ability to generalize beyond a set of observations to some universal statement. According to the Dictionary of Epidemiology, “A study is externally valid, or generalizable, if it allows unbiased inferences regarding some other target population beyond the subjects in the study.”1

The basic process of generalizing study results is neither mechanical nor sta- tistical, for one must understand which conditions are relevant or irrelevant to the generalization. Although representativeness of the sample is a condition of external validity, generalizability also is independent of how representative of the target population the study groups are; this statement applies particularly when one uses epidemiology to improve understanding about the biologic basis of a disease.

An example is a feeding study designed to evaluate the utility of plasma carotenoids (compounds found in plant foods thought to have anticarcinogenic properties) as a marker of vegetable intake.2 Subjects (volunteers who agreed to participate) were randomized into a crossover feeding study of 4 experimental diets of 9 days each. Thus, after spending 9 days consuming a particular experi- mental diet, the participants were “crossed over” to one of the other diets. It should be noted that volunteers for health studies tend to be more educated and health conscious than the general population. In this particular study, how- ever, 11 exclusion criteria were applied to restrict the pool of volunteers. These included a medical history of gastrointestinal disorders, food allergies, weight loss or gain greater than 4.5 kg within the past year, major changes in eating hab- its within the past year, exercise regimens requiring significant short-term dietary changes, antibiotic use within the past 3 months, body weight greater than 130% of ideal, current treatment for a diagnosed disease, alcohol intake greater than 2 drinks per day, oral contraceptive use, and unwillingness to consume all foods provided in the study. Although these exclusion criteria greatly improved

51589_CH10_Printer.indd 439 09/02/13 12:53 PM

440 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

the internal validity of the study, they might have decreased the generalizability of the study if the subjects enrolled were physiologically atypical with respect to how their plasma carotenoid levels responded to a high-vegetable diet.

The preceding example reaffirms the notion that clinical trials are often ini- tially based on a highly selected subgroup of patients. Nonetheless, the infor- mation gleaned from such trials often can be generalized to a much broader category of patients.

Sources of Error in Epidemiologic Research

In the context of epidemiologic research, one should consider two categories of error: random and systematic. Random errors reflect fluctuations around a true value of a parameter (such as a rate or a relative risk) because of sampling vari- ability. They can occur as a result of poor precision, sampling error, or variability in measurement. Systematic errors refer to measurement biases.

Factors That Contribute to Random Error Poor precision This type of random error occurs when a study factor is not measured sharply. Consider the analogy of aiming a rifle at a target that is not in focus. The target may correctly yield the proper direction in which one should be aiming, but the blurry picture makes it difficult to hit the bull’s-eye, causing bullets to scatter all over the target. Increasing the sample size of a study or the number of measure- ments will yield greater precision. For example, in the Bogalusa Heart Study, a prospective study of the early natural history of cardiovascular disease in a small, rural Louisiana community, an average of six blood pressure readings was used to characterize an individual child’s blood pressure.3 Each child was randomly assigned to two of three trained observers who each made three independent blood pressure measurements. By taking the average of six readings, the random error was reduced, thereby improving precision.

Sampling error Sampling error is a type of error that arises when obtained sample values (statis- tics) differ from the values (parameters) of the parent population. Sampling error is relevant to all types of epidemiologic studies: cross-sectional, case-control, cohort, or intervention.

In epidemiologic research, one wishes to make inferences about a target popula- tion without necessarily having to measure each member of the target population.

51589_CH10_Printer.indd 440 09/02/13 12:53 PM

s o u r C e s o f e r r o r I n e p I D e m I o l o g I C r e s e a r C h 441

The target population may be the general population of the entire United States or a specified subset (e.g., residents of California; children aged 5–9; African Americans; or Hmong residents of the Minneapolis-St. Paul area of Minnesota). For this reason, one typically selects samples from the target population that are of a more manageable size for study than would occur if every member of the target population was examined.

When one conducts a case-control study of colorectal cancer in the state of Utah, the study group of cases may be considered a sample of all cases of colorec- tal cancer in the United States. When one draws a sample from a larger popula- tion, the possibility always exists that the sample selected is not representative of the target population. Nonrepresentative samples may occur without any inten- tion or fault of the investigators even if subjects are randomly selected. To a certain extent, sampling error may be thought of as just plain bad luck of the draw, just as there can be an unusual run of cards in poker or run of colors in a roulette game. Although there is no way to prevent a nonrepresentative sample from occurring, increasing the size of the sample can reduce the likelihood of its happening.

Variability in measurement The validity of a study will be enhanced greatly if the data that are collected are objective, reliable, accurate, and reproducible. Even under the best of circum- stances, however, errors in measurement can and do occur. For example, the Bogalusa Heart Study investigators were concerned about the stability of labora- tory measures over long periods of time.3 To determine consistency in measure- ment, a blind sample from randomly selected individuals was included when samples of blood were sent to the laboratory for analysis. In fact, perfect agree- ment was rarely achieved despite the fact that the same procedures were used and the samples were from the same individuals and collected at the same time. The lack of agreement in results from time to time reflects random error inherent in the type of measurement procedure employed.

Factors That Contribute to Systematic Errors A much more serious problem for the validity of a study than random errors is systematic errors, or bias. As the definition of bias implies, systematic errors can be introduced at any point in an investigation. These errors can be conve- niently grouped into three broad categories: selection bias, information bias, and confounding.

51589_CH10_Printer.indd 441 09/02/13 12:53 PM

442 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

Selection bias Selection bias refers to “[d]istortions that result from procedures used to select subjects and from factors that influence participation in the study.”1 Selection bias arises when the relation between exposure and disease is different for those who participate and those who would be theoretically eligible for study but do not participate.4 Such bias may occur during the follow-up period of a study, during the period of recruitment for the study, or even before the study begins. For example, the healthy worker effect represents a source of bias that may occur when only employed individuals, such as an occupational cohort, are eligible for a study. Workers typically are relatively healthy people who have had the opportunity to find and maintain employment. Physically or mentally disabled individuals may not have enjoyed a similar opportunity and thus would not be represented in the study.

One type of selection bias results from nonresponse. Given that it is unusual for a study to have the ability to compare responders and nonresponders, a useful exercise is to estimate how the results might have changed if all of the nonrespon- dents were from the exposed group. Researchers were able to estimate the effects of nonresponse bias in the Iowa Women’s Health Study.

The target population in the Iowa Women’s Health Study consisted of women between the ages of 55 and 69.5 From the total eligible pool of licensed female driv- ers, a 50% random sample of women in this age range was selected. Not all age- eligible subjects were identified using this method because only 94% of women in this age category actually had a valid Iowa driver’s license and thus the potential to participate. Data were available on the self-reported height and weight of all partic- ipants and nonparticipants from the driver’s license data tape provided by the state motor vehicle agency. These data from respondents and nonrespondents provided

The term bias denotes “Systematic deviation of results or inferences from the truth. Processes leading to such deviation. An error in the conception and design of the study or in the collection, analysis, inter- pretation, reporting, publication, or review of data leading to results or conclusions that are systematically (as opposed to randomly) different from truth.”1 n

Bias (Systematic Errors)

51589_CH10_Printer.indd 442 09/02/13 12:53 PM

s o u r C e s o f e r r o r I n e p I D e m I o l o g I C r e s e a r C h 443

a rare opportunity to examine possible selection bias resulting from nonresponse. It was found that respondents were on average 3 months younger and 0.38 kg/m2 lighter than nonrespondents. Based on 1980 census data, respondents were slightly more likely than nonrespondents to live in rural, less affluent counties. One advan- tage of conducting the study in Iowa is that one’s driver’s license (DL) number is linked to one’s Social Security (SS) number. This circumstance facilitated efficient record linkage with the State Health Registry for documentation of subsequent cancer occurrence. Presently, the SS number is not shown on a DL in Iowa.

Bisgard et al.6 compared the rates of cancer incidence and mortality between respondents and nonrespondents in the same study population. Results sug- gested that smoking-associated diseases were more frequent among the nonre- spondents than among respondents, a finding consistent with a lower response rate of smokers to a health survey.

Information bias Information bias is a kind of bias introduced as a result of measurement error in assessment of both exposure and disease. One example of information bias is recall bias, which denotes a phenomenon whereby cases may be more likely to recall past exposures than controls. Suppose that in a study of childhood leukemia, mothers are interviewed regarding drug use during pregnancy. Mothers of cases are likely to have spent considerable periods of time pondering their children’s illness. Although the frequency of exposure may actually be equivalent in both study groups, better recall among the cases than among the controls would yield positive evidence of an association. A special case of recall bias is recollection of a family history of disease, labeled family recall bias.7 Cases learn of a family his- tory of a disease from relatives after the diagnosis is made. As a result of family recall bias, data on the occurrence of the same disease among family members is likely to be more complete among cases than among controls. Although the true prevalence of the disease among family members of cases and controls may be similar, this bias would give the appearance of a higher prevalence among cases.

Interviewer/abstractor bias can occur when well-intentioned interviewers probe more thoroughly for an exposure in a case than in a control. Similarly, an abstractor may pore over records more thoroughly to identify an exposure in a case than in a control.

Prevarication (lying) bias is a type of information bias that may occur when participants have ulterior motives for answering a question and thus may under- estimate or exaggerate exposures. For example, questions asked of married, apparently heterosexual men with acquired immune deficiency syndrome may

51589_CH10_Printer.indd 443 09/02/13 12:53 PM

444 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

not necessarily reveal information about sexual contacts with other men. Surveys of individuals who have drinking disorders or members of religious groups that disallow alcohol use may yield false responses. Studies testing interventions to eliminate cigarette use often supplement self-reported measures of smoking with biochemical measures of nicotine metabolites in urine.

Information bias may occur also in relation to ascertainment of health out- comes. Consider again the Iowa Women’s Health Study cohort of postmeno- pausal women. One of the exposures of interest was a positive family history of selected cancers, especially breast cancer. The National Cancer Institute (NCI) recommends the use of screening mammography for the early detection of breast cancer; the NCI argues that although mammography is imperfect, it is the best available tool for breast cancer screening. Nevertheless, according to data from the Centers for Disease Control and Prevention, the frequency of women receiv- ing mammograms did not exceed 80% between 2000 and 2010, with the highest participation among women between the ages of 50 and 74 years; participation declined among women between 40 and 49 years of age and was lowest among women over the age of 75.8 (See Figure 10–1.)

FigurE 10–1 Mammography use in the past two years among women 40 years of age and over, by age: United States, 2000–2010. Source: Reproduced from National Center for Health Statistics. Health, United States, 2011: With Special Feature on Socioeconomic Status and Health. Hyattsville, MD. 2012, p. 16.

100

P er

ce nt

80 50–64 years

65–74 years

75 years and over

0 2000 2003 2005

Year 2008 2010

40–49 years

51589_CH10_Printer.indd 444 09/02/13 12:53 PM

s o u r C e s o f e r r o r I n e p I D e m I o l o g I C r e s e a r C h 445

A positive family history of breast cancer is an established risk factor for the disease, however, and some data suggest that women with a known family his- tory of the disease are more likely to be administered mammograms than women with a negative history of the disease.9 Accordingly, a physician who knows of a patient’s family history of breast cancer may refer the patient for a mammogram, increasing the likelihood of detecting a malignancy in the exposure group of interest.

Confounding Confounding is the term used to describe distortion of the estimate of the effect of an exposure of interest because it is mixed with the effect of an extraneous factor. According to Susser, a confounding variable is “an independent variable that varies systematically with the hypothetical causal variable under study. When uncontrolled, the effects of a confounding variable cannot be distinguished from those of the study variable.”10(p 95) According to a good working definition, con- founding occurs when the crude and adjusted measures of effect are not equal. Formal statistical tests can be performed to evaluate the statistical significance of a confounder. A reasonable rule of thumb is that a change in the estimate of effect by at least 10% when crude and adjusted measures of effect are compared suggests the influence of a confounder.

Although the categories of error (selection bias, information bias, and con- founding) are not mutually exclusive, practically speaking, only confounding can be controlled in the data analysis. To be a confounder, the extraneous factor must satisfy the following three criteria:

1. be a risk factor for the disease (not necessarily causal, but at least a marker for the actual cause of the disease)

2. be associated with the exposure under study in the population from which the cases derive (e.g., smoking would not be a confounder in an occupational cohort study if unassociated with occupational exposure; on the other hand, age might be a confounder because older employees will have had more opportunities than younger employees to be in a job category that would result in an occupational exposure)

3. not be an intermediate step in the causal path between exposure and disease

An excellent illustration of confounding is known as Simpson’s paradox,11 which means that an association in observed subgroups of a population may be reversed in the entire population. (Refer to Table 10–1 and Figure 10–2.) As an example, suppose a person enters a shop to buy a hat and sees 2 tables, each

51589_CH10_Printer.indd 445 09/02/13 12:53 PM

446 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

Table 10–1 An Analogy to Confounding: Simpson’s Paradox (Data on Day 1)

Table Hat Color Total Number Number That Fit Percent That Fit

Black

Gray

Black

Gray

90%

85%

15%

10%

FigurE 10–2 Simpson’s paradox: an illustration of confounding. Hats shown with an X do not fit.

Day 1, Table 1

Day 2, All Hats Combined on a Single Table

Day 1, Table 2

51589_CH10_Printer.indd 446 09/02/13 12:53 PM

s o u r C e s o f e r r o r I n e p I D e m I o l o g I C r e s e a r C h 447

with 30 hats. At the first table 90% of the black hats fit but only 85% of the gray hats fit. Over at the second table the person notices that, similar to the first table, a greater proportion of black hats than gray hats fits (15% versus 10%). Unfortunately, the shop is closing and the customer is forced to return the next day. Much to the customer’s chagrin, on the following day, the store clerk has placed all 60 of the hats on a single table. Although on the previous day the greatest proportion of hats that fit at each table was black in color, the customer soon discovers that, now that all the hats have been mixed together on the same table, 60% (18 of 30) of the gray hats fit but only 40% (12 of 30) of the black hats fit.

This intriguing example is neither obvious nor intuitive. Confounding can be equally vexing; sometimes associations can be so distorted that even the direction is reversed.12 An example of confounding is the positive association between air pollution and bronchitis. Air pollution varies directly and systemati- cally with urban density and overcrowding, factors that may facilitate the spread of respiratory-associated diseases, such as bronchitis. In this situation, crowding represents a confounding variable.

Another example is the inverse relation between coronary heart disease (CHD) mortality and altitude described by Buechley et al.13 Some investigators had reported that populations residing at high altitudes had lower heart disease mortality because of the protective effect of adaptation to reduced oxygen tension. A confounding variable that had not been previously accounted for was ethnicity: Hispanics in New Mexico tended to live at higher altitudes and to have lower CHD rates than other ethnic groups. Thus, there was an apparent associa- tion between altitude and CHD mortality because of unrecognized differences in ethnic composition of the regions being compared.

A more complicated example is obesity and lung cancer. Obesity has been asso- ciated with an increased risk of cancer at a number of sites. A notable exception appears to be lung cancer, for which several studies have suggested a modest inverse association.14,15 Cigarette smoking is directly associated with lung cancer risk, how- ever, and inversely associated with body mass index, which is a measure of obesity.16 A careful analysis of the obesity–lung cancer association with proper control for the confounding effect of tobacco exposure suggested that the previous observations were spurious.17 A pictorial representation of the model is presented in Figure 10–3.

Summary To recapitulate, error can be introduced into an epidemiologic study at many stages. An overview of these sources of error is depicted in Figure 10–4.

51589_CH10_Printer.indd 447 09/02/13 12:53 PM

448 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

EXTERNAL POPULATION

TARGET POPULATION

STUDY POPULATION

STUDY PARTICIPANTS

COMPARISON GROUPS

RANDOM SYSTEMATIC

Sampling

Request participation

Measurement, classification

Sampling error

Random variability (within subject)

(within observer) (between observers)

(test)

Misclassification (subject bias)

(observer bias) (measurement bias)

Selection bias (e.g., nonparticipation)

Selection bias (e.g., survivor bias)

Selection bias (bad choice of populations)

Nonrandom sampling

Confounding

FigurE 10–4 Sources of error and bias in epidemiologic studies.

FigurE 10–3 Graphic representation of how smoking confounds the body weight–lung cancer association.

Smoking

Body Weight Lung Cancer

51589_CH10_Printer.indd 448 09/02/13 12:53 PM

t e C h n I q u e s t o r e D u C e B I a s 449

Techniques to Reduce Bias

A variety of methods are available to reduce or prevent the occurrence of bias in epidemiologic research. Some guidelines that may help prevent selection bias are as follows:

●● Develop an explicit (objective) case definition. ●● Enroll all cases in a defined time and region. ●● Strive for high participation rates (incentives). ●● Take precautions to ensure representativeness. ●● For cases:

1. Ensure that all medical facilities are thoroughly canvassed. 2. Develop an effective system for case ascertainment. 3. Consider whether all cases require medical attention; consider possible

strategies to identify where else the cases might be ascertained. ●● For controls:

1. Try to compare the prevalence of the exposure with other sources to evaluate credibility.

2. Attempt to draw controls from a variety of sources.

One way to prevent nonrepresentative sampling of eligible cases is to develop—before data collection—explicit definitions about what constitutes a case. Study personnel should be trained to follow the guidelines irrespective of the exposure status of the case. There can be no doubt about the repre- sentativeness of the cases if all cases in the target population are selected for study. Definition of the number of cases eligible by time period and geographic region, for example, a 3-year period in a 5-state area, gives precision to the denominator. An established case registry facilitates the identification of cases; if a registry is not available, a surveillance network of medical facilities where patients would be seen should be established, and all such facilities should be enrolled. Health conditions that do not universally motivate afflicted persons to seek medical attention raise a special concern for the investigator; those indi- viduals who present for medical care may represent only the most severe cases. If they are atypical with respect to their exposure patterns, severe cases may bias the sample.

Low participation rates always raise concerns about the validity and generaliz- ability of a study. One approach to enhance participation is to use incentives. These may take the form of T-shirts, key chains, buttons, stickers, coupons for discounts on healthful food choices, free medical evaluation, and even monetary compensation.

51589_CH10_Printer.indd 449 09/02/13 12:53 PM

450 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

Techniques to reduce information bias include the following:

●● Use memory aids and validate exposures. ●● Blind interviewers as to subjects’ study status. ●● Provide standardized training sessions and protocols. ●● Use standardized data collection forms. ●● Blind participants as to study goals and participants’ classification status. ●● Try to ensure that questions are clearly understood through careful wording

and pretesting.

The problem of recall bias can be reduced by using memory aids to prompt for responses. For example, when one is conducting an interview of subjects to determine foods eaten over the previous day (a 24-hour recall), it is helpful to structure the interview to refer to particular meals and snack times. The use of food models to indicate portion sizes can help quantify intakes, and posters of commonly eaten snack foods are useful reminders. Studies of oral contraceptive use have utilized pictures of the pill dispensers to help subjects identify brand names and formulations.

Although a study staff committed to the research is an asset, well-intentioned interviewers or abstractors may introduce bias in data collection. Whenever pos- sible, staff should be blinded as to the status of the study subjects: case versus noncase, exposed versus nonexposed. In some situations, it also may be desirable to blind the subjects themselves to the true goals of the study. This strategy can help reduce the likelihood of subjects providing responses to please the investiga- tor, attempting to anticipate the “correct” answer to a given question, or produc- ing what they consider socially desirable answers. Clearly, biases introduced by the study subjects are less of an issue when the exposure is a biologic factor that cannot be purposely changed by the subject. One situation in which biases from study subjects might be an issue, however, would be if the biases influenced their decision to participate. Although ethical conduct of research on humans dictates that subjects be informed of the reason for the study and the basis for their invi- tation to participate, the specific hypothesis to be tested need not be revealed. Development of standardized data collection forms and survey instruments helps ensure that complete data are collected on all subjects in a uniform manner.

Methods to Control Confounding

There are two general approaches to control for confounding. Prevention strate- gies represent an attempt to control confounding through the study design itself.

51589_CH10_Printer.indd 450 09/02/13 12:53 PM

m e t h o D s t o C o n t r o l C o n f o u n D I n g 451

Analysis strategies seek to control confounding through the use of statistical analysis methods.

Prevention strategies ●● Randomization ●● Restriction ●● Matching

Analysis strategies ●● Stratification ●● Multivariate techniques n

Control of Confounding

Three Prevention Strategies to Control Confounding The first prevention strategy is randomization of study subjects. The intended net effect is to ensure equal distributions of the confounding variable in each exposure category. This strategy is an extremely efficient approach with a num- ber of clearly defined advantages. Randomization of subjects, if the sample size is sufficiently large, provides control of all factors known and unknown. It is a fairly convenient method, is inexpensive, and permits straightforward data analysis. The primary disadvantages are that randomization can be applied only to inter- vention studies when investigators have control over the exposure and are able to assign subjects to study groups. Even then, randomization works well only for large sample sizes. If the number of subjects is small, a chance remains that the distribution of confounding variables will be dissimilar across study groups.

The second prevention strategy, restriction of admission criteria, may prohibit variation of the confounder in the study groups. For example, if age is thought to be a potential confounder, the study could simply be restricted to subjects within a narrow age category. Restriction is extremely effective in providing complete control of known confounding factors; it shares with randomization the virtues of being convenient and inexpensive and permitting relatively easy data analy- sis compared with some of the alternatives. The difficulties encountered with restriction include the distinct possibility that there may still be residual con- founding if restrictions are not sufficiently narrow. Moreover, restriction will not address problems created by unknown confounding. From a practical standpoint,

51589_CH10_Printer.indd 451 09/02/13 12:53 PM

452 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

restriction may shrink the pool of available subjects to an unacceptably low level. Depending on the health problem studied, one may not be able to generalize the results beyond restricted categories.

Matching of subjects in the study groups according to the value of the sus- pected or known confounding variable to ensure equal distributions is the third prevention strategy; an example of a potential confounding variable that might be controlled using matching is age. Several types of matching are available. In frequency matching, the number of cases with the particular matching charac- teristics is tabulated. For example, if one is matching on 5-year age groups, a frequency distribution of the cases by age group would be generated. Each 5-year age group is called an age stratum. A stratum is a homogeneous population sub- group, such as that characterized by a narrow age range (e.g., a 5-year age group). Controls are then selected until the required number of controls for each stratum has been acquired. If the controls are to be studied concurrently with the cases, one can generate an expected frequency of cases for each matching stratum based on previously observed rates. Another type of matching is individual matching, the pairing of one or more controls to each case based on similarity of one or several variables, for example, sex or race.

The use of matching to control confounding has a number of clear advan- tages. In terms of sample size requirements for follow-up studies, matching is efficient in that fewer subjects are required than in unmatched studies of the same hypothesis.11 Matching also may enhance the validity of a follow-up study. Despite these advantages, matching can be costly, often requiring exten- sive searching and record keeping to find matches. For example, Ross and col- leagues18 conducted a case-control study of renal cancer to evaluate the potential role of analgesics in carcinogenesis. A total of 314 cases of incident cancer of the renal pelvis were identified through the Cancer Surveillance Program. Of these, 61 died before contact, 20 refused to be interviewed, and another 30 were pro- hibited from participating by the attending physician, leaving 203 cases. Con- trols were matched to cases on birth date (65 years), race, sex, and neighborhood. A predetermined walking algorithm, starting with the residence of the case, was applied for the selection of controls. The procedure continued until a suitable control was found or until 40 houses had been approached. Successful matches were found for 187 (92%) of the cases, and an average of 22 household units were approached per case to find an appropriate control.

For case-control studies, matching may introduce confounding rather than control for it. Confounding typically occurs by matching subjects on factors associated with exposure but then ignoring the matching in the analysis stage.

51589_CH10_Printer.indd 452 09/02/13 12:53 PM

m e t h o D s t o C o n t r o l C o n f o u n D I n g 453

The result is often an estimate of effect that is biased toward the null value (i.e., an odds ratio of 1). When one matches subjects on a potential confounder, that particular exposure variable can no longer be evaluated with respect to its contribution to risk; the distribution of the exposure variable is constrained to be similar (perhaps even identical) in both groups. When one matches on several factors, such as age, sex, and neighborhood, there is a danger of making the study groups so similar that one essentially has matched oneself out of business; this problem sometimes is called overmatching. That is, the exposure that is associ- ated with the outcome of interest (not the confounder) becomes equivalent in the two study groups, and no meaningful associations are identified. Depending on how imprecise the matching variables are, one also can have residual con- founding. For example, in a nested case-control analysis of alcohol and lung cancer risk in the Iowa Women’s Health Study cohort, cases and noncases were matched on pack-years of smoking.19 Careful inspection of the data revealed that within strata of pack-years cases tended to have slightly higher mean exposure levels than noncases.

Two Analysis Strategies to Control Confounding Although it can be somewhat more comforting to conduct a study in which the potential for known confounding can be addressed in the design phase, issues of cost and feasibility may make it necessary instead to address confounding in the analysis stage. Furthermore, attempts to minimize confounding in the design phase obviously can be done only for known confounders. Often the presence of a confounding factor is not observed or detected until analyses are underway, making analysis strategies for dealing with confounding important tools indeed.

The first analysis strategy, stratification, occurs when analyses are performed to evaluate the effect of an exposure within strata (levels) of the confounder. A general approach is to define homogeneous categories or narrow ranges of the confounding variable. One can then combine stratum-specific effects into an overall effect by standard statistical principles and methods (e.g., the Mantel- Haenszel procedure).20 There are three advantages to this approach. First, per- forming analyses within strata is a direct and logical strategy. Second, there are minimum assumptions that must be satisfied for the analyses to be appropriate. Third, the computational procedure is quite straightforward.

The difficulties with stratification arise in several areas. The basic process of stratification of the data may result, unfortunately, in small numbers of obser- vations in some strata. When dealing with a continuous variable or an ordinal

51589_CH10_Printer.indd 453 09/02/13 12:53 PM

454 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

variable with a relatively large number of categories, one is faced with a variety of ways to form strata. Knowing or deciding which cut points are most appropriate may be difficult. If several confounding factors must be evaluated, necessitating stratification across two or more variables, each of which may have multiple lev- els, one can easily run into difficulty in interpretation. Finally, from a statistical standpoint, categorization almost always results in loss of information.

The second analysis strategy involves multivariate techniques in which com- puters are used to construct mathematical models that describe simultaneously the influence of exposure and other factors that may be confounding the effect. This strategy tends to be more feasible with smaller numbers of study subjects than stratification, although multivariate techniques generally require large sam- ple sizes. An advantage of this tactic pertains to the use of continuous versus categorical variables in the analysis of confounding. When factors that must be controlled are entered into the models as continuous variables, the problem of creating categorical variables is obviated. Continuous variables may be converted to categorical variables by establishing cut points for categories. The epidemi- ologist may be faced with theoretical difficulties in knowing or deciding where to form such cut points. Another major advantage of multivariate modeling is that it allows for simultaneous control of several exposure variables in a single analysis.

The main disadvantage of multivariate techniques is their great potential for misuse. There are some restrictive assumptions about the distribution of the data that should be, but are not always, examined. The choice of the model may be difficult, especially when the investigator is faced with a large number and wide variety of variables that could be selected. The widespread availability and user- friendly nature of commercial computer software make the method accessible to some data analysts who may not have had adequate instruction in its appropri- ate applications. When they are misapplied, multivariate techniques have the potential to contribute to incorrect model development, misleading results, and inappropriate interpretation of the effects of hypothesized confounders.

Bias in Analysis and Publicat ion

Although not a reflection of errors in selection, measurement, or analysis (con- founding), it is important to consider another source of bias that can have a profound influence. This type of bias is presented in a separate section of the chapter because it is generally outside the control of individual investigators.

51589_CH10_Printer.indd 454 09/02/13 12:53 PM

B I a s I n a n a l y s I s a n D p u B l I C a t I o n 455

Publication bias is a phenomenon that occurs because of the influence of study results on the chance of publication. In particular, studies with positive results are more likely to be published than studies with negative results,21 and to be published more quickly.22,23 The net effect is a preponderance of false-positive results in the literature. The bias is compounded when published studies on a topic are subjected to meta-analysis. These tend to give an even greater air of importance because of the intent to summarize and synthesize a large number of studies on a given topic. Moreover, meta-analyses of observational studies may be used as the rationale for large clinical trials. Recent comparisons of results from meta-analyses and subsequent randomized clinical trials confirm that results can- not be accurately predicted roughly 35% of the time.24 This deficiency raises the question of whether meta-analyses should include unpublished data. Crowther and Cook25 pointed out that the inclusion of unpublished data in systematic literature reviews such as meta-analyses is controversial. Although these reviews may be affected by publication bias, inclusion of unpublished data may lower the quality of the meta-analysis itself.

Although each of us as public health professionals adheres to a professional code of conduct and integrity in the work that we do, we cannot control the actions of our peers. In these days of “publish or perish” pressures within aca- demia, and the increasing collaboration of universities with private businesses in sponsored research, not all actions are above reproach even though they are not necessarily fraudulent. Thus one must have a dose of skepticism about what appears in the literature. For example, one may have difficulty in ascer- taining whether or not study results from an observational study were tests of a priori hypotheses or post hoc analyses masquerading as such. This conundrum is particularly worrisome in these days of large data sets, powerful analysis packages, and high-speed computers. Readers of the literature may have dif- ficulty telling whether or not reported results were merely the product of “data dredging” and actually only a chance observation that would be most difficult to replicate.

Further raising the cloud of suspicion is the habit of deleting certain study subjects because they are considered outliers, or when a number of cut points are considered in the analysis and only the ones that yield statistically signifi- cant results are presented. Unfortunately, although most scientific journals have a peer review process, such actions are exceedingly difficult to detect. Moreover, there are tendencies for editors to make their decisions on suitability for publica- tion based on the direction or strength of the study findings.26

51589_CH10_Printer.indd 455 09/02/13 12:53 PM

456 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

Conclusion

The ability to evaluate critically sources of error in epidemiologic research is nec- essary not only to interpret properly the plethora of media reports, but also to design and analyze studies. Even if one does not conduct one’s own epidemiolog- ical research, competence in interpretation of empirical studies is essential if pub- lic health interventions are to be enacted based on epidemiologic observations. Two main types of research errors must be considered: random errors, which occur because of sampling error, lack of precision, and variability in measure- ment; and systematic errors (bias), which occur through selection of subjects, collection of information about exposure and disease, and confounding. This chapter presented a number of techniques to reduce bias and introduced some helpful methods to control confounding. Prevention strategies include random- ization of subjects into exposure groups, restriction of admission criteria, and matching subjects on the potential confounder. Analysis strategies include strati- fication and multivariate modeling.

Study Questions and Exercises

1. In a study to determine the incidence of a chronic disease, 150 people were examined at the end of a 3-year period. Twelve cases were found, giving an incidence rate of 8%. Fifty other members of the initial cohort could not be examined; 20 of these 50 could not be examined because they died. Does this loss of subjects to follow-up represent a source of bias that may have affected the study results?

2. A case-control study was carried out in which 120 of 200 cases of stomach cancer and 50 of 200 control subjects gave a history of exposure to radiation. In further analysis, however, the investigators noticed that 50% of the cases but only 25% of the controls were men. What would be a practical and efficient way to eliminate differences between cases and controls with respect to sex?

3. Two automated blood cell counters are tested twice using a prepared suspension of leukocytes containing 8,000 cells/mm3. The cell counts by device A are 8,400 cells/mm3 the first time and 8,350 cells/mm3 the second. Device B’s counts are 8,200 and 7,850 cells/mm3, respectively. Which device (A or B) gives leukocyte counts with greater validity? Which device gives leukocyte counts with greater reliability?

51589_CH10_Printer.indd 456 09/02/13 12:53 PM

s t u D y q u e s t I o n s a n D e x e r C I s e s 457

4. You are planning a case-control study of lung cancer to test the hypothesis that vegetable consumption is protective against lung cancer. Would you match on smoking? Explain your answer.

5. A follow-up study was conducted of 3,000 military troops deployed at an atomic test site in Nevada to detect the occurrence of leukemia. A total of 1,870 persons were successfully traced by the investigators, and an addi- tional 443 contacted the investigators on their own as a result of public- ity about the study. Four cases of leukemia occurred among the 1,870 individuals traced by the investigators, and an additional four occurred among those individuals who contacted the investigators on their own. Could interpretation of the study results be subject to bias?

6. You are conducting a study of insulin resistance and its relationship to body weight. Using a weight scale as your instrument to measure body weight, you find the scale always reads 30 kg regardless of who is stand- ing on it. Discuss the validity and reliability of the scale. Questions 7 through 10 are multiple choice items. Select the correct answer from the options that follow each question.

7. You are investigating the role of physical activity in heart disease, and your data suggest a protective effect. While presenting your findings, a colleague asks whether you have thought about confounders, such as fac- tor X. Under which of the following conditions could this factor have confounded your interpretation of the data? a. It is a risk factor for some other disease, but not heart disease. b. It is a risk factor associated with the physical activity measure and

heart disease. c. It is part of the causal pathway by which physical activity affects heart

disease. d. It has caused a lack of follow-up of test subjects. e. It may have blinded your study.

8. Surgeons at hospital A report that the mortality rate at the end of a 1-year follow-up after a new coronary bypass procedure is 15%. At hospital B, the surgeons report a 1-year mortality rate of 8% after the same procedure. Before concluding that the surgeons at hospital B had vastly superior skill, which of the following possible confounding factors would you examine? a. the severity (stage) of disease of the patients at the two hospitals at

baseline b. the start of the one-year follow-up at both hospitals (after operation

versus after discharge)

51589_CH10_Printer.indd 457 09/02/13 12:53 PM

458 C h a p t e r 1 0 D ata I n t e r p r e tat I o n I s s u e s

c. differences in postoperative care at the two hospitals d. equality of follow-up for mortality e. all of the above

9. Which of the following is not a method to control for the effects of confounding? a. randomization b. stratification c. matching d. blinding

10. The strategy that will not help reduce selection bias is: a. development of an explicit case definition b. the use of incentives to encourage high participation c. a standardized protocol for structured interviews d. enrollment of all cases in a defined time and region

References

1. Porta M. A Dictionary of Epidemiology. 5th Ed. New York: Oxford University Press, 2008.

2. Martini MC, Campbell DR, Gross MD, et al. Plasma carotenoids as biomarkers of vegetable intake: the University of Minnesota Cancer Prevention Research Unit Feeding Studies. Cancer Epidemiol Biomarkers Prev. 1995;4:491–496.

3. Berenson GS, McMahan CA, Voors AW, et al. Cardiovascular Risk Factors in Children: The Early Natural History of Atherosclerosis and Essential Hypertension. New York, NY: Oxford University Press; 1980.

4. Greenland S. Response and follow-up bias in cohort studies. Am J Epidemiol. 1977;106:184–187.

5. Folsom AR, Kaye SA, Potter JD, et al. Association of incident carcinoma of the endometrium with body weight and fat distribution in older women: early findings of the Iowa Women’s Health Study. Cancer Res. 1989;49:6828–6831.

6. Bisgard KM, Folsom AR, Hong CP, Sellers TA. Mortality and cancer rates in nonrespondents to a prospective study of older women: 5-year follow-up. Am J Epidemiol. 1994;139:990–1000.

7. Sackett DL. Bias in analytic research. J Chronic Dis. 1979;32:51–63. 8. National Center for Health Statistics. Health, United States, 2011: With Special

Feature on Socioeconomic Status and Health. Hyattsville, MD. 2012. 9. Cook NR, Rosner BA, Hankinson SE, Colditz GA. Mammographic screening and

risk factors for breast cancer. Am J Epidemiol. 2009; 170(11):1422–1432. 10. Susser M. Causal Thinking in the Health Sciences. New York, NY: Oxford University

Press; 1973. 11. Rothman KJ. Modern Epidemiology. Boston, MA: Little, Brown; 1986.

51589_CH10_Printer.indd 458 09/02/13 12:53 PM

r e f e r e n C e s 459

12. Rothman KJ. A pictorial representation of confounding in epidemiologic studies. J Chronic Dis. 1975;28:101–108.

13. Buechley R, Key C, Morris D, et al. Altitude and ischemic heart disease in tricultural New Mexico: an example of confounding. Am J Epidemiol. 1979;109:663–666.

14. Kabat GC, Wynder EJ. Body mass index and lung cancer risk. Am J Epidemiol. 1992;135:769–774.

15. Knekt P, Heliovaara M, Rissanen A, et al. Leanness and lung cancer risk. Int J Cancer. 1991;49:208–213.

16. Rigotti NA. Cigarette smoking and body weight. N Engl J Med. 1989;320:931–933. 17. Drinkard CR, Sellers TA, Potter JD, et al. Association of body mass index and

body fat distribution with risk of lung cancer in older women. Am J Epidemiol. 1995;142:600–607.

18. Ross RK, Paganini-Hill A, Randolph J, et al. Analgesics, cigarette smoking, and other risk factors for cancer of the renal pelvis and ureter. Cancer Res. 1989;49:1045–1048.

19. Potter JD, Sellers TA, Folsom AR, McGovern PG. Alcohol, beer and lung can- cer in postmenopausal women: the Iowa Women’s Health Study. Ann Epidemiol. 1992;2:587–595.

20. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst. 1959;22:719–748.

21. Begg CB, Berlin JA. Publication bias and dissemination of clinical research. J Natl Cancer Inst. 1989;81:107–115.

22. Stern JM, Simes RJ. Publication bias: evidence of delayed publication in a cohort study of clinical research projects. Br Med J. 1997;315:640–645.

23. Ioannidis JP. Effect of the statistical significance of results on the time to completion and publication of randomized efficacy trials. JAMA. 1998;279:281–286.

24. LeLorier J, Gregoire G, Benhaddad A, et al. Discrepancies between meta-analyses and subsequent large randomized, controlled trials. N Engl J Med. 1997;337:536–542.

25. Crowther MA, Cook DJ. Trials and tribulations of systematic reviews and meta- analyses. Hematology. 2007:493–497.

26. Dickersin K. The existence of publication bias and risk factors for its occurrence. JAMA. 1990;263:1385–1389.

51589_CH10_Printer.indd 459 09/02/13 12:53 PM

51589_CH10_Printer.indd 460 09/02/13 12:53 PM

3chapte r

461

11chapte r

Screening for Disease in

the Community

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● define and discuss reliability and validity, giving differentiating characteristics and interrelationships

●● identify sources of unreliability and invalidity of measurement ●● define the term screening and list desirable qualities of screening tests ●● define and discuss sensitivity and specificity, giving appropriate

formulas and calculations for a sample problem ●● identify a classification system for a disease

CHAPTER OUTLINE

I. Introduction II. Screening for Disease

III. Appropriate Situations for Screening Tests and Programs IV. Characteristics of a Good Screening Test V. Evaluation of Screening Tests

VI. Sources of Unreliability and Invalidity VII. Measures of the Validity of Screening Tests

VIII. Effects of Prevalence of Disease on Screening Test Results IX. Relationship Between Sensitivity and Specificity X. Evaluation of Screening Programs

XI. Issues in the Classification of Morbidity and Mortality XII. Conclusion

XIII. Study Questions and Exercises

51589_CH11_Printer.indd 461 18/02/13 9:59 PM

462 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

Introduction

Increasingly, the public health field has recognized the importance of screening programs for the secondary prevention of morbidity and mortality. Efforts to control diseases by early detection through screening have led to a basic change in the nature of medical practice from an exclusive focus upon a small number of ill persons to a targeting of large numbers of asymptomatic persons.1 At the same time, patients and healthcare providers alike find recommendations for use of screening tests such as the prostate-specific antigen (PSA) test and screening mammography to be confusing.

Screening programs for coronary heart disease risk factors have incited the public’s awareness of hypertension control and dietary components of hypercho- lesterolemia. Breast cancer screening by mammography for early malignancies combined with effective cancer therapies have contributed to the high 5-year survival rates for this cancer site. According to research findings, breast can- cer screening is efficacious for women 50 years of age and older. Exhibit 11–1 describes one opinion regarding the current debate over the effectiveness of screening women who are in the 40- to 49-year-old age group. This chapter dis- cusses screening for disease in the population, including reliability and validity of measures, concepts and terminology of screening, sensitivity, and specificity as well as positive and negative predictive values.

Content removed due to copyright restrictions

Should Women Aged 40 Through 49 Years Receive Routine Mammography Screening?

E x

h ib

continues

51589_CH11_Printer.indd 462 20/02/13 12:51 AM

i n t r o D u C t i o n 463

Content removed due to copyright restrictions

Exhibit 1 continued

continues

51589_CH11_Printer.indd 463 20/02/13 12:51 AM

464 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

Screening for Disease

A tenet of public health is that primary prevention of disease is the best approach. If all cases of disease cannot be prevented, however, then the next best strategy is early detection of disease in asymptomatic, apparently healthy individuals. Screening is defined as the presumptive identification of unrecognized disease or defects by the application of tests, examinations, or other procedures that can be applied rapidly. The qualifier presumptive is included in the definition to emphasize the preliminary nature of screening; diagnostic confirmation is required, usually with the benefit of more thorough clinical examination and additional tests. As an illustration of screening, Figure 11–1 demonstrates a mammography (part A) and a blood pressure screening event (part B).

Some screening programs are conducted in order to screen interested and con- cerned individuals for specific health problems, such as hypertension, cervical cancer, or sickle-cell disease. An example of this type of screening program would be administration of a free thyroid test (serum level of thyroxine) to passersby in a shopping center or members of a senior citizens center.5 Other screening programs

Content removed due to copyright restrictions

Exhibit 1 continued

51589_CH11_Printer.indd 464 20/02/13 1:09 AM

S C r e e n i n g f o r D i S e a S e 465

Figure 11–1 Mammography (part A) and a blood pressure screening event (part B). Source: Reproduced from Centers for Disease Control and Prevention. Public Health Image Library. Image numbers 8295 and 7874. Available at http://phil.cdc.gov/phil/. Accessed April 19, 2012.

may be applied on a mass basis to almost all individuals in the population; an example is screening for phenylketonuria (PKU) among all neonates.

It should be noted that screening differs from diagnosis, which is the process of confirming an actual case of a disease.6,7 As a result of diagnosis, medical interven- tion, if appropriate, is initiated. Diagnostic tests are used in follow-up of positive screening test results (e.g., phenylalanine loading test in children positive on PKU screening) or directly for screening (e.g., fetal karyotyping in prenatal screening for Down syndrome). For example, if a thyroid test is administered to determine an exact cause of a patient’s illness, it would then be a diagnostic test.5 The thyroid test also could be a screening test, however, as will be demonstrated subsequently.

Screening for three types of cancer (breast, cervical, and colorectal) could go a long way in reducing mortality from these malignancies. In a typical year, 350,000 persons in the United States are diagnosed with these forms of cancer; 100,000 persons die from them each year. The U.S. Preventive Services Task

51589_CH11_Printer.indd 465 18/02/13 9:59 PM

466 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

Force (USPSTF) advocates for screening for these three types of cancer in order to reduce morbidity and mortality from them. Healthy People 2020 has estab- lished national targets for population levels of participation in screening tests. Figure 11–2, based on data from the National Health Interview Survey (NHIS), shows the percentage of men and women between the years 2000 and 2010 who were up-to-date on screening for breast cancer, cervical cancer, or colorectal cancer. Among the factors related to higher screening participation rates were education, screening availability, use of health care, and length of U.S. residence. Low rates of participation occurred among Asians in comparison with whites and blacks. In addition, persons of Hispanic descent were less likely than other groups to be screened for cervical and colorectal cancer.8

Figure 11–2 Percentage of men and women up-to-date on screening for breast, cervical, or colorectal cancer, via type of test, sex, and year—United States, 2000–2010. Source: Reproduced from Centers for Disease Control and Prevention. Cancer screening—United States, 2010. MMWR. 2012; 61:42.

Pap test∗

Mammogram†

Any CRC test (male)§

Any CRC test (female)§

100

0 2000

Abbreviations: CRC = colorectal cancer; Pap = Papanicolaou. ∗Among women aged 21–65 years with no hysterectomy. †Among women aged 50–74 years. §Among persons aged 50–75 years.

% u

p- to

-d at

e fo

r sc

re en

in g

20052003 2008

Year

2010

51589_CH11_Printer.indd 466 18/02/13 9:59 PM

S C r e e n i n g f o r D i S e a S e 467

Multiphasic Screening Although screening programs can be restricted to early detection of a single disease, a more cost-effective approach is to screen for more than one disease. Multiphasic screening is defined as the use of two or more screening tests together among large groups of people.9 The multiphasic screening examination may be administered as a pre-employment physical, and successfully passing the exami- nation may be a necessary condition for employment in the organization. As an employee benefit, some companies repeat the screening examination on an annual basis and direct suggestive findings to the employee’s own physician while maintaining confidentiality of the results. Typical multiphasic screening pro- grams assess risk factor status as well as individual and family history of illness, and they also collect physiologic and health measurements. Multiphasic screen- ing also is a cornerstone of health maintenance organizations, such as Kaiser Permanente and Group Health Incorporated.

Mass Screening and Selective Screening Mass screening (also known as population screening) refers to screening of total population groups on a large scale, regardless of any a priori information as to whether the individuals are members of a high-risk subset of the population. Selective screening, sometimes referred to as targeted screening, is applied to sub- sets of the population at high risk for disease or certain conditions as the result of family history, age, or previous exposures. It is likely to result in the greatest yield of true cases and represents the most economical utilization of screening measures. For example, screening tests for Tay-Sachs disease might be applied to individuals of Jewish extraction whose ancestors originated in Eastern Europe because this group has a higher frequency of the genetic alteration.

Mass Health Examinations Several other activities are similar to screening but differ in one or more critical respects. Population or epidemiologic surveys aim to elucidate the natural his- tory, prevalence, incidence, and duration of health conditions in defined popu- lations.9 The purpose of these surveys is to gain new knowledge regarding the distribution and determinants of diseases in carefully selected populations. Thus, they are not considered screening because they imply no immediate health ben- efits to the participants.10

Epidemiologic surveillance aims at the protection of community health through case detection and intervention (e.g., tuberculosis control).11 It refers

51589_CH11_Printer.indd 467 18/02/13 9:59 PM

468 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

to the continuous observation of the trends and distribution of disease incidence in a community or other population over time to prevent disease or injury.12 Sources of data for surveillance include morbidity and mortality reports, for example, those reported by the Centers for Disease Control and Prevention. Around the early 1990s, surveillance activities detected an increase in tuberculo- sis in the United States as well as an increase in measles cases; subsequently, the latter disease was brought under control by stepped-up immunization of chil- dren. Surveillance programs are used for detection and control of conditions ranging from infectious diseases to injuries to chronic diseases.

Case finding, also referred to as opportunistic screening, is the utilization of screening tests for detection of conditions unrelated to the patient’s chief com- plaint.5,13 An example would be administration of a screening for colon cancer to a patient who came to a physician complaining of pharyngitis.

Appropriate Situations for Screening Tests and Programs

A number of criteria must be considered carefully before a decision is made to implement a screening program.9 Although the ideal situation is one in which all criteria are satisfied, numerous examples can be cited to illustrate how screening programs that violate one or more of these issues can still be extremely valuable (Exhibit 11–2).

appropriate Situations for Screening

Social: The health problem should be important for the individual and the community. Diagnostic follow-up and intervention should be available to all who require them. There should be a favorable cost-benefit ratio. Public acceptance must be high. Scientific: The natural history of the condition should be adequately understood. Identification should occur during prepathogenesis with sufficient lead time (see text for definition of lead time). There

is sound case definition in addition to a policy regarding whom to treat as patients. A knowledge base exists for the efficacy of prevention and the occurrence of side effects. The prevalence of the disease or condition is high.

e x

h ib

it 1

1 –2

continues

51589_CH11_Printer.indd 468 18/02/13 9:59 PM

a p p r o p r i a t e S i t u a t i o n S 469

Social Of major importance is the magnitude of the health problem for which screen- ing is being considered. Magnitude is relevant in a number of dimensions: to the community, in terms of economics, and medically. From the community per- spective, the disease or outcome must be viewed as a major health problem. This means that there is general consensus that the health problem is of sufficiently high priority as to justify the commitment of resources to implement and carry out the program. Furthermore, acceptance of the program by the public must be high. For example, an effective screening test for a major health problem will not necessarily result in an effective screening program if the public refuses to participate.

Although tempted to do so, one must not automatically assume that screen- ing programs are beneficial. To be successful over the long run early detection efforts must be cost-effective. Thus, one must consider the costs of the test itself, the costs of follow-up examinations, and the costs of treatments avoided. The most clear-cut evidence of cost-effectiveness manifests itself when the cost of the program itself is more than offset by the savings of more expensive treatment that would have been necessary had the condition advanced to a more serious stage. Oftentimes this may not be the case, however, and one must consider as benefits improvements in quality of life and the value of years of life saved. Negative costs should be considered also: There are emotional costs to healthy individuals who are falsely labeled as ill by a screening test and emotional costs to individuals (and their loved ones) who are diagnosed early and yet die quickly anyway.

An obvious determinant of the cost–benefit ratio of a screening program is the current cost to the medical community in the absence of screening. How much

Ethical: The provider initiates the service and, therefore, should have evi- dence that the program can alter the natural history of the condition in a significant proportion of those screened. Suitable, acceptable tests for screening and diagnosis of the condition as well as acceptable, effective methods of prevention are available.

Source: Data are from Wilson JMG, Jungner F. Principles and practice of screening for disease, Public Health Papers, No. 34, World Health Organization, 1968; and from Cochrane AL, Holland WW. Validation of screening procedures. British Medical Bulletin, Vol 27, pp. 3–8, Churchill Livingstone; 1971.

exhibit 11–2 continued

51589_CH11_Printer.indd 469 18/02/13 9:59 PM

470 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

money is being spent to treat individuals with the disease? How many hospital beds are being utilized? What is the number of health personnel assigned to the problem? Diseases and conditions that are costly to treat may still be considered for early detection even if the scientific justification for screening is weaker than for a disease that represents less of a medical burden.

Scientific Early detection efforts are most likely to be successful when the natural history of the disease is known. This knowledge permits identification of early stages of dis- ease and appropriate biologic markers of progression. For example, it is known that individuals with high cholesterol and high blood pressure are at increased risk for coronary heart disease. Because these risk factors precede onset of an acute myocardial infarction, identification of such high-risk individuals may lead to medical intervention (changes in diet, exercise, weight loss, or use of drugs) to prevent the disease. This example illustrates that there also should be good tests (screening and diagnostic) to measure blood pressure and blood cholesterol and that effective treatment should be available.

Ethical It is most desirable to implement screening programs for diseases that—when diagnosed early—have their natural history altered, that is, for which effective treatment is available. Note, however, that screening is sometimes done for dis- eases for which effective treatment is not available. For example, we are yet with- out a cure for infection with the human immunodeficiency virus. Screening is nonetheless important to prevent spread of the disease from infected to unin- fected individuals and to improve the prognosis of those who may be affected by initiating appropriate treatments. For those diseases for which effective treatments are available, it is important to consider the capacity of the medical community to handle the increased number of individuals requiring definitive diagnoses. Suppose a volunteer organization decides to offer a free health screen- ing for high cholesterol at the local community center and that 10,000 citizens attend. Suppose further that 1,000 citizens are found to have high cholesterol. These individuals are mailed a letter informing them of their results with the suggestion to see their physician for further evaluation. A number of ethical issues can be envisioned. What if physicians in the local medical community are unable to accommodate the sudden increased demand for their services? What if these individuals lack medical insurance and have no physician?

51589_CH11_Printer.indd 470 18/02/13 9:59 PM

e v a l u a t i o n o f S C r e e n i n g t e S t S 471

Characterist ics of a Good Screening Test

There are five attributes of a good screening test: simple, rapid, inexpensive, safe, and acceptable9,10,14:

1. Simple: The test should be easy to learn and perform. One that can be administered by nonphysician medical personnel will necessarily cost less than one that requires years of medical training.

2. Rapid: The test should not take long to administer, and the results should be available soon. The amount of time required to screen an individual is directly related to the success of the program: If a screening test requires only 5 minutes out of a person’s schedule, it is likely to be perceived as being more valuable than one that requires an hour or more. Further- more, immediate feedback is better than a test in which results may not be available for weeks or months. Results of a blood pressure screening are usually known immediately; results of a screen for high cholesterol must await laboratory analysis. Fortunately, much progress is being made in the development of rapid screening tests for many conditions.

3. Inexpensive: As discussed earlier, the cost–benefit ratio is an important criterion to consider in the evaluation of screening programs. The lower the cost of a screening test, the more likely it is that the overall program will be cost beneficial.

4. Safe: The screening test should not carry potential harm to screenees. 5. Acceptable: The test should be acceptable to the target group. An effective

protocol has been developed to screen for testicular cancer, but accep- tance rates among men have not been as high as for a similar procedure, mammography, among women.

Evaluation of Screening Tests

Recall that the purpose of a screening test is to classify individuals as to whether they are likely to have disease or be disease-free. To do this classification, a mea- suring instrument or combination of instruments is required. Examples of such instruments are clinical laboratory tests, a fever thermometer, weighing scales, and standardized questionnaires. The preceding section made no mention of the important issue of how well the screening test should actually work. This com- plex subject requires the introduction of several new concepts. The first and sec- ond of these concepts are reliability and validity.

51589_CH11_Printer.indd 471 18/02/13 9:59 PM

472 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

Reliability Reliability, also known as precision, is the ability of a measuring instrument to give consistent results on repeated trials. According to Morrison, reliability of a test refers to “its capacity to give the same result—positive or negative, whether correct or incorrect—on repeated application in a person with a given level of disease. Reliability depends on the variability in the manifestation on which the test is based (e.g., short-term fluctuation in blood pressure), and on the variability in the method of measurement and the skill with which it is made.”1(p 10)

Repeated measurement reliability refers to the degree of consistency between or among repeated measurements of the same individual on more than one occasion. For example, if one were to measure the height of an adult at differ- ent times, one would expect to observe similar results. That is because, in part, one’s true value of height is relatively constant (although we are actually slightly shorter at the end of the day than we were at the beginning!). There also might be slight errors in measurement from one occasion to another, however; some measurements overestimate and others underestimate the true value. Although one might expect to measure height reliably, other measures, such as blood pres- sure, may be much more unreliable than height. Technicians’ skills in the mea- surement of blood pressure, slight variations in the calibration of the manometer cuff, and variability in subjects’ true blood pressure levels from one occasion to another all affect the reliability of blood pressure measurements.

Internal consistency reliability evaluates the degree of agreement or homo- geneity within a questionnaire measure of an attitude, personal characteristic, or psychological attribute. For example, a researcher may be interested in the relationship between general anxiety level and peptic ulcer. A multi-item paper- and-pencil measure for general anxiety may be utilized in the research. The Kuder–Richardson reliability coefficient measures the internal consistency reli- ability of this type of measure.15 It is based on the average intercorrelation of a set of items in a multi-item index. Chronbach’s α coefficient is used also to mea- sure internal consistency reliability; a value of 0.7 or greater is generally accepted as satisfactory reliability and suggests that a set of items is measuring a common dimension.16 These two reliability measures are particularly applicable to epi- demiologic research that uses survey measures, such as interviews or self-report questionnaires.

Interjudge reliability refers to reliability assessments derived from agree- ment among trained experts. The ratings of psychiatrists in psychiatric research, for example, may be used to measure an individual’s degree of psychiatric

51589_CH11_Printer.indd 472 18/02/13 9:59 PM

e v a l u a t i o n o f S C r e e n i n g t e S t S 473

impairment. To obtain an estimate of the reliability of the rating procedure, the average percentage of agreement of the judges who are rating an attribute may be calculated.

There are several ways to express the reliability (precision) of a set of measure- ments.14 One is to obtain repeated measurements of an attribute for a single person and then obtain the standard deviation of the measurements, known as the standard error of measurement. A second is the reliability coefficient, which is an indicator of repeated measurement reliability. It is a correlation coefficient that quantifies the degree of agreement between measurements taken on two dif- ferent occasions.

Reliability types Validity types Repeated measurements Content Internal consistency Criterion-referenced Interjudge Predictive

Concurrent Construct

Validity Also known as accuracy, validity is the ability of a measuring instrument to give a true measure (i.e., how well it measures what it purports to measure).17 Validity can be evaluated only if an accepted and independent method exists for confirm- ing the test measurement. Validity is an important component of epidemiologic research, including areas outside screening. Variations on the theme of validity are presented in the next few sections. The issues discussed extend beyond the role of validity in screening for a disease. However, they may be applicable to screening for high-risk behaviors.

Content validity is defined as “[t]he extent to which the measurement incorpo- rates the domain of the phenomenon under study. For example, a measurement of functional health status should embrace activities of daily living: occupational, family, and social functioning, etc.”7 Often, content validity is used to measure the validity of survey instruments or paper-and-pencil measures. In this context, content validity “refers to how much a measure covers the range of meanings included within a concept.”18(p. 152) It concerns the extent to which the items in a questionnaire seem to be valid for measuring the domain of the phenomenon that they are supposed to measure; that is, the measurement includes and fully covers all the aspects of the dimension being measured. For example, the content

51589_CH11_Printer.indd 473 18/02/13 9:59 PM

474 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

validity of a test of mechanical aptitude would measure whether the test contains items that cover a full range of mechanical abilities. This type of validity also is referred to as rational or logical validity.14

Criterion-referenced validity generally refers to validity that is found by correlating a measure with an external criterion of the entity being assessed.7 The external standard used to assess validity is called the validity criterion. The two types of criterion-referenced validity are predictive validity and concurrent validity.

Predictive validity denotes the ability of a measure to predict some attribute or characteristic in the future. An illustration of this type of validity is the asso- ciation between type A behavior (coronary prone behavior) and coronary heart disease. Researchers attempted to demonstrate the validity of the type A measure through its positive correlations with future incidence of coronary heart disease (CHD). The future outcome (CHD) that was predicted by the type A mea- sure was the validity criterion. A second example is measuring the validity of a scholastic aptitude test by demonstrating that it predicts success in school, the predictive validity criterion.7 Predictive validity also is called empirical or statisti- cal validity.

Similar to predictive validity is concurrent validity. This type of validity refers to establishing the validity of a measure by correlating it with an alternative mea- sure of the same phenomenon taken at the same point in time; typically the con- current validating criterion is more cumbersome than the new measure. Much work has been devoted to self-administered measures of mental health charac- teristics for use in epidemiologic studies. An example would be the validation of a self-administered depressive symptoms questionnaire against the criterion of psychiatric diagnosis. A medical example would be validation of test for bacterial activity against clinical evidence of infection.7

Construct validity refers to the degree to which the measurement agrees with the theoretical concept being investigated.7 As an illustration, if a theory sug- gests that some phenomenon should change as a person ages, then the measure- ments obtained by the measure should reflect changes with age.7 Suppose that a test is being developed to assess age-related changes in bone density and muscle mass. A test that has construct validity would pick up these changes as a person grows older.

Another example of construct validity involves the confirmation of a the- oretical construct, such as anxiety. In designing a paper-and-pencil test of anxiety, the investigator would first need to specify what types of behavior

51589_CH11_Printer.indd 474 18/02/13 9:59 PM

e v a l u a t i o n o f S C r e e n i n g t e S t S 475

are associated with anxiety. Then the investigator would compose items that measure these behaviors and demonstrate that they are consistent with a theo- retical conception of anxiety. Construct validity is concerned primarily with the meaning of the items in a measure15 and whether the measure is associated logically with other variables specified in a theoretical framework.18 Construct validity is important to epidemiologic measures, such as scales of depressive symptomatology.

Interrelationships Between Reliability and Validity Figure 11–3 is designed to assist the reader in differentiating between reli- ability and validity, and in understanding how the two terms are interrelated. Figure 11–3 has three parts. In part A, the periphery of the target shows a measure that is highly reliable but invalid. Bullets have hit the target in the same general area (i.e., have clustered around the same general area) but have missed the bull’s-eye (center) of the target. Part B shows a measure that is nei- ther reliable nor valid. The bullets have scattered randomly around the target and have not consistently hit the bull’s-eye. Part C illustrates a measure that is both reliable and valid. The bullets have consistently hit the bull’s-eye and also cluster in the same general area. Thus, it is possible for a measure to be highly reliable but invalid. Reliability means only that the same measurement results are being reproduced on repeated occasions. Conversely, however, it is not possible for a measure to be valid but unreliable. If the measure consistently hits the bull’s-eye on repeated occasions, then the measure is, by definition, both reliable and valid.

Figure 11–3 Graphic representation of reliability and validity.

*** *** ***

*** *

* *

Reliable, Valid

Reliable, Invalid

Unreliable, Invalid

51589_CH11_Printer.indd 475 18/02/13 9:59 PM

476 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

Sources of Unrel iabi l i ty and Inval idity

This section covers three sources of unreliability and invalidity that can affect mea- sures used in screening tests and other applied epidemiologic settings; these sources are measurement bias, the halo effect, and social desirability effects. In review, the term bias is defined as the “[s]ystematic deviation of results or inferences from truth.”7 Bias can inject unreliability and invalidity into measures used for screen- ing tests. Measurement bias refers to constant errors that are introduced by a faulty measuring device. For example, a miscalibrated blood pressure manometer may consistently underestimate or overestimate true blood pressure values. Psychiatrists and clinicians also might introduce measurement biases in their judgments, which would be revealed if one rater had an average group of judgments that was consis- tently higher (or lower) than the mean ratings of other judges.

The halo effect is “[t]he influence upon an observation of the observer’s per- ception of the characteristics of the individual observed (other than the charac- teristic under study). The influence of the observer’s recollection or knowledge of findings on a previous occasion.”7 A hypothetical illustration would be a health- care provider’s tendency to rate a patient’s sexual behavior or substance use in a particular manner, based on a general opinion about a patient’s characteristics without obtaining specific information about past sexual behavior or substance abuse. With respect to screening for disease, the net effect might be to underes- timate the patient’s risk of a sexually transmitted infection (e.g., infection with the human papilloma virus) or adverse lifestyle characteristic (e.g., excessive use of alcohol or other substance).

Social desirability effects are introduced when a respondent answers questions in a manner that corresponds to the prevailing socially acceptable norms instead of giving a true answer. For example, teenage boys might respond to a screening interview about sexual behavior (and risk of sexually transmitted diseases) by exaggerating their frequency of sexual activities because these behaviors might be perceived as socially desirable among some male peer groups. A person being screened for substance use might deny using substances because many individu- als in society regard this behavior as unacceptable (and illegal).

Measures of the Validity of Screening Tests

In the context of screening, there are four measures of validity that must be considered: sensitivity, specificity, predictive value (+), and predictive value (−). Figure 11–4 represents a sample of individuals who have been examined with

51589_CH11_Printer.indd 476 18/02/13 9:59 PM

m e a S u r e S o f t h e v a l i D i t y o f S C r e e n i n g t e S t S 477

F ig

u r

e 1

1 –4

u rf

ol d

(2 ×

2 )

ta bl

e fo

r cl

as si

fi ca

ti on

o f

sc re

en in

g te

st r

es u

lt s.

D ef

in it

io n

T ru

e po

si ti

ve s

ar e

in di

vi du

al s

w h

o bo

th h

av e

be en

s cr

ee n

ed p

os it

iv e

an d

tr u

ly h

av e

th e

co n

- di

ti on

; fa

ls e

po si

ti ve

s ar

e in

di vi

du al

s w

h o

h av

e be

en s

cr ee

n ed

p os

it iv

e bu

t do

n ot

h av

e th

e co

n di

ti on

; fa

ls e

n eg

at iv

es a

re i

n di

vi du

al s

sc re

en ed

n eg

at iv

e w

h o

tr u

ly h

av e

th e

co n

di ti

on ;

an d

tr u

e n

eg at

iv es

a re

i n

di vi

du al

s w

h o

bo th

h av

e be

en s

cr ee

n ed

n eg

at iv

e an

d do

n ot

h av

th e

co n

di ti

on .

a =

T ru

e po

si tiv

es b

= F

al se

po si

tiv es

a +

b P

re di

ct iv

e V

al ue

( +

) a

a +

P os

iti ve

Te st

R es

ul t

c =

F al

se ne

ga tiv

es d

= T

ru e

ne ga

tiv es

c +

d P

re di

ct iv

e V

al ue

( –)

d c

+ d

N eg

at iv

a +

c b

+ d

To ta

G ra

nd To

ta l

a +

b +

c +

d S

en si

tiv ity

a a

+ c

S pe

ci fic

ity d

b +

P re

se nt

A bs

en t

To ta

C on

di tio

n A

cc or

di ng

to G

ol d

S ta

nd ar

51589_CH11_Printer.indd 477 18/02/13 9:59 PM

478 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

both a screening test for disease (rows) and a definitive diagnostic test (columns). Thus, we are able to determine how well the screening test performed in identi- fying individuals with disease.

●● Sensitivity: the ability of the test to identify correctly all screened individu- als who actually have the disease. In Figure 11–4, a total of a + c individu- als were determined to have the disease, according to some established gold standard, a definitive diagnosis that has been determined by biopsy, surgery, autopsy, or other method5 and has been accepted as the standard. Sensitivity is defined as the number of true positives divided by the sum of true positives and false negatives. Suppose that in a sample of 1,000 individuals there were 120 who actually had the disease. If the screening test correctly identified all 120 cases, the sensitivity would be 100%. If the screening test was unable to identify all individuals who should be referred for definitive diagnoses, then sensitivity would be less than 100%.

●● Specificity: the ability of the test to identify only nondiseased individuals who actually do not have the disease. It is defined as the number of true negatives divided by the sum of false positives and true negatives. If a test is not specific, then individuals who do not actually have the disease will be referred for additional diagnostic testing.

●● Predictive value (+) [synonym, positive predictive value]: the proportion of individuals screened positive by the test who actually have the disease. In Figure 11–4, a total of a + b individuals were screened positive by the test. Predictive value (+) is the proportion a/(a + b) who actually have the condition, according to the gold standard.

●● Predictive value (−) [synonym, negative predictive value]: an analogous measure for those screened negative by the test; it is designated by the formula d/(c + d).

Note that the only time these measures can be estimated is when the same group of individuals has been examined using both the screening test and the gold standard. According to McCunney:

False positive results are inherent in most laboratory reference lim- its, simply because of the manner by which those limits are estab- lished. People whose results are beyond 2 standard deviations from the mean are by definition “abnormal.” In general, 1 out of 20 “well people” have an abnormal test result—without evidence of illness . . . the rates of false positive results reported from a variety of health fairs [range from approximately 3% for blood chemistry tests for iron to over 20% for triglycerides].19(p 299)

51589_CH11_Printer.indd 478 18/02/13 9:59 PM

e f f e C t S o f p r e v a l e n C e o f D i S e a S e 479

The accuracy of a screening test is found by the following formula: (a + d)/ (a + b + c + d). Accuracy measures the degree of agreement between the screen- ing test and the gold standard. A sample calculation for accuracy as well as sensi- tivity, specificity, and predictive value is shown in Exhibit 11–3.

Effects of Prevalence of Disease on Screening Test Results

Sensitivity and specificity are stable properties of screening tests and, as a result, are unaffected by the prevalence of a disease. Predictive value, however, is very much affected by the prevalence of the condition being screened. Many screen- ing tests are validated upon groups that have a contrived prevalence of disease (e.g., approximately 50%). This prevalence would usually be higher than what is found in clinical practice.5

In Table 11-1, the cells of a 2 by 2 table have been arranged horizontally. Sen- sitivity and specificity are stable properties of screening tests that remain constant across groups that have different prevalences of disease. The data from Exhibit 11 –3 have been transposed to row A. Previously sensitivity and specificity were calculated to be 94.1% and 89.8%, respectively. The prevalence of disease was 51%. In row B, the prevalence of disease is 10%. The number of cases of disease (a + c) in row

Sample calculation of Sensitivity, Specificity, and predictive Value

Suppose the following data are obtained from a screening test applied to 500 people:

Condition According to Gold Standard

Screening Test Result Positive Negative Total

Positive a = 240 b = 25 a + b = 265 Negative c = 15 d = 220 c + d = 235 Total a + c = 255 b + d = 245 a + b + c + d = 500

Sensitivity = a/(a + c) = (240/255) × 100 = 94.1% Specificity = d/(b + d) = (220/245) × 100 = 89.8% Predictive Value (+) = a/(a + b) = (240/265) × 100 = 90.6% Predictive Value (−) = d/(c + d) = (220/235) × 100 = 93.6% Prevalence = (a + c)/(a + b + c + d) = (255/500) × 100 = 51% Accuracy = (a + d)/(a + b + c + d) = (460/500) × 100 = 92%

e x

h ib

it 1

1 –3

51589_CH11_Printer.indd 479 18/02/13 9:59 PM

480 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

Table 11-1 Effects of Disease Prevalence on the Predictive Value of a Screening Test

Cell Values Total

(a + b + c + d) Prevalence

No. of Cases of Disease

(a + c) Predictive Value (+)

Predictive Value (−)Row a b c d

A 240 25 15 220 500 51% 255 90.6% 93.6% B 47 45 3 405 500 10% 50 51.1% 99.3%

B is found by multiplying 500 × 0.10 = 50. The number of true positives (a) is the number of cases of disease multiplied by the sensitivity of the test: 50 × 0.94 = 47. The number of true negatives (d) is the number of diagnosed negatives multiplied by the specificity of the test: 450 × 0.90 = 405. The numbers of false negatives (c) and false positives (b) are found by subtraction (3 and 45, respectively). For row B, the predictive value (+) is [a/(a + b)] × 100 = 51.1%, and the predictive value (−) is [d/(c + d)] × 100 = 99.3%. Thus, when the values for sensitivity and specificity found in row A are applied to the data in row B, among a group of people who have a 10% prevalence of disease, the predictive value (+) decreases to 51.1% and the predictive value (−) increases to about 99%.

When the prevalence of a disease decreases the predictive value (+) falls and the predictive value (−) rises. The clinical implications of low predictive value (+) are that any individual who has a positive screening test would have low probability of having the disease; an invasive diagnostic procedure would prob- ably not be warranted for this patient. (Refer to Exhibit 11–4 for more infor- mation in the context of the positive value [+] of the prostate cancer screening test.) Table 11-1 demonstrates the effects of changing the prevalence of disease upon predictive values. When the prevalence of disease drops from about 51% to 10%, the predictive value (+) drops from about 91% to 51%, and the predictive value (−) increases from about 94% to 99%.

the importance of positive predictive Value for prostate cancer Screening

An excellent illustration of the importance of positive predictive value [predictive value (+)] is the prostate cancer screening contro- versy. Prostate cancer is the most common cancer among men in

e x

h ib

it 1

1 –4

continues

51589_CH11_Printer.indd 480 18/02/13 9:59 PM

e f f e C t S o f p r e v a l e n C e o f D i S e a S e 481

the United States; according to the National Cancer Institute over 240,000 men will be diagnosed with the disease in 2012 and 28,000 will die from it. The disease can be detected early through a simple blood test to mea- sure levels of prostate-specific antigen (PSA). PSA is a molecule found in cells that make up the prostate gland and is released when tumors disrupt the prostate cells. The PSA test was originally developed to tell whether prostate cancer was coming back in men already treated for prostate can- cer. However, doctors began giving the test to healthy men with no symp- toms of prostate cancer. Routine PSA screening became widespread in the United States by 1991, a year before the start of the first large clinical trial designed to determine if PSA screening actually saved lives. Although the sensitivity of the PSA test is good, it is not perfect as not all men with pros- tate cancer have elevated PSA levels. The specificity is more problematic because PSA blood levels can be elevated for other reasons besides cancer. For example, as men age, their prostate glands tend to enlarge and even benign conditions of the prostate can cause elevated PSA levels, as can an infection of the prostate, a condition called prostatitis. This means that a lot of men will have definitive diagnostic tests that turn out to reveal that cancer doesn’t exist. Indeed, although PSA screening can detect prostate cancer in its early, curable stages, the positive predictive value is low and 1,000 men must be screened to save 1 man’s life from prostate cancer. Men with suspicious PSA levels may go on to have a prostate biopsy. This is done with a needle; usually about a dozen small “cores” are taken. It’s unpleasant, but usually uneventful. Even so, about 70 out of 10,000 biop- sies result in infection, bleeding, or urinary difficulties. Men found to have prostate cancer—about 25% to 35% of men biopsied—have a number of options. One is to closely watch the cancer to see if it gets worse. In this case, the harm is anxiety and possibly waiting too long to get treatment. In the United States, most men opt for one of several available treatments for prostate cancer. These treatments are very effective at curing the cancer but they have a high rate of side effects including impotence, incontinence, heart attacks, and occasionally even death from treatment of tiny tumors that never would have killed them. After comparing those harms to the benefit of saving one life, the U.S. Preventive Services Task Force calculated that the harms of PSA screening outweigh the benefits.

exhibit 11–4 continued

51589_CH11_Printer.indd 481 18/02/13 9:59 PM

482 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

Relat ionship Between Sensit ivi ty and Specif ic i ty

Figure 11–5 illustrates the relationship between sensitivity and specificity. When the screening test result is a continuous or ordered variable with several levels, then the choice for a cut point that discriminates optimally between suspected diseased and normal individuals is a trade-off. Figure 11–5 demonstrates the effects of choosing various cut points. The figure shows a hypothetical distribu- tion of trait values (e.g., fasting blood glucose, an indicator of diabetes) for nor- mal individuals and a distribution curve for the diseased population that overlaps the curve for the normal population. For example, fasting blood glucose levels may approximate the normal distribution with a mean of 100 mg/dL. A sub- ject may have an elevated glucose level in the high range for a population (e.g., 120 mg/dL) and not be diabetic. Some diabetic individuals who are at the lower end of the curve for the diseased group also may have glucose levels in the high normal range. Thus, the two distributions may overlap: Some nondiseased indi- viduals may have elevated glucose levels, and some diseased individuals may have glucose levels in the lower ranges for the abnormal group. The cut point may be set at B to maximize both sensitivity and specificity. If the cut point is moved to A by lowering the specific blood glucose level that is to be classified as abnormal, almost all of the individuals who have the disease will be screened as positive, and sensitivity will approach 100%. Specificity will be lowered because more of the nondiseased individuals will be classified as diseased. By moving the cut point

Figure 11–5 Interrelationship between sensitivity and specificity.

DiseasedNormal

A B C

51589_CH11_Printer.indd 482 18/02/13 9:59 PM

e v a l u a t i o n o f S C r e e n i n g p r o g r a m S 483

to C, which represents a higher blood glucose level than point A or B, specificity will be increased at the expense of sensitivity.

Another example of establishing a cut point to distinguish between diseased and nondiseased people is setting the referral criteria for screening for glaucoma.20 By using the criterion of 15 mm Hg intraocular pressure, the sensitivity of the screening test would be high, and few persons with glaucoma would be missed. At the same time, many persons who did not have the disease would be incorrectly classified. If a high referral point were selected (e.g., 33 mm Hg), the majority of those without glaucoma would not be referred, but many with the disease would be missed by the screening test. Thus, this example demonstrates that sensitivity and specificity are complementary. “The key to a successful screening is to balance the referral criteria so that both the overreferrals and underreferrals are minimized.”20(p 360)

In summary, if one wishes to improve sensitivity, the cut point used to classify individuals as diseased should be moved farther in the range of the nondiseased. To improve specificity, the cut point should be moved farther in the range typi- cally associated with the disease. There are a number of additional procedures that can improve both sensitivity and specificity:

●● Retrain screeners: If the test requires human assessment (e.g., blood pressure readings), then improving the precision of measurement through addi- tional training sessions will reduce the amount of misclassification.

●● Recalibrate screening instrument: For those tests that utilize technology (e.g., a weighing device or a densitometer), it may be possible to reduce the amount of imprecision through refinement of the methodology.

●● Utilize a different test: In some situations there may be more than one way to measure the outcome of interest. Suppose there are two laboratory assays available to quantify serum cholesterol. If one assay performs poorly (low reliability and validity), it may be possible to replace it with a better assay.

●● Utilize more than one test: Because of the variability in some measures, it is easy to misclassify an individual as high or low. By taking more than one measure of blood pressure and averaging the results, the ability to label an individual correctly as hypertensive will be improved, resulting in better measures of sensitivity and specificity.

Evaluation of Screening Programs

Despite the intuitive appeal of screening programs, their utility should never be assumed. Rather, it is imperative that they be evaluated with the same rigor used to identify risk factors in the pathogenesis of disease. The ideal design is a

51589_CH11_Printer.indd 483 18/02/13 9:59 PM

484 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

randomized controlled trial. Under this approach, subjects are randomized either to receive the new screening test or program or to receive usual care. If the dis- ease of interest is fatal, then the appropriate end point would be differences in mortality between the two groups. For nonfatal diseases (e.g., cataracts), differ- ences in incidence between the screened and nonscreened populations should be estimated. Another approach, although less rigorous, is to conduct ecologic time trend studies in geographic areas with and without screening programs. Finally, the case-control method can be applied also: Cases are fatal (or likely to be fatal) cases of the disease, controls are nonfatal cases of the disease, and the exposure is participation in a screening program. Regardless of the approach that is taken, evaluation of screening programs is subject to several types of bias that have not yet been described fully in this chapter. Figure 11–6 depicts the natural history of disease in relation to the time of diagnosis.

Suppose the disease begins at time A and results in death at time D. A case detected as the result of screening may be picked up at time B, whereas a case that is picked up as a result of clinical signs and symptoms may not be detected until time C.

●● Lead time bias: the perception that the screen-detected case has a longer survival simply because the disease was identified earlier in the natural history of the disease. Thus, although these two individuals had identical dates of onset and death, there is an apparent increase in survival for the screen-detected case. The extent of this bias is estimated as the difference between time periods B and C.

●● Length bias: used particularly with respect to certain cancer screening pro- grams. In illustration, tumors that are detected by a screening program

Figure 11–6 Natural history of disease in relation to time of diagnosis. C to D, survival time for unscreened case; B to D, survival time for screened case; B to C, lead time.

SCREENED

NOT SCREENED

B C D

DEATH

Disease begins

Screen diagnosis

Diagnosis because of signs and symptoms

51589_CH11_Printer.indd 484 18/02/13 9:59 PM

i S S u e S i n t h e C l a S S i f i C a t i o n 485

tend to be slower growing and hence have an inherently better prognosis than tumors that are more rapidly growing and are detected as a result of clinical manifestation.

●● Selection bias: Although this topic was covered as an aspect of study designs, selection bias also is relevant to the evaluation of screening programs. In particular, individuals who are motivated enough to participate in screen- ing programs may have a different probability of disease (as a result of other healthy behaviors) than individuals who refuse participation.

In conclusion, the foregoing section has discussed a number of factors that need to be taken into account in the evaluation of screening tests. Indeed, many of the issues remain controversial, as noted in Exhibit 11–1 for mammography screening and Exhibit 11–4 for the prostate-specific antigen (PSA) test.

Issues in the Classi f icat ion of Morbidity and Mortal i ty

The theme of this chapter has been screening for disease in the community and the related topics of reliability and validity of measurement. Schemes for the nomenclature and classification of disease are central to the reliable measurement of the outcome variable in epidemiologic research. The terms nomenclature and classification are defined as follows: A nomenclature is a highly specific set of terms for describing and recording clinical or pathologic diagnoses to classify ill persons into groups. A system of nomenclature must be extensive so that all con- ditions encountered by the practitioner in a particular health discipline can be recorded. Classification, in contrast, lends itself to the statistical compilation of groups of cases of disease by arranging disease entities into categories that share similar features.21

The classification systems used are in some cases purely arbitrary because they are determined by the function that is to be served by classification; neverthe- less, all practitioners in a discipline need to have at their disposal a standardized system for classification of diseases. Many classification systems for diseases are theoretically possible; they might be based upon age, circumstance of onset, geo- graphic location, or some other factor connected with the purpose for which they are to be used. The categories of disease should be general so that there will be a limited number of categories that take into account all the diseases that might be encountered. The use of general categories facilitates the epidemio- logic study of disease phenomena by giving rise to groups of interrelated morbid

51589_CH11_Printer.indd 485 18/02/13 9:59 PM

486 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

conditions. The classifications of disease must be distinct so that a disease falls into only one category of the classification system. Each category of the clas- sification system must refer to diseases that are sufficiently frequent to permit several cases of disease to fall into the category. Otherwise, there would be an excessively detailed list of categories to contain the range of morbid conditions. Finally, a well-devised classification system permits standardization across differ- ent agencies and even countries so that comparisons in morbidity and mortality from disease can be made.

Two types of criteria are used for the classification of ill persons: causal and manifestational.21 It is possible to classify cases of disease according to a causal basis (e.g., tuberculosis or syphilis) or according to manifestation (e.g., affected anatomic site: hepatitis or breast cancer). Epidemiologic research relies primarily on manifestational criteria for classification in the hope that there will be a strong enough connection with causal factors to make possible etiologic studies.21

One example of a classification system is the Diagnostic and Statistical Manual of Mental Disorders-IV-TR, now in its fourth edition (text revision); it provides for standardization of the classification of psychiatric diagnoses.22 Clinicians, researchers, insurance companies, and other personnel who work in the mental health field make use of the DSM-IV-TR for classifying mental disorders.

A second example, the International Statistical Classification of Diseases and Related Health Problems, is one of the most widely used systems for the classifica- tion of diseases and is now in its 10th revision (ICD-10).23 The ICD is sponsored by the World Health Organization (Collaborative Centers for Classification of Diseases). It is designed for varied uses: for both clinical and general epidemio- logic purposes and for the evaluation of health care. The ICD-10 spans three vol- umes; volume one provides classification of diseases into three- and four- character levels (an alphanumeric coding scheme replaces the previous numeric one).

Conclusion

This chapter discussed terminology related to the quality of measures employed in epidemiology. Measurement is a crucial issue because even the most care- fully designed study may yield spurious results if premised upon faulty mea- sures. Topics covered in this chapter included reliability and validity, screening for disease, and methods for the classification of diseases. Formulas and examples for calculation of sensitivity, specificity, and predictive value were provided. The effect of prevalence of disease upon predictive value was discussed also.

51589_CH11_Printer.indd 486 18/02/13 9:59 PM

S t u D y Q u e S t i o n S a n D e x e r C i S e S 487

Study Questions and Exercises

1. Are you able to define the following? a. reliability b. validity c. precision d. accuracy e. sensitivity f. specificity g. predictive value (+) and predictive value (−)

2. What factors should govern the selection and use of a screening instru- ment by a health clinic?

3. What is the relationship between reliability and validity? Is it possible for a measure to be reliable and invalid? Conversely, is it possible for a mea- sure to be unreliable and valid?

4. Assume that the fasting blood level of a lipid is normally distributed in the population of people who do not have disease “X.” There is a smaller distribution curve of the fasting blood levels of this lipid, which also is normal in shape, for the population of persons who have disease “X,” and the curve overlaps the upper end (right side) of the curve for people without the disease. Draw distribution curves for the diseased and non- diseased populations and discuss the effects upon sensitivity and specific- ity of setting the cut point for disease and nondisease at various positions on the two overlapping curves.

5. How does the predictive value of a screening test vary according to the prevalence of disease?

6. A serologic test is being devised to detect a hypothetical chronic disease. Three hundred individuals were referred to a laboratory for testing. One hundred diagnosed cases were among the 300. A serologic test yielded 200 positives, of which one-fourth were true positives. Calculate the sen- sitivity, specificity, and predictive value of this test. (Hint: After setting up the appropriate 2 by 2 table, find missing data by subtraction. The numbers for the cells should then correspond to the numbers shown in Appendix 11.)

51589_CH11_Printer.indd 487 18/02/13 9:59 PM

488 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

7. A new test was compared with a gold standard measurement with the following results:

Gold Standard

New Test + − + 18 2 − 8 72

What are the sensitivity and specificity? 8. Using the data from question 7, what is the predictive value (+) and the

predictive value (−)? 9. A test-retest reliability study of the new test was conducted with the fol-

lowing results:

Test

Retest + − + 80 9 − 8 3

What is the percentage agreement (accuracy)? 10. The prevalence of undetected diabetes in a population to be screened

is approximately 1.5%, and it is assumed that 10,000 persons will be screened. The screening test will measure blood serum glucose content. A value of 180 mg% or higher is considered positive. The sensitivity and specificity associated with this screening test are 22.9% and 99.8%, respectively. a. What is the predictive value of a positive test? b. What is the predictive value of a negative test?

References 1. Morrison AS. Screening in Chronic Disease. New York, NY: Oxford University Press;

1985. 2. Shapiro S, Venet W, Strax P, Venet L. Periodic Screening for Breast Cancer: The Health

Insurance Plan Project and Its Sequelae, 1963–1986. Baltimore, MD: Johns Hopkins University Press; 1988.

3. Hurley SF, Kaldor JM. The benefits and risks of mammographic screening for breast cancer. Epidemiol Rev. 1992;14:101–129.

4. Fletcher SW, Black W, Harris R, et al. Report of the International Workshop on Screening for Breast Cancer. J Natl Cancer Inst. 1993;85:1644–1656.

5. Haynes RB. How to read clinical journals, II: to learn about a diagnostic test. Can Med Assoc J. 1981;124:703–710.

51589_CH11_Printer.indd 488 18/02/13 9:59 PM

r e f e r e n C e S 489

6. Commission on Chronic Illness. Chronic Illness in the United States: Prevention of Chronic Illness. Cambridge, MA: Harvard University Press; 1957:1.

7. Porta M., ed. A Dictionary of Epidemiology. 5th ed. New York, NY: Oxford University Press; 2008.

8. Centers for Disease Control and Prevention. Cancer screening–United States, 2010. MMWR. 2012;61:41–45.

9. Wilson JMG, Jungner F. Principles and Practice of Screening for Disease. Public Health Paper 34. Geneva, Switzerland: World Health Organization; 1968.

10. Sackett DL, Holland WW. Controversy in the detection of disease. Lancet. 1975;2:357–359.

11. World Health Organization. Mass Health Examinations. Geneva, Switzerland: World Health Organization; 1971. Public Health Paper 45.

12. Halperin W, Baker EL Jr. Public Health Surveillance. New York, NY: Van Nostrand Reinhold; 1992.

13. Beaglehole R, Bonita R, Kjellström T. Basic Epidemiology. Geneva, Switzerland: World Health Organization; 1993.

14. Cochrane AL, Holland WW. Validation of screening procedures. Br Med Bull. 1971;27:3–8.

15. Thorndike RL, Hagen E. Measurement and Evaluation in Psychology and Education. 2nd ed. New York, NY: Wiley; 1961.

16. Abramson JH. Survey Methods in Community Medicine. 4th ed. New York, NY: Churchill Livingstone; 1991.

17. Weiss NS. Clinical Epidemiology: The Study of the Outcome of Illness. New York, NY: Oxford University Press; 1986.

18. Babbie E. The Practice of Social Research. 13th ed. Belmont, CA: Wadsworth; 2013. 19. McCunney RJ. Medical surveillance: principles of establishing an effective program.

In: McCunney RJ, ed. Handbook of Occupational Medicine. Boston, MA: Little, Brown; 1988:297–309.

20. Myrowitz E. A public health perspective on vision screening. Am J Optom Physiol Opt. 1984;61:359–360.

21. MacMahon B, Pugh TF. Epidemiology Principles and Methods. Boston, MA: Little, Brown; 1970.

22. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Text Revision: DSM-IV-TR. Washington, DC: American Psychiatric Association; 2000.

23. World Health Organization. International Statistical Classification of Diseases and Related Health Problems. 2nd ed. 10th revision. Geneva, Switzerland: World Health Organization; 2004.

51589_CH11_Printer.indd 489 18/02/13 9:59 PM

490 C h a p t e r 1 1 S C r e e n i n g f o r D i S e a S e i n t h e C o m m u n i t y

appenDix

11 Data for Problem 6

Given Find by Subtraction

Total = 300 Total − (TP + FN) = FP + TN = 300 − 100 = 200 TP + FN = 100 FP = (TP + FP) − TP = 200 − 50 = 150 TP + FP = 200 FN = (TP + FN) − TP = 100 − 50 = 50

TP = 50 TN = (FP + TN) − FP = 200 − 150 = 50

TP, true positive; FN, false negative; FP, false positive; TN, true negative.

51589_CH11_Printer.indd 490 18/02/13 9:59 PM

3chapte r

491

12chapte r

Epidemiology of Infectious Diseases

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● state modes of infectious disease transmission ●● define three categories of infectious disease agents ●● identify the characteristics of agents, such as infectivity,

pathogenicity, virulence, and incubation period ●● define quantitative terms used in infectious disease outbreaks ●● describe the procedure for investigating a disease outbreak

CHAPTER OUTLINE

I. Introduction II. Agents of Infectious Disease

III. Characteristics of Infectious Disease Agents IV. Host V. The Environment

VI. Means of Transmission: Directly or Indirectly from Reservoir VII. Measures of Disease Outbreaks

VIII. Procedures Used in the Investigation of Infectious Disease Outbreaks

IX. Epidemiologically Significant Infectious Diseases in the Community

X. Conclusion

XI. Study Questions and Exercises

51589_CH12_Printer.indd 491 15/02/2013 3:19 AM

492 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Introduction

Controlling infectious diseases is one of the most familiar applications of epidemiology at work in the community. Despite many advances in the pre- vention and treatment of infectious diseases, they remain significant causes of morbidity and mortality for the world’s population in developed as well as developing countries. In the United States, pneumonia–influenza was the eighth leading cause of death in 2008.1 From the world perspective, infections are the leading cause of death of children and young adults.2 Additionally, diseases such as cancer (e.g., cervical cancer, some forms of liver cancer, and bladder cancer) are associated with infectious agents. Due to increasing world travel, passengers who are infected with a dangerous or exotic communicable disease can transmit the condition from a remote corner of the globe to a crowded city within the time span of a long-distance plane flight. The 2011 film Contagion gave a fictionalized account of a deadly and mysterious infec- tious disease that crossed international borders, creating a global threat. Con- tagion portrayed events that were riveting because of their plausibility in the contemporary era of the global village.

Institutional settings (e.g., hospitals, day care centers, and facilities for the developmentally disabled) are important venues for infectious disease outbreaks. Estimates from 2002 indicated that approximately 4.5 per 100 patients admitted to U.S. hospitals experienced nosocomial (hospital- or healthcare unit-acquired) infections, which ranged from wound infections to pneumonia to bloodstream infections.3 About 100,000 of the estimated 1.7 million hospital-acquired infec- tions were fatal.3 An example of an infectious disease that affects children in child care centers is giardiasis, a gastrointestinal illness.4 In addition, outbreaks of hepatitis B occur in children’s day care centers and residential settings for the developmentally disabled. Also commanding the attention of public health prac- titioners is the spread of hospital-acquired antibiotic-resistant bacterial infections into the community.

To begin this discussion of infectious disease epidemiology, we will explore one of the models—the epidemiologic triangle—that is used to explain the occurrence of disease outbreaks such as those that occur in institutional set- tings and the community at large. The epidemiologic triangle recognizes three major factors—agent, environment, and host—in the pathogenesis of disease. Epidemiologists refer frequently to this venerable model,5 which has been used for many decades. The epidemiologic triangle (Figure 12–1) provides one of the fundamental public health conceptions of disease causality.

51589_CH12_Printer.indd 492 15/02/2013 3:19 AM

a g e n t s o f i n f e C t i o u s d i s e a s e 493

The three-factor epidemiologic triangle is particularly well suited to explaining the etiology of infectious diseases, although the current view regarding etiology of infectious diseases involves more complex multivariate causality as well. This chapter discusses how agent, host, and environment relate to the key topics in infectious disease epidemiology: methods for transmission of disease agents and specific outcomes, including foodborne illness and the major infectious diseases. Methods for investigation and control of epidemics also include examination of agent, host, and environment factors. Figure 12–2 presents one approach to categorizing the specific infectious diseases to be examined in this chapter.

Agents of Infect ious Disease

The study of biologic agents is the province of microbiology and will not be reviewed in detail in this book. Rather, the authors present a brief overview of some of the major biologic agents and the diseases associated with them. Our goal is to demonstrate how epidemiologists describe the frequency of diseases caused by infectious disease agents in populations and how they attempt to dis- cover and control mechanisms of transmission. The first component of the epi- demiologic triangle is an agent, which must be present for an infection to occur. Microbial agents include the following:

●● Bacteria: In the United States and Europe, bacterial diseases were among the leading killers during the 19th century. Antibiotics and improvements in medical care have helped to control some of these killers. Nevertheless, bacteria remain significant causes of human illness. Examples of diseases caused by bacteria are tuberculosis (TB), salmonellosis, and streptococcal

Figure 12–1 Epidemiologic triangle.

Host

Agent Environment

51589_CH12_Printer.indd 493 15/02/2013 3:19 AM

494 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

infections (e.g., strep throat and flesh-eating disease–necrotizing fasciitis). Of particular concern is the growing emergence of bacterial strains that are resistant to antibiotics (e.g., methicillin-resistant Staphylococcus aureus–MRSA).

●● Viruses: A virus is “[a] microorganism composed of a piece of genetic mate- rial (RNA or DNA) surrounded by a protein coat. To replicate, a virus must infect a living cell. . . .”9 Examples of diseases caused by viruses

Figure 12–2 Epidemiologically significant infectious diseases (a partial list). The term vaccine-preventable diseases is used by the CDC. Although some categories overlap, they are helpful for didactic purposes. Sources: Data are from Wallace RB, ed., Wallace/Maxcy-Rosenau-Last: Public Health and Preventive Medicine, 15th ed. pp. vii–ix, New York: McGraw-Hill Medical, 2008; Centers for Disease Control and Prevention, Recommended immunization schedules for persons aged zero through 18 years—United States, 2012. MMWR. Vol 61, p. 2, February 10, 2012; Sexually transmitted diseases-information from CDC. Available at http://www.cdc.gov/std/. Accessed June 28, 2012; CDC and fungal diseases. Available at http://www.cdc.gov/ncezid/ dfwed/mycotics/. Accessed August 31, 2012.

Botulism

Amebiasis Cholera Dracunculiasis E. coli Diarrhea Giardiasis Legionellosis Shigellosis

PoliomyelitisPertussisMumpsMeaslesInfluenzaDiphtheriaChickenPox

Chlamydial Infections Genital Herpes Gonorrhea HIV/AIDS

Human Papillomavirus Infections Syphilis Trichomoniasis

Viral HepatitisTuberculosis Respiratory InfectionsLeprosy

Herpes Simplex

Group A Streptococcal Diseases

Aseptic Meningitis

Dengue Fever Lyme Disease Malaria Plague

Rocky Mountain Spotted Fever

Viral Encephalitis West Nile Virus

TularemiaRabiesQ FeverPsittacosisLeptospirosisBrucellosisAnthrax

Aspergillosis Blastomycosis Candidiasis Coccidioido- mycosis Cryptococcosis Histoplasmosis Tinea Pedis

Norovirus Salmonellosis Staphyloccal Disease

Campylobacter Enteritis Trichinosis

Clostridium perfringens Food Intoxication

Foodborne

Water- and Foodborne

Vaccine Preventable

Sexually Transmitted

Person-to- Person

Arthropod- Borne

Zoonotic

Fungal

51589_CH12_Printer.indd 494 15/02/2013 3:19 AM

a g e n t s o f i n f e C t i o u s d i s e a s e 495

include viral hepatitis A, herpes simplex, influenza, and viral meningitis (aseptic meningitis). A family of RNA viruses is known as retroviruses. The human immunodeficiency virus (HIV) is a retrovirus.

●● Rickettsia: Rickettsia is a genus of bacteria that can grow within cells. Ectoparasites (e.g., fleas, lice, and ticks) transmit the majority of rickett- sial agents, which cause a variety of diseases. These include typhus fever, Q fever, Rocky Mountain spotted fever, and rickettsialpox.

●● Fungi: An example of a fungal disease is coccidioidomycosis (also called San Joaquin Valley fever), which is an endemic mycosis; the term mycosis refers to a disease caused by fungi. Other examples are blastomycosis, ring- worm (tinea capitis), and athlete’s foot (tinea pedis). Opportunistic myco- ses, which impose an increasing threat to immunocompromised patients, include candidiasis, cryptococcosis, and aspergillosis.6 During fall 2012, a major outbreak of fungal meningitis was associated with a contaminated steroid medication shipped by a Massachusetts compounding pharmacy. More than 400 cases and 30 deaths in at least 19 states were linked to injections of the medication.

●● Protozoa: These are microscopic single-cell organisms, which are responsible for diseases such as malaria, amebiasis, babesiosis, cryptosporidiosis, and giardiasis. For example, malaria is transmitted by mosquitoes in endemic areas; one of the modes for acquiring giardiasis is through ingestion of contaminated water that contains cysts of the protozoa Giardia lamblia.

●● Helminths: These organisms (found most frequently in moist, tropical areas) include intestinal parasites—roundworms (which produce asca- riasis), pinworms, and tapeworms—as well as the organisms that cause trichinellosis (trichinosis; infectious agent Trichinella spiralis). Another well-known helminth, Schistosoma mansoni (and several other species), is responsible for schistosomiasis, sometimes known as snail fever, which occurs in Africa (e.g., along the Nile River) and in South America (Brazil, Surinam, and Venezuela), China, Japan, and many other areas. The infec- tious agents responsible for schistosomiasis are not indigenous to North America.7

●● Arthropods: One of the largest classes of living things, arthropods act as insect vectors that carry a disease agent from its reservoir to humans. Mos- quitoes, ticks, flies, mites, and other insects of this type are examples of arthropod vectors that transmit a number of significant human diseases, such as Dengue fever, Lyme disease, viral encephalitis, Rocky Mountain spotted fever, trypanosomiasis, and leishmaniasis.

51589_CH12_Printer.indd 495 15/02/2013 3:19 AM

496 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Characterist ics of Infect ious Disease Agents

●● Infectivity ●● Toxigenicity ●● Pathogenicity ●● Resistance ●● Virulence ●● Antigenicity

The following characteristics influence when an infectious disease agent will be transmitted to a host, whether it will produce disease, the severity of disease, and the outcome of infection.

●● Infectivity refers to the capacity of the agent to enter and multiply in a susceptible host and thus produce infection or disease. Polio and measles are diseases of high infectivity. The secondary attack rate (discussed later in this chapter) is used to measure infectivity.

●● Pathogenicity refers to the capacity of the agent to cause overt disease in the infected host. Measles is a disease of high pathogenicity (few subclinical cases), whereas polio is a disease of low pathogenicity (most cases of polio are subclinical). A measure of pathogenicity is the ratio of the number of individ- uals with clinically apparent disease to the number exposed to an infection.

●● Virulence refers to an agent’s capacity to induce disease in the host.9 Virulence is sometimes used as a synonym for pathogenicity. A measure of virulence is the ratio formed by the number of total cases with overt infection divided by the total number of infected cases. If the disease is fatal, virulence can be measured by the case fatality rate (CFR). The rabies virus, which almost always produces fatal disease in humans, is an extremely virulent agent with a high CFR.

●● Toxigenicity refers to the capacity of the agent to produce a toxin or poison. The pathologic effects of agents for diseases such as botulism and shell- fish poisoning result from the toxin produced by the microorganism rather than from the microorganism itself.

●● Resistance (of agent) refers to the ability of the agent to survive adverse environmental conditions. Some agents are remarkably resistant (e.g., the agents responsible for coccidioidomycosis and hepatitis) and others are extremely fragile (e.g., the gonococcus and influenza viruses). Note: The term resistance also is applied to the host.

●● Antigenicity refers to the ability of the agent to induce antibody production in the host. A related term is immunogenicity, which refers to an infection’s abil- ity to produce specific immunity.5 Agents may or may not induce long-term

51589_CH12_Printer.indd 496 15/02/2013 3:19 AM

h o s t 497

immunity against infection. For example, repeated reinfection is common with gonococci, whereas reinfection with measles virus is thought to be rare.

Host

Although it was stated earlier that an agent must be present for an infectious disease to develop, it is not a sufficient cause. That is, the agent must be capable of infecting a host. The host, after exposure to an infectious agent, may progress through a chain of events leading from subclinical (inapparent) infection to an active case of the disease. The end result may be complete recovery, permanent disability, disfigurement, or death. For example, the common cold is usually self- limiting, and a complete recovery can be expected. Smallpox at one time was greatly feared because of its high morbidity and mortality. A small proportion of untreated cases of group A streptococcal infection (B hemolytic) may produce the incapacitating sequelae of rheumatic fever and nephritis. Other examples of variation in the severity of illness are shown in Figure 12–3, which demonstrates that the largest proportion of TB cases are inapparent; a small proportion are fatal. Measles virus produces a large proportion of cases with moderately severe illness and somewhat more fatalities than TB. Some of the highly infectious virulent agents, such as the rabies virus, almost invariably cause death.

Figure 12–3 Variation in the severity of illness. Infectious diseases can result in a variety of effects ranging from no clinically detectable disease to fulminating symptoms and death. Source: Modified from JS Mausner and S Kramer, Epidemiology: An Introductory Text, 2nd ed. p. 265, with permission of WB Saunders, © 1985.

inapparent mild moderate severe fatal

TUBERCULOSIS

MEASLES

RABIES

51589_CH12_Printer.indd 497 15/02/2013 3:19 AM

498 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

The ability to cause infection is determined by a number of factors; some are properties of the host. An important determinant of the degree of infec- tion and the corresponding disease severity is the host’s ability to fight off the infectious agent. This ability comprises two broad categories: nonspecific defense mechanisms and disease-specific defense mechanisms.

Nonspecific Defense Mechanisms The human body is equipped with a number of means to reduce the likelihood that an agent will penetrate and cause disease. Most environmental agents are unable to enter the body because of the protection afforded by our skin. Simi- larly, the mucosal surfaces also afford protection against foreign invaders. Tears and saliva can be thought of as a means to wash away would-be infectious agents. The high pH of our gastric juices is lethal to many agents that manage to enter the body via ingestion. The immune system is also highly developed to ingest, via phagocytes and macrophages, infectious agents.

Although the foregoing examples of nonspecific mechanisms are important in determining host susceptibility, several other factors influence host responses to infectious agents. As we age, the ability of our nonspecific defense mechanisms to fend off agents may decrease (e.g., through reduced immune function). The nutritional status of the host also may be critical because, in comparison to those who have adequate nutrition, malnourished individuals may be less able to fight off infections. Genetic factors are involved also, as illustrated by the clear differ- ences in individuals’ reactions to a mosquito bite, for example. Some may dem- onstrate little or no reaction, whereas others may develop a large welt at the site.

Disease-Specific Defense Mechanisms Disease-specific defenses include immunity against a particular agent. Immunity refers to the resistance of the host to a disease agent. Immunity to a disease may be either natural or artificial and active or passive:

●● Active: A disease organism stimulates the potential host’s immune system to create antibodies against the disease.8 Active immunity is long lasting but requires time to develop.

●● Passive: A preformed antibody is administered to a recipient; the immu- nity is usually of short duration (half-life, 3–4 weeks) for immune globulin (gamma globulin) derived from the pooled plasma of adults.8

●● Natural, active: This type of immunity, also called natural immunity, results from an infection by the agent. For example, a patient develops long-term immunity to measles because of a naturally acquired infection.8

51589_CH12_Printer.indd 498 15/02/2013 3:19 AM

t h e e n v i r o n m e n t 499

●● Artificial, active: This type of immunity, also called vaccine-induced immunity, results from an injection with a vaccine that stimulates anti- body production in the host.8 All or part of a microorganism or a modi- fied part of that microorganism is administered to invoke an immunologic response. The response mimics the natural infection but presents little or no risk to the recipient.

●● Natural, passive: Preformed antibodies during pregnancy are transferred across the placenta to the fetal bloodstream to produce short-term immu- nity in the newborn.

●● Artificial, passive: Preformed antibodies against a specific disease are admin- istered to an exposed individual to confer protection against a disease. An example is prophylaxis against hepatitis by administration of immune globulin to individuals who have been exposed.

The Environment

The environment refers to the domain in which the disease-causing agent may exist, survive, or originate. The external environment is the sum total of influences that are not part of the host and comprises physical, climatologic, biologic, social, and economic components. The physical environment includes weather, temperature, humidity, geologic formations, and similar physical dimensions. Contrasted with the physical environment is the social environment, which is the totality of the behavioral, personality, attitudinal, and cultural characteristics of a group of peo- ple. Both these facets of the external environment have an impact on agents of dis- ease and potential hosts because the environment may either enhance or diminish the survival of disease agents and may serve to bring agent and host into contact.

The environment can serve as a reservoir or niche that fosters the survival of infectious disease agents. The reservoir may be a part of the physical envi- ronment or may reside in animals or insects (vectors) or other human beings (human reservoir hosts). As an example of an environmental reservoir, contam- inated water supplies or food may harbor infectious disease agents that cause typhoid, cholera, and many other illnesses. Fungal disease agents that may reside in the soil produce coccidioidomycosis (San Joaquin Valley fever). Some infec- tious diseases have vertebrate animal reservoirs. Zoonoses are diseases that are potentially transmissible from animal reservoirs to humans under natural condi- tions,9 noteworthy examples of zoonoses being rabies and plague. Some diseases have only humans as the reservoir; notable among these is smallpox, which has been successfully eradicated because the virus apparently does not survive outside

51589_CH12_Printer.indd 499 15/02/2013 3:19 AM

500 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

the human reservoir. Other diseases, for example typhoid fever, may induce a chronic carrier state in some individuals who are not symptomatic for the dis- ease but who have the capacity to transmit it to other susceptible individuals. A famous case was Typhoid Mary, a New York City area cook in the early 20th century who was a notorious and unwitting typhoid carrier.

Means of Transmission: Direct ly or Indirect ly from Reservoir

Figure 12–4 illustrates the phases involved in the transmission of an infectious disease, which may be transmitted either directly or indirectly. For example, por- tals of exit and entry of infectious agents are required for disease transmission to take place.

Direct Transmission Direct transmission of diseases refers to spread of infection through person-to- person contact wherein transmission happens without indirect contact with an intermediate contaminated object. Direct transmission may occur, for example, from contact with the blood or bodily fluids of an infected person as in the spread of sexually transmitted diseases.

Figure 12–4 The chain of infection. Source: Reproduced from the Centers for Disease Control and Prevention. Principles of Epidemiology. 2nd ed. Atlanta, GA:CDC; 1998, p. 45.

RESERVOIR MODE OF TRANSMISSION SUSCEPTIBLE

HOST

PORTALS OF ENTRY

DROPLET

DIRECT CONTACT

VECTOR

VEHICLE AIRBORNE

AGENT FOMITE

51589_CH12_Printer.indd 500 15/02/2013 3:19 AM

m e a n s o f t r a n s m i s s i o n : d i r e C t l y o r i n d i r e C t l y 501

Indirect Transmission Indirect transmission involves the spread of infection through an intermediary source: vehicles, fomites, or vectors. Examples of vehicles that can be linked to the transmission of infectious diseases are contaminated water, infected blood on used hypodermic needles, and food. A fomite is an inanimate object—such as a doorknob or clothing—laden with disease-causing agents. Contamination refers to the presence of a living infectious agent in or on an inanimate object. Other means of indirect transmission include the unwashed hands of healthcare workers, microbe-laden patient-care devices, shared toys in a pediatric ward, and inadequately sterilized medical instruments.

A vector is an animate, living insect or animal involved with transmission of the disease agent. Arthropod vectors, such as flies and mosquitoes, sometimes form a component of the life cycle of the disease agent. For example, the Anoph- eles mosquito is essential to the survival of infectious agents for malaria (e.g., Plasmodium vivax); if the mosquito population is eradicated, the frequency of malaria cases diminishes. Control of arthropod vectors can be an effective means of limiting outbreaks of vector-borne diseases, such as malaria.

Portals of Exit and Entry Portals of exit, defined as sites where infectious agents may leave the body, include the respiratory passages, the alimentary canal, the openings in the geni- tourinary system, and skin lesions (Table 12–1). Additional portals of exit may be made available through insect bites, the drawing of blood, surgical procedures,

Table 12–1 Correspondence Between Portal of Exit (Escape) Mode of Transmission and Portal of Entry

Portal of Exit Mode of Transmission Portal of Entry Type of Disease

Respiratory secretions Airborne droplets, fomites

Respiratory tract

Common cold, measles

Feces Water, food, fomites, flies

Alimentary tract Typhoid, poliomyelitis

Lesions, exudate Direct contact, fomites, sexual intercourse

Skin, genital membranes

Carbuncles, syphilis, gonorrhea

Conjunctival exudate Fomites, flies Ocular mucous membrane

Trachoma

Blood Bloodsucking arthropod vector

Skin (broken) Malaria, yellow fever, epidemic typhus

Source: Modified from Fox JP, Hall CE, Elvebach LR. Epidemiology: Man and Disease. New York: The Macmillan Company, 1970, p. 63.

51589_CH12_Printer.indd 501 15/02/2013 3:19 AM

502 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

and injuries. For the chain of transmission to be continued, the portal of exit must be appropriate to the particular agent. To produce infection, the agent must exit the source in sufficient quantity to survive in the environment and to overcome the defenses surrounding the portals of entry in the new host.

Transmission of an infectious agent requires a locus of access to the human body known as the portal of entry. Examples are the respiratory system (e.g., for diseases such as influenza and the common cold); the mouth and digestive sys- tem (e.g., for diseases such as hepatitis A or staphylococcal food poisoning); and the mucous membranes or wounds in the skin (e.g., for bacterial diseases such as staphylococcal infections).

Other Concepts Related to Disease Transmission Other concepts related to disease transmission shown in the following box are covered in this section.

●● Inapparent infection ●● Incubation period ●● Herd immunity ●● Generation time

●● Colonization and infestation ●● Iceberg concept of infection ●● Inapparent/apparent case ratio ●● Indirect transmission

Inapparent infection An inapparent (subclinical) infection is one that has not yet penetrated the clini- cal horizon (i.e., does not have clinically obvious symptoms). Nevertheless, inap- parent infections can be of major epidemiologic significance: Asymptomatic individuals could transmit the disease to other susceptible hosts, some of whom might develop a severe illness. Isolation of infected individuals is more likely to occur when the infectious disease is clinically apparent (i.e., when the ratio of apparent to inapparent cases is high). Examples of communicable diseases that require isolation are cholera, infectious tuberculosis, and the plague.

To determine whether an infection has taken place in both symptomatic and asymptomatic individuals, one may search for serologic evidence of infection by immunoassay. An elevated blood antibody level (elevated antibody titer) sug- gests previous exposure and infection by the disease agent. For example, hepatitis A (infectious hepatitis) often is manifested as a subclinical infection in nursery school children, who may transmit the disease even though they do not have clinical symptoms. The infectious process may be tracked by monitoring blood

51589_CH12_Printer.indd 502 15/02/2013 3:19 AM

m e a n s o f t r a n s m i s s i o n : d i r e C t l y o r i n d i r e C t l y 503

antibody and enzyme responses to infection with hepatitis A virus (HAV). The epidemiologist and clinician may conclude that infection has taken place by noting antibody and enzyme increases for hepatitis A after an appropriate incubation period.

Incubation period The incubation period is the time interval between exposure to an infectious agent and the appearance of the first signs or symptoms of disease.9 The incubation period, when the infectious organism replicates within the host, is often a fixed period of hours, days, or weeks for each disease agent. Frequently, epidemiolo- gists take into account the incubation period when attempting to fix the source of an infectious disease outbreak with respect to the time and circumstance of exposure as well as type of agent. For example, the incubation period for measles (rubeola) is most commonly 10 days, but ranges from 7 to 18 days and requires about 2 weeks for a rash to appear. Another application of incubation periods is to determine the causes of outbreaks of foodborne illness; calculation of the incubation period helps to narrow down possible etiologic agents, each of which has characteristic incubation periods.

Herd immunity The term herd immunity refers to the immunity of a population, group, or com- munity against an infectious disease when a large proportion of individuals are immune (through either vaccinations or past infections). Herd immunity can occur when immune persons prevent the spread of a disease to unimmunized individuals; herd immunity confers protection to the population even though not every single individual has been immunized. For example, herd immunity against rubella may require that 85–90% of community residents are immune; for diphtheria it may be only 70%.

Generation time The term generation time relates to the time interval between lodgment of an infectious agent in a host and the maximal communicability of the host. The generation time for an infectious disease and the incubation time may or may not be equivalent. For some diseases (e.g., mumps), the period of maximum communicability precedes the development of active symptoms. The period of maximum communicability for mumps precedes the swelling of salivary glands by about 48 hours.5 There is another distinction between incubation period and

51589_CH12_Printer.indd 503 15/02/2013 3:19 AM

504 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

generation time. The term incubation period applies only to clinically apparent cases of disease, whereas the term generation time applies to both inapparent and apparent cases of disease. Thus, generation time is utilized for describing the spread of infectious agents that have a large proportion of subclinical cases.

Colonization and infestation It is important to emphasize that not all exposures to agents lead to illness. Colonization refers to the situation where an infectious agent may multiply on the surface of the body without invoking tissue or immune response. Infestation describes the presence of a living infectious agent on the body’s exterior surface, upon which a local reaction may be invoked. Thus, the full spectrum of disease in the community setting may involve much more than individuals presenting with clinical symptoms.

Iceberg concept of infection The iceberg concept of infection posits that the tip of the iceberg, which cor- responds to active clinical disease, accounts for a relatively small proportion of hosts’ infections and exposures to disease agents. Figure 12–5 demonstrates that most infections are subclinical and that in a substantial number exposure to a disease agent may not produce any infection or cell entry.

Figure 12–5 Iceberg concept of infection. Source: Adapted from Evans AS, Kaslow RA. (editors) Viral Infections of Humans: Epidemiology and Control, 4th ed. New York: Plenum Medical Book Company, 1997.

Lysis of cell Fatal

Clinical and severe disease

Moderate severity Mild illness

Infection without clinical illness

Exposure without infection

Exposure without cell entry

Incomplete viral maturation

Cell transformation or

cell dysfunction

Discernible effect

Below visual change

Clinical disease

Subclinical disease

CELL RESPONSE HOST RESPONSE

51589_CH12_Printer.indd 504 15/02/2013 3:19 AM

m e a n s o f t r a n s m i s s i o n : d i r e C t l y o r i n d i r e C t l y 505

Table 12–2 Variations in the Clinical Presentations of Viral Infections with Comparative Rank in Inapparent/Apparent Case Ratios

Comparative Rank

Inapparent/ Apparent

Case Ratio Name of Virus Primary Symptoms

Approximate Percentage of

Asymptomatic Cases (Ratio of Inapparent

to Apparent Cases Shown in

Parentheses)

+++++ High Hepatitis B (children)

Similar to other forms of acute viral hepatitis

Almost all cases are asymptomatic

+++++ High Human papillomavirus (HPV)

Anogenital warts, cancers

Almost all cases are asymptomatic

+++++ High Poliomyelitis Minor, nonspecific illness (4%–8%); nonparalytic aseptic meningitis (1%–2%); flaccid paralysis (<1%)

95% (ratio of inapparent to paralytic illness, 50:1 to 1,000:1)

++++ Very high Hepatitis A (children younger than 6)

Jaundice 70% (somewhat more than 2:1)

+++ Medium Hepatitis B (adults)

Similar to other forms of acute viral hepatitis

50% (about 1:1)

+++ Medium Rubella Maculopapular rash about 2 weeks after exposure

Up to 50% (about 1:1)

++ Low Hepatitis A (adults)

Jaundice 30% (about 1:2)

+ Very low Rabies Delirium, hallucinations Virtually no cases are asymptomatic; nearly always a fatal disease

Sources: Data from Atkinson W, Hamborsky J, Stanton A, Wolfe C. Epidemiology and Prevention of Vaccine-Preventable Diseases, 12th ed. Education, Information and Partnership Branch. National Center for Immunization and Respiratory Diseases, Centers for Disease Control and Prevention; 2012 and CDC Signs and Symptoms—Rabies. http://www.cdc.gov. Accessed October 17, 2012.

Inapparent/apparent case ratio Table 12–2 illustrates the wide variation in clinical presentation of viral infec- tions. Although the vast majority of polio infections do not produce severe dis- ease, the opposite is true for rabies, a condition for which only a few infected cases (in the absence of timely administration of rabies prophylaxis) have survived. The percentage of apparent clinical cases for hepatitis A increases from childhood to adulthood. About half of rubella and hepatitis B infections (among adults) are manifested as clinical cases.

51589_CH12_Printer.indd 505 15/02/2013 3:19 AM

506 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Measures of Disease Outbreaks

Attack Rate An attack rate is “[t]he proportion of a group that experiences the outcome under study over a given period (e.g., the period of an epidemic).”9 An attack rate, a variant of an incidence rate, is used when the occurrence of disease among a population at risk increases greatly over a short period of time, often related to a specific exposure. In addition, the disease rapidly follows the exposure during a fixed time period because of the nature of the disease process.11 Thus, the term attack rate is frequently applied to the occurrence of acute infectious disease out- breaks (e.g., microbial foodborne illness and other acute infectious diseases) and also can be used for other acute health-related events (e.g., acute exposures of large groups to toxic agents).

An attack rate is often expressed as a percentage; the formula is:

The numerator consists of the number of people who develop an illness as a result of exposure to the suspected agent, and the denominator consists of all people, whether well or ill, who were exposed to the agent during a time period. The time interval during which exposure occurred is an important element of the definition, but is often defined arbitrarily or is uncertain; hence an attack rate is not a true rate.

Table 12–3 provides a calculation example for attack rates associated with an outbreak of foodborne illness. The example illustrates how attack rates helped to identify food items that might have caused the outbreak; in the hypothetical out- break, several foods were implicated. Table 12–3 shows the method to calculate the attack rate for a specific food item. Food X demonstrated a 77% attack rate among those who ate and a 64% attack rate among those who did not eat the food. In order to identify foods suspected of producing an outbreak, the follow- ing procedure is recommended.

First, compile a list of all foods consumed during the outbreak. Next, categorize the persons involved in the outbreak in two columns: A (ate the food)

Attackrate Ill

(Ill Well) 100 duringa=

+ × ttime period

51589_CH12_Printer.indd 506 15/02/2013 3:19 AM

m e a s u r e s o f d i s e a s e o u t b r e a k s 507

and B (did not eat the food). One calculates the attack rate among those in categories A and B by dividing the number of ill persons by the total number of persons and multiplying the result by 100. For example, the attack rate in column A is (10/13) × 100 = 77%. After calculating the attack rate, one finds the difference in attack rates (A – B) between those who ate and those who did not eat the food, in this case 13% (77% – 64%). One would repeat this process for each of the foods that were suspected in the outbreak of foodborne illness. Those foods that have the greatest difference in attack rates may be the foods that were responsible for illness. To complete the investigation, additional studies, including cultures and laboratory tests, might be required. In the fore- going situation, “time” is the estimated time interval during which the out- break occurred—from exposure through appearance of all associated cases of disease. An example is given in the study questions and exercises at the end of the chapter.

Secondary Attack Rate The secondary attack rate yields an index of the spread of a disease within a household or similar circumscribed unit (Exhibit 12–1). The secondary attack rate is defined as “[T]he number of cases of an infection that occur among contacts within the incubation period following exposure to a primary case in relation to the total number of exposed contacts; the denominator is restricted to susceptible contacts when these can be determined. The second- ary attack rate is a measure of contagiousness and is useful in evaluating con- trol measures.”9

As a hypothetical example of a secondary attack rate, suppose measles spreads from two initial cases of measles brought into a barracks to other resi- dents. Table 12–4 provides data for calculation of a secondary attack rate for the hypothetical spread of measles (rubeola) in a military barracks housing Reserve Officers’ Training Corps summer cadets. The two initial cases occurred

Table 12–3 Data to Illustrate Calculation of Attack Rates for Food X

A (ate the food) B (did not eat the food)

Ill Not Ill A Total Attack Rate Ill Not Ill B Total Attack Rate

10 3 13 77% 7 4 11 64%

51589_CH12_Printer.indd 507 15/02/2013 3:19 AM

508 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

at the beginning of the military summer program and were presumed to have resulted from exposure outside the military base. Of the two initial cases, the case that came to the attention of public health authorities is called the index case.5 The other case could be considered a coprimary case. This term refers to a case that is related to the index case so closely in time that it is thought to belong to the same generation of cases as the index case. The eight sec- ondary cases occurred approximately 10–12 days after measles symptoms were observed in the index cases. The total number of new cases (initial cases plus secondary cases) in the group was 10. The secondary attack rate was [(10 − 2)/ (14 − 2)] × 100 = 66.7%.

Secondary attack rate

The secondary attack rate refers to the spread of disease in a family, household, dwelling unit, dormitory, or similar circumscribed group.

Secondary AttackRate (%)

Number ofnewc

= aasesin

group initial case(s) Number ofsusc

− eeptible persons

in the group initial case(s− ))

Initial case(s) = Index case(s) + coprimaries Index case(s) = Case that first comes to the attention of public health

authorities Coprimaries = Cases related to index case so closely in time that they are

considered to belong to the same generation of cases

e x

h ib

it 1

2 –1

Table 12–4 Hypothetical Secondary Attack Rate Data for Military Cadets

Number of Cadets In Barracks Number of Initial Cases Number of Secondary Cases

Unimmunized Immunized Unimmunized Immunized Unimmunized Immunized

14 6 2 0 8 0

51589_CH12_Printer.indd 508 15/02/2013 3:19 AM

m e a s u r e s o f d i s e a s e o u t b r e a k s 509

“Seven cases of hepatitis A occurred among 70 children attending a child care center. Each infected child came from a different family. The total number of persons in the 7 affected families was 32. One incuba- tion period later, 5 family members of the 7 infected children also devel- oped hepatitis A.”12(p. 89) By applying the formula shown in Exhibit 12–1 to these data, we obtain the following information:

Numerator = 5 (number of cases of hepatitis A among family contacts with hepatitis)

Denominator = 25 = 32 − 7 (total number of family members minus children already infected)

The secondary attack rate is (5/25) × 100 = 20%. Note: In this example, no subtraction is necessary in calculating the numerator because the initial cases have already been removed from the number of new cases in the group; the problem specified that five family members developed hepatitis one incubation period later.

Here is a second example of how to calculate a secondary attack rate:

For diseases such as measles, which confer prolonged immunity, the index cases (and coprimaries) are excluded (subtracted) from the denominator.13 If means are available to determine immune status, any other immune persons also would be excluded from the denominator (as implied by the definition).12 In addition to assessing the infectivity of an infectious disease agent, the secondary attack rate may be used to evaluate the efficacy of a prophylactic agent (e.g., a vaccine or gamma globulin). It also may be used to trace the secondary spread of a disease of unknown etiology to determine whether there is a transmissible agent.14

Here is an example of a calculation of a secondary attack rate for an outbreak of pandemic influenza. Communicable disease experts estimated the secondary attack rate of pandemic influenza during a 2009 outbreak in Western Australian households.15 The researchers studied a total of 595 households in which pan- demic influenza A (H1N1) 2009 had occurred. A household consisted of at least two people residing together in a dwelling, but excluded residential institutional settings. The definition of an index case was the first symptomatic case of influ- enza in a household. Anyone who developed influenza within 1–7 days following the index case’s development of symptoms was called a secondary case. Among the 1,589 household contacts of the index cases, 231 persons developed influ- enza. The secondary attack rate was 14.5% [(231/1,589) × 100].

51589_CH12_Printer.indd 509 15/02/2013 3:19 AM

510 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Case Fatality Rate The case fatality rate (CFR) refers to a proportion formed by the number of deaths caused by a disease (or, more generally, health condition) among those who have the disease during a time interval; the CFR is expressed as a percent- age. (Refer to Exhibit 12–2.) The CFR provides an index of the virulence of a particular disease within a specific population.

Sometimes confusion exists regarding the distinction between a CFR and a cause-specific mortality rate. One important distinction is that the CFR differs from a cause-specific mortality rate for a disease with respect to the denominator used. The denominator for a CFR for a specific disease is the number of cases of that disease. The denominator for a cause-specific mortality rate is the size of the population in which mortality from the disease occurs.

Let us compare the kinds of information yielded by a cause-specific death rate and the CFR. Recent epidemiologic surveillance demonstrates that mortal- ity from human rabies is very uncommon in the United States. The fact that rabies is uncommon in humans may be attributed to its confinement to wildlife (epizootic rabies) and to post-exposure prophylaxis (passive immunization and vaccines) among those who have been potentially exposed to rabies. Therefore, the cause-specific death rate for any given recent year due to rabies would be low.

case Fatality rate (cFr)

CFR (%) Number ofdeathsdue to disease "= XX"

Number ofcasesofdisease "X" 100 durin× gga time period

Note that the numerator and denominator refer to the same time period.

Sample calculation: Assume that an outbreak of plague occurs in an Asian country during the month of January. Health authorities

record 98 cases of the disease, all of whom are untreated. Among these, 60 deaths are reported.

CFR = (60/98) × 100 = 61.2%

Examples of diseases with a high CFR are rabies and untreated bubonic plague. n

e x

h ib

it 1

2 –2

51589_CH12_Printer.indd 510 15/02/2013 3:19 AM

p r o C e d u r e s u s e d i n t h e i n v e s t i g a t i o n 511

The reason is that the CFR for this condition has a small numerator and uses the total population as the denominator. The cause-specific mortality rate due to rabies therefore would be a small number. In contrast, the CFR for rabies would be high. The CFR reflects the fatal outcome of disease, which is affected by effi- cacy of treatment. Among the cases of human rabies that may occur as a result of failure to receive post-exposure prophylaxis, mortality remains almost invariably certain, as has been historically true.

Basic Reproductive Rate (R0) The basic reproductive rate (R0) is “[a] measure of the number of infections pro- duced on average by an infected individual in the early stages of an epidemic when virtually all contacts are susceptible.”9 Among its possible applications, R0 can be used as a measure of the transmissibility of influenza. As an example, Kenah et al. developed a model to explain influenza transmission patterns dur- ing the 2009–2010 pandemic of influenza A (H1N1) 2009 and to project the possible future global pandemics.10 Their model included factors such as the cities where influenza originated, linkages among cities through air transporta- tion networks, time of the year, and R0. As a reflection of seasonality, the basic reproductive rate was hypothesized to be higher during the influenza season than at other times of the year.

Procedures Used in the Invest igat ion of Infect ious Disease Outbreaks

These include many of the techniques developed by John Snow16: mapping and tabulation of cases, identification of agents, and clinical observation. The investi- gation of an outbreak can be logically divided into five basic steps:

1. Define the problem. It is important to determine from the outset whether the outbreak or epidemic is real. For example, suppose a restaurant patron claimed that a gastrointestinal illness she developed was caused by the food she ate in the restaurant. The epidemiologist would need to verify that this was a case of foodborne illness and not a sporadic case of stom- ach upset. Other cases of the same illness reported to the health depart- ment would increase the index of suspicion that the illness originated at the restaurant.

2. Appraise existing data. This step includes evaluation of known distri- butions of cases with respect to person, place, and time. Examples of

51589_CH12_Printer.indd 511 15/02/2013 3:19 AM

512 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

activities performed at this stage include case identification, making clinical observations, and generation of tables and spot maps.

●● Case identification. This step includes tracking down all cases of disease potentially involved in the outbreak, for instance, in a mass occurrence of foodborne illness. Examples of foodborne illnesses include staphy- lococcal food poisoning and trichinosis. For communicable diseases, such as TB, contacts of cases need to be identified.

●● Clinical observations. The epidemiologist records the number, types, and patterns of symptoms associated with the disease (e.g., whether the symptoms are primarily gastrointestinal, respiratory, or febrile). Additional clinical information may come from stool samples, cul- tures, or antibody assays.

●● Tabulation and spot maps. Cases of disease may be plotted on a map to show geographic clustering (as in an outbreak of TB in a high school). Cases may be tabulated by date or time of onset of symptoms, by demographic characteristics (i.e., age, sex, and race), or by risk catego- ries (e.g., intravenous drug use or occupational exposure). This method is similar to Snow’s mapping of cholera cases in the Broad Street district of London in the mid-1800s.16 It remains an important epidemiologic technique. Graphing cases according to time of onset helps determine the modal incubation period or other aspects of the outbreak.

●● Identification of responsible agent. The epidemiologist may be able to determine the agent or other factors responsible for the disease out- break by estimating the incubation period, by reviewing the specific symptoms, and by noting evidence from cultures and other laboratory tests. In some outbreaks of foodborne illness, it may be possible to link an etiologic agent to cultures of food specimens and stool samples.

3. Formulate a hypothesis. What are the possible sources of infection? What is the likely agent? What is the most likely method of spread? What would be the best approach for control of the outbreak?

4. Test the hypothesis. Collect the data necessary to confirm or refute your initial suspicions. At this stage it is important to continue to search for additional cases, evaluate alternative sources of data, and begin laboratory investigations to identify the causative agent.

5. Draw conclusions and formulate practical applications. Based on the results of your investigation, it is likely that programs, policies, or procedures will need to be implemented to facilitate long-term surveillance and ulti- mate prevention of the recurrence of similar outbreaks.

51589_CH12_Printer.indd 512 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 513

Epidemiological ly Signif icant Infect ious Diseases in the Community

The following discussion illustrates major outbreaks of infectious diseases and demonstrates how the foregoing methods are utilized. A partial list of epidemio- logically significant infectious diseases is shown in Figure 12–2. These include the broad categories of foodborne, waterborne, vaccine preventable, sexually transmitted, person-to-person, arthropod-borne, zoonotic, and fungal diseases.

Foodborne Illness in the Community Table 12–5 provides names of agents and diseases that are associated with food contamination, which is one of the most common infectious disease problems in the community. Among the agents of foodborne illness noted in Table 12–5 is Staphylococcus aureus, which can cause outbreaks when food is stored at improper temperatures. A documented historical example occurred on Febru- ary 3, 1975, during the flight of a jumbo jet from Japan. Omelets prepared in Alaska, held over for the flight, and served onboard sickened almost 200 passengers. This episode demonstrated that contamination of food can occur during any one of the settings for food production. Figure 12–6 illustrates components of the food production chain. Each of the venues (from farm to home) shown in the figure provides an opportunity for dissemination of infec- tious foodborne agents.

Another foodborne illness is trichinosis (Exhibit 12–3). Trichinosis is most commonly associated with consumption of pork products that have not been adequately cooked.

trichinosis associated with Meat from an alaskan Grizzly bear

Eight cases of trichinosis reported from Barrow, Alaska, were associated with a dinner on December 20, 1980. The 12 persons who attended were served a meal that included maktak (whale blubber), ugruk (bearded-seal meat, dried and stored in seal oil), fresh raw whitefish and grayling, and quaq (raw frozen meat), thought by the participants to be caribou but later discovered to have been grizzly bear.

e x

h ib

it 1

2 –3

continues

51589_CH12_Printer.indd 513 15/02/2013 3:19 AM

514 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Five men and three women, ranging in age from 32 to 76 years, became ill 2 to 16 days after the meal (mean, 8.6 days). All eight reported eat- ing quaq; the four who denied doing so remained well, a statistically sig- nificant difference. Quaq was the only meat eaten by all the persons who became ill. Thirty other family members who were not present at the din- ner also remained well, which again is a statistically significant difference.

Signs and symptoms of illness included edema (100%), fatigue (100%), myalgia (87.5%), rash (87.5%), fever (87.5%), chills (75%), periorbital edema (62.5%), headache (50%), visual disturbance (37.5%), diarrhea (37.5%), abdominal cramps (25%), nausea (25%), and vomiting (25%).

None of the ill persons had notable pulmonary, neurologic, or cardiac complications. Five were hospitalized. Five received steroids, and two received anthelmintic therapy.

The grizzly bear from which the meat came had been shot the previ- ous autumn at the family’s summer camp, 140 miles inland from Barrow. At that time, parts had been cooked thoroughly and consumed without adverse effects. The hindquarters were included in a large cache of moose and caribou meat that was returned to Barrow and stored frozen in the family’s cold cellar. None of the bear meat had been eaten in Barrow before the dinner on December 20, and none had been given away. The remains of the hindquarter eaten at the dinner were fed to dogs; the other hindquarter remained in cold storage. A sample taken from the digestive tract of one of the patients contained 70 Trichinella larvae per gram of meat. n

Source: Adapted from Centers for Disease Control and Prevention. Trichinosis associated with meat from a grizzly bear—Alaska. MMWR. Vol 30, pp. 115–116, March 20, 1981.

exhibit 12–3 continued

Table 12–5 Partial List of Infectious Agents That Cause Foodborne Illness

Disease/Agent Usual Incubation Period and

Syndrome Mode of Transmission

Classic botulism/ Clostridium botulinum

12–36 hours, classic syndrome compatible with botulism

Contaminated food containing toxins (e.g., home-canned foods)

Salmonellosis/various species of Salmonella (e.g., S. typhimurium and S. enteritidis)

12–36 hours, gastrointestinal syndrome

Contaminated food that contains Salmonella organisms (e.g., undercooked chicken, eggs, meat; raw milk)

continues

51589_CH12_Printer.indd 514 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 515

Table 12–5 continued

Disease/Agent Usual Incubation Period and

Syndrome Mode of Transmission

Staphylococcal food poisoning/Staphylococcus aureus (see Fig. 12–6)

2–4 hours, gastrointestinal syndrome; majority of cases with vomiting

Contaminated food that contains staphylococcal enterotoxin

Cholera/Vibrio cholerae 2–3 days, profuse watery diarrhea (painless)

Contaminated water that contains infected feces or vomitus; also contaminated food

Clostridium perfringens food poisoning

10–12 hours, diarrhea Heavily contaminated food (e.g., meats and gravies inadequately heated or stored at temperatures that permit multiplication of bacteria)

Campylobacter enteritis/ Campylobacter jejuni

2–5 days, diarrhea, abdominal pain, malaise, fever

Undercooked chicken or pork, contaminated food and water, raw milk

Source: Data are from Heymann DL. Control of Communicable Diseases Manual. 19th ed. Washington, DC: American Public Health Association; 2008.

Figure 12–6 The food production chain. Source: Reproduced from CDC. The Food Production Chain—How Food Gets Contaminated. Available at: http://www.cdc .gov/outbreaknet/investigations/figure_food_production.html Accessed: November 3, 2012.

51589_CH12_Printer.indd 515 15/02/2013 3:19 AM

516 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Water- and foodborne diseases Water- and foodborne diseases include amebiasis, cholera, giardiasis, legionellosis, and schistosomiasis. These conditions are responsible for a significant burden of morbidity and mortality in areas where they occur. This section covers schistoso- miasis (river fever, bilharzia), cholera, and amebiasis.

One location for the transmission of schistosomiasis is along the Nile River in Africa. The Nile is the site of many human activities (e.g., recreation, tasks related to personal hygiene, and sewage disposal). Deposition of human wastes into the river’s water causes contamination by schistosomes, the parasitic worms that cause schistosomiasis. Those who come into contact with the water are at increased risk of schistosomiasis. After the building of the Aswan Dam, rates of schistosomiasis increased due to human intervention in the life cycle of schistosomiasis.

Figure 12–7 illustrates the life cycle of schistosomiasis.17 Transmission of schistosomiasis requires Biomphalaria glabrata, the intermediate snail host for schistosomes. The three major species that infect human beings are Schistosoma haematobium, S. intercalulatum, and S. mansoni. The species that is the major cause of schistosomiasis in Africa is S. mansoni. The life cycle of schistosomes entails a complex web that involves alternate human and snail hosts. Refer to the numbers shown in the figure:

1. Eggs are illuminated with feces or urine. 2. Under optimal conditions, the eggs hatch and release miracidia. 3. The miracidia swim and penetrate specific snail intermediate hosts. 4. This stage in the snail includes two generations of sporocysts. 5. The snails produce cercariae. 6. Upon release from the snail the infective cercariae swim and penetrate the

skin of the human host. 7. In the human host, they shed their tails, becoming schistosomulae. 8. The schistosomulae migrate through several tissues and stages to their

residence in the veins. 9. The schistosomulae migrate to the portal blood in the liver and mature

into adults. 10. Paired adult worms migrate to mesenteric venules of bowel/rectum (lay-

ing eggs that circulate in the liver and shed in stools).

Source: Adapted from Centers for Disease Control and Prevention. Parasites and health. Schistosomiasis. http://dpd.cdc.gov/dpdx/html/schistosomiasis.htm. Accessed July 1, 2012.

51589_CH12_Printer.indd 516 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 517

Cholera is another example of a water- or foodborne disease. It is characterized as an acute enteric disease with sudden onset, occasional vomiting, rapid dehy- dration, acidosis, and circulatory collapse.7 Caused by the bacterial agent Vibrio cholerae, it still occurs in many parts of the developing world. During 1991 and 1992 cholera epidemics spread throughout South America, eventually reaching Central America and Mexico. More than 700,000 cholera cases and 6,000 deaths from cholera-related causes were reported from 21 countries in the western hemi- sphere during this time period. In 1992, 102 cholera cases were reported in the United States; this figure exceeded the number of cases in any previous year since cholera surveillance began in 1961.18 Presently, cholera is an infrequent condi- tion in the United States, with 5 to 10 cases reported during each year between 2005 and 2009 and 8 cases reported as of December 2010.19 Since 2000, the incidence of cholera has shown an increasing trend with approximately 317,000 cases and 7,500 deaths globally.20 Endemic areas include developing countries in Africa and Asia; often these countries are crowded, have poor sanitation, and

Figure 12–7 Life cycle of Schistosoma mansoni. Source: Reproduced from Centers for Disease Control and Prevention. Parasites and health. Schistosomiasis. Available at: http://dpd.cdc.gov/dpdx/html/schistosomiasis.htm. Accessed July 1, 2012.

Paired adult worms migrate to: mesenteric venules of bowel/rectum (laying eggs that circulate to the liver and shed in stools) venous plexus of bladder

S. haematobiumS. mansoni

in feces in urine

Miracidia penetrate snail tissue

Penetrate skin

Cercariae released by snail into water and free-swimming

Cercariae lose tails during penetration and become schistosomulae

Circulation

Migrate to portal blood in liver and mature into adults

Sporocysts in snail (successive generations)

= Infective Stage

= Diagnostic Stage

S. japonicum

4 5

51589_CH12_Printer.indd 517 15/02/2013 3:19 AM

518 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

unsafe water supplies. An outbreak during August 2007 affected 30,000 people in Kirkuk, Iraq.

On January 12, 2010, a severe earthquake caused severe destruction to Haiti including its drinking water treatment facilities. In October 2010, the Haitian Ministry of Public Health and Population was informed of numerous patients affected with symptoms of cholera, subsequently confirmed by labora- tory tests.21 By November 2010, a total of 16,111 cases of hospitalization for cholera symptoms (e.g., acute watery diarrhea) had occurred. Haiti is subdi- vided into 10 departments (départements in French), which are administra- tive districts. Figure 12–8 presents the number of hospitalizations by Haitian department; seven of the ten Haitian departments were affected. Artibonite was the department that reported the largest number of cases (10,230). Before the 2010 outbreak, Haiti had been free from cholera outbreaks for more than a century.

Figure 12–8 Number of persons hospitalized with cholera, by department [administrative division]—Haiti, October 20— November 13, 2010. Source: Reproduced from Centers for Disease Control and Prevention. Update: cholera outbreak—Haiti 2010, MMWR. 2010; 59 (45):1475.

D o m i n i c a n

R e p u b l i c

(Nnmber of Cases)

Grande Anse

Nippes

(1,086) Centre

Artibonite (10,230)

Nord’Est (1)

Nord’Ouest

(1,548)

Nord

Artibonite River

Sud’ Est

(875) Port-au-Prince Ouest (794)

Sud (45)

(1,548) > 9,000 hospitalizations

< 1,000 hospitalizations

No laboratory-confirmed cases

1,001-9,000 hospitalizations

51589_CH12_Printer.indd 518 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 519

A third example of a water- or foodborne disease agent is Entamoeba histolytica, a fairly common parasitic organism associated with amebiasis, which in many cases manifests as intestinal disease. The following excerpt from a report by the Centers for Disease Control and Prevention demonstrates the transmission of E. histolytica through colonic irrigation:

The Colorado State Department of Health has reported an outbreak of amebiasis that occurred in the period December 1977– November 1980 and was associated with a chiropractic clinic. All of the cases had received colonic irrigation—a series of enemas performed by machine to “wash out” the colon—a practice that has been gain- ing popularity recently among some chiropractors, naturopaths, and nutritional counselors. Thirteen cases were confirmed by biopsy review or serologic tests. Seven cases were fatal.22(p 101)

Stool cultures, blood tests, and biopsy studies strongly suggested amoebas as the etiologic agent. Specimens from the colonic irrigation machine were heavily contaminated with fecal bacteria.

Sexually Transmitted Diseases A major public health problem from this category is the HIV/AIDS epidemic, which has had major domestic and international consequences for economic activi- ties and utilization of health resources. In 1981, the Morbidity and Mortality Weekly Report published descriptions of five cases of Pneumocystis carinii pneumonia among previously healthy men.23 The disease became known as acquired immune deficiency syndrome (AIDS) caused by the human immunodeficiency virus (HIV).

After 1981 the number of diagnosed AIDS cases among persons 13 years of age and older increased rapidly until peaking in 1992 when 75,457 cases were diagnosed. In 1995 the number of deaths from AIDS reached 50,624. A highly active antiretroviral therapy has been associated with a significant decline in the number of AIDS diagnoses and deaths. (Refer to Figure 12–9.) As of 2008 the CDC estimated that 1,178,350 persons were living with HIV, including approx- imately one-fifth who were undiagnosed.23

Most HIV infections (75.0%) occur among men. High-risk populations for HIV infection include men who have sex with men (MSN), African-Americans, and Hispanics or Latinos. HIV infections are most likely to be undiagnosed among younger persons (13–34 years of age), males with high-risk heterosexual contacts, MSM, and certain ethnic and racial groups (e.g., Asian or Pacific Islanders; Amer- ican Indians or Alaska natives). Refer to the text box for more information.

51589_CH12_Printer.indd 519 15/02/2013 3:19 AM

520 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Figure 12–9 Estimated number of AIDS diagnoses and deaths and estimated number of persons living with AIDS diagnosis and living with diagnosed or undiagnosed HIV infection among persons aged ≥13 years– United States, 1981–2008. Source: Reproduced from Centers for Disease Control and Prevention. HIV surveillance—United States, 1981—2008, MMWR. 2011; 60(21):691.

AIDS diagnoses

N o.

of A

ID S

d ia

gn os

es /d

ea th

s (in

th ou

sa nd

AIDS surveillance case definition expanded

Introduction of highly active antiretroviral therapy

AIDS deaths

Living with HIV infection

Living with AIDS diagnosis

1,200

1,100

1,000

900

800

700

600

500

400

300

200

100

N o.living w

ith A ID

S diagnosis/H

IV infection

(in thousands)

1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 2007

Year

General U.S. population: 157.7 per 100,000 population Adult and adolescent males: 297.3 per 100,000 Adult and adolescent females, 86.5 per 100,000 Children (younger than 13 years old at the time of diagnosis): 7.3 per 100,000 Blacks/African-Americans: 560.0 per 100,000 (the highest estimated prevalence among racial/ethnic groups.)24 n

Estimated Prevalence of AIDS Diagnoses at the End of 2008

The human immunodeficiency virus (HIV) epidemic is particularly acute from a worldwide perspective. The World Health Organization estimates that approximately 34.2 million people were living with HIV in 2011. Approximately 23.5 million HIV-infected adults and children were estimated to be living in the sub-Saharan region of Africa. Other high prevalence regions included South/ Southeast Asia, 4.2 million persons; Eastern Europe/Central Asia, 1.5 million persons; and Latin America and North America, 1.4 million persons each.25 (Figure 12–10).

51589_CH12_Printer.indd 520 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 521

Vaccine-Preventable Diseases In the United States, healthcare providers administer routine vaccina- tions to children aged 0–6 years for the prevention of several diseases. Among these conditions are diphtheria, pertussis, tetanus, Haemophilus influenzae type B infections, hepatitis A, hepatitis B, measles, mumps, rubella, paralytic poliomyelitis, influenza, meningococcal meningitis, chicken pox, pneumococcal disease, and rotaviral enteritis. As a result of public health efforts and availabil- ity of effective vaccines, many vaccine-preventable diseases have low incidence in the U.S. According to data reported by the CDC for 2010,19 the following are the reported numbers of cases of several vaccine-preventable conditions that reached very low frequency:

●● congenital rubella syndrome, 0 ●● rubella, 6 ●● diphtheria, 0 ●● paralytic poliomyelitis, 0

The incidence of measles in the United States has tended to remain low fol- lowing the licensing of the measles vaccine in 1963. However, a major resurgence of measles occurred during the period 1989 to the early 1990s. The respective numbers were 27,786 (1990); 9,643 (1991); and 2,231 (1992).26 A total of

Figure 12–10 Adults and children estimated to be living with HIV in 2011. Source: Courtesy of UNAIDS. Joint United Nations Programme on HIV/AIDS (UNAIDS) and World Health Organization (WHO), July 2012 Core Epidemiology Slides, Slide #5.

Total: 34.2 million [31.8 million – 35.9 million]

Western & Central Europe

860,000 [780,000 – 960,000]

Middle East & North Africa 330,000

[250,000 – 450,000]

Sub-Saharan Africa 23.5 million

[22.2 million – 24.7 million]

Eastern Europe & Central Asia 1.5 million

[1.3 million – 1.8 million]

South & South-East Asia 4.2 million

[3.1 million – 4.7 million]

Oceania 53,000

[47,000 – 60,000]

North America 1.4 million

[1.1 million – 2.0 million]

Latin America 1.4 million

[1.1 million – 1.7 million]

East Asia 830,000

[590,000 – 1.2 million]Caribbean 230,000

[200,000 – 250,000]

51589_CH12_Printer.indd 521 15/02/2013 3:19 AM

522 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

138 measles cases were reported in 1997; this number represented a dramatic drop after the early 1990s. The numbers of cases continued to decline in the late 1990s and thereafter; the incidence was 61 cases in 2010.19 Figure 12–11 illustrates the number of reported cases of measles from 1950 to 2009. Part A

Figure 12–11 Measles. (Part A) Reported measles cases, by year–United States, 1950–1997. (Part B) Measles incidence (per 100,000 population), by year—United States, 1974–2009. Sources: (Part A) Reproduced from From Centers for Disease Control and Prevention. Measles—United States, 1997, MMWR. 1998;47(14): 273. (Part B) Reproduced from Centers for Disease Control and Prevention. Summary of Notifiable Diseases—United States, 2009, MMWR. Vol 58, p. 67, May 13, 2011.

1000

800

600

400

200

0 1950 1954 1958 1962 1966 1970 1974 1978 1982 1986 1990 1994

Year

Vaccine Licensed

1992 1993 1994 1995 1996 1997

C as

es (

T ho

us an

ds )

C as

es (

T ho

us an

ds )

Year

1994 1999 2004 Year

2009

* Per 100,000 population. Y–axis is log scale.

In ci

de nc

MEASLES.Incidence, *by year–United States, 1994–2009

10.00

1.00

0.10

0.01

In ci

de nc

0 1974 1979 1984 1989 1994 1999 2004 2009

Part B

51589_CH12_Printer.indd 522 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 523

(1950–1994) shows the steep drop in the number of cases after licensing of the measles vaccine in 1963. Part B (1974–2009) shows a stable trend in measles incidence since 1994.

Formerly, measles outbreaks occurred in institutional settings, such as univer- sities or colleges and other places where susceptible persons congregated. Nowa- days, measles outbreaks are unlikely to occur in the United States because most people have been immunized. The epidemiology of measles supports the conclu- sion that endemic transmission of measles has been eliminated in the United States. Sporadic cases that have occurred in recent years were imported from for- eign countries. Because 100% coverage of U.S. residents might never be achieved, accurate surveillance and rapid response to outbreaks are essential to prevention of widespread transmission of imported measles. Public health practitioners will need to continue advocating for high levels of immunization. Exhibit 12–4, which typifies measles outbreaks that can occur in contemporary times, describes a measles outbreak associated with an arriving refugee during 2011.

Measles Outbreak associated with an arriving refugee—Los angeles county, california, august–September 2011

Background: Each year, on average, 60 people in the United States are reported to have measles. But, in 2011, the number of reported cases was higher than usual—222 people had the disease. Nearly 40% of these people got measles in other countries, including countries in Europe and Asia. They brought the disease to the United States and spread it

to others. This caused 17 measles outbreaks in various U.S. communities. Measles was declared eliminated from the United States in 2000.

Widespread use of measles-mumps-rubella (MMR) vaccine has resulted in elimination of indigenous measles circulation in the United States. So, the disease no longer spreads year round in this country. However, spo- radic outbreaks of measles continue to occur in the United States, typically linked to imported cases from countries where measles remains endemic.

But, the disease is still common throughout the world, including some coun- tries in Europe, Asia, the Pacific, and Africa. Anyone who is not protected against measles is at risk of getting infected when they travel internationally. They can

e x

h ib

it 1

2 –4

continues

51589_CH12_Printer.indd 523 15/02/2013 3:19 AM

524 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

bring measles to the United States and infect others. Unvaccinated people put themselves and others at risk for measles and its serious complications.

The 2011 refugee-associated outbreak: An ill passenger arriving in Los Angeles from Malaysia was linked to cases of measles in two passengers on the same flight and a U.S. Customs and Border Protection officer. The index patient had never been vaccinated against measles and was immigrating to the United States. A total of 50 health officials interviewed 298 contacts in the resulting investigation. Delayed diagnosis and notification of health officials precluded the use of MMR vaccination for outbreak containment.

Implications for public health practice: Measles should be considered in the differential diagnosis of any febrile rash illness in a patient with recent international travel; in suspected cases, health authorities should be noti- fied immediately and the patient isolated. Widespread MMR vaccination is a highly effective way to limit illness and complications from measles.

Sources: Adapted and reprinted from Centers for Disease Control and Prevention. Measles out- break associated with an arriving refugee–Los Angeles County, California, August–September 2011. MMWR. 2012;61(21):388; Centers for Disease Control and Prevention. Measles out- breaks. http://www.cdc.gov/measles/outbreaks.html. Accessed August 5, 2012. n

Diseases Spread by Person-to-Person Contact Examples of infectious diseases in this category are aseptic meningitis, group A streptococcal diseases, and respiratory infections. (Refer to Figure 12–2.) This section covers two examples: Tuberculosis (TB) and viral hepatitis.

Tuberculosis Tuberculosis is a significant cause of morbidity and mortality throughout the world. However, TB was uncommon in many developed countries, including the United States. Beginning in the late 1980s, TB incidence increased greatly in the United States and then declined toward the end of the 20th century. Rea- sons for the resurgence of TB included the increasing prevalence of HIV infec- tion, growing numbers of homeless persons, and the importation of cases from endemic areas. From 1984 to 1992, the number of reported TB cases trended upward from 22,201 to 26,673. During this time period 51,700 excess cases were reported, according to statistical models of expected cases in comparison with observed cases. (Figure 12–12 presents a historically important chart that shows this upward trend.) Between 1985 and 1992, the number of cases increased in all racial/ethnic groups except non-Hispanic whites and Native Americans.

51589_CH12_Printer.indd 524 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 525

All age groups except for the over-65 age group showed increases, with the largest increase occurring among the 25- to 44-year age group.27

By 2010 TB incidence had declined to 11,181 cases (3.6 per 100,000 popula- tion).28 The most affected groups in the United States were foreign-born indi- viduals and racial and ethnic minorities. At present, two high-risk populations are migrant farm workers and homeless persons. Even though the number of TB cases has declined over time, this communicable disease remains a signifi- cant public health threat. In 2007, a suspected case of extensively drug-resistant tuberculosis (XDR TB) commanded media attention (refer to the case study).

Viral hepatitis Viral hepatitis includes several forms and associated viruses, which are discussed in Table 12–6. Figure 12–14 shows hepatitis virions. The most common types are hepatitis A, hepatitis B, and hepatitis C. The incidence of hepatitis A and

Figure 12–12 Expected and observed number of tuberculosis cases— United States, 1980–1992. Source: Reprinted from Centers for Disease Control and Prevention. Tuberculosis morbidity—United States, 1992. MMWR, vol 42, p 696, September 17, 1993.

1992199019881986198419821980 0

10,000

15,000

20,000

25,000

30,000

35,000

40,000

45,000

50,000

Observed Cases

Expected Cases

51,700 Excess Cases

Year

C as

es (

Lo g

S ca

le )

51589_CH12_Printer.indd 525 15/02/2013 3:19 AM

526 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

On May 12, 2007, a suspected XDR TB patient traveled to Europe for his honeymoon. The traveler boarded a commercial flight, Air France number 385, to Paris. His itinerary took him to several European coun- tries. On May 24, he returned to Canada on Czech Air flight num- ber 0104 from Prague, Czech Republic. During his voyage, the CDC issued an advisory that warned of the traveler’s infection with XDR TB. He returned to the United States from Canada via automobile with- out being detained by U.S. officials at the border. Upon arrival in the United States, he was hospitalized in a respiratory isolation unit for evaluation.

The form of TB known as XDR TB resists treatment with first-line drugs (e.g., isoniazid) and second-line drugs. Fellow passengers aboard flight 385 and other flights the patient took, as well as other persons who came into contact with the patient during his 14-day trip, poten- tially were exposed to XDR TB. (Later the traveler’s diagnosis was down- graded to multidrug-resistant TB: MDR TB). Refer to Figure 12–13 for a timetable of events. n

Sources: Data from Centers for Disease Control and Prevention. Health Alert Update. Corrected: Investigation of U.S. traveler with extensively drug-resistant tuberculosis (XDR TB). http://www.bt.cdc.gov/HAN/han00262.asp. Accessed July 2, 2012; Centers for Disease Control and Prevention. Extensively drug-resistant tuberculosis–United States, 1993–2006, MMWR. 2007;(56)11:250–253.

Case Study: Extensively Drug-Resistant Tuberculosis (XDR TB)

hepatitis B generally has been declining since the late 1980s. One of the reasons for reductions in the incidence of HAV infections is the vaccination of children in states that have high rates of such infections. Reduction in the incidence of HBV infections has been linked to routine vaccination of infants.29

Hepatitis A is transmitted by ingestion of fecal matter (fecal–oral dissemina- tion). Hepatitis B can be spread by contact with infected body fluids (e.g., among adults who engage in risky sexual practices; hence, HBV infection remains more common in this subpopulation than in other groups). Among other means for spread of HBV are contacts with blood and blood products, such as through accidental needle sticks.

51589_CH12_Printer.indd 526 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 527

F ig

u r

e 1

2 –1

3 (

co nt

in ue

Ja n

‘0 7

3/ 8

3/ 26

4/ 10

4/ 25

4/ 27

4/ 30

5/ 4

5/ 9

5/ 10

5/ 11

5/ 12

A C

T IO

N S

t o

P R

O T

E C

T P

U B

L IC

H E

A L

T H

: X

D R

T B

C A

S E

T E

S T

IN G

IN T

E R

IM *

T IM

E T

A B

L E

Ja n

‘0 7

P at

ie nt

’s c

he st

X -R

ay a

nd C

T 6

sc an

a re

ab no

rm al

Ja n

‘0 7

S pu

tu m

sm ea

r ne

ga tiv

e fo

r T

4/ 25

P hy

si ci

an as

ks fo

r ex

pe di

te d

su sc

ep tib

ili ty

te st

in g

of is

ol at

e cu

ltu re

4/ 26

G A

P H

fin ds

su ffi

ci en

t gr

ow th

to be

gi n

su sc

ep tib

ili ty

te st

in g

4/ 25

P at

ie nt

ad vi

se s

cl in

ic ph

ys ic

ia n

of ov

er se

as tr

av el

( no

t iti

ne ra

ry )

4/ 25

P at

ie nt

re po

rt s

to th

e F

ul to

n C

o. T

B C

lin ic

4/ 27

C D

C re

ce iv

es pa

tie nt

sa m

pl e

fo r

su sc

ep tib

ili ty

te st

in g

5/ 10

F ul

to n

C o.

H D

3 be

gi ns

re vi

ew o

f l eg

al re

m ed

ie s

to re

st ric

t p at

ie nt

m ov

em en

5/ 10

–5 /1

1 G

A D

P H

a sk

s C

D C

fo r

op tio

ns th

at c

ou ld

be u

se d

to re

st ric

t t ra

ve l i

n a

pe rs

on w

ith un

tr ea

te d

M D

R 7 T

B .

C D

C r

es po

nd s

w ith

o pt

io ns

5/ 11

P at

ie nt

ch an

ge s

fli gh

t de

pa rt

ur e

fr om

5/ 14

to 5

/1 2

(u nk

no w

n to

he al

th au

th or

iti es

) 5/ 12

P at

ie nt

de pa

rt s

U .S

. on

A ir

F ra

nc e

#3 85

(u nk

no w

n to

he al

th au

th or

iti es

)

4/ 30

G A

P H

pr el

im in

ar y

su sc

ep tib

ili ty

te st

r es

ul ts

in di

ct at

e M

D R

7 T

5/ 9

G A

P H

re pe

at ed

su sc

ep tib

ili ty

te st

r es

ul ts

in di

ct at

e M

D R

7 T

5/ 10

F am

ily m

ee tin

g he

ld at

P M

D 2

of fic

e w

ith F

ul to

n C

o. T

B c

lin ic

re pr

es en

te d

5/ 4

G A

P H

fin al

su sc

ep tib

ili ty

te st

r es

ul ts

in di

ct at

e M

D R

7 T

3/ 8

D ia

gn os

tic br

on ch

os co

3/ 26

S pe

ci m

en cu

ltu re

p os

iti ve

fo r

M .T

B 5

at ho

sp ita

l l ab

3/ 26

re sc

rib ed

1s t l

in e,

4 d

ru g

T B

th er

ap y

fo r

se lf

ad m

in is

tr at

io n

4/ 10

S pe

ci m

en co

nf irm

ed fo

r M

.T B

5 at

ho sp

ita l l

4/ 10

G A

P H

re ce

iv es

is ol

at e

fo r

M .T

B 5

su sc

ep tib

ili ty

te st

in g

an d

re in

cu ba

tio n

L O

C A

T E

X D

R T

B P

A T

IE N

1 P H

L =

P ub

lic H

ea lth

L ab

or at

or y

2 P M

D =

P riv

at e

D oc

to r

3 H D

= H

ea lth

D ep

ar tm

en t

4 D P

H =

D iv

is io

n of

P ub

lic H

ea lth

5 M .T

B =

M yc

ob ac

te riu

m tu

be rc

ul os

is 6 C

T =

C om

pu te

d T

om og

ra ph

y (C

A T

s ca

n) 7 M

D R

= M

ul tid

ru g

R es

is ta

*6 /1

9/ 07

, S ub

je ct

to U

pd at

es A

ll da

te s

re fle

ct E

D T

5/ 11

–5 /1

2 F

ul to

n C

o. H

D 3

at te

m pt

s to

h an

d de

liv er

w rit

te n

ad vi

so ry

to pa

tie nt

. P at

ie nt

ca nn

ot b

e lo

ca te

F ig

u r

e 1

2– 13

ct io

n s

to p

ro te

ct p

u bl

ic h

ea lt

h :

X D

R T

B c

as e.

51589_CH12_Printer.indd 527 15/02/2013 3:19 AM

528 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

F ig

u r

e 1

2– 13

ct io

ns to

p ro

te xt

p ub

lic h

ea lth

: X D

R T

B c

as e.

( C

on tin

ue d)

S ou

rc e:

D at

a fr

om C

en te

rs fo

r D

is ea

C on

tr ol

a nd

P re

ve nt

io n,

E xt

en si

ve ly

d ru

g- re

si st

an t t

ub er

cu lo

si s,

I nt

er im

ti m

el in

e— A

ct io

ns to

p ro

te ct

p ub

lic h

ea lth

: I nv

es ti

ga ti

on o

f U

.S . t

ra ve

le r

w it

h X

D R

T B

— Ju

ne 1

9, 2

00 7.

A va

ila bl

e at

: h tt

p: //

w w

w .c

dc .g

ov /t

b/ xd

rb /t

im el

in e.

ht m

. A cc

es se

d Ju

ne 2

1, 2

00 7.

5/ 18

5/ 22

5/ 24

5/ 25

5/ 26

5/ 28

5/ 29

5/ 30

5/ 31

6/ 2

A C

T IO

N S

t o

P R

O T

E C

T P

U B

L IC

H E

A LT

H : X

D R

T B

C A

S E

IN V

E S

T IG

A T

E S

P R

E A

D O

F IN

F E

C T

IO N

IN T

E R

IM *

T IM

E T

A B

L E

5/ 18

G A

D P

H n

ot ifi

ed C

D C

-D G

M Q

th at

p at

ie nt

w ith

M D

R 1

T B

tr av

el ed

in te

rn at

io na

lly ;

D G

M Q

in iti

at es

e ffo

rt to

lo ca

te p

at ie

nt a

nd tr

av el

it in

er ar

5/ 18

–5 /2

5 C

D C

c on

ta ct

s fa

m ily

, a cq

ua in

ta nc

es to

fi nd

p at

ie nt

5/ 25

P at

ie nt

is is

ol at

ed in

N Y

C ho

sp ita

l u nd

er fe

de ra

l a ut

ho rit

5/ 26

–5 /2

7 C

D C

te am

in te

rv ie

w s

pa tie

5/ 28

C D

C fl

ie s

pa tie

nt to

A tla

nt a

5/ 29

–5 /3

0 C

D C

d is

cu ss

es op

tio n

fo r

sa fe

tr av

el w

ith pa

tie nt

5/ 30

–5 /3

1 C

D C

r ec

ei ve

s pa

ss en

ge r

co nt

ac t i

nf o

fr om

m ul

tip le

so ur

ce s

fo r

co nt

ac t t

ra ci

ng to

fa ci

lit at

e te

st in

5/ 31

C D

C e

sc or

ts pa

tie nt

to D

en ve

5/ 25

C D

C o

bt ai

ns iti

ne ra

ry fr

om pa

tie nt

a nd

re qu

es ts

pa ss

en ge

r in

fo fo

r co

nt ac

t t ra

ci ng

5/ 25

C D

C co

nt ac

ts p

at ie

nt in

U .S

5/ 25

–5 /2

7 C

D C

r eq

ue st

s pa

tie nt

e va

lu at

io n

an d

la b

te st

s in

N Y

5/ 25

C B

P 5

va lid

at es

U S

re en

tr y

of p

at ie

nt ,

no tif

ie s

C D

6/ 2

C D

re sc

in ds

fe de

ra l

is ol

at io

n or

de r

5/ 22

C D

C te

st s

co nf

ir m

X D

R 2

T B

5/ 22

B as

ed o

n C

D C

in fo

, C B

P 5

fla gs

p at

ie nt

fo r

na tio

nw id

e bo

rd er

lo ok

ou t

5/ 22

–5 /2

3 C

D C

s pe

ak s

to pa

tie nt

in R

om e

to d

is cu

ss di

ag no

si s

an d

op tio

ns fo

r sa

fe re

tu rn

5/ 24

P at

ie nt

lo ca

tio n

un kn

ow n;

C D

C re

qu es

ts T

S A

is su

e or

de r

to pr

ev en

t p at

ie nt

fr om

b oa

rd in

g U

.S . b

ou nd

fli gh

5/ 24

C D

C no

tif ie

s W

H O

in G

en ev

fo rm

al IH

R 7

no tif

ic at

io n

by H

H S

o n

5/ 25

. C

D C

n ot

ifi es

Ita lia

n M

in is

tr y

of H

ea lth

a ut

ho rit

ie s

5/ 29

–o n

g o

in g

C D

C e

ng ag

es n

ew s

m ed

ia to

n ot

ify p

as se

ng er

s of

p os

si bl

e ex

po su

re a

nd d

ire ct

th em

to C

D C

h ot

lin e.

1 M D

R =

M ul

tid ru

g R

es is

ta nt

2 X D

R =

E xt

en si

ve ly

D ru

g R

es is

ta nt

3 r el

ia bl

e m

et ho

ds fo

r 2n

d lin

e dr

ug su

sc ep

tib ili

ty te

st s

re qu

ire 2

1– 28

d ay

4 D G

M Q

= D

iv is

io n

of G

lo ba

l M ig

ra tio

an d

Q ua

ra nt

in e

5 C B

P =

C us

to m

s an

d B

or de

r P

ro te

ct io

6 W H

O =

W or

ld H

ea th

O rg

an iz

at io

n 7 I

H R

= In

te rn

at io

na l H

ea lth

R eg

ul at

io ns

*6 /1

9/ 07

, S ub

je ct

to U

pd at

es A

ll da

te s

re fle

ct E

D T

5/ 29

–o n

g o

in g

C D

C c

on ta

ct s

th os

e se

lf- id

en tif

ie d

pa ss

en ge

rs w

ho c

al le

d C

D C

IS O

L A

T E

X D

R T

B P

A T

IE N

5/ 25

–o n

g o

in g

C D

C c

ol la

bo ra

te s

w ith

in te

rn at

io na

l, fe

de ra

l, st

at e,

a nd

lo ca

l h ea

lth au

th or

iti es

o n

co nt

ac t t

ra ci

ng . A

s of

6 /1

1/ 07

, 9 5%

o f U

.S . p

as se

ng er

s co

nt ac

te d

by C

D C

IS O

L A

T E

X D

R T

B P

A T

IE N

T L

O C

A T

E X

D R

T B

P A

T IE

N T

51589_CH12_Printer.indd 528 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 529

Figure 12–14 Hepatitis virions of an unknown strain of the organism. Source: Reproduced from Centers for Disease Control and Prevention, Public Health Image Library, ID number 8153. Available at: http://phil.cdc.gov/phil/. Accessed April 21, 2012.

Hepatitis C virus, carried primarily in blood, can be transmitted through injection drug use and sexual contact. The virus is also transmitted during the perinatal period from mother to child. Approximately 80% of HCV infec- tions are asymptomatic; HCV infection is a leading cause of liver transplan- tation. Refer to Figure 12–15 for trends in viral hepatitis incidence in the United States.

51589_CH12_Printer.indd 529 15/02/2013 3:19 AM

530 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

T ab

le 1

2– 6

F iv

e Ty

pe s

of V

ir al

H ep

at it

H ep

at it

is A

H ep

at it

is B

H ep

at it

is C

H ep

at it

is D

H ep

at it

is E

D es

cr ip

tio n

A n

ac ut

e liv

er d

is ea

ca us

ed b

y th

e he

pa tit

is A

ru s

(H A

V) , l

as tin

g fr

a fe

w w

ee ks

t o

se ve

ra l

m on

th s;

d oe

s no

t le

ad t

o ch

ro ni

c in

fe ct

io n.

A li

ve r

di se

as e

ca us

by t

he h

ep at

iti s

B v

iru s

(H B

V) . S

ev er

ity r

an ge

s fr

om a

m ild

il ln

es s,

la st

in g

a fe

w w

ee ks

( ac

ut e)

, to

a s

er io

us lo

ng -t

er m

hr on

ic )

ill ne

ss t

ha t

ca n

le ad

t o

liv er

d is

ea se

o r

liv er

nc er

A li

ve r

di se

as e

ca us

by t

he h

ep at

iti s

C v

iru s

(H C

V) . S

om et

im es

r es

ul ts

a n

ac ut

e ill

ne ss

, b ut

os t

of te

n be

co m

es a

ro ni

c co

nd iti

on t

ha t

ca n

le ad

t o

ci rr

ho si

s of

t he

li ve

r an

d liv

er c

an ce

A s

er io

us li

ve r

di se

as e

ca us

ed b

y th

e he

pa tit

D v

iru s

(H D

V) ; r

el ie

s on

B V

to r

ep lic

at e.

It is

co m

m on

in t

he U

.S .

A s

er io

us li

ve r

di se

as e

ca us

ed b

y th

e he

pa tit

E vi

ru s

(H EV

); u

su al

re su

lts in

a n

ac ut

e in

fe ct

io n

an d

do es

t le

ad t

o a

ch ro

ni c

in fe

ct io

n. R

ar e

in t

U .S

., bu

t co

m m

on in

an y

pa rt

s of

t he

w or

ld .

Tr an

sm is

si on

In ge

st io

n of

fe ca

l m at

te r,

en in

m ic

ro sc

op ic

ou nt

s, fr

om c

lo se

rs on

-t o-

pe rs

co nt

ac t

or in

ge st

io n

of c

on ta

m in

at ed

fo od

d rin

ks . U

su al

as ym

pt om

at ic

in c

hi ld

re n

yo un

ge r

th an

6 .

C on

ta ct

w ith

in fe

ct io

bl oo

d, s

em en

, a nd

he r

bo dy

f lu

id s

fr om

vi ng

s ex

w ith

a n

in fe

ct ed

p er

so n,

s ha

rin g

co nt

am in

at ed

n ee

dl es

t o

in je

ct d

ru gs

, o r

fr om

a n

in fe

ct ed

m ot

he r

to h

ne w

bo rn

C on

ta ct

w ith

t he

b lo

of a

n in

fe ct

ed p

er so

pr im

ar ily

t hr

ou gh

s ha

rin g

co nt

am in

at ed

n ee

dl es

t o

in je

ct d

ru gs

C on

ta ct

w ith

in fe

ct io

bl oo

d, s

im ila

r to

h ow

H B

V is

s pr

ea d.

In ge

st io

n of

fe ca

l m at

te r,

en in

m ic

ro sc

op ic

ou nt

s; o

ut br

ea ks

a re

ua lly

a ss

oc ia

te d

w ith

nt am

in at

ed w

at er

pp ly

in c

ou nt

rie s

w ith

or s

an it

at io

V ac

ci na

tio n

R ec

om m

en de

d fo

r al

l ch

ild re

n st

ar tin

g at

a ge

ye ar

, t ra

ve le

rs t

o ce

rt ai

n co

un tr

ie s,

a nd

o th

er s

at r

is k.

R ec

om m

en de

d fo

r al

l in

fa nt

s, o

ld er

c hi

ld re

n an

d un

va cc

in at

ed a

do le

sc en

ts ,

an d

ad ul

ts a

t ris

k fo

r H

B V

in fe

ct io

Th er

e is

n o

va cc

in e

fo r

he pa

tit is

C .

Th er

e is

n o

va cc

in e

fo r

he pa

tit is

D .

Th er

e is

c ur

re nt

ly n

o FD

A -a

pp ro

ve d

va cc

in e

fo r

he pa

tit is

E .

So ur

ce : A

da pt

ed a

nd r

ep rin

te d

fr om

C en

te rs

fo r

D is

ea se

C on

tr ol

a nd

P re

ve nt

io n.

V ira

l H ep

at iti

s. h

tt p:

// w

w w

.c dc

.g ov

/h ep

at iti

s/ Pu

bl ic

In fo

.h tm

#w ha

tI sH

ep .

A cc

es se

d A

pr il

21 , 2

01 2.

51589_CH12_Printer.indd 530 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 531

Zoonotic Diseases A zoonosis is a disease that, under natural conditions, can be spread from ver- tebrate animals to humans. Zoonotic diseases may be either enzootic (similar to endemic in human diseases) or epizootic (similar to epidemic in human dis- eases).7 Refer to Figure 12–2 for a brief list of zoonotic diseases. Noteworthy examples are anthrax, brucellosis, leptospirosis, Q fever, and rabies.

Anthrax (agent, Bacillus anthracis) is uncommon in industrialized countries, but is a hazard for workers who come into contact with wool, hair, and hides; it has been distributed intentionally via the U.S. mail system as a bioterrorism agent. Several people died as a result of their exposure during a 2001 bioterrorism attack.

Brucellosis (undulant fever) is caused by Brucella abortus. Those primarily at risk are farmers and others who work with infected cattle, swine, goats, and sheep. Some- times raw (unpasteurized) milk is a vehicle associated with outbreaks of brucellosis.

Figure 12–15 Incidence of viral hepatitis* by year–United States, 1979 to 2009. Source: Reproduced from Centers for Disease Control and Prevention. Summary of notifiable diseases—United States, 2009. MMWR. Vol 58, No 53, p. 61, May 13, 2011. *Per 100,00 population.

1979 0

In ci

de nc

1984 1989

†Hepatitis A vaccine was first licensed in 1995. §Hepatitis B vaccine was first licensed in June 1982. ¶An anti-hepatitis C virus (HCV) antibody test first became available in May 1990.

1994 1999 2004 Year

2009

Hepatitis A, viral, acute†

Hepatitis B, viral, acute§

Hepatitis C, viral, acute¶

51589_CH12_Printer.indd 531 15/02/2013 3:19 AM

532 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Leptospirosis (agent [a type of spirochete] leptospira bacteria) is an infrequent condition that can occur among persons who swim in freshwater bodies of water (e.g., rivers and lakes), especially in tropical and sub-tropical regions. Also at risk are rice farmers, fish harvesters, and veterinarians. The bacteria can infect domes- tic and wild animals; sometimes urine from these infected animals finds its way into freshwater where the bacteria can be hazardous.

Q fever (agent, Coxiella burnetii) is a fourth example of a zoonosis. The res- ervoir for C. burnetii is infected livestock (cattle, sheep, and goats). The spec- trum of infection in humans ranges from mild to severe, debilitating illness with symptoms similar to those of influenza or pneumonia. Those at high risk for infection include workers and others who come into contact with infected live- stock: veterinarians, farmers, agricultural employees, and laboratory personnel. For example, Q fever occurred among laboratory personnel who used sheep in their research at a medical center in San Francisco, California. A more ordi- nary source of transmission is infected raw milk. See Figure 12–16 for data on the distribution of Q fever cases in the United States. A total of 131 cases were reported in 2010, with the greatest number occurring in California and Texas.30

Figure 12–16 Q fever, acute and chronic, number of reported cases*– United States and U.S. territories, 2010. Source: Reproduced from Centers for Disease Control and Prevention, Summary of notifiable diseases—United States, 2010. MMWR. Vol 59, No 53, p. 77, 2012. *Number of Q fever acute cases/number of Q fever chronic cases.

NYC AS

CNMI

0 ≥1

1/2

4/0

3/0

26/0

4/0

4/0 4/0

10/2

4/03/1

3/1 0/3

2/1

4/2

2/0

4/2

5/1

N/N

0/1

1/1

1/0

0/1

N/N 0/1

6/2

1/0

1/2

N/N

O/N

1/0

51589_CH12_Printer.indd 532 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 533

Rabies (agent, rabies virus) is an acute and highly fatal disease of the central nervous system. The rabies virus is transmitted most often through saliva from the bites of infected animals. Dog bites are the principal source of transmission of rabies to humans.7 In the United States, human cases of rabies are rare because of vaccination programs for domestic animals, measures to control animals, and public health laboratories for conducting rabies tests. The vast majority of rabies cases occur in wild animals, raccoons and bats being the most commonly affected species. Read the news headline box about a male U.S. Army soldier who developed rabies while serving abroad.

Fungal Diseases Fungal diseases (mycoses) encompass three major types: opportunistic infec- tions among persons who have weakened immune systems, hospital-associated infections, and community-acquired infections. An example of the third type of infection is coccidioidomycosis (San Joaquin Valley fever), which often

On August 19, 2011, a male U.S. Army soldier with progressive right arm and shoulder pain, nausea, vomiting, ataxia, anxiety, and dyspha- gia was admitted to an emergency department (ED) in New York for suspected rabies. Rabies virus antigens were detected in a nuchal skin biopsy, rabies virus antibodies in serum and cerebrospinal fluid (CSF), and rabies viral RNA in saliva and CSF specimens by state and CDC rabies laboratories. An Afghanistan canine rabies virus variant was iden- tified. The patient underwent an experimental treatment protocol (1) but died on August 31. The patient had described a dog bite while in Afghanistan. However, he had not received effective rabies post-exposure prophylaxis (PEP). In total, 29 close contacts and healthcare personnel (HCP) received PEP after contact with the patient. This case highlights the continued risks for rabies virus exposure during travel or deployment to rabies-enzootic countries, the need for global canine rabies elimina- tion through vaccination, and the importance of following effective PEP protocols and ensuring global PEP availability. Source: Centers for Disease Control and Prevention. Imported human rabies in a U.S. Army’s soldier–New York, 2011. MMWR. 2012;61:302–305. n

News Headline: Human Rabies Case

51589_CH12_Printer.indd 533 15/02/2013 3:19 AM

534 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

manifests as a lung disease and is caused by a fungus, Coccidioides immitis. The agent is endemic to the San Joaquin Valley in California, and to the lower Sonoran life zone, which covers parts of California, Arizona, Texas, and Mexico. Cases of infection have been associated with those occupations or activities that bring susceptible persons in contact with contaminated soil (e.g., construction, archeology, and dirt bike riding). Merely driving through an endemic area may result in infection. In December 1977, a coccidioido- mycosis epidemic occurred after exposure to a severe dust storm at a naval air station in Lemoore (Kings County), California. Williams et al.31 concluded that a temporal relationship existed between the dust storm and the marked increase in the incidence of symptomatic San Joaquin Valley fever. In 1991, an outbreak of coccidioidomycosis (1,208 cases) was reported in California. The majority of cases (80%) occurred in Kern County, where coccidioidomycosis is endemic.32

The numbers of reported cases of coccidioidomycosis in the United States for year 2009 are shown in Figure 12–17. California and Arizona reported the largest numbers of cases; the incidence of reported U.S. cases of coccidioidomy- cosis has shown an increasing trend during this century. The CDC estimates that about 60% of coccidioidomycosis cases occur in Arizona.

Arthropod-Borne Diseases Arthropod-borne diseases are defined as diseases transmitted by insect vec- tors such as sand flies, ticks, and mosquitoes. Some arthropod-borne diseases are known as arboviral diseases. Examples of arthropod-borne diseases are Dengue fever, Lyme disease, malaria, viral encephalitis, West Nile virus, and plague. This section provides information on arboviral diseases and Lyme disease.

Arboviral diseases Arboviral diseases are a diverse group of diseases that involve transmission of arboviruses (arthropod-borne viruses) between vertebrate hosts (e.g., from ani- mal to animal or from animal to human) by blood-feeding arthropod vectors,9 the last of which are responsible for transmission of the encephalitis virus. Dur- ing 2010 (as of December 25, 2010), the following types and numbers of cases of encephalitis were reported: St. Louis encephalitis, 8 reported cases; Califor- nia serogroup, 71 reported cases; Eastern equine encephalitis, 10 reported cases;

51589_CH12_Printer.indd 534 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 535

and Western equine encephalitis, 0 reported cases.19 Enzootic arboviruses afflict animals in a particular locality. Viral isolates or antigens of arboviruses may be found in wild birds, sentinel birds, and captured mosquitoes. Transmission cycles involve interactions between arthropods and vertebrates (birds or small mammals); human beings are incidental or dead-end hosts.33

Lyme disease Lyme disease, a second example of a vector-borne illness, is transmitted by ticks. The reservoir for Lyme disease consists of various species of vertebrate host ani- mals (e.g., mice, squirrels, and shrews). Lyme disease cases are distributed across the United States and showed a steady increase from 1992 to 2006. During this period, the CDC received aggregate reports of 248,074 cases. In 1992 a total of 9,908 cases were reported; in 2006 the number of reported cases was

Figure 12–17 Coccidioidomycosis. Number of reported cases— United States* and U.S. territories, 2009. Source: Reproduced from Centers for Disease Control and Prevention. Summary of notifiable diseases—United States, 2009. MMWR. Vol 58, No 53, p. 52, May 13, 2011.

NYC AS

CNMI

0 ≥1

N N

61 39

2,488

N N

4710,233

N N

NN N

N 1

51589_CH12_Printer.indd 535 15/02/2013 3:19 AM

536 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

19,931—an increase of 101%.34 Figure 12–18 shows a map of the distribution of cases during 2009, with cases heavily concentrated in New England and the upper Midwest.

Emerging Infections Emerging infections denote “[i]nfectious diseases that have recently been identi- fied and taxonomically classified. Many of them are capable of causing dangerous epidemics.” The term reemerging infections refers to “. . . certain ‘old’ diseases, such as tuberculosis and syphilis, that have experienced a resurgence because of a changed host-agent-environment conditions.”35(p 85) Examples of emerging dis- eases are HIV/AIDS, hepatitis C virus infections, Lyme disease, E. coli O157:H7 foodborne illnesses, and hantavirus pulmonary syndrome. Many emerging infec- tions are not caused by sudden mutations in a pathogen; instead they appear when an existing pathogen gains access to new host populations. Changes in climate, human activities such as farming or reforestation, technologic changes such as air travel and organ transplantation, and demographic changes such

Figure 12–18 Lyme disease. Incidence per 100,000 population of reported confirmed cases, by county—United States, 2009. Source: Reproduced from Centers for Disease Control and Prevention. Summary of notifiable diseases—United States, 2009. MMWR. Vol 58, No 53, p. 66, May 13, 2011.

1.01–10.00 10.01–100.00 >–100.01

51589_CH12_Printer.indd 536 15/02/2013 3:19 AM

e p i d e m i o l o g i C a l l y s i g n i f i C a n t i n f e C t i o u s 537

as migration to cities may all contribute to the emergence of seemingly new and deadly infections. Other potential sources of infection are bush meat (meat obtained by hunting wildlife) and companion animals (cats and dogs; pet rats).36 For example, from time to time cats are associated with the transmission of toxoplasmosis. Table 12–7 summarizes examples of emerging infections.

Hantavirus pulmonary syndrome The hantavirus pulmonary syndrome is an example of the emergence of seemingly new infectious disease that challenged epidemiologists who first encountered the virus. A young, physically fit Navajo man appeared in an emergency room in New Mexico with fever and acute respiratory distress and died a short time later. After ruling out several known diseases, virologists linked the case to a previously unknown type of hantavirus.35 This syndrome, identified initially in the Four Corners region of the United States, claimed 20 lives out of 36 patients: a greater than 50% case fatality rate. Subsequently it has been identified in California and other states. A hantavirus outbreak occurred in Yosemite National Park during late summer and fall 2012. Hanta- virus infection had been described previously in other parts of the world, but appeared suddenly in a slightly altered form in several locations in the United States.

Escherichia coli E. coli foodborne illness outbreaks have received dramatic and continuing atten- tion in the media. During 2006, an outbreak of illnesses caused by E. coli was linked to fresh spinach. Two outbreaks were associated with the same fast-food restaurant chain in 1982. A multistate outbreak in the western United States produced 700 cases of illness and four deaths in 1993.37 The pathogen respon- sible for the vast majority of cases of illness that present as severe bloody diar- rhea (hemolytic uremic syndrome) is called E. coli O157:H7. This organism is regarded as an emerging pathogen because of the occurrence of major food- borne illness outbreaks that have been distributed over a wide geographic area. Such large-scale outbreaks were not generally recognized before the early 1980s. Although other foods have been associated with E. coli O157:H7, the vehicle implicated most frequently in foodborne illness outbreaks is ground beef. When hamburgers made with contaminated ground beef are eaten rare or are inad- equately cooked, infections may occur.

51589_CH12_Printer.indd 537 15/02/2013 3:19 AM

538 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

T ab

le 1

2– 7

E xa

m pl

es o

f E m

er gi

ng In

fe ct

io ns

P at

ho ge

n D

is ea

se D

is tr

ib ut

io n

N at

ur al

H os

t Tr

an sm

is si

on N

ot ab

le F

ac to

rs in

E m

er ge

nc e

B un

ya vi

rid ae

H an

ta vi

ru se

H em

or rh

ag ic

fe ve

r A

si a

R od

en ts

In ha

la tio

n of

in fe

ct ed

cr et

a C

on ta

ct w

ith r

od en

t ho

st d

ur in

g ric

e ha

rv es

t, d

is se

m in

at io

n by

s hi

pb or

in fe

ct ed

r at

s R

es pi

ra to

ry d

is tr

es s

sy nd

ro m

e U

.S .,

pr ob

ab ly

E ur

op e

an d

So ut

h A

m er

ic a

R od

en ts

In ha

la tio

n of

in fe

ct ed

cr et

a Ec

ol og

ic c

ha ng

es c

au si

ng in

cr ea

se d

co nt

ac t

w ith

r od

en ts

C ry

pt os

po rid

iu m

C ry

pt os

po rid

io si

s Pr

ob ab

ly w

or ld

w id

e ?U

ng ul

at es

* In

ge st

io n

co nt

am in

at ed

w at

er Fa

ul ty

w at

er p

ur if

ic at

io n

Fi lo

vi rid

ae (

Eb ol

M ar

bu rg

) H

em or

rh ag

ic fe

ve r

A fr

ic a

U nk

no w

n C

on ta

ct w

ith u

nk no

w n

ho st

, n os

oc om

ia l

(i nf

ec te

d ne

ed le

U nk

no w

n (i

n Eu

ro pe

a nd

U .S

., im

po rt

at io

n of

m on

ke ys

)

G ro

up A

s tr

ep to

co cc

i N

ec ro

tiz in

g fa

sc iit

is ;

ot he

r in

va si

ve in

fe ct

io ns

Pr ob

ab ly

w or

ld w

id e

?H um

an s

In ha

la tio

n (c

la ss

ic s

tr ep

w ou

nd in

fe ct

io n;

s ki

n co

nt am

in at

io n

U nk

no w

n, p

os si

bl y

ne w

s tr

ai n

st ra

in n

ot r

ec en

tly p

re va

le nt

H ep

at iti

s C

v iru

s H

ep at

iti s

W or

ld w

id e

?H um

an s

C on

ta m

in at

ed IV

n ee

dl es

, tr

an sf

us io

ns , o

rg an

an sp

la nt

at io

n, p

ro ba

bl y

se xu

Tr an

sf us

io ns

La ss

a vi

ru s

La ss

a fe

ve r

W es

t A

fr ic

a R

od en

ts C

on ta

ct w

ith in

fe ct

se cr

et io

ns (

of n

at ur

al h

os t

or in

fe ct

ed p

er so

ns )

U rb

an iz

at io

V ib

rio ch

ol er

ae C

ho le

ra A

si a;

r ec

en tly

in tr

od uc

ed t

o So

ut h

A m

er ic

a, A

fr ic

?Z oo

pl an

kt on

In ge

st io

n of

in fe

ct ed

m an

e xc

re ta

in fo

or w

at er

Tr an

so ce

an ic

s hi

pp in

g, la

ps ed

p ub

lic

he al

th m

ea su

re s,

t ra

ve l

*H oo

fe d

m am

m al

s su

ch a

s sh

ee p,

g oa

ts , a

nd r

ei nd

ee r.

So ur

ce : R

ep ro

du ce

d w

ith p

er m

is si

on , S

S M

or se

. P at

te rn

s an

d pr

ed ic

ta bi

lit y

in e

m er

gi ng

in fe

ct io

ns . H

os pi

ta l P

ra ct

ice , 1

99 6,

3 1(

4) , p

. 8 5.

99 6

Th e

M cG

ra w

-H ill

om pa

ni es

, I nc

51589_CH12_Printer.indd 538 15/02/2013 3:19 AM

s t u d y Q u e s t i o n s a n d e x e r C i s e s 539

West Nile virus West Nile virus (WNV) is an arboviral disease transmitted from infected birds (the reservoir) to humans by mosquitoes. The majority of cases, approximately four-fifths, are asymptomatic; most of the remaining one-fifth experience mild illnesses (e.g., fever, headache, body aches, and skin rashes). However, among a few patients (1 in 150 cases) the disease progresses to serious symptoms that can include high fever, coma, paralysis, and death. The incubation period for WNV is approximately 3 to 14 days after a mosquito bite. In the United States, the first domestically acquired human case of WNV occurred in 1999. Outbreaks of WNV in humans and equines have occurred in Africa, Asia, and Europe. West Nile virus was the most common arboviral disease in the United States during 2010 with a total of 1,021 reported cases.33

Conclusion

Infectious disease epidemiology has developed a body of methods for investi- gating and controlling infectious diseases in the community. The reservoir for infectious disease may consist of humans, animals, arthropods, or the physical environment. Transmission of disease may be direct (person to person) or indi- rect. Noteworthy terms used to describe infectious disease outbreaks include attack rate, secondary attack rate, and case fatality rate. Agents for infectious disease encompass a broad range of microbial agents, from bacteria to viruses to protozoa. Infectious diseases remain major causes of morbidity and mortality. Examples of significant problems include foodborne illness, vaccine-preventable diseases such as measles, diseases spread from person to person such as TB, and sexually transmitted diseases, notably AIDS.

Study Questions and Exercises

1. Define and give the formulas for attack rate, secondary attack rate, and CFR. 2. A flu outbreak occurred in a military barracks that housed 20 soldiers.

Case A began on October 1 and case B was diagnosed on October 2. After approximately 10 days, 12 additional cases occurred during approx- imately a one-week time span. Military epidemiologists believed that this second group of cases represented the same generation of cases, and was

51589_CH12_Printer.indd 539 15/02/2013 3:20 AM

540 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

in the second incubation period after the occurrence of cases A and B; none of the 20 soldiers was known to be immune. Calculate the second- ary attack rate using the foregoing data.

3. Explain the etiology of TB, measles, and rabies by applying the epide- miologic triangle (Figure 12–1).

4. Give one example of each type of prevention—primary, secondary, and tertiary—for foodborne salmonellosis, malaria, and AIDS. To answer this question you will need to review the chapters where the three types of prevention are discussed and apply the methods to a new situation.

5. What is the epidemiologic importance of an inapparent infection? 6. Name two examples of a disease spread from person to person, and sug-

gest methods for the control of such a disease in the community. 7. When is isolation for an infectious disease not likely to be an effective

means of control of the disease in the community? 8. Discuss host responses to infectious disease agents. Be sure to include

herd immunity as a community health concept. 9. Discuss the following statement: High cooking temperature will sanitize

food even after it has been stored improperly (e.g., at room temperature for 6 hours); one can be certain that there will be no remaining hazard to human health and that all the pathogenic material has been destroyed.

10. A local health department epidemiologist investigated an outbreak of gastrointestinal illness thought to be associated with a college cafete- ria. There were many complaints about the quality of the cafeteria’s offerings, and it appeared that the students’ worst expectations were confirmed when several students visited the college’s infirmary during the middle of the night and the following day complaining of nausea, diarrhea, fever, vomiting, and cramps. The health department’s investi- gation revealed that 24 students had eaten in the cafeteria immediately before the outbreak. The times between eating in the cafeteria and the development of active symptoms ranged from 20 to 36 hours. A list of foods eaten, the number of persons eating the foods, and tabula- tions of illness are presented in Appendix 12. Fill in the attack rates where indicated. On the basis of your calculations, answer the follow- ing questions: a. What food or foods would you suspect caused the problem? b. Based on the description of the clinical symptoms and the list of

infectious disease agents presented in Table 12–5, which agent(s) do you think was (were) responsible?

51589_CH12_Printer.indd 540 15/02/2013 3:20 AM

s t u d y Q u e s t i o n s a n d e x e r C i s e s 541

Figure 12–19 Rabies, number of reported cases among wild and domestic animals,* by year–United States and Puerto Rico, 1975–2005. Source: Reproduced from Centers for Disease Control and Prevention, Summary of notifiable diseases—United States, 2005, MMWR, Vol 54, No. 53, p. 63, 2007. *Data from the National Center for Zootic, Vector-Borne, and Enteric Diseases (proposed).

N um

be r

(in th

ou sa

nd s)

Total Domestic Wild

1975 1980 1985 1990 1995 20052000 Year

Periods of resurgence and decline of rabies incidence are primarily the result of cyclic reemergence. As populations are decimated by epizootics, numbers of reported cases decline until populations again reach levels to support epizootic transmission of disease. Recent declines in the number of reported cases among terrestrial reservoir species (raccoons, skunks, and foxes) have been offset by increases in testing and the subsequent detection of rabid bats. In addition, interventions such as the oral vaccination of wildlife species might contribute to the decreasing trend in recent years.

11. Sharpen your skills in interpreting charts. Figure 12–19 is a chart that shows data on rabies for 1975 through 2005. The x-axis (abscissa) is the horizontal axis. The y-axis (ordinate) is the vertical axis. a. What are the labels of the x and y axes? b. Describe the overall trends reflected in the chart. c. Why does the line for domestic animals diverge from that of wild

animals? d. What type of disease is rabies?

12. Figure 12–20 is a chart that presents data on the incidence of pertussis (whooping cough) for 1975 through 2005. a. What are the labels of the x and y axes? b. Describe the overall trends reflected in the chart.

51589_CH12_Printer.indd 541 15/02/2013 3:20 AM

542 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

Figure 12–20 Pertussis. Incidence* by year–United States, 1980-2010. Source: Reproduced from Centers for Disease Control and Prevention, Summary of notifiable diseases—United States, 2010. MMWR. Vol 59, No 53, p. 75, 2012. *Per 100,000 population

9.0

8.0

7.0

6.0

5.0

4.0I nc

id en

3.0

2.0

1.0

0.0 1980

Pertussis continues to have cyclic peaks every 3 to 5 years. Incidence in 2010 surpassed the previous peaks in 2004 and 2005.

1985 1990 1995 Year

2000 2005 2010

10.0

References 1. Heron MP. Deaths: Leading causes for 2008. National Vital Statistics Reports.

2008;60(6). Hyattsville, MD: National Center for Health Statistics. 2. World Health Organization. Infectious diseases are the biggest killer of the young.

In: Infectious Diseases Report. http://www.who.int/infectious-disease-report/pages/ ch1text.html. Accessed June 26, 2012.

3. Klevens RM, Edwards JR, Richards CL Jr., et al. Estimating health care-associated infections and deaths in U.S. hospitals, 2002. Public Health Rep. 2007;122:160–166.

4. Yoder JS, Beach MJ. Giardiasis surveillance—United States, 2003–2005. MMWR. Surveillance Summaries. 2007;56(SS-7):11–18.

5. Mausner JS, Kramer S. Mausner & Bahn Epidemiology—An Introductory Text. 2nd ed. Philadelphia, PA: Saunders; 1985.

6. Centers for Disease Control and Prevention. WHO Collaborating Center for the Mycoses. http://www.cdc.gov/Ncidod/dbmd/mdb/index.htm. Accessed June 28, 2012.

51589_CH12_Printer.indd 542 15/02/2013 3:20 AM

r e f e r e n C e s 543

7. Heymann DL, ed. Control of Communicable Diseases Manual. 19th ed. Washington, DC: American Public Health Association; 2008.

8. Centers for Disease Control and Prevention. Vaccines and Immunizations: Immunity Types. http://www.cdc.gov/vaccines/vac-gen/immunity-types.htm. Accessed June 28, 2012.

9. Porta M, ed. A Dictionary of Epidemiology. 5th ed. New York, NY: Oxford University Press; 2008.

10. Kenah E, Chao DL, Matrajt L, et al. The global transmission and control of influ- enza. Plos ONE. 2011;6(5). e19515. Doi:10.1371/journal.ppone.0019515.

11. Hennekens CH, Buring JE. Epidemiology in Medicine. Boston, MA: Little, Brown; 1987.

12. Centers for Disease Control and Prevention. Principles of Epidemiology. 2nd ed. Atlanta, GA: CDC; 1992. (Reviewed October 1998.)

13. MacMahon B, Pugh TF. Epidemiology Principles and Methods. Boston, MA: Little, Brown; 1970.

14. Lilienfeld AM, Lilienfeld DE. Foundations of Epidemiology. 2nd ed. New York, NY: Oxford University Press; 1980.

15. Carcione D, Giele CM, Goggin LS, et al. Secondary attack rate of pandemic influ- enza A(H1N1)2009 in Western Australian households. Euro Surveill. 2009;29:1–8.

16. Snow J. Snow on Cholera. Cambridge, MA: Harvard University Press; 1965. 17. Centers for Disease Control and Prevention. Parasites and health. Schistosomiasis.

http://dpd.cdc.gov/dpdx/html/schistosomiasis.htm. Accessed July 1, 2012 18. Centers for Disease Control and Prevention. Update: cholera—Western Hemisphere,

1992. MMWR. 1993;42:89–91. 19. Centers for Disease Control and Prevention. Table: Provisional cases of infrequently

reported notifiable diseases (<1,000 cases reported during the preceding year)— United States, week ending December 25, 2010 (51st Week). MMWR. 2011;59:1690.

20. Centers for Disease Control and Prevention. Update: cholera outbreak—Haiti 2010, MMWR. 2010; 59: 1473–1479.

21. World Health Organization. Weekly epidemiological record. Cholera, Iraq—update. 2007;82(41):357–360.

22. Centers for Disease Control and Prevention. Amebiasis associated with colonic irrigation—Colorado. MMWR. 1981;30:101–102.

23. Centers for Disease Control and Prevention. HIV surveillance—United States, 1981–2008. MMWR. 2011;60: 689–693.

24. Centers for Disease Control and Prevention. HIV/AIDS Surveillance Report. 2011;21. Published February 2011. Available at: http://www.cdc.gov/hiv/topics/surveillance/ resources/reports/. Accessed November 12, 2012.

25. World Health Organization. UNAIDS. July 2012. Core Epidemiology Slides. Geneva, Switzerland: World Health Organization; 2012.

26. Centers for Disease Control and Prevention. Measles surveillance—United States, 1991. MMWR. 1992;41:1–11.

27. Centers for Disease Control and Prevention. Tuberculosis morbidity—United States, 1992. MMWR. 1993;42:696–697, 703–704.

28. Centers for Disease Control and Prevention. Trends in tuberculosis—United States, 2010. MMWR. 2011;60:333–337.

51589_CH12_Printer.indd 543 15/02/2013 3:20 AM

544 C h a p t e r 1 2 e p i d e m i o l o g y o f i n f e C t i o u s d i s e a s e s

29. Centers for Disease Control and Prevention. Summary of notifiable diseases–United States, 2005. Hepatitis, viral. Incidence, by year—United States, 1975–2005. MMWR. 2007;54.

30. Centers for Disease Control and Prevention. Summary of notifiable diseases–United States, 2010. MMWR. 2012;59.

31. Williams PL, Mendez P, Smyth LT. Symptomatic coccidioidomycosis following a severe natural dust storm. Chest. 1979;76:566–569.

32. Centers for Disease Control and Prevention. Coccidioidomycosis—United States, 1991–1992. MMWR. 1993;42:21–24.

33. Centers for Disease Control and Prevention. West Nile virus disease and other arbovi- ral diseases—United States, 2010. MMWR. 2011; 60: 1009–1013.

34. Centers for Disease Control and Prevention. Surveillance for Lyme disease—United States, 1992–2006. MMWR. 2008;57:(SS-10).

35. Morse SS. Patterns and predictability in emerging infections. Hosp Pract. 1996;31:85–91, 96–101, 104.

36. Cutler SJ, Fooks AR, van der Poel WHM. Public health threat of new, reemerg- ing, and neglected zoonoses in the industrialized world. Emerging infectious diseases. 2010;16(1):1–7.

37. Armstrong GL, Hollingsworth J, Morris JG. Emerging foodborne pathogens: Escherichia coli O157:H7 as a model of entry of a new pathogen into the food supply of the developed world. Epidemiol Rev. 1996;18:29–51.

51589_CH12_Printer.indd 544 15/02/2013 3:20 AM

d a t a f r o m a f o o d b o r n e i l l n e s s 545

Data from a Foodborne I l lness Outbreak in a College Cafeteria

Number of Persons Who Ate Number Who Did Not Eat

Food Items Served Ill Not Ill Total Attack Rate

Ill Not Ill Total Attack Rate

Three-bean salad 10 3 13 7 4 11 Beef, rare 17 6 23 0 1 1 Beef, specified well cooked 3 6 9 5 10 15 Potato salad 12 6 18 4 2 6 Macaroni salad 11 5 16 5 3 8 Tuna salad* 13 1 14 3 7 10 Cold cuts and cheese plate 10 6 16 5 3 8

Rolls and butter 13 4 17 4 3 7

*The tuna salad was prepared from fresh ingredients approximately 1 hour before consumption and stored under refrigeration.

appendix

51589_CH12_Printer.indd 545 15/02/2013 3:20 AM

51589_CH12_Printer.indd 546 15/02/2013 3:20 AM

3chapte r

547

13chapte r

547

Epidemiologic Aspects of Work in the

Environment

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● define the term environmental epidemiology ●● give examples of environmental agents that are associated with

human health effects ●● provide examples of study designs used in environmental

epidemiology ●● state methodologic difficulties with research on environmental

health effects ●● list some of the terms used to characterize environmental exposure

and human responses to exposure ●● cite health outcomes studied in relation to environmental agents

CHAPTER OUTLINE

I. Introduction II. Health Effects Associated with Environmental Hazards

III. Study Designs Used in Environmental Epidemiology IV. Toxicologic Concepts Related to Environmental Epidemiology V. Types of Agents

51589_CH13_Printer.indd 547 09/02/13 2:25 PM

548 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

Introduction

During the 21st century adverse human environmental impacts, such as global warming, have risen to the forefront as pressing concerns for global society. Consider two examples of environmental catastrophes: the Deepwater Horizon Oil Spill (April 20 to July 15, 2010) in the Gulf of Mexico and the devastating earthquake (March 11, 2011) accompanied by a nuclear reactor meltdown in Fukushima, Japan. (A later section will cover the reactor meltdown.) The Deepwater Horizon oil spill was history’s largest accidental marine oil spill. On the night of April 20, 2010, an explosion caused by methane gas, followed by a conflagration, devastated the British Petroleum Deepwater Horizon oil platform. The platform was located approximately 50 miles southwest of Venice, Louisiana. The explosion and fire injured 17 crew members; an additional 11 workers were unaccounted for and were presumed to have perished at sea. Up to 60,000 barrels of oil per day spewed from the wellhead located about one mile below the water surface. The unabated flow continued for 86 days until workers capped the wellhead on July 15. Oil slicks from the gushing well covered thousands of square miles of pristine Gulf waters and drifted into ecologically sensitive coastal areas of Louisiana and states bordering the Gulf. Impacted species of wildlife included birds, dolphins, blue crabs, and turtles. Government authorities prohibited fishing at the mouth of the Mississippi River and nearby affected areas. Although the Gulf has shown many positive indicators of recovery, the full effects of the oil spill will not be known for many years.

Human activities cause environmental hazards that portend enor- mous ramifications for society, health, and the economy. The World Health Organization estimates that environmental factors are linked to as much as 24% of the global burden of disease and 23% of all deaths.1 From the worldwide per- spective, cancer is the leading killer and accounted for 7.6 million deaths in 2008. Environmental exposures (including those in the occupational environment) are likely to be responsible for 19% of all cancers globally.2 The causes of environ- mentally related morbidity and mortality include exposure to toxic chemicals,

VI. Environmental Hazards Found in the Work Setting VII. Noteworthy Community Environmental Health Hazards

VIII. Conclusion

IX. Study Questions and Exercises

51589_CH13_Printer.indd 548 09/02/13 2:25 PM

i n t r o d u C t i o n 549

indoor and outdoor air pollution, dangerous conditions found in the workplace, and land use policies that encourage the survival of microbes and disease vectors.

What is the potential role for epidemiologists in environmental health? Through epidemiologic research, it may be possible to control or prevent environmentally associated adverse health outcomes by identifying relevant expo- sures, demonstrating how the exposures are associated with the outcomes, and suggesting methods for control of exposures and remediation of hazards. Among the more notable of human exposures to environmental hazards are the following:

●● chemical agents (from chemical spills, pesticides, and hazardous wastes) ●● electromagnetic radiation from high-tension wires ●● ionizing radiation from natural and synthetic sources ●● heavy metals ●● air pollution (See Figure 13–1, which shows fumes issuing from an electric

generating plant.) ●● temperature increases from global warming and climate change

Environmental epidemiology is one of the disciplines that have the potential to provide insight into health effects associated with the environment. The term environmental epidemiology refers to the study of diseases and conditions (occurring in the population) that are linked to environmental factors. Generally speaking,

Figure 13–1 Fumes issuing from an electric generating plant.

51589_CH13_Printer.indd 549 09/02/13 2:25 PM

550 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

exposures are involuntary—outside of the control of the exposed individual. Environmental epidemiology applies standard epidemiologic methods to the study of health outcomes hypothesized to be associated with the environment. Other closely allied disciplines include toxicology and molecular epidemiology.3

Health Effects Associated with Environmental Hazards

Health effects (both morbidity and mortality) attributed to environmental exposures encompass a wide range of conditions, including cancer, infertility, reproductive impacts (e.g., congenital malformations and low birth weight), and infectious diseases. In the work environment, ionizing radiation, infectious agents, toxic substances, and drugs may pose unique health risks for pregnant workers and the unborn fetus.4 Other adverse health outcomes associated with the occupational environment include lung diseases (e.g., brown lung disease, coal workers’ pneumoconiosis, and asbestosis), dermatologic problems, neurotoxicity, coronary heart disease, injuries and trauma, and various psychological conditions (e.g., work absenteeism and stress at work). Researchers have been concerned about the possible link between occupational exposure to carcinogenic agents and certain forms of cancer (e.g., bladder cancer in dye workers and leukemia among workers exposed to benzene). Finally, deaths from malaria are increasing as a result of deforestation and inadequate water management practices.1

Study Designs Used in Environmental Epidemiology

Environmental epidemiology employs a wide range of study designs, including both descriptive and analytic approaches. Definitions of descriptive and etiologic studies applied to the epidemiology of occupational diseases apply equally well to the broader category of environmental epidemiology. With respect to occu- pational epidemiology, Wegman states that “[d]escriptive studies provide infor- mation for setting priorities, identifying hazards, and formulating hypotheses for new occupational risk.”5(p 944) A historical example is William Farr’s work showing that Cornwall metal miners (1848–1853) had higher mortality from all causes than the general population.5 “Etiologic studies are planned examinations of causality and the natural history of disease. These studies have required increas- ingly sophisticated analytic methods as the importance of low-level exposures is explored and greater refinement in exposure-effect relationships is sought.”5(p 945) The authors will next give examples of different types of study designs that are

51589_CH13_Printer.indd 550 09/02/13 2:25 PM

s t u d y d e s i g n s u s e d i n e n v i r o n m e n t a l e p i d e m i o l o g y 551

employed in environmental and occupational epidemiology an effort to present a broad overview of the field.

Retrospective Cohort Studies In evaluating the health effects of occupational exposures to toxic agents, research- ers are able to select one or more possible end points. For example, in some stud- ies self-reported symptom rates are used as a measure of the effects of low-level chemical exposure. Occupational health investigators can design and administer self-report questionnaires inexpensively. Self-reports to questionnaires, however, may not always be reliable, and although they often correlate with clinical diagno- ses they also may differ markedly.6 Physiologic or clinical examinations are other means to evaluate adverse health effects. For example, in a study of respiratory diseases, pulmonary function tests, such as forced expiratory volume, may be an appropriate indicator. While clinical examinations may provide “harder” evidence of health effects than self-reports, they may be expensive or impractical to collect in the case of workers who have left employment. In other studies, mortality is the outcome of interest; research on mortality frequently uses a retrospective cohort study design.7 Mortality experience in an employment cohort can be compared with the expected mortality in the general population (national, regional, state, or county) by using the standardized mortality ratio. One also can contrast the mor- tality experience of exposed workers with the mortality rate of nonexposed work- ers in the same industry. For example, production workers might be compared with drivers or office workers. Another option is to identify a second industry or occupation that is comparable in terms of skill level, educational requirements, or geographic location, but in which the exposure of interest is not present.

Use of mortality as a study end point has several advantages, including the fact that it may be relevant to agents that have a subtle effect over a long time period. Although any fatal chronic disease may be investigated, mortality from cancer is often studied as an outcome variable in occupational exposures. According to Monson, “cancer specifically tends to be a fatal illness; its presence is usually indicated on the death certificate. Also, cancer is a fairly specific disease and is less subject to random misclassification than, say, one of the cardiovascular diseases.”7(p 106)

Methods for Selection of a Research Population and Collection of Exposure Data Investigators select research populations from a variety of sources (e.g., residents of geographic areas affected by air pollution, people who live near nuclear power plants, populations exposed to contaminated water supplies, and employees of

51589_CH13_Printer.indd 551 09/02/13 2:25 PM

552 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

industries that use toxic chemicals). For occupational health studies, researchers might choose study groups based on information from industrial personnel records or documentation regarding chemicals used, exposure levels from internal monitoring devices, and methods of storage and elimination. If a company has retained the records of former and retired workers, a complete data set spanning long time periods may be available. Ideally, every previous and current worker exposed to the factor should be included in the database. Selection bias may occur if some workers are excluded because their records have been purged from the company’s database.7 Examples of research data collected from employment records include:

●● personal identifiers to permit record linkage to Social Security Administration files and retrieval of death certificates

●● demographic characteristics, length of employment, and work history with the company

●● information about potential confounding variables, such as the employee’s medical history, smoking habits, lifestyle, and family history of disease

One of the most crucial issues for environmental epidemiology is exposure assessment. Many environmental health studies involve the potential effects of low-level exposures, which present several unique sets of challenges. The ideal approach is direct measurement in the work environment. However, if a particular exposure is suspected of being harmful, obtaining compliance from company leaders to permit such a study is increasingly restricted because of potential legal liabilities. Thus, such work is increasingly conducted outside the United States. Another challenge is that occupational exposures can change over time with new manufacturing procedures, making it extremely difficult to quan- tify exposures. Additional methodologic challenges arise when researchers employ indirect exposure measurements. In research on low-level exposures, high quality data are required to differentiate the effects of an exposure of interest from other environmental exposures. Examples of indirect exposure measurements used in environmental epidemiology include taking samples of toxic fumes from a manu- facturing plant, collecting air pollution readings in the community, and measuring distances of housing tracts from power lines that emit electromagnetic radiation. Such measures are prone to error, because one cannot be certain about any par- ticular individual’s actual exposure to the environmental hazard being studied. An illustration would be community studies of air pollution in which monitoring stations are placed throughout the community in order to assess levels of particu- late pollution. With this type of indirect exposure assessment, the researcher does

51589_CH13_Printer.indd 552 09/02/13 2:25 PM

s t u d y d e s i g n s u s e d i n e n v i r o n m e n t a l e p i d e m i o l o g y 553

not know how much time a given person spends in the community (e.g., some residents are absent during work hours); also, study participants might be exposed to other forms of air pollution such those at work and from cigarette smoke. Consequently, one would have difficulty differentiating the effects of exposure to particulate pollution from the other type of exposures.

Confounding and Bias in Environmental and Occupational Epidemiology Previously, confounding was defined as the masking of the association between an exposure and an outcome of interest because of the influence of a third vari- able that was not considered in the study design or analysis. Bias refers to sys- tematic departures of measures from their true measures as in the example of a miscalibrated blood pressure manometer that under- or overestimates a person’s blood pressure.

Socioeconomic status and health: an example of confounding Among the many possibilities for confounding in environmental epidemiology studies is the relationship between low socioeconomic status and adverse health outcomes. Low socioeconomic status (SES) may be associated with hazardous environmental conditions such as exposures (e.g., air pollution or secondhand cigarette smoke), which in turn may impact health. In this situation, exposures to environmental hazards confound the relationship between SES and adverse health outcomes.

The healthy worker effect: an example of bias The healthy worker effect refers to the “observation that employed populations tend to have a lower mortality experience than the general population.”7(p. 114) The healthy worker effect is one of the forms of bias that may reduce the validity of exposure data in occupational health research, thereby affecting occupational health studies. The term by bias refers to “systematic deviation of results or infer- ences from truth. Processes leading to such deviation.”8

The healthy worker effect may impact on occupational mortality studies in several ways. People whose life expectancy is shortened by disease are less likely to be employed than healthy persons. One consequence of this phenomenon would be a reduced (or attenuated) measure of effect for an exposure that increases morbidity or mortality. That is, because the general population includes both employed and unemployed individuals, the mortality rate of that population

51589_CH13_Printer.indd 553 09/02/13 2:25 PM

554 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

may be somewhat elevated compared with a population in which everyone is healthy enough to work. As a result, any excess mortality associated with a given occupational exposure is more difficult to detect when the healthy worker effect is operative. The healthy worker effect is likely to be stronger for nonmalignant causes of mortality, which usually produce worker attrition during an earlier career phase, than for malignant causes of mortality, which typically have longer latency periods and occur later in life. In addition, healthier workers may have greater total exposure to occupational hazards than those who leave the work force at an earlier age because of illness.

Ecologic Study Designs Studies of the health effects of air pollution have used ecologic analyses to cor- relate air pollution with health effects. Instead of correlating individual exposure to air pollution with mortality, the researcher measures the association between average exposure to air pollution within a census tract and the average mortality in that census tract. Other types of geographic subdivisions besides census tracts may be used as well. This type of study attempts to demonstrate that mortality is higher in more polluted census tracts than in less polluted census tracts. A major problem of the ecologic technique for the study of air pollution, however, stems from uncontrolled factors. Examples include individual levels of smoking and smoking habits, occupational exposure to respiratory hazards and air pollution, differences in social class and other demographic factors, genetic background, and length of residence in the area.9 Nonetheless, ecologic studies may open the next generation of investigations. Future studies will probably attempt to mea- sure the relevant potential confounders in more rigorous analytic study designs.

Cross-Sectional Studies If an occupational exposure has a demonstrable biological effect on the body, simple cross-sectional studies can provide insight and direct evidence to that end. Consider the study conducted by the National Cancer Institute in Tianjin, China. Benzene is known to cause toxicity to the hematopoietic system (hematotoxicity). Leukemia and exposure to benzene occurs among workers in oil, shipping, auto- mobile repair, shoe manufacture, and other industries and in the general public from cigarette smoke, gasoline, and automobile emissions. Lan and colleagues compared 250 benzene-exposed shoe workers with 140 unexposed age- and sex- matched controls who worked in three clothes-manufacturing factories in the same region near Tianjin, China.10 Subjects were on average 30 years of age and

51589_CH13_Printer.indd 554 09/02/13 2:25 PM

t o x i C o l o g i C C o n C e p t s r e l a t e d t o e n v i r o n m e n t a l 555

primarily female (two-thirds); shoe workers had been employed an average of 6 years. For each subject, individual benzene and toluene exposure was moni- tored repeatedly up to 16 months before a blood draw to obtain a blood count. Post-shift urine samples were collected from each subject. Subjects were catego- rized into four groups by mean benzene levels measured during the month before the blood draw. The investigators found that total white blood counts (WBCs), granulocytes, lymphocytes, B cells, and platelets declined significantly with increasing benzene exposure and were lower in workers exposed to benzene at air levels of 1 ppm or less compared with controls. This elegant small study nonethe- less has impacted policy decisions on acceptable exposure limits to Benzene.

Case-Control Studies Case-control studies collect information on past exposures to environmental hazards and toxins among persons who have or do not have a health outcome of interest. In comparison with ecologic study designs, case-control studies may provide more complete exposure data, especially when the exposure information is collected from the friends and relatives of cases who died of a particular cause. Nevertheless, some unmeasured exposure variables as well as confounding vari- ables may remain in case-control studies. For example, in studies of health and air pollution, precise quantitation of both air pollution exposure and unobserved confounding factors, including smoking habits and occupational exposure to air pollution, may be difficult to achieve.9

Toxicologic Concepts Related to Environmental Epidemiology

The terms dose–response curve, threshold, latency, and synergism, which are from toxicology, characterize exposure to hazardous agents. For more informa- tion about toxicology see Casarett and Doull’s Essentials of Toxicology.11

Dose–Response Curve A dose–response relationship refers to a type of correlative association between an exposure (e.g., a toxic chemical) and effect (e.g., a biologic outcome such as cell death). A dose–response curve maps this association and is used to assess the effect of exposure to a chemical or toxic substance upon an organism (e.g., an experimental animal). A typical dose–response curve is shown in Figure 13–2. The dose is indicated along the x-axis, and the response is shown along the y-axis.

51589_CH13_Printer.indd 555 09/02/13 2:25 PM

556 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

The response could be measured as the percentage of exposed animals showing a particular effect, or it could reflect the effect in an individual subject. The dose–response curve, which has a sigmoid shape, is also a cumulative percentage response curve. At the beginning of the curve, there is a flat portion suggesting that at low levels an increase in dosage produces no effect. This is also known as the subthreshold phase. After the threshold is reached, the curve rises steeply and then progresses to a linear phase, where an increase in response is proportional to the increase in dose. When the maximal response is reached, the curve flattens out. A dose–response relationship is one of the indicators used to assess a causal effect of a suspected exposure upon a health outcome.

Threshold The threshold refers to the lowest dose at which a particular response may occur. It is unclear whether exposure (especially long-term exposure) to toxic chemicals at low (subthreshold) levels is sufficient to produce any health response. The effects of low-level exposures within the population are difficult to assess. Nevertheless, some occupational health specialists voice their increasing concerns over the long-term effects of low-level exposures to toxic substances in the workplace.7 Although some researchers hypothesize that the effects of low-level exposures may be pathologic, other investigators express a point of view known as hormesis—the belief that low-level exposures may induce protective effects for

Figure 13–2 Illustration of the dose–response curve.

X Variable (Dose)

R es

po ns

51589_CH13_Printer.indd 556 09/02/13 2:25 PM

t y p e s o f a g e n t s 557

high-level exposures. Hormesis is controversial and some experts believe that it may disregard established science.12

Latency Latency refers to the time period between initial exposure and a measurable response. The latency period can range from a few seconds (in the case of acutely toxic agents) to several decades. For example, mesothelioma (a rare form of can- cer) has a latency period as long as 40 years between first exposure to asbestos and subsequent development of the condition. The long latency for many of the health events studied in environmental research makes detection of hazards a methodologically difficult problem. Because multiple exposures (often at low levels) may have occurred during the latency period, the epidemiologist may be unable to sort out which exposures are salient for a disease of interest from those that are not important.

Synergism Synergism refers to a situation in which the combined effect of several exposures is greater than the sum of the individual effects. The synergistic relationship between asbestos and smoking in causing lung cancer is an example. A classic study of lung cancer risk among asbestos insulation workers was reported by Selikoff and colleagues.13 A total of 370 workers were studied between 1963 and 1967. The occurrence of lung cancer in this occupational group was seven to eight times greater than that expected for the general white population of the United States. It was apparent that exposure to asbestos was not the entire expla- nation, however. No lung cancer deaths were observed among the 87 workers who did not smoke, but 24 of 283 workers who smoked died of lung cancer, a 92-fold greater risk than that for workers who did not smoke and were not exposed to asbestos as part of their occupation.

Types of Agents

A partial list of potential disease agents found in the environment (work, home, and external) is shown in Table 13–1. The nature and significance of these agents are discussed below. Although there are thousands of possible agents of environ- mental disease, the broad categories include toxic chemicals, metals (especially heavy metals [e.g., lead]), electric and magnetic radiation, ionizing radiation, allergens and molds, dusts (e.g., silica and coal dust), asbestos, and mechanical and physical energy.

51589_CH13_Printer.indd 557 09/02/13 2:25 PM

558 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

Chemical Agents Although chemicals are essential to the functioning of modern society, their use raises concerns about their possible short-term and long-term health effects. Human beings are exposed to chemical agents from a variety of sources in the home and at work. And, as the use of chemicals continues to grow exponentially, possibilities for exposure to them increase. A wide range of potential adverse impacts on human health have been studied in relationship to chemical agents (e.g., acute toxicity, direct skin irritation, pulmonary diseases, and long-term effects such as cancer).

Sources of chemical exposure include foods, household cleaning agents, automotive chemicals, paints, and pesticides. Present in our food are chemi- cal additives that increase shelf life, preserve quality, enhance taste and appear- ance, and increase nutritional value. Some kinds of plastics as well as packaging used for foods and drinks contain Bisphenol A (BPA), a chemical used to make polycarbonate plastics and epoxy resins. Increasingly, the public has expressed concerns about potential adverse effects that might be associated with BPA exposure,14 which is widespread.

Table 13–1 Selected List of Environmental Disease Agents

Type of Agent Examples Health Effects Studied

Chemical Pesticides, organochlorides Cancers

Vinyl chloride Angiosarcoma

Benzene Leukemia

Heavy metals and metallic compounds

Mercury Minamata disease

Lead Neurologic impairment

Cadmium, manganese Cancers

Arsenic Cancers

Electric and magnetic radiation Radiation from high-tension power lines

Leukemia

Ionizing radiation g-rays Cancers

X-rays Cancers

Radon Lung cancer

Allergens and molds Animal fur and dander; pollen Allergic responses

Asbestos Brake linings Lung cancer, mesothelioma

Construction materials

Dusts Coal dust Pneumoconiosis

Silica Silicosis Physical/mechanical energy Industrial machinery, high

ambient temperature Noise-induced hearing loss, mortality

51589_CH13_Printer.indd 558 09/02/13 2:25 PM

t y p e s o f a g e n t s 559

Among the household products that make use of chemicals are disinfectant soaps, detergents, air fresheners, and many others. Some brands of auto body and engine cleaners and related products contain volatile and corrosive solvents. Many types of paints and solvents produce vapors that are potentially hazardous when inhaled or when these products come into direct contact with the skin. Other sources of chem- ical exposure at home are construction materials that emit formaldehyde-containing vapors. Finally, several types of pesticides—used routinely to control insects, rodents, and weeds—are highly toxic to human beings and animals.

Pesticides Pesticides are substances (e.g., insecticides, herbicides, and rodenticides) used to control pests. The four major classes of insecticides are organophosphates, organocarbamates, pyrethroids, and organochlorides (also known as organo- chlorines), which include dichlorodiphenyltricloroethane (DDT), lindane, and chlordane. Polychlorinated biphenyls and dioxins are also from the organochlo- ride family. Direct exposure to pesticides may follow mishandling, improper disposal in landfills, releases from chemical plants, accidents, or spills; indirect pesticide exposure may result from groundwater contamination through agricul- tural use of pesticides.

When pesticides are applied to agricultural fields, their vapors may drift, exposing both workers and nearby community residents. Researchers analyzed almost 3,000 cases of exposures to agricultural pesticide drift that occurred between 1998 through 2006. About half of the persons exposed lived nearby in the community; the remainder were exposed on the job.15 Figure 13–3 presents the distribution of acute health effects associated with cases of pesticide drift.

An example of a pesticide that has aroused the concern of public health offi- cials is the organochloride insecticide DDT, which was used widely after intro- duction in the 1940s for the control of mosquitoes and other insect vectors. Because of concerns about DDT’s toxicity to wildlife and its persistence in the environment, its use was banned in the United States in 1972. DDT has a long half-life (2–15 years) in the environment and tends to increase in concentration as a result of biomagnification, a process whereby DDT levels strengthen as pes- ticide residues move up the food chain from lower to higher organisms. DDT also binds with fat molecules in animals and, hence, has the highest concentra- tions in adipose tissue in comparison with other bodily tissues.16 Consequently, even though application of DDT has been banned, human exposure still may occur via the consumption of contaminated fatty foods derived from animal

51589_CH13_Printer.indd 559 09/02/13 2:25 PM

560 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

sources as well as by eating contaminated fish. Although some of the effects (e.g., neurotoxicity and changes in the liver) of acute and long-term occupational exposure to organochlorine pesticides have been observed, the health effects of low-level exposure of the general population to DDT and similar agents have not been established definitively. Some studies have suggested that DDT can mimic some of the effects of the sex hormone estrogen, posing a hazard to aquatic life (e.g., feminization) when present as a contaminant in water. Several laboratory studies have reported a link between estrogen-like forms of DDT and enhance- ment of growth of certain types of mammary tumors.17

Asbestos Formerly, asbestos was used widely in shipbuilding, automotive brake linings, insulation, and construction, resulting in the widespread contamination of schools, homes, and public buildings. Before the United States curtailed use of

Figure 13–3 Health effects associated with pesticide drift (n = 2,945). Percentages do not add up to 100% because of multiple mentions. Source: Data from Lee, S-J, Mehler L, Beckman J, et al. Acute pesticide illnesses associated with off-target pesticide drift from agricultural applications: 11 states, 1998–2006. Environ Health Perspect. 119 (8), 2011:1162–1169.

Ey e

pr ob

lem s

Ne ur

olo gic

Re sp

ira to

Ga str

oin te

sti na

l Sk

in Ot

he r

Ca rd

iov as

cu lar

P er

ce nt

ag e

60 58.2

52.8

47.8

41.5

14.7

Health Effect

11.4

5.1

51589_CH13_Printer.indd 560 09/02/13 2:25 PM

t y p e s o f a g e n t s 561

asbestos in the 1980s, billions of tons of the substance were added to consumer products and applied to construction sites. Despite a 1991 court action that permitted many uses of the mineral, it is found less commonly in consumer products than it once was.

Classic epidemiologic studies determined that asbestos exposure was associated with asbestosis, malignant mesothelioma, and lung cancer. The research of Selikoff stands out as a pioneering effort to explore the epidemiol- ogy of asbestos-related health effects. Selikoff et al.18 reported unexpectedly high death rates due to cancer of the lung or pleura; mesothelioma; and cancer of the stomach, colon, or rectum among building trade insulation workers who had relatively light exposure to asbestos.

Metallic Compounds Metallic compounds that pose an environmental hazard include aluminum, arsenic, antimony, beryllium, cadmium, chromium, lead, mercury, nickel, and tin. This section highlights three of these: arsenic, mercury, and lead.

Arsenic Arsenic in its pure form is a crystalline metalloid (an element with properties that are intermediate between those of a metal and a nonmetal) capable of combining with other substances. For example, arsenic in the environment exists as inor- ganic compounds composed of arsenic plus oxygen, chlorine, or sulfur.19 Arsenic unites with carbon and hydrogen found in plants and animals creating organic forms of the element. A poisonous material that is ubiquitous in nature—in soils and water—arsenic varies in toxicity depending upon its chemical form; it is also a by-product of refining gold and other metals. Before World War II, clinicians employed Salvarsan (an arsenic-based preparation) for treatment of syphilis until antibiotics became available.20 Among the current medical applications of arse- nic is in chemotherapy for leukemia and in some traditional medicines.20 At one time, lumber mills treated wood with the arsenic-based preservatives; this prac- tice became illegal in December 31, 2003, for treatment of dimensional lumber destined for residential settings. Figure 13–4 shows wood in playground equip- ment treated with copper chromate arsenic. A current application for arsenic is in semiconductor devices manufactured by hi-tech industries.20

Several massive arsenic poisoning incidents, due to the contamination of food- stuffs with arsenic, have been documented.21 Arsenic-contaminated beer killed 6,000 people in England in 1900; several thousand people were poisoned in Japan

51589_CH13_Printer.indd 561 09/02/13 2:25 PM

562 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

in the mid-20th century due to arsenic-contaminated dry milk and soy sauce. Particularly at risk of arsenic exposure are some groups of mining and smelter workers as well as agricultural workers who come into contact with pesticides.

Accumulated evidence suggests that arsenic is a carcinogen. It is a cause of skin and lung cancer, when ingested or inhaled, and has been linked to internal cancers, such as bladder, kidney, and liver cancers.22 Geographic regions that have high levels of naturally occurring arsenic in the drinking water (e.g., Taiwan and Argentina) also have elevated levels of bladder cancer mortality.23 A dose– response relationship between exposure to inorganic arsenic in drinking water and bladder cancer has been reported. However, because many of the studies that report this association have used ecologic study designs, additional analytical epidemiologic research would be helpful in clarifying the relationship.

Mercury A naturally occurring chemical that is highly toxic, mercury has been used medically to treat syphilis, as an agricultural fungicide, and in dental amalgams. In 1956, an environmental catastrophe occurred in Minamata Bay, Japan, where approximately 3,000 cases of neurologic disease resulted among people who ate fish contaminated with methyl mercury.24 The neurologic condition, which became known as Minamata disease, was characterized by numbness of the extremities, deafness, poor vision, and drowsiness; the condition was unrespon- sive to medical intervention and frequently culminated in death. The cause was attributed to discharges of mercury compounds into the bay by a plastics factory. Mercury contamination of local waterways is a legacy of mining operations in some areas (e.g., California).

Figure 13–4 Portions of wood darkened by chromated copper arsonate. Source: Reproduced from U.S. Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry, ATSDR Case Studies in Environmental Medicine. Available at: http://www.atsdr.cdc.gov/csem/ arsenic/docs/arsenic.pdf. Accessed: June 17, 2012.

51589_CH13_Printer.indd 562 09/02/13 2:25 PM

t y p e s o f a g e n t s 563

Nowadays, consumers are advised to avoid certain species of ocean fish (e.g., swordfish and canned albacore) that may contain high levels of mercury. Aquatic organisms that live in the ocean and other bodies of water bioaccumu- late contaminants such as mercury, pesticides, and PCBs. Subsequently, these contaminants pass up the food chain resulting in concentrations in top predator fish that are much higher than in the water in which they live. As a public health measure, the EPA issues advisories (recommendations) not to eat fish caught in certain U.S. freshwater lakes and rivers. More than 80% of advisories in 2010 were for mercury contamination. The EPA issued mercury advisories for 16.4 million acres of lakes and 1.1 million miles of rivers during 2010. Figure 13–5 indicates the number of lake acres under advisory for mercury plus five other types of contaminants during 2010.25

Lead Exposure to lead, which was once widely used in paint, motor vehicle fuels, and industrial processes, is associated with grave central nervous system effects (e.g., intellectual impairment and deficits) among children and hypertension

Figure 13–5 Total lake acres under advisory for Mercury during 2010. Source: U.S. Environmental Protection Agency. 2010 biennial national listing of fish advisories. EPA-820-F-11–014. November 2011, p. 4.

M er

cu ry

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

1996

1995

1994

1993

Millions of Lake Acres Under Advisory

0 2 4 6 8 10 12 14 16 18

51589_CH13_Printer.indd 563 09/02/13 2:25 PM

564 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

and kidney disease among adults.26 Very high blood lead levels (BLLs), defined as greater than 70 μg/dL of blood, are associated with acute effects (e.g., seizure, coma, and death).27 However, even at low levels exposure to lead is thought to be harmful, because the lowest threshold for adverse effects is unknown. Fetuses and young children are particularly at risk of the harmful effects of lead exposure.

Exposure to lead is an important public health concern. According to Healthy People 2020, the national objective of the United States is to “[r]educe blood lead levels in children”28(p 92) (Objective EH-8). Objective EH-8 includes sub- objective EH-8.1, “[e]liminate elevated blood levels in children”28(p 92) and sub- objective EH-8.2, “[r]educe the mean blood levels in children.”28(p 92) During 2005 through 2008 the average blood lead level of children aged 1 to 5 years was 1.5 μg/dL. The target for the mean reduction in blood lead level is 10%.28

Findings of the National Health and Nutrition Examination Surveys (NHANES) indicate progress in reducing children’s blood lead levels. The per- centage of children who had BLLs that exceeded the standard of 10 μg/dL of blood declined steeply among four NHANES data collection periods (1976– 1980, 1988–1991, 1991–1994, and 1999–2002). However, approximately 1.6% of children aged 1 to 5 years remained at risk for elevated BLLs as of 1999–2002.26 BLLs declined greatly in all groups in recent years but remained highest among non-Hispanic black children in comparison with non-Hispanic white and Mexican-American children. (Refer to Figure 13–6.) In 2006, the prevalence of elevated BLLs in the population younger than 6 years of age was 1.21%.29 Risk factors for lead exposure include living in cities, occupying struc- tures built before the 1950s, having low family income, and being a member of a minority racial or ethnic group.

A frequent source of children’s exposure to lead is paint. The use of lead- containing paint in residential construction was banned in 1978. Nevertheless, children may come into contact with lead in paint chips and dust from buildings constructed before the ban was introduced. Lead exposure from buildings is of particular concern to children and their families who reside in low-income areas. Some imported toys, jewelry, and lunchboxes for children have been demon- strated to contain possibly dangerous lead levels.

Electric and Magnetic Fields Electric and magnetic fields are forms of energy known as nonionizing radiation. Examples of sources of electric and magnetic fields are high-voltage electric lines, microwave ovens, stoves, clocks, electric blankets, toasters, and cellular telephones. Epidemiologic research has found an association between residential proximity

51589_CH13_Printer.indd 564 09/02/13 2:25 PM

t y p e s o f a g e n t s 565

to high-tension wires and childhood cancers.30–33 London et al.34 conducted a case-control study of leukemia risk among children in Los Angeles as a function of measured magnetic or electric fields and wiring configuration (e.g., overhead electric transmission and distribution facilities). They reported an association for wiring configuration and childhood leukemia risk but not for measured magnetic and electric fields. A second case-control study investigated everyone in Sweden under age 16 years who had lived on property within 300 meters of high-tension power lines during a 25-year period.35 A total of 142 cancer cases were identi- fied (including 39 leukemia cases and 33 central nervous system tumors) from the Swedish Cancer Registry. Models of historical exposure to magnetic fields

Figure 13–6 Percentage of children aged 1–5 years with blood lead levels ≥ ug/dL, by race/ethnicity and survey period–National Health and Nutrition Examination Surveys, United States, 1988–1991, 1991–1994, and 1999–2002. Source: Reproduced from the Centers for Disease Control and Prevention. MMWR. 2005, Vol. 54, No 20, p. 515.

0 1988–1991

*95% confidence interval.

1991–1994

Survey period

1999–2002

P er

ce nt

ag e

Black, non-Hispanic

Mexican-American

White, non-Hispanic

51589_CH13_Printer.indd 565 09/02/13 2:25 PM

566 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

were used. The estimated relative risk for childhood leukemia increased at higher magnetic field exposure levels. Other research found no significant association between childhood leukemia and power lines.36 Regarding other carcinogenic effects of exposure to electromagnetic radiation, research conducted in the United States and Norway reported an increased risk of male breast cancer (which is rare) among male electrical workers potentially exposed to electromagnetic fields.37

This field of epidemiologic inquiry, which has numerous methodologic chal- lenges, provides a rich opportunity to address an important public policy issue. Among the methodologic issues is the precise measurement of residential and other exposure to electromagnetic radiation.38 The issues of the association of electromagnetic fields with outcomes such as breast cancer, heart disease, suicide, and depression are controversial. With the expanding use of cellular telephones, some experts have explored the relationship between exposure to radiofrequency fields and cancer, especially brain tumors. Despite concerns about the possible carcinogenic effects of cellular telephone use, one of the major health effects of their use seems to be automobile accidents caused by inattentive drivers who do not observe the road while talking on the telephone.

Ionizing Radiation Ionizing radiation, a more intense form of energy than nonionizing radiation, consists of either particle energy (e.g., highly energetic protons, neutrons, and α and β particles) or electromagnetic energy (e.g., γ-rays and X-rays). U.S. radia- tion sources comprise two main categories: natural radiation, which is respon- sible for the majority of the annual radiation exposure of the human population, and synthetic radiation. Natural radiation consists of radon, cosmic rays from outer space, and radiation from geologic sources. Synthetic sources are medical X-rays and agents used in nuclear medicine, consumer products, nuclear genera- tors, and nuclear weapons explosions.

Examples of topics in the field of ionizing radiation of interest to epidemiologists are:

●● Health effects associated with ionizing radiation from nuclear facilities and disposal of nuclear waste (discussed later in this chapter).

●● Long-term consequences of exposure to radiation from nuclear weapons detonated as part of aboveground testing or used during warfare. For example, radiation from atomic bombs dropped in Japan was associated with increases in the risk of breast cancer, especially if exposure occurred between the ages of 10 and 19 years.39,40

51589_CH13_Printer.indd 566 09/02/13 2:25 PM

t y p e s o f a g e n t s 567

●● Intentional exposure of the population to ionizing radiation through terrorists’ detonation of dirty bombs. In this scenario, terrorists would use conventional explosives to disperse radioactive materials. Emergency response agencies would need to have a protocol in place to deal with such an act. Although the radiation from a terrorist act would be confined to a limited geographic area, this event would cause panic among the general population.

●● Health effects associated with low-level exposures as in the case of patients who receive X-rays and diagnostic nuclear medicine tests. The dose of radi- ation received from medical tests is generally low and carries more benefits than risks. Nevertheless, the possibility for adverse health effects exists.

●● Adverse impacts of occupational exposure to ionizing radiation. Important epidemiologic research questions include the effects of occupational expo- sure upon pregnant women and health outcomes associated with continu- ous long-term exposures at low levels.

●● Effects of exposure to naturally occurring radiation such as that emitted by radon gas, as noted below.

Environmental radon produces one of the largest sources of human exposure to ionizing radiation. Radon, a documented human carcinogen,41 is an inert gas produced by the decay of radium and uranium. Both elements occur universally in the earth’s crust in varying amounts. Globally, estimates suggest that radon induces 3–14% of lung cancers.42 In the United States radon-associated lung can- cer mortality mirrors global experience despite evidence that cigarette smoking, asbestos exposure, and urban air pollution are the leading causes of lung cancer. Refer to Figure 13–7 for data regarding estimated U.S. mortality from selected forms of cancer including radon-associated lung cancer. Of the approximately 21,000 U.S. lung cancer deaths from radon exposure annually, 2,900 occur among nonsmokers.43 Thus radon is the leading cause of lung cancer among nonsmokers. Radon gas builds up in the basements of homes located in some regions of the United States. Proper construction techniques help to prevent the intrusion of radon into homes; ventilation removes radon from basements and other sections of buildings where the gas accumulates. Radon found in residences is linked to lung cancer and should be reduced to the lowest possible levels.44

Allergens and Molds Allergens, found in ambient air and elsewhere in the environment, are substances that provoke an allergic reaction in susceptible persons. If susceptible, individu- als can demonstrate a great deal of variation in their sensitivity and responses.

51589_CH13_Printer.indd 567 09/02/13 2:25 PM

568 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

The allergenic stimulus may consist of fur, pollen, or any of numerous other substances in the environment. Allergic reactions range from dermatitis, asthma, itchy eyes, and other uncomfortable sensations to anaphylactic shock.

Mold proliferates in moist environments and is omnipresent in the environ- ment. In sensitive individuals molds produce health effects such as irritation of the respiratory system; symptoms may include wheezing and coughing. Some persons who have mold allergies may undergo more serious reactions. Patients who are immunocompromised can develop serious lung infections. Although reported in some research, other serious health effects, including memory loss and pulmonary hemorrhage among infants, have not been substantiated.45

While employed by a local health department, Robert Friis once investigated a community outbreak of respiratory disease symptoms alleged to be an allergy associated with molds that were growing in a certain housing tract. A visit to the area disclosed a large number of expensive homes that had been constructed with elegant sunken living rooms. After intense rainfall, the sunken areas flooded and remained permanently saturated with water. Multicolored molds flourished on the nearby walls and carpets and resisted all attempts at elimination. Some of the residents began to complain of increased numbers of colds, respiratory illnesses, and allergic symptoms. The health department’s serologic tests and questionnaire

Figure 13–7 SEER estimated 2010 U.S. mortality for selected cancers. Source: Reproduced from US Environmental Protection Agency. Health risks. Radon. Available at: http://www.epa.gov/radon/health risks.html. Accessed March 15, 2012.

1,460

0 5,000 10,000

# of annual deaths

15,000 20,000 25,000

1,690

3,600

7,880

8,700

10,570

12,200

13,040

13,140

13,850

14,500

18,910

21,000

21,530

21,840

Bone/joint

Thyroid

Larynx

Oral/pharynx

Melanoma

Stomach

Cervix

Kidney and renal pelvis

T yp

es o

f c an

ce r

Brain/nervous system

Ovary

Esophagus

Liver and intrahepatic bile duct

Lung cancer (Radon)

Lymphoma

Leukemia

51589_CH13_Printer.indd 568 09/02/13 2:25 PM

t y p e s o f a g e n t s 569

studies, however, did not reveal any increase in illness or symptoms above what might usually be expected. Investigators concluded that, in this instance, there was no statistically significant relationship between exposure to molds and lung diseases or other health conditions examined.

The 2005 Hurricanes Katrina and Rita caused severe flooding in the New Orleans area. One of the consequences was widespread mold growth, which cre- ated the specter of residents and workers being exposed to high levels of mold in buildings that had been inundated.46 It is likely that the flooding associated with Hurricane Sandy in 2012 will confront some residents of affected areas in the Northeastern United States with potentially hazardous mold proliferation.

Physical and Mechanical Energy In addition to extremes of temperature, agents of physical and mechanical energy include traumatic forces, noise, and vibration. Traumatic forces are associated with unintentional injuries, as in automobile crashes or mishaps in the home and workplace. Among the adverse effects of noise and vibration are hearing loss and musculoskeletal disorders.

Morbidity and mortality from the effects of physical and mechanical energy may be studied epidemiologically. Here is an example of the descriptive epide- miology of unintentional injury: In the United States between 1999 and 2004, unintentional injuries were the leading cause of death within the age group 1–44 years. Such injuries caused approximately 625,000 fatalities. The distribution of the respective percentages of deaths was as follows: motor vehicle-traffic injuries (41.0%), poisonings (15.5%), falls (15.0%), and suffocation (5.4%).47 In view of the significance to society of deaths from unintentional injuries, epidemiolo- gists should accord priority to this topic.

Another type of physical energy is high heat such as that associated with the pos- tulated effects of global warming. One of the outcomes attributed to global warm- ing (the gradual increase in the earth’s surface temperature over time) has been extreme climatic conditions and high temperatures, for example, the heat waves observed in recent decades. Figure 13–8 illustrates the projected effects of global warming in the eastern United States. By the latter third of this decade (2070– 2099), summer in New Hampshire may resemble summer in North Carolina dur- ing the latter part of the 20th century. Climate trends suggest that global warming will accelerate under the scenario of high emissions of greenhouse gases.

Exhibit 13–1 reports on high temperature and mortality in the United States. One example of the impact of high temperatures was the July 1995 episode in Chicago, during which an excess of at least 700 deaths occurred at the same time

51589_CH13_Printer.indd 569 09/02/13 2:25 PM

570 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

as a record-breaking heat wave; most of this excess mortality was classified as heat-related. Semenza et al. conducted a case-control study of risk factors associ- ated with heat-related and cardiovascular mortality.48 The investigators reported that those at greatest risk of dying from heat-related causes did not have access to air conditioning or were socially isolated. Yet another example was the August 2003 calamity in Europe that caused at least 35,000 deaths.

Figure 13–8 Effects of global warming in the northeastern United States by the end of the 21st century. Source: Reproduced from U.S. Environmental Protection Agency. Northeast impacts and adaptation. Climate change. Available at http://www.epa.gov/climatechange/ impacts-adaptation/ northeast.html. Accessed June 11, 2012.

1961–1990

2010–2039

2040–2069

2070–2099

Higher Emissions Scenario

Lower Emissions Scenario

51589_CH13_Printer.indd 570 09/02/13 2:25 PM

e n v i r o n m e n t a l h a z a r d s f o u n d i n t h e W o r k s e t t i n g 571

Environmental Hazards Found in the Work Sett ing

This section covers procedures for monitoring work-related environmental hazards. Examples of monitoring techniques are surveillance programs that aid in the pre- vention of occupational illness. The programs can be used to identify occurrences of illness or injury in the workplace and to monitor trends in illnesses or injuries. The trends may vary by industry, geographic area, and over time, and suggest spe- cific industries to be targeted for further investigation or intervention. Additional topics of this section are specific hazards found in the work environment.

Monitoring and Surveillance of Exposure to Occupational Hazards Two types of surveillance are hazard surveillance and health surveillance. Hazard surveillance refers to the characterization of known chemical, physical, and bio- logic agents in the workplace. If measurements demonstrate that the hazards are present in sufficient quantities to affect health, strategies can be used to reduce exposure of workers to those hazards.49

high temperature and Mortality in the United States

High temperatures contribute to a significant number of deaths in the United States. When the ambient temperatures increase to high levels, the body may develop cramping, fainting, heat exhaustion, and heat stroke. Those who are most susceptible to heat-induced illness are the elderly, young children, and individuals with certain preexisting medical conditions. Within the 5-year span of 1999 to

2003, the death certificates of 3,442 (annual mean = 688) persons listed excessive heat as either an underlying or contributing cause of death. Of the 3,442 deaths, excessive heat was listed as an underlying cause for 2,239 (65%), and hyperthermia was noted as a contributing cause for 1,203 (35%). For 681 (57%) of those with hyperthermia as a contributing cause of death, the underlying cause of death was cardiovascular disease. n

Source: Data from Centers for Disease Control and Prevention. Heat-related deaths—United States, 1999–2003. MMWR. Vol 55, No 29, pp. 796–798, July 28, 2006.

e x

h ib

it 1

3 –1

51589_CH13_Printer.indd 571 09/02/13 2:25 PM

572 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

Related to heath surveillance is the concept of the sentinel health event, popularized by Rutstein and colleagues.50 “A sentinel health event is a case of unnecessary disease, unnecessary disability, or untimely death whose occurrence is a warning signal that the quality of preventive or medical care may need to be improved.”50(p 1054) Health surveillance is concerned with the health of individual workers or groups of workers. The rationale for health surveillance is to “detect adverse health effects resulting from occupational exposures at as early a stage as possible, so that appropriate preventive measures can be instituted promptly. This is a form of secondary prevention.”51(p 706)

Specific Hazards: Biologic Agents, Mineral and Organic Dusts, and Industrial Chemicals Biologic Hazards Estimates indicate that more 300,000 workers worldwide die annually from infectious diseases acquired on the job.52 The two main groups of biological agents linked to work-related infections are allergenic and/or toxic agents that form bioaerosols (e.g., from bacteria and fungi) and zoonotic agents spread by vectors, the airborne route, or contact with the skin.53 Healthcare employees, workers who come into contact with animals, laboratory personnel, and refuse workers are examples of persons who are potentially exposed to risk of infection from biologic agents of disease.52 Sewage workers may be exposed to biohazards carried in raw sewage; agricultural workers are at increased risk of exposure to zoonotic diseases from farm animals and disease agents contained in the soil. Physicians and nurses come into direct contact with patients who may be affected by a communicable disease. Accordingly, they are at increased risk of hepatitis B and other diseases carried in bodily fluids. Sources of infection include possible accidental needle punctures and errors in routine laboratory procedures. Sometimes installations such as dialysis centers bring employees into direct contact with blood and blood products. Of great concern is the possible transmission of the human immunodeficiency virus through accidental needle sticks that might occur in these facilities. Also at risk of work-related infections are laboratory personnel who work with biohazards.

Mineral and organic dusts One of the important contributions of epidemiology is the identification and control of health risks of occupational exposure to dusts. Prolonged and even short-term exposure to dusts in the work environment can pose major health hazards. Dusts from metals such as aluminum, wood used in the manufacture

51589_CH13_Printer.indd 572 09/02/13 2:25 PM

e n v i r o n m e n t a l h a z a r d s f o u n d i n t h e W o r k s e t t i n g 573

of consumer products, cotton fabrics, fiberglass, and plastics (e.g., Styrofoam) present hazards in the occupational environment. Potential adverse health out- comes associated with dust exposure include silicosis (from exposure to dusts from sand), pulmonary emphysema, lung cancer, and chronic obstructive pul- monary disease (COPD). For example, COPD can result from exposure to rub- ber dust.54

Among the notorious examples of occupational diseases resulting from expo- sure to dusts is black lung disease, which also is known as coal workers’ pneu- moconiosis (CWP). Often, beginning phases of CWP are asymptomatic, but later stages may cause disability and premature mortality among miners.55 From 1968 to 2006, deaths from CWP declined 73% from 1,106.2 per year to 300.0 per year. Figure 13–9 provides annual data on years of potential life lost (YPLL) from CWP before age 65 years. The annual YPLL from CWP decreased by more than 90% between 1968 and 2006.

As a general principle, miners’ contact with dust in coal mines leads to increased death rates.56 A cohort life-table analysis of mortality followed 8,899 coal miners over more than two decades from initial examination in 1969–1971. In a detailed examination of causes of mortality among the cohort, researchers found that mortality was elevated for nonviolent causes, nonmalignant respira- tory disease, and unintentional injuries.

Figure 13–9 Years of potential life lost (YPLL) before age 65 years and mean YPLL per decedent for decedents aged ≥25 years with coal workers pneumoconiosis as the underlying cause of death, United States 1968–2006. Source: Reproduced from Centers for Disease Control and Prevention. Coal workers’ pneumoconiosis-related years of potential life lost before age 65 years—United States, 1968–2006. MMWR. Vol 58, No 50, p. 1414, December 25, 2009, p. 1414.

2000 12

M ean Y

P LL

YPLL Mean YPLL

1800

Y P

1600

1400

1200

1000

800

600

400

200

1968 1972 1976 1980 1984 1988

Year

1992 1996 2000 2004 2006 0

51589_CH13_Printer.indd 573 09/02/13 2:25 PM

574 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

Industrial chemicals Exposures to vastly higher concentrations of chemicals may occur among workers than among the general population. Exposure to chemicals in the occupational setting averages 1 to 100 times higher than in the ambient environment.57 Each year, an estimated 50,000–70,000 U.S. workers develop chronic occupational diseases produced by exposure to toxic chemicals in the workplace.57

Vapors and fumes represent a significant occupational hazard in many types of work. Given the yearly increase in the numbers of chemical substances and solvents that are utilized in the work environment, fumes and vapors are likely to become an increasing hazard to the health of workers. Various organic solvents, such as benzene and vinyl chloride, may act as carcinogenic agents. In addition, organic solvents may damage internal organs of the body, the liver being espe- cially vulnerable.

Exposure to vinyl chloride, a solvent used in the plastics industry, is associ- ated with a rare tumor known as angiosarcoma of the liver.58 Vinyl chloride is described also as a carcinogenic substance related to lung cancer59 and central nervous system tumors.58 Workers in employment settings who are occupation- ally exposed to vinyl chloride have experienced the specific hazard of hepatic angiosarcoma. However, research evidence has not supported vinyl chloride as responsible for other cancers or nonmalignant diseases.60

Occupational exposure to pesticides such as organophosphates is of great potential concern for up to 5 million farm workers in the United States. Some of the acute effects of such exposure are nausea, vomiting, and vertigo.61 Research investigations have suggested that the long-term effects of pesticide exposure include several types of cancer, miscarriages and teratogenic effects, sterility, spontaneous abortions, skin problems, respiratory difficulties, and cognitive deficits. However, a comprehensive literature review concluded that aside from the acute effects of pesticide exposure, clearly delineated long-term health effects have not been identified.61 Some of the long-term effects may be nonspecific (e.g., fatigue, poor concentration, and dizziness). McCauley et al. argued that challenges to epidemiologic research in examining the association between farm workers’ exposure to pesticides and health outcomes “. . . are determination of the population at risk; a valid determination of exposure; verification of diag- nosis, symptom, or biological marker of a health effect among the populations being studied; methods to link individual exposure to health effects; and the ability to establish a temporal relationship between the exposure and the health effect. In attempts to study farmworker populations, these tools are often incom- plete, dysfunctional, or nonexistent.”61(p 954)

51589_CH13_Printer.indd 574 09/02/13 2:25 PM

n o t e W o r t h y C o m m u n i t y e n v i r o n m e n t a l h e a l t h 575

Noteworthy Community Environmental Health Hazards

Not confined to the workplace, environmental hazards also impinge upon the community at large. Such potential hazards include chemicals from industrial sources and toxic waste dumps, air pollution, emissions of ionizing radiation from nuclear power facilities, and degradation of water quality. Examples of two specific community environmental health issues are the sick building syndrome

and Gulf War syndrome (see Exhibits 13–2 and 13–3, respectively).

Hazardous Waste Sites Globally and in the United States, hazardous waste sites represent a potential source of human exposure to toxic chemicals. Outside of the United States, the volume of hazardous wastes is increasing with growing industrialization of less developed regions of the world. Disposal of hi-tech trash has become the burden of developing countries. In the United States, more than 750 million tons of toxic chemical substances had been discarded into as many as 50,000 hazardous waste sites as of the late 1980s.70 Notorious toxic waste sites in the United States and the dates when they gained public attention include Love Canal, New York (1978); the Valley of the Drums in Kentucky (1967); Times Beach, Missouri (1985); the Stringfellow acid pits in California (1980s), and the Casmalia Waste Disposal Facility in California (1990s). A major public health concern is the potential impact upon human health of waste leachates emitted by disposal sites into community water supplies. Most communities in the United States receive water supplies from underground aquifers and surface water, which are at poten- tial risk of contamination by toxic wastes. One study reported a statistically sig- nificant excess of some forms of cancer mortality among residents of counties that contain hazardous waste sites.70

Epidemiologic research into the health effects associated with hazard- ous waste sites confronts several methodologic difficulties. Because hazardous wastes involve a complex mixture of substances, it is difficult not only to sort out which chemicals affect human health, but also to determine how best to measure specific exposures in a valid and reliable manner. Some studies may not control adequately for potentially confounding factors and thus need to be interpreted with caution.71 Measurement of the long-term effects of continu- ous exposure is difficult. Some of the research in this field is based upon small study samples, relatively few health events, statistically nonsignificant findings, and inadequate assessment of exposures. A technique for reducing the danger of

51589_CH13_Printer.indd 575 09/02/13 2:25 PM

576 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

the Sick building Syndrome (SbS)

SBS has enormous potential for great economic impact from lost workdays, litigation, and workers’ compensation claims. It is a condition that has several hypothesized etiologies, one being poor indoor air quality, which may affect health adversely. Indoor air quality became an issue during the 1970s when, in response to an energy crisis, buildings were made more airtight and ventilation rates were reduced.62 Cases of SBS frequently occur in heavily

populated, carpeted, and poorly cleaned buildings. The term sick building syndrome was coined in 1983 at a World

Health Organization meeting in Geneva to describe illnesses of office employees that included “dryness of the skin and mucous membranes, mental fatigue, headaches, general pruritis, and airway infections.”63 The prevalence of SBS has not been determined accurately because of the lack of a precise case definition and the absence of specific biologic markers.

Not only is the precise definition of SBS unclear, but it also bears some similarity to three other controversial disorders: multiple chemical sensitiv- ity, chronic fatigue syndrome, and fibrositis. For example, all have fatigue as a component as well as sharing a nonspecific quality. One of the note- worthy features of SBS, in contrast to these other disorders, is temporality of symptoms; they begin when the affected person enters the building and diminish when the person leaves. One study showed that the prevalence of SBS symptoms decreased greatly when employees were exposed to an improved ventilation system.64

In addition to poor indoor air quality and inadequate ventilation, other proposed causes of SBS include exposure to volatile organic compounds (e.g., solvents), low humidity, airborne microbial agents— bacterial endotoxins, molds, dust mites—and tobacco smoke and odors.62 Several psychosocial factors for SBS have been linked to the nature of some work environments: monotonous, authoritarian, and overly demanding.62 Organizational stress—poor worker cooperation and physical and mental stress—may be an important component of the SBS.65 n

e x

h ib

it 1

3 –2

51589_CH13_Printer.indd 576 09/02/13 2:25 PM

n o t e W o r t h y C o m m u n i t y e n v i r o n m e n t a l h e a l t h 577

the Gulf War Syndrome (GWS)

Following the conclusion of the 1991 war in the Persian Gulf, veterans began to report symptoms of GWS.66 As of the late 1990s, nearly 15% of the almost 700,000 U.S. veterans who served in this conflict complained of health conditions that they believed were the result of the war. Some of the veterans’ health problems could be accounted for by standard medical or psychiatric diagnoses that were not connected with the war. Nevertheless, for many of

the remaining ill veterans, GWS emerged as a common pattern or array of nonspecific, vague symptoms with uncertain etiology.67

An influential Presidential Advisory Committee on Gulf War Veterans’ Illnesses suggested that GWS might be explained by wartime stress. GWS appeared to bear similarities to the post-traumatic stress disorder that was said to follow other wars. However, some reports have been critical of this explanation.67

Among other causes suggested for GWS were chronic fatigue syn- drome, fibromyalgia, or exposure to nerve gas, which is known to be a cholinesterase inhibitor.66 Possible contact with nerve gas might have occurred when U.S. troops destroyed an ammunition dump that was later found to contain traces of mustard gas and the potent nerve agent sarin.68 A large number of troops were downwind from the dump. Many of the symptoms of GWS would be consistent with neurotoxicity from low-grade nerve gas poisoning, pesticides, or other cholinesterase inhibi- tors. Exposure to these agents might account for the muscle weakness, sweating, wheezing, and other symptoms that the affected Gulf War vet- erans reported.

An extensive review of U.S. Gulf War veterans’ mortality and utiliza- tion of healthcare services noted that many large controlled studies “. . . revealed an increased post-war risk for mental health diagnoses, multi- symptom conditions and musculoskeletal disorders. Again, these data failed to demonstrate that Gulf War veterans suffered from a unique Gulf War-related illness.”69(p. 553) A thorny issue for epidemiologic research on GWS is that the questions that must be asked are not clearly delineated, nor are the possible exposures involved. n

e x

h ib

it 1

3 –3

51589_CH13_Printer.indd 577 09/02/13 2:25 PM

578 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

misinterpretation inherent in a single study is the use of meta-analysis, which allows for the pooling of the results of all available research. In summary, epide- miologists face challenges in quantifying risks associated with hazardous waste sites. Not only is accurate exposure information difficult to obtain, but also the effects of low-level exposures are difficult to demonstrate.72

Reported adverse effects of hazardous waste exposure include birth outcomes (low birth weight and occurrence of congenital malformations or other birth defects), neurologic disease, cancer, synergistic effects, illness symptoms, and other adverse health conditions. A study of residents of Love Canal (which is located in upstate New York near Niagara Falls) showed an excess of low birth weights as well as growth retardation for those exposed compared with the gen- eral population and an excess of birth defects in residents closest to the site. The rate of respiratory cancer was similar to that for the Niagara Falls area, however.71 An extensive case-control study of over 9,000 newborns with congenital mal- formations living in proximity to hazardous waste sites in New York state found a small but statistically significant risk for birth defects.73 A population-based case-control study that reviewed vital records in Washington state for the years 1987–2001 suggested that residence near hazardous waste sites was not associ- ated with fetal deaths, although living in proximity to sites that contained pesti- cides increased risk of fetal death.74

Air Pollution Not only does air pollution lower our quality of life by obscuring the natural envi- ronment and, in some instances, by being malodorous, but it also has been impli- cated in adverse human health impacts. Many cities in the developed world have made strides in reducing air pollution levels. Unfortunately, severe episodes of air pollution are becoming increasingly common in the rapidly industrializing cities of developing countries. Noteworthy are the high levels in some of the large cities in China, the Middle East, South Asia, and Latin America. For example, air pollution associated with heart attacks exacts a significant toll in Iran; the residents of cities in India breathe excessive amounts of toxic pollutants. Regrettably, air quality standards in many developing countries are nonexistent or poorly enforced. Hazards from pol- luted air are not confined to the developing world. Several regions of the United States such as the Los Angeles Basin continue to have excessive air pollution levels.

Among the constituents of air pollution (although these vary from one location to another depending upon the types of fuels in use) are sulfur oxides, particles, oxidants (including ozone, carbon monoxide, hydrocarbons, and nitro- gen oxides), lead, and some other heavy metals.75 Indoor air pollution from

51589_CH13_Printer.indd 578 09/02/13 2:25 PM

n o t e W o r t h y C o m m u n i t y e n v i r o n m e n t a l h e a l t h 579

cigarette smoke, gas stoves, and formaldehyde may pose a risk for respiratory illness.76 Fine airborne particles can bypass the body’s defenses, be inhaled deeply into the respiratory system, enter the circulatory system, and be distributed throughout the body. Another example related to indoor air pollution is the SBS,77,78 described earlier in this chapter.

Major lethal air pollution episodes include those in Donora, Pennsylvania (1948); London, England (1952); and Meuse Valley in Western Europe (1930). Air pollution levels, which in many urban areas (e.g., Mexico City) are alarm- ingly high, are related to human mortality. Individuals who have preexisting heart and lung disease may be at particular risk for the fatal or aggravating effects of air pollution. Cigarette smoking and air pollution may act synergistically in aggravation of lung diseases, such as emphysema.75

A venerable example of research on air pollution’s health effects is that con- ducted by Henderson et al.79 Census tracts in Los Angeles were aggregated into 14 study areas that represented homogeneous air pollution profiles. The study reported a correlation between the geographic distribution of lung cancer cases and the general location of emission sources for hydrocarbons. More recently, compelling evidence that diesel exhaust is a risk factor for lung cancer after care- ful adjustment for potential confounding has been reported.80

Other epidemiologic analyses have shown a correlation between increases in total daily mortality and increased air pollution in New York City.81 Daily mortal- ity also was related to gaseous and particulate air pollution in St. Louis, Missouri, and the counties in eastern Tennessee.82 Researchers estimated the daily mortality rate associated with inhalable particles, fine particles, and aerosol acidity. The total mortality rate was found to have increased by 16% in St. Louis, Missouri, and by 17% in the eastern Tennessee counties. The data further suggested that the mass concentrations of particles have an association with daily mortality.

Carbon monoxide arises from cigarette smoking, automobile exhaust, and certain types of occupational exposures. An investigation into the possible asso- ciation between angina pectoris and heavy freeway traffic found no direct asso- ciation between myocardial infarction and ambient carbon monoxide. It was hypothesized that there was an indirect association between exposure to carbon monoxide in the ambient air and acute myocardial infarction through smoking, which is associated with elevated blood carbon monoxide levels.83 An analysis of data from European cities demonstrated significant associations of carbon mon- oxide levels with total mortality and cardiovascular deaths.84

The air of one large metropolitan city exposed pedestrians and outdoor work- ers to carbon monoxide levels that ranged from 10 to 50 ppm; there were even

51589_CH13_Printer.indd 579 09/02/13 2:25 PM

580 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

higher levels in poorly ventilated areas of the city.85 The recommended standard for carbon monoxide in ambient air is a maximum of 40 ppm, not to be exceeded more than once a year. Pedestrians and workers who have heart problems may be at increased risk of aggravation of their condition when exposed to high levels of ambient carbon monoxide in the urban environment.

In addition to all-cause mortality and chronic diseases such as lung cancer and heart disease, another condition linked to air pollution is COPD, which is among the top sources of morbidity and mortality in the United States and the world. A major investigation, the Tucson Epidemiological Study of Airways Obstructive Diseases, has tracked the etiology and natural history of obstructive lung disease.86 Based on a multistage stratified cluster sample of white households in the Tucson, Arizona, area, the study has generated many important findings, including the effects of passive smoking in children.87 Another report on the same topic comes from the Renfrew and Paisley (MIDSPAN) Study, a 25-year prospective investigation of 15,411 residents of west central Scotland. Information on respiratory symptoms and function was collected at baseline between 1972 and 1976. A total of 4,064 married couples from the original cohort were recontacted in 1995. Allowing for subjects lost to follow-up and nonresponse, data were collected in 1996 from 1,477 families and their offspring (1,040 males and 1,298 females). The MIDSPAN study has yielded noteworthy insights into the natural history of COPD including a familial association of cigarette smoking and passive smoking with lowered cardiorespiratory health.88

The term passive smoking (also known as secondhand or sidestream exposure to cigarette smoke) refers to the involuntary breathing of cigarette smoke by non- smokers in an environment where there are cigarette smokers present. In restau- rants, waiting rooms, international airliners, and other enclosed areas, cigarette smokers can expose nonsmokers unwillingly (and, perhaps, unwittingly) to a potential health hazard. The effects of chronic exposure to cigarette smoke in the work environment were examined in a cross-sectional study of 5,210 cigarette smokers and nonsmokers. Nonsmokers who did not work in a smoking environ- ment were compared with nonsmokers who worked in a smoking environment as well as with smokers. Exposure to smoke in the work environment among the nonsmokers was associated with a statistically significant reduction in pul- monary function test measurements in comparison with the nonsmokers in the smoke-free environment.89

A 1992 report from the U.S. Environmental Protection Agency (EPA) con- cluded that environmental tobacco smoke is a human lung carcinogen responsible for approximately 3,000 lung cancer deaths annually among U.S. nonsmokers.90

51589_CH13_Printer.indd 580 09/02/13 2:25 PM

n o t e W o r t h y C o m m u n i t y e n v i r o n m e n t a l h e a l t h 581

Among children, passive smoking is associated with bronchitis, pneumonia, fluid in the middle ear, asthma incidence, and aggravation of existing asthma. A review of data on 53,879 children from 12 cross-sectional studies concluded that parental smoking is related to respiratory symptoms such as wheezing, asthma, bronchitis, and nocturnal cough.91 The 2006 U.S. Surgeon General’s92 report on the health consequences of involuntary exposure to tobacco con- cluded that “Secondhand smoke exposure causes disease and premature death in children and adults who do not smoke.” In addition, the report noted that “Exposure of adults to secondhand smoke has immediate adverse effects on the cardiovascular system and causes coronary heart disease and lung cancer.”

Research on passive smoking presents several methodologic difficulties.93 Relatively small increases in risk of death from passive smoking are difficult to demonstrate in those situations where exposure assessment has not been well- developed, as in the use of questionnaires to quantify exposure. Studies need to account for long- and short-term variability in exposures to cigarette smoke from many sources (e.g., those at workplaces, restaurants, and entertainment venues). The long latency period between exposure to cigarette smoke and onset of disease contributes to the methodologic difficulties of this field of research. Additional research will require improved methods for assessing exposure to cigarette smoke, such as the use of biologic markers (e.g., cotinine).

Nevertheless, despite the complexity of assessing the relationship between exposure to environmental tobacco smoke and adverse health outcomes, the weight of the evidence supports the view that secondhand cigarette smoke poses health risks to children and adults. This inference is consistent with findings of research conducted in the United States and elsewhere. For example, EPIC is a massive European prospective cohort study of 500,000 participants.94 A nested case-control study that incorporated information from more than 300,000 per- sons in the EPIC database demonstrated that exposure to environmental tobacco smoke increased the risk for respiratory diseases and lung cancer. Many coun- tries, states, and localities now restrict cigarette smoking in areas such as the workplace, bars and restaurants, and airliners.

Nuclear Facilities Nuclear facilities include weapons production plants, test sites for detonation of nuclear weapons, and nuclear power plants. Many of the former nuclear weapons production plants have been decommissioned, but the legacy of the environmental contamination they caused remains. Although the United States has stopped atmospheric testing of nuclear weapons, ongoing research tracks

51589_CH13_Printer.indd 581 09/02/13 2:25 PM

582 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

the health effects of earlier population exposures to nuclear fallout. Finally, the possible dissemination of radioactive materials from nuclear power plants continues to be an ongoing concern in the United States and worldwide.

Nuclear power plants Refer to Figure 13–10 for a picture of the San Onofre Nuclear Power Plant in Southern California. On January 31, 2010, officials shut down the plant temporarily because several defective reactor tubes emitted small amounts of radioactivity.

Noteworthy releases of radioactivity from nuclear power plants were those that occurred at Three Mile Island in Pennsylvania, Chernobyl in Ukraine, and Fukushima in Japan. Epidemiologists have studied the health effects of a nuclear accident at Three Mile Island, Pennsylvania, that occurred on March 26, 1979. One study reported that there was a modest association between postaccident cancer rates and proximity to the power plant. There was a postaccident increase in cancer rates during 1982 and 1983, which subsequently declined. Radiation emissions from the plant did not appear to account for the observed increase in cancer rates, however.95

The nuclear power plant accident at Chernobyl, Ukraine, in April 1986 was a major public health disaster that produced massive exposure of European populations to ionizing radiation. Radioactive materials, primarily iodine and cesium, were dispersed over the eastern part of the former Soviet Union, Sweden, Austria, Switzerland, and parts of Germany and northern Italy. Despite this wide

Figure 13–10 The nuclear power plant at San Onofre, California.

51589_CH13_Printer.indd 582 09/02/13 2:25 PM

n o t e W o r t h y C o m m u n i t y e n v i r o n m e n t a l h e a l t h 583

distribution of radionuclides, cancer rates (with the exception of thyroid cancer rates in the areas that were most highly contaminated) do not seem to have increased in Europe.96

Closer to the Chernobyl nuclear facility, there were marked increases in thyroid cancer as soon as four to five years after the accident. The Gomel region of Belarus in the former Soviet Union lies immediately to the north of Chernobyl. There was a sharp increase in thyroid cancer cases among children in Gomel from 1 to 2 per year during 1986–1989 to 38 in 1991.97 These increases suggested that the carcinogenic effect of radioactive fallout, particularly among young children and fetuses, is much greater than previously believed.98 Persons exposed to radiation as children had increased risk of thyroid cancer,98,99 possibly through uptake of iodine-131. Adults who resided in these same areas did not appear to have elevated risks of thyroid cancer, leukemia, and other cancers; workers who were exposed to radiation while cleaning up had increased risk of leukemia.99 An assessment made 20 years after the accident reaffirmed that the most significant consequence of Chernobyl was an increase in thyroid cancer cases among those exposed during childhood, especially among children who lived close to the reactor.100

Another health effect of the Chernobyl accident could be the production of birth defects, such as congenital anomalies. Data from the EUROCAT epidemi- ologic surveillance of congenital anomalies did not suggest an increase in central nervous system anomalies or Down syndrome in Western Europe approximately 5 years after the incident.101

With the exception of cleanup workers directly exposed to radiation from Chernobyl, increases in incidence of cancers such as leukemia among other per- sons who resided far away from the reactor and in other parts of Europe may be difficult to demonstrate. Currently, no method exists to distinguish between cases of leukemia that resulted from the Chernobyl disaster and cases that might be linked to other sources of ionizing radiation. In addition to radiation from the Chernobyl release, European populations also are exposed to natural background radiation and medical radiation.

On March 11, 2011, a 9.0-magnitude earthquake followed by a devastating tsu- nami struck northern Japan.102 The unprecedented natural disaster killed approxi- mately 16,000 people and displaced 340,000.103 Cooling systems at the Fukushima Daiichi Nuclear Power Plant shut down, causing three reactors to experience par- tial meltdowns.102 Additional radiation escaped from spent fuel rods at another reactor when it caught on fire and exploded. Resulting explosions and fires spewed huge amounts of radiation into the atmosphere. The Japanese government advised

51589_CH13_Printer.indd 583 09/02/13 2:25 PM

584 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

residents who lived within 19 miles of the reactor to evacuate.104 The American Embassy urged all American citizens within 50 miles of the plant to leave.

A year following the release of radioactive materials, experts concurred that physical health effects associated with the incident are likely to be negligible.103 The total amount of materials released was smaller than in Chernobyl and limited to few types of radioactive materials that have relatively short half-lives and can be detected easily. An example is iodine-131, which has an 8-day half-life. The most highly exposed persons were members of the first cleanup crew who received a lifetime dose of radiation within a few hours; none has reported instances of acute radia- tion sickness. Soil, food, and seafood from the area do not appear to show evidence of harmful radiation levels. Authorities in the Fukushima Prefecture will conduct surveillance for adverse health effects for 20 years in order to discern any adverse outcomes. Perhaps the most significant impacts of the disaster will be upon the mental health of the Japanese citizens who were traumatized severely by this event.

In terms of cancer rates and proximity to other nuclear power plants, studies with conflicting nonsupportive results have emerged. One study reported an excess of leukemia in a five-town area in Massachusetts in which one of the towns was the site of a commercial nuclear power plant. The excess cases were found mostly in adults and the elderly, but the authors believed that the results from this descriptive study were suggestive and warranted additional, more inten- sive follow-up investigations.105 A systematic literature review concluded that, although many studies of the health effects of community exposure to radiation were statistically sound, they did not provide adequate quantitative estimates of radiation dose that could be used to assess dose–response relationships.106 A nationwide cohort study carried out in Switzerland reported that there was lit- tle evidence of a relationship between living near a nuclear power plant and risk of leukemia and other childhood cancers.107 The KiKK (Epidemiologic Case- Control Study of Childhood Cancer and Nuclear Power Plants) study conducted in Germany reported a relationship between proximity of residence within 5 km of a nuclear power plant and children’s risk of developing leukemia before 5 years of age.108 Given the conflicting findings, some experts believe that cancer risk among persons who live near nuclear power plants is an open question.109

Nuclear weapons plants In the United States, nuclear weapons have been produced at Oak Ridge, Tennessee; Rocky Flats near Denver, Colorado; and Hanford, Washington. The Hanford, Washington, facility produced plutonium for nuclear weapons beginning in 1943 as part of the Manhattan Project and continued operation

51589_CH13_Printer.indd 584 09/02/13 2:25 PM

n o t e W o r t h y C o m m u n i t y e n v i r o n m e n t a l h e a l t h 585

through the cold war. Production ceased in 1989, and a massive cleanup effort commenced. The EPA noted that “the operations at Hanford created one of the largest and most complex cleanup projects in the United States. Weapons pro- duction resulted in more than 43 million cubic yards of radioactive waste, and over 130 million cubic yards of contaminated soil and debris. Approximately 475 billion gallons of contaminated water was discharged to the soil. Some of the contaminants have made it to groundwater above the site; over 80 square miles of groundwater is contaminated to levels above groundwater protection standards.”110 Epidemiologists have conducted ongoing investigations of can- cer among Hanford workers in association with radiation exposure. Evidence suggests that radiation is associated with cancer mortality among the cohort of workers and consists primarily of lung cancer mortality related primarily to radi- ation exposures that occurred among older age groups.111

The Department of Energy owns the Rocky Flats plant, which is known pres- ently as the Rocky Flats Environmental Technology Site (RFETS). From 1951 to 1989, the RFETS produced components of nuclear weapons until it stopped oper- ating in 1989 as a result of concerns about its environmental impacts. Large quan- tities of petroleum and other hazardous substances were stored in the installation. A major fire in 1957 exposed nearby residents to plutonium.112 They were also exposed during routine operations at the plant and from plutonium contamination of the soil. In 1989 the EPA placed the site on the Comprehensive Environmen- tal Response, Compensation, and Liability Act (CERCLA) national priorities list. Subsequently, the DOE cleaned up and restored the site, although some contami- nation remains in the core production areas. However, according to research this contamination does not present threats to human health and the environment.113

Nuclear weapons testing Numerous studies have examined the health effects of aboveground atmo- spheric testing of nuclear weapons at the Nevada test site in the United States; testing occurred during 1951 to 1958. Among the components of fallout from weapons testing is radioactive iodine, which as noted previously may become concentrated in the thyroid gland, producing thyroid cancer. An ongoing epide- miologic project is a cohort study of young people who lived in proximity to the test site during infancy and childhood.114 Three cohorts, similar in demographic and lifestyle characteristics, were selected. Two cohorts were from Washington County, Utah, and Lincoln County, Nevada, both on the west side of the test site, close to the site, and in the pathway of the heaviest fallout. A third cohort (unexposed) was selected from Graham County, far to the south of the test

51589_CH13_Printer.indd 585 09/02/13 2:25 PM

586 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

site. At 12 to 15 years and 30 years after the heaviest fallout, there was a slight but nonsignificant increase in rates of thyroid cancer among the two exposure cohorts in comparison with the control cohort. Thus, it was concluded that liv- ing near the Nevada test site did not produce a statistically significant increase in thyroid neoplasms.

Drinking Water Public health experts increasingly have been concerned about possible degrada- tion of the quality of water supplies in the United States due to development of polluting hi-tech industries, urbanization, and population growth. In fact, unavailability of adequate, reliable, and safe water resources is currently an issue for the residents of many countries worldwide, especially those in arid parts of the globe. Approximately one-sixth of the world’s population lacks safe drink- ing water, and many of these who are most vulnerable are children. Sources of groundwater contamination include industrial facilities (e.g., chemical plants and nuclear installations), new human habitation, and runoff from growing urbanized areas, all of which threaten the supply of potable water with perma- nent contamination from pesticides, industrial chemicals and solvents, radioac- tivity, and pathogenic microorganisms. The roles of environmental epidemiology in water quality include monitoring and control of infectious disease outbreaks and study of the health effects of low levels of toxic agents that may be present in the water supply. The problem of health hazards associated with low water qual- ity is one that requires much additional epidemiologic study.

In the United States, water quality is regulated by the EPA, which sets drink- ing water standards for more than 80 contaminants that may be implicated in human health.115 The contaminants are categorized according to whether they are responsible for acute effects (occurring within a few hours or days) or chronic effects (long-term after consumption for many years). Generally, microorgan- isms are responsible for the former, whereas contaminants that increase risk of chronic effects fall into the following classes: chemicals (pesticides, solvents, and by-products of the disinfection of water), radionuclides, and minerals (arsenic and lead). Epidemiologic research has examined the relationship between expo- sure to the foregoing types of contaminants and outcomes such as cancer and disruption of reproductive processes. Much is unknown about the effects of organic compounds, such as pesticides, in water.

In addition to the types of contaminants noted in the preceding paragraph, particles (i.e., finely divided solids) may be found in drinking water supplies, especially those not treated by filtration. Although many of these particulate

51589_CH13_Printer.indd 586 09/02/13 2:25 PM

n o t e W o r t h y C o m m u n i t y e n v i r o n m e n t a l h e a l t h 587

contaminants are not believed to be harmful, asbestos fibers may pose a hazard to health. The evidence regarding the toxicity of waterborne asbestos particles is not conclusive, however.

The use of chlorination and standard drinking water treatments adopted early in the 20th century have led to a decline in the incidence of gastroenteric dis- eases; implementation of EPA water quality regulations in the late 1980s has led to further reduction in the number of cases of waterborne diseases. The number of outbreaks from 1971 to 2008 by etiology (e.g., Legionella spp., chemical, viral, bacterial, and parasitic) is shown in Figure 13–11.

Figure 13–11 Number of waterborne disease outbreaks associated with drinking water (n = 818),* by year and etiology–waterborne disease and outbreak surveillance system, United States, 1971–2008. Source: Reproduced from Centers for Disease Control and Prevention. Surveillance for waterborne disease outbreaks associated with drinking water—United States, 2007–2008. MMWR. Surveillance Summaries, September 23, 2011, Vol 60 (SS12) p. 48.

† Legionnaires’ disease (LD) was reported to the Waterborne Disease and Outbreak Surveillance System (WBDOSS) beginning in 2001. A review of publications and CDC-led investigations during 1971–2000 resulted in the addition of 17 historic LD drinking water outbreaks to WBDOSS.

§ Includes all bacteria except Legionella.

N o.

of o

ut br

ea ks

0 1971 1974 1977 1980 1983 1986 1989 1992 1995 1998

Multiple Legionella spp.†

Unidentified Chemical Viral Bacterial§

Parasitic

2001 2004 2007

* Some outbreaks from prior reporting periods were added, reclassified, or excluded during an extensive review (Craun GF, Brunkard JM, Yoder JS, et al. Causes of outbreaks associated with drinking water in the United States from 1971 to 2006. Clin Microbiol Rev 2010;23:507–28); therefore, data are not comparable to figures in previous report.

51589_CH13_Printer.indd 587 09/02/13 2:25 PM

588 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

For the period 2007–2008, the CDC reported a total of 48 disease outbreaks that were waterborne. These outbreaks occurred in 24 states and Puerto Rico. According to the CDC,

Of these 48 outbreaks, 36 were associated with drinking water, 8 with WNID [water not intended for drinking], and 4 with WUI [water use of unknown intent]. The 36 drinking water–associated outbreaks caused illness among at least 4,128 persons and were linked to 3 deaths. Etiologic agents were identified in 32 (88.9%) of the 36 drinking water–associated outbreaks; 21 (58.3%) outbreaks were associated with bacteria, 5 (13.9%) with viruses, 3 (8.3%) with parasites, 1 (2.8%) with a chemical, 1 (2.8%) with both bacteria and viruses, and 1 (2.8%) with both bacteria and parasites. Four outbreaks (11.1%) had unidentified etiologies. Of the 36 drinking water–associated outbreaks, 22 (61.1%) were outbreaks of acute gas- trointestinal illness, 12 (33.3%) were outbreaks of acute respiratory illness (ARI), 1 (2.8%) was an outbreak associated with skin irrita- tion, and 1 (2.8%) was an outbreak of hepatitis. All outbreaks of ARI were caused by Legionella spp.116

As noted previously, one type of chemical contaminant of drinking water is disinfection by-products, which may be associated with adverse health effects. Epidemiologic research has suggested that tap water consumption might endan- ger pregnant women by causing spontaneous abortions. In response to media reports of epidemiologic research that linked water disinfection to adverse preg- nancy outcomes, panicked citizens besieged public health officials with their concerns about the safety of the water supply. A flurry of reports and an acerbic public debate ensued for several weeks (see Exhibit 13–4).

Conclusion

Environmental epidemiology is the study of the impact of the environment on human health in populations. Epidemiologic methods are used to investigate a wide variety of health outcomes hypothesized to be associated with physical and work environments. Suspected health outcomes include morbidity and mortality from cancer, lung disease, birth defects, injuries and trauma, neurologic disease, and dermatologic problems. Epidemiologic researchers employ many of the traditional study designs to investigate environmentally associated health problems. Toxicologic concepts also play a central role in this field of

51589_CH13_Printer.indd 588 09/02/13 2:25 PM

C o n C l u s i o n 589

Findings that Link tap Water to Miscarriages evoke panic: Los angeles area residents Urged to Show prudence

“Reacting to a new study that suggests a possible link between drinking tap water and miscarriages by pregnant women in their first trimester, Los Angeles officials said Tuesday that the city’s water often contains amounts of the suspect contaminants exceed- ing the levels that have triggered concern. But the officials advised

prudence rather than panic.”117

During early 1998, the media reported a possible association between use of tap water from public drinking water supplies and increased risk of miscarriages. Physicians were besieged by calls from anxious patients, and some young expectant couples agonized over how they would pay the cost of bottled water.

With increasing industrialization, urbanization, and population growth, some residential water supplies may carry dangerous pollution levels. Industrial solvents and chemicals as well as household pesticides may leach into aquifers or contaminate groundwater that is used in tap water. Even when the public water supply is free from these sources of pollution, potentially dangerous chemicals may be released when the chlorine used to sanitize the water and kill microbial agents reacts with organic materials in the water. The resulting end products are known as trihalomethanes. The abbreviation TTHM is used to denote total trihalomethanes.

Epidemiologists Swan and colleagues, affiliated with California’s Department of Health Services, published findings on potential reproductive health effects of drinking water use.118 The investigators noted that TTHMs are found in nearly all U.S. drinking water supplies. TTHMs, which include chemicals such as chloroform, have possible associations with reproductive abnormalities in animal studies. Human exposure to TTHMs may be related to adverse pregnancy outcomes, such as spontaneous abortions.

In their research, Swan et al. addressed the specific issue of whether use of tap water was associated with adverse pregnancy outcomes in humans. The investigators’ previous research (based on retrospective stud- ies) reported an association between spontaneous abortion rates and use

e x

h ib

it 1

3 –4

continues

51589_CH13_Printer.indd 589 09/02/13 2:25 PM

590 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

of tap water in comparison with bottled water among pregnant women in a single California county. In subsequent research (a prospective study), they followed pregnant women who resided in three California regions, two in the north (regions I and II) and one in the south (region III), served by the Kaiser Permanente Medical Care Program. The respective regions were served by one of three types of water: groundwater mixed with sur- face water; primarily surface water; or primarily groundwater. A total of 5,342 pregnant women completed interviews during the first trimester of their pregnancy. Information was collected on a range of topics, including medical history, lifestyle and psychosocial factors, and water consump- tion practices during the week beginning with the last menstrual period and the week before the interview. For example, information regarding use of cold tap water, heated tap water, noncarbonated bottled water, and carbonated water was assessed. Water consumption was quantitated in glasses per day. Finally, pregnancy outcomes were ascertained from vari- ous sources that included review of medical and vital records and tele- phone interviews.

The study demonstrated that spontaneous abortion rates were slightly different from one region to another, ranging from 9.2% to 10.1%. In region I only, consumption of cold tap water was associated with the occurrence of spontaneous abortions. High levels of tap water consump- tion were even more strongly related to the rate of spontaneous abortions. High consumption levels of bottled water in comparison to tap water seemed to be associated with a reduced rate of spontaneous abortions. Type of water consumption did not alter the risk of spontaneous abortion in the other two regions studied. Further research showed that one type of TTHM (bromodichloromethane) found in cold tap water was associated with spontaneous abortions.119 n

exhibit 13–4 continued

research. Agents of environmentally associated disease include toxic chemicals, dusts, metals, and electromagnetic and ionizing radiation. Workers in many occupational settings are at risk of exposure to a wide variety of hazardous agents, justifying public health efforts to minimize exposure and monitor for adverse effects. Environmental health hazards that may affect the community include toxic waste dumps, air pollution, ionizing radiation from power plants and weapons testing, and polluted drinking water. Growth of the human population

51589_CH13_Printer.indd 590 09/02/13 2:25 PM

s t u d y Q u e s t i o n s a n d e x e r C i s e s 591

and increasing urbanization are certain to perpetuate society’s concern about health hazards potentially associated with the environment.

Study Questions and Exercises 1. Define the following terms:

a. environmental epidemiology b. latency c. synergism d. threshold e. dose–response curve f. ionizing radiation g. hazard surveillance

2. A hypothetical community located near a large military base is suspected of having a toxic chemical present in the groundwater. Propose the design of a case-control and a cohort study to examine the impacts on humans of exposures to the toxic chemicals.

3. An ecologic study reports an increase in mortality in census tracts that have high levels of air pollution in comparison with less polluted census tracts. What are some possible alternative explanations for the findings of the study?

4. What is meant by end points in occupational health studies? Discuss the advantages and disadvantages of using each of the following end points: self-reported symptoms, results of clinical examinations, and mortality.

5. How does the “healthy worker effect” influence the interpretation of findings from occupational health research? An epidemiologic researcher finds that mortality for assembly line workers in an automobile factory is slightly higher than the mortality of the general population. Assume that the healthy worker effect is operative. Would it tend to decrease or increase mortality differences between the workers and the general population?

6. Name the four major classes of pesticides. What are some of the possible hazards associated with the organochloride pesticide DDT?

7. Discuss the possible long-term and short-term effects that farm workers may experience when they are exposed to organophosphate pesticides.

8. List examples of metallic compounds that pose environmental hazards. What are some examples of health effects thought to be associated with arsenic, mercury, and lead?

51589_CH13_Printer.indd 591 09/02/13 2:25 PM

592 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

9. Discuss the possible health effects that have been associated with exposure to electric and magnetic fields and ionizing radiation. What methodo- logic difficulties exist with respect to investigations of the health effects of these forms of radiation?

10. Examples of environmental hazards found in the work setting include biologic agents, mineral and organic dusts, vapors, and occupational stress. What are possible roles for epidemiologists in designing research studies to investigate and control exposures to these hazards?

11. Identify challenges and opportunities for epidemiology from the following environmental health problems: hazardous waste sites, air pollution, and nuclear electricity-generating plants.

12. Name three historically important lethal air pollution episodes. What countries or regions of the world are currently faced with extremely poor air quality?

13. Describe two major adverse health outcomes associated with air pollution. Give your own opinions about what can be done to control air pollution.

14. Define the term passive smoking. State why public health officials are concerned about exposure to secondhand tobacco smoke. What is being done to control secondhand tobacco smoke?

15. Describe the 1986 incident in Chernobyl. What health effects have been studied in relation to this event? Have similar events occurred in the United States?

16. Why is the availability of a reliable and safe water source a concern to officials in the United States and the rest of the world? What types of challenges impact the availability of a safe water supply? What is meant by the term TTHM, and why is it relevant to water quality?

17. Regarding the various environmental epidemiologic topics covered in this chapter, name three common challenges that relate to exposure assessment.

References 1. Prüss-Üstün A, Corvalán C. Preventing Disease through Healthy Environments:

Towards an Estimate of the Environmental Burden of Disease. Geneva, Switzerland: World Health Organization; 2006.

2. World Health Organization. Environmental and Occupational Cancers. Fact sheet number 350, March 2011.

51589_CH13_Printer.indd 592 09/02/13 2:25 PM

r e f e r e n C e s 593

3. Friis RH. Essentials of Environmental Health, 2nd ed. Sudbury, MA: Jones and Bartlett Learning; 2012.

4. National Institute for Occupational Safety and Health (NIOSH). The Effects of Workplace Hazards on Female Reproductive Health. DHHS (NIOSH) Publication No. 99–104. Cincinnati, OH: NIOSH; 1999.

5. Wegman DH. The potential impact of epidemiology on the prevention of occupa- tional disease. Am J Public Health. 1992;82:944–954.

6. Lebowitz MD, Burrows B, Traver GA, et al. Methodological considerations of epide- miological diagnoses in respiratory diseases. Eur J Epidemiol. 1985;1:188–192.

7. Monson RR. Occupational Epidemiology, 2nd ed. Boca Raton, FL: CRC Press; 1990. 8. Porta M. A Dictionary of Epidemiology, 5th ed. New York: Oxford University

Press; 2008. 9. Lave LB, Seskin EP. Air pollution and human health. Science. 1970;169:723–733.

10. Lan Q, Zhang L, Li G, et al. Hematotoxicity in workers exposed to low levels of Benzene. Science. 2004;306:1774–6.

11. Klaassen CD, Watkins JB III, eds. Casarett and Doull’s Essentials of Toxicology, 2nd ed. New York, NY: The McGraw-Hill Companies, Inc.; 2010.

12. Thayer KA, Melnick R, Burns K, et al. Fundamental flaws of hormesis for public health decisions. Environ Health Perspect. 2005;113:1271–1276.

13. Selikoff IJ, Cuyler HE, Churg J. Asbestos exposure, smoking, and neoplasia. JAMA. 1968;204:106–112.

14. National Institute of Environmental Health Sciences, National Toxicology Program. Bisphenol A (BPA). Fact sheet, August 2010.

15. Lee, S J, Mehler L, Beckman J, et al. Acute pesticide illnesses associated with off- target pesticide drift from agricultural applications: 11 states, 1998–2006. Environ Health Perspect. 119(8),2011:1162–1169.

16. Centers for Disease Control and Prevention. Spotlight on Organochlorine Pesticides. CDC’s Fourth National Report on Human Exposure to Environmental Chemicals. Atlanta, GA: CDC, 2009.

17. Gray, J. State of the Evidence: The Connection between Breast Cancer and the Environment, 6th ed. 2010. Breast Cancer Fund. http://www.breastcancerfund.org/ assets/pdfs/publications/state-of-the-evidence-2010.pdf. Accessed June 16, 2012.

18. Selikoff IJ, Churg J, Hammond EC. Asbestosis exposure and neoplasia. JAMA. 1964;188:142–146.

19. U.S. Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry, Division of Toxicology and Environmental Medicine. ToxFAQs™. Arsenic CAS number 7440-38-2, August 2007.

20. U.S. Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry. ATSDR Case Studies in Environmental Medicine. http://www.atsdr.cdc.gov/csem/arsenic/docs/arsenic.pdf. Accessed June 17, 2012.

21. Peters GR, McCurdy RF, Hindmarsh JT. Environmental aspects of arsenic toxicity. Crit Rev Clin Lab Sci. 1996;33:457–493.

22. Cantor KP. Arsenic in drinking water: how much is too much? Epidemiology. 1996;7:113–115.

23. Hopenhayn-Rich C, Biggs ML, Fuchs A, Bergoglio R, Tello EE, Nicolli-Smith AH. Bladder cancer mortality associated with arsenic in drinking water in Argentina. Epidemiology. 1996;7:117–124.

51589_CH13_Printer.indd 593 09/02/13 2:25 PM

594 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

24. Powell PP. Minamata disease: a story of mercury’s malevolence. South Med J. 1991;84:1352–1358.

25. U.S. Environmental Protection Agency. 2010 biannual national listing of fish adviso- ries. EPA-820-F-11–014, November 2011.

26. Centers for Disease Control and Prevention. Blood lead levels—United States, 1999–2002. MMWR. 2005;54(20):513–516.

27. Centers for Disease Control and Prevention. Surveillance for elevated blood lead levels among children—United States, 1997–2001. MMWR Surveill Summ. 2003;52(No. SS-10):1–8.

28. Healthy People.gov. Healthy People 2020. Topics and objectives index, environmental health, toxics and waste. http://www.healthypeople.gov/2020/topicsobjectives2020/ pdfs/HP2020objectives.pdf. Accessed June 18, 2012

29. Centers for Disease Control and Prevention, Department of Health and Human Services. Fourth National Report on Human Exposure to Environmental Chemicals, 2009.

30. Draper G, Vincent T, Kroll ME, Swanson J. Childhood cancer in relation to distance from high voltage power lines in England and Wales: a case-control study. Br Med J. 2005;330:1290–1293.

31. Wertheimer N, Leeper E. Electrical wiring configurations and childhood cancer. Am J Epidemiol. 1979;109:273–284.

32. Tomenius L. 50-Hz electromagnetic environment and the incidence of childhood tumors in Stockholm County. Bioelectromagnetics. 1986;7:191–207.

33. Savitz DA, Wachtel H, Barnes FA, et al. Case-control study of childhood cancer and exposure to 60-Hz magnetic fields. Am J Epidemiol. 1988;128:21–38.

34. London SJ, Thomas DC, Bowman JD, et al. Exposure to residential electrical and magnetic fields and risk of childhood leukemia. Am J Epidemiol. 1991;134:923–937.

35. Feychting M, Ahlbom A. Magnetic fields and cancer in children residing near Swedish high-voltage power lines. Am J Epidemiol. 1993;138:467–481.

36. Fulton JP, Cobb S, Preble L, et al. Electrical wiring configurations and childhood leukemia in Rhode Island. Am J Epidemiol. 1980;111:292–296.

37. Thomas DB. Breast cancer in men. Epidemiol Rev. 1993;15:220–231. 38. Ahlbom A, Cardis E, Green A, et al. Review of the epidemiologic literature on EMF

and health. Environ Health Perspect. 2001;109 (Suppl 6):911–933. 39. Tokunaga M, Norman JE, Asano M, et al. Malignant breast tumors among

atomic bomb survivors, Hiroshima and Nagasaki, 1950–1974. J Natl Cancer Inst. 1979;62:1347–1359.

40. Tokunaga M, Land CE, Tokuoka A, et al. Incidence of female breast cancer among atomic bomb survivors, 1950–1985. Radiation Research. 1994;138:209–23.

41. Samet JM, Eradze GR. Radon and lung cancer: taking stock at the millennium. Environmental Health Perspectives. 2000;108:635–641.

42. Zeeb H, Shannon F. eds., WHO Handbook on Indoor Radon: A Public Health Perspective. Geneva, Switzerland: World Health Organization, 2009.

43. U.S. Environmental Protection Agency. Radon health risks. http://www.epa.gov/ radon/healthrisks.html. Accessed June 18, 2012.

44. Turner MC, Krewski D, Chen Y, et al. Radon and lung cancer in the American Cancer Society cohort. Cancer Epidemiol Biomarkers Prev 2011; 20:438–448.

45. Centers for Disease Control and Prevention. Mold: Facts about Mold and Dampness. http://www.cdc.gov/mold/dampness_facts.htm. Accessed June 16, 2012.

51589_CH13_Printer.indd 594 09/02/13 2:25 PM

r e f e r e n C e s 595

46. Centers for Disease Control and Prevention. Health concerns associated with mold in water-damaged homes after Hurricanes Katrina and Rita—New Orleans area, Louisiana, October 2005. MMWR. 2006;55(2):41–44.

47. Centers for Disease Control and Prevention. State-specific unintentional-injury deaths—United States, 1999–2004. MMWR. 2007;56(43):1137–1140.

48. Semenza JC, Rubin CH, Falter KH, et al. Heat-related death during the July 1995 heat wave in Chicago. N Engl J Med. 1996;335:84–90.

49. Baker EL, Matte TP. Surveillance of occupational illness and injury. In: Halperin W, Baker EL Jr, Monson RR, eds. Public Health Surveillance. New York, NY: Van Nostrand Reinhold; 1992.

50. Rutstein DD, Mullan RJ, Frazier TM, et al. Sentinel health events (occupational): a basis for physician recognition and public health surveillance. Am J Public Health. 1983;73:1054–1062.

51. Koh D., Aw T-C. Surveillance in occupational health. Occup Environ Med. 2003;60:705—710.

52. Haagsma JA, Tariq L, Heederik DJ, Havelaar AH. Infectious disease risks associated with occupational exposure: a systematic review of the literature. Occup Environ Med. 2012;69(2):140–146.

53. Dutkiewicz JE, Cisak E, Sroka J, et al. Biological agents as occupational hazards— selected issues. Ann Agric Environ Med. 2011;18(2), 286–293.

54. Lednar WM, Tyroler HA, McMichael AJ, Shy CM. The occupational determi- nants of chronic disabling pulmonary disease in rubber workers. J Occup Med. 1977;19:263–268.

55. Centers for Disease Control and Prevention. Coal workers pneumoconiosis-related years of potential life lost before age 65 years—United States, 1968–2006. MMWR. 2009;58(50): 1412–1415.

56. Attfield MD, Kuempel ED. Mortality among U.S. underground coal miners: a 23-year follow-up. Am J Ind Med. 2008; 51: 231–245.

57. Landrigan PJ. Commentary: environmental disease—a preventable epidemic. Am J Public Health. 1992;82:941–943.

58. Buffler PA, Wood S, Eifler C, et al. Mortality experience of workers in a vinyl chlo- ride monomer production plant. J Occup Med. 1979;21:195–203.

59. Gennaro V, Ceppi M., Crosignani P, Montanaro F. Reanalysis of updated mortality among vinyl and polyvinyl chloride workers: confirmation of historical evidence and new findings. BMC Public Health 2008;8:21.

60. Doll R. Effects of exposure to vinyl chloride: an assessment of the evidence. Scand J Work Environ Health. 1988;14:61–78.

61. McCauley LA, Anger WK, Keifer M, et al. Studying health outcomes in farmworker populations exposed to pesticides. Environ Health Perspect. 2006;114:953–960.

62. Horvath EP. Building-related illness and sick building syndrome: from the specific to the vague. Cleve Clin J Med. 1997;64:303–309.

63. Bardana EJ. Sick building syndrome—a wolf in sheep’s clothing. Ann Allergy Asthma Immunol. 1997;79:283–293.

64. Bourbeau J, Brisson C, Allaire S. Prevalence of the sick building syndrome symptoms in office workers before and six months and three years after being exposed to a build- ing with an improved ventilation system. Occup Environ Med. 1997;54:49–53.

65. Ooi P, Goh K. Sick building syndrome: an emerging stress-related disorder. Int J Epidemiol. 1997;26:1243–1249.

51589_CH13_Printer.indd 595 09/02/13 2:25 PM

596 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

66. Roberts J. New U.S. theory on Gulf War syndrome. Br Med J. 1996;312:1058. 67. Haley RW. Is Gulf War syndrome due to stress?: the evidence reexamined. Am J

Epidemiol. 1997;146:695–703. 68. Roberts J. US responds to new suggestion of Gulf War syndrome. (News) Br Med J.

1996;312:1629. 69. Gray GC, Kang HK. Healthcare utilization and mortality among veterans of the Gulf

War. Phil Trans R Soc B. 2006;361:553–569. 70. Griffith J, Riggan WB. Cancer mortality in U.S. counties with hazardous waste sites

and ground water pollution. Arch Environ Health. 1989;44:69–74. 71. Najem GR, Cappadona JL. Health effects of hazardous chemical waste disposal sites

in New Jersey and in the United States: a review. Am J Prev Med. 1991;7:352–362. 72. Vrijheid M. Health effects of residence near hazardous waste landfill sites: a review of

epidemiologic literature. Environ Health Perspect. 2000;108(Suppl 1):101–112. 73. Geschwind SA, Stolwijk JAJ, Bracken M, et al. Risk of congenital malfor-

mations associated with proximity to hazardous waste sites. Am J Epidemiol. 1992;135:1197–1207.

74. Mueller BA, Kuehn CM, Shapiro-Mendoza CK, Tomashek KM. Fetal deaths and proximity to hazardous waste sites in Washington state. Environ Health Perspect. 2007;115:776–780.

75. Mitchell RS, Judson FN, Moulding TS, et al. Health effects of urban air pollution. JAMA. 1979;242:1163–1168.

76. Lebowitz MD, Holberg CJ, Boyer B, Hayes C. Respiratory symptoms and peak flow associated with indoor and outdoor air pollutants in the southwest. J Air Pollut Control Assoc. 1985;35:1154–1158.

77. Jaakkola JJK, Tuomaala P, Seppänen O. Air recirculation and sick building syndrome: a blinded crossover trial. Am J Public Health. 1994;84:422–428.

78. Mendell MJ, Fine L. Building ventilation and symptoms—where do we go from here? Am J Public Health. 1994;84:346–348.

79. Henderson BE, Gordon RJ, Menck H, et al. Lung cancer and air pollution in south- central Los Angeles County. Am J Epidemiol. 1975;101:477–488.

80. Silverman DT, Samanic CM, Lubin JH, et al. The diesel exhaust in miners study: a nested case-control study of lung cancer and diesel exhaust. J Natl Cancer Inst. 2012;104:855–868.

81. Schimmel H, Greenberg L. A study of the relation of pollution to mortality: New York City, 1963–1968. J Air Pollut Control Assoc. 1972;22:607–616.

82. Dockery DW, Schwartz J, Spengler JD. Air pollution and daily mortality: associa- tions with particulates and acid aerosols. Environ Res. 1992;59:362–373.

83. Kuller LH, Radford EP, Swift DP, et al. Carbon monoxide and heart attacks. Arch Environ Health. 1975;30:477–482.

84. Samoli E, Touloumi G, Schwartz J, et al. Short-term effects of carbon monox- ide on mortality: an analysis within the APHEA Project. Environ Health Perspect. 2007;115:1578–1583.

85. Wright GR, Jewizyk S, Onrot J, et al. Carbon monoxide in the urban atmosphere. Arch Environ Health. 1975;30:123–129.

86. Lebowitz MD, Holberg CJ, Knudson RJ, Burrows B. Longitudinal study of pulmo- nary function development in childhood, adolescence, and early adulthood. Am Rev Respir Dis. 1987;136:69–75.

51589_CH13_Printer.indd 596 09/02/13 2:25 PM

r e f e r e n C e s 597

87. Lebowitz MD. The relationship of socio-economic factors to the prevalence of obstructive lung diseases and other chronic conditions. J Chronic Dis. 1977;30: 599–611.

88. Mannino DM, Watt G, Hole D, et al. The natural history of chronic obstructive pulmonary disease. Eur Respir J. 2006;27:627–643.

89. White JR, Froeb HF. Small-airways dysfunction in nonsmokers chronically exposed to tobacco smoke. N Engl J Med. 1980;302:720–723.

90. US Environmental Protection Agency (EPA). Respiratory Health Effects of Passive Smoking: Lung Cancer and Other Disorders. EPA publication 600/6-90/006F. Washington, DC: EPA; 1992.

91. Pattenden S, Antova T, Neuberger M, et al. Parental smoking and children’s respi- ratory health: independent effects of prenatal and postnatal exposure. Tob Control. 2006;15:294–301.

92. U.S. Department of Health and Human Services. The Health Consequences of Involuntary Exposure to Tobacco Smoke: A Report of the Surgeon General. 6 Major Conclusions of the Surgeon General Report. U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 2006.

93. Boyle P. The hazards of passive—and active—smoking. N Engl J Med. 1993;328:1708–1709.

94. Vineis P, Airoldi L, Veglia F, et al. Environmental tobacco smoke and risk of respira- tory cancer and chronic obstructive pulmonary disease in former smokers and never smokers in the EPIC prospective study. Br Med J. 2005;330:277–280.

95. Hatch M, Wallenstein S, Beyea J, et al. Cancer rates after the Three Mile Island nuclear accident and proximity of residence to the plant. Am J Public Health. 1991;81:719–724.

96. Cardis E, Krewski D, Boniol M, et al. Estimates of the cancer burden in Europe from radioactive fallout from the Chernobyl accident. Int J Cancer. 2006;119(6)):1224–1235.

97. Kazakov VS, Demidchik EP, Astakhova LN. Thyroid cancer after Chernobyl. (Letter) Nature. 1992;359:21.

98. Baverstock K, Egloff B, Pinchera A, et al. Thyroid cancer after Chernobyl. (Letter) Nature. 1992;359:21–22.

99. Hatch M, Ron E, Bouville A, et al. The Chernobyl disaster: cancer following the acci- dent at the Chernobyl nuclear power plant. Epidemiol Rev. 2005;27:56–66.

100. Baverstock K, Williams D. The Chernobyl accident 20 years on: an assessment of the health consequences and the international response. Environ Health Perspect. 2006;114 (9):1312–1317.

101. Dolk H, Lechat MF. Health surveillance in Europe: lessons from EUROCAT and Chernobyl. Int J Epidemiol. 1993;22:363–368.

102. Japan. New York Times. Updated June 17, 2012. http://topics.nytimes.com/top/news/ international/countriesandterritories/japan/index.html. Accessed June 21, 2012.

103. Harman K. Japan’s post-Fukushima earthquake health woes go beyond radiation effects. Scientific American. March 2, 2012. http://www.scientificamerican.com/ article.cfm?id=japans-post-fukushima-earthquake-health-woes-beyond-radiation. Accessed June 21, 2012.

51589_CH13_Printer.indd 597 09/02/13 2:25 PM

598 C h a p t e r 1 3 e p i d e m i o l o g i C a s p e C t s o f W o r k

104. The evacuation zones around the Tokushima Daiichi nuclear plant. New York Times. Updated March 25, 2011. http://www.nytimes.com/interactive/2011/03/16/world/ asia/japan-nuclear-evaculation-zone.html. Accessed June 21, 2012.

105. Clapp R, Cobb S, Chan C, Walker B. Leukemia near Massachusetts nuclear power plant. (Letter) Lancet. 1987;2:1324–1325.

106. Shleien B, Ruttenber AJ, Sage M. Epidemiologic studies of cancer in populations near nuclear facilities. Health Phys. 1991;61:699–713.

107. Spycher BD, Feller M, Zwahlen M, et al. Childhood cancer and nuclear power plants and Switzerland: a census-based cohort study. International Journal of Epidemiology 2011;40:1247–1260.

108. Kaatsch P, Spix C, Jung I, Blettner M. Childhood leukemia in the vicinity of nuclear power plants in Germany. Dtsch Arztebl Int. 2008;105(42):725–732.

109. Ghirga G. Cancer in children residing near nuclear power plants: an open question. Ital J Pediatr. 2010;36:60.

110. United States Environmental Protection Agency. Hanford-Washington. http:// yosemite.epa.gov/r10/cleanup.nsf/0/2f133ac95a7d2684882564ff0078b367!Open Document. Accessed June 22, 2012.

111. Wing S, Richardson DB. Age at exposure to ionizing radiation and cancer mortal- ity among Hanford workers: follow-up through 1994. Occup Environ Med. 2005; 62:465–472.

112. Till JE, Rood AS, Voilleque PG, et al. Risks to the public from historical releases of radionuclides and chemicals from the Rocky Flats Environmental Technology Site. J Expo Anal Environ Epidemiol. 2002;12:355–72.

113. U.S. Environmental Protection Agency. Region 8 Superfund Program. Rocky Flats Plant (USDOE). http://www.epa.gov/region8/superfund/co/rkyflatsplant/. Accessed June 22, 2012.

114. Rallison ML, Lotz TM, Bishop M, et al. Cohort study of thyroid disease near the Nevada test site: a preliminary report. Health Phys. 1990;59:739–746.

115. United States Environmental Protection Agency. Drinking Water and Health: What You Need to Know! EPA 816-K-99–001. Washington, DC: Office of Water; 1999.

116. Brunkard JM, Ailes E, Roberts VA, et al. Surveillance for waterborne-disease out- breaks associated with drinking water—United States, 2007–2008. MMWR Surveill Summ. 2004;60(SS-12):38–68.

117. Newton J, Maugh TH II. Water officials suggest prudence, not panic. Los Angeles Times. February 11, 1998,:1.

118. Swan SH, Waller K, Hopkins B, et al. A prospective study of spontaneous abor- tion. Relation to amount and source of drinking water consumed in early pregnancy. Epidemiology. 1998;2:126–133.

119. Waller K, Swan SH, DeLorenze G, Hopkins B. Trihalomethanes in drinking water and spontaneous abortion. Epidemiology. 1998;9:134–140.

51589_CH13_Printer.indd 598 09/02/13 2:25 PM

3chapte r

14chapte r

599

Molecular and Genetic Epidemiology

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● state the fundamental differences between molecular and genetic epidemiology

●● describe the basic principles of inheritance and sources of genetic variation

●● identify at least three reasons for the familial aggregation of a given disease

●● define epidemiologic approaches for the identification of genetic components to disease

●● explain the basic principles of segregation and linkage analysis ●● state research applications of molecular and/or genetic epidemiol-

ogy in infectious diseases, cancer, and other chronic diseases

CHAPTER OUTLINE

I. Introduction II. Definitions and Distinctions: Molecular Versus Genetic

Epidemiology III. Epidemiologic Evidence for Genetic Factors IV. Causes of Familial Aggregation V. Shared Family Environment and Familial Aggregation

VI. Gene Mapping: Segregation and Linkage Analysis

51589_CH14_Printer.indd 599 09/02/13 2:38 PM

600 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

Introduction

Mapping of the human genome and the subsequent advances in molecular biology forever changed epidemiologic research on disease etiology. (Refer to Figure 14–1 for a photograph of the deoxyribonucleic acid [DNA] helix.) Gone are the days when measurements of exposure are limited to simple interview data, mailed questionnaires, inspection of secondary data, and surrogate mea- sures of the primary exposure of interest. The value of descriptive epidemiol- ogy and disease monitoring and surveillance remain important applications of epidemiology. However, modern epidemiologists find themselves armed with several new strategies to assess precursors of disease, identify biologic markers of exposure, and search for the biologic bases for responses. The wide differential in human responses to the same environmental exposure is an intriguing issue for epidemiology that may be explored by using these advanced techniques.

The traditional epidemiologic approach—characterized by examination of the distribution of health conditions in populations and discerning risk factors for them—has proved useful for generating hypotheses and unraveling disease etiologies. However, suppose that it were possible to go beyond these methods and look inside the “black box” of disease processes. If this black box were to become transparent, epidemiologists would be able to change the definition of risk factors or clarify their location in a causal model.1

This chapter presents an overview of fundamental principles of molecular and genetic epidemiology. For readers without a strong background in biology, we begin with a review of basic principles of human genetics (Exhibit 14–1). Next, we will define the terms genetic epidemiology and molecular epidemiology and distinguish between them. We will then present several epidemiologic approaches to identify genetic components of disease, provide an overview of some strate- gies to identify genes in studies of families, and present a description and recent results from genome-wide association studies (GWAS). The chapter includes

VII. Genome-Wide Association Studies (GWAS) VIII. Linkage Disequilibrium Revisited: Haplotypes

IX. Application of Genes in Epidemiologic Designs X. Genetics and Public Health

XI. Conclusion XII. Study Questions and Exercises

51589_CH14_Printer.indd 600 09/02/13 2:38 PM

i n t r o d u C t i o n 601

Figure 14–1 Model of the DNA helix.

Basic principles of human Genetics

Readers with a prior background in molecular biology and genetics (MBG) may be tempted to skip this section. It is primarily intended to be a refresher for readers who have some familiarity with MBG or who have been away from the topic. However, we include some dis- coveries about the nature of the genetic code with which you may not be familiar. For readers with limited grounding in MBG, this brief review may be insufficient. You are encouraged to pursue addi-

tional details in one of several fine textbooks on human genetics. In order of simplest to most complex, we recommend the following texts:

Mange EJ, Mange AP. Basic Human Genetics, 2nd ed. Sunderland, MA: Sinauer Associates, Inc; 1999.

Speicher M, Antonarakis SE, Notulsky AG, eds. Vogel and Motulsky’s Human Genetics: Problems and Approaches, 4th ed. Heidelberg, Germany: Springer; 2010.

e x

h iB

it 1

4 –1

continues

51589_CH14_Printer.indd 601 09/02/13 2:38 PM

602 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

Singer M, Berg P. Genes & Genomes: A Changing Perspective. Mill Valley, CA: University Science Books; 1991.

The Genetic Code The genetic code is the blueprint of instructions to make our body. These instructions, coded in the form of DNA, are passed on from parents to their children at the time of conception and are recorded using a very sim- ple “alphabet” of only four “letters:” A, C, G, and T. These represent four different nucleic acids—adenine, cytosine, guanine, and thymine—that are used to spell “words,” called codons. Each codon (“word”) contains only three “letters” and represents the codes to construct amino acids.

Although the genetic alphabet contains only four letters (deoxyribonu- cleic acids), they can be combined to code for 64 different codons (amino acids). However, the human body is composed of only 20 amino acids, some of which can be encoded in more than one way. This possibility means that our genetic code is degenerate, meaning that most amino acids can be specified in several ways. The only exceptions to coding in several ways are methionine (TAC) and the least frequent amino acid, tryptophan (ACC), which are encoded in one unique way. In fact, nine amino acids can be coded in two different ways, one can be coded three different ways, five can be coded four different ways, and three can be coded with six different codons. Three codons do not code for an amino acid, but rather signal the end of the gene (ACT, ATT, ATC). The amino acids (“words”) are strung together in long sequences to form “sentences.” These sentences are the complete instructions to make a specific protein in a part of our body— such as skin, hair, red blood cells, bone, nerve spindles—or the enzymes, hormones, and growth factors that regulate our body and make us what we are. A gene is the genetic code corresponding to one “sentence.”

Physical Arrangement of DNA Although the coded instructions for making our bodies are contained in linear sequences of DNA (representing roughly 3 billion bases), these sequences are not one long garbled string of genes. Rather, the units of DNA themselves are organized onto chromosomes. Each human should

exhibit 14–1 continued

continues

51589_CH14_Printer.indd 602 09/02/13 2:38 PM

i n t r o d u C t i o n 603

have 23 pairs of chromosomes: one pair to determine sex and 22 pairs of autosomes.

Each chromosome differs in size and is numbered from 1 to 22. Women have two X chromosomes (one from the mother and one from the father) and men have one X from their mother and a Y chromosome from their father. Transmission of chromosomes from parents to offspring occurs through the formation of gametes during a process called meiosis. For men, meiosis is the production of sperm and occurs throughout life. For women, meiosis leads to the production of oocytes, or eggs. As opposed to men, women are born with their full complement of oocytes, but only one or two are allowed to mature each month during their years of reproduc- tive potential. Normal sperm and oocytes contain only one copy of each chromosome, so at the time of conception a full complement of 23 pairs of chromosomes is formed.

From DNA to Protein There are several important things to know about the human genome. The first point is that not all DNA contained in our cells is transcribed into protein. Second, within a region of DNA on a particular chromosome that codes for a gene, only certain segments are transcribed—the process by which DNA is copied into RNA (ribonucleic acid)—and translated—the pro- cess by which RNA is read and proteins are assembled. That is, the sequence of nucleic acids that determine the order and length of amino acids needed to build a certain protein is not necessarily a straight run. Certain stretches of DNA will be copied (called exons or expressed sequences) and other stretches of DNA will be essentially ignored (called introns or intervening sequences). As an extreme example, the gene that codes for clotting factor VIII (and is mutated in persons with hemophilia) has 26 exons that code for about 2,000 amino acids. However, these codes represent only about 4% of the total length of the gene! The final, and most important, point is that individuals differ from one another in terms of their DNA.

Although all humans can be thought of as having essentially the same number of genes, they clearly do not have identical sequences of DNA. Sev- eral recent discoveries have revealed that the genome is far more complex.

exhibit 14–1 continued

continues

51589_CH14_Printer.indd 603 09/02/13 2:38 PM

604 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

For example, the amount of DNA individuals carry is not identical, with the difference in genomic size between two individuals being as large as 9 million base pairs.2

Genetic Variation How is DNA different from one person to another? Although the complete story is beyond the scope of this chapter, suffice it to say that changes can occur in a wide variety of ways. A mutation is defined as a change in DNA that may adversely affect the host. One category of mutations, known as frameshift mutations, is the result of deletions or insertions of one or more DNA bases. These mutations not only alter the codon in which they occur, but also may shift the reading frame of all successive three-letter words.

Another type of mutation is one that changes the chemical structure of one nucleic acid to that of another. Because most amino acids can be formed from more than one combination of nucleic acids, sometimes mutations can be “silent” and not result in a change in amino acid. For example, a mutation from AAA to AAG would still lead to the incorpora- tion of phenylalanine. Thus, although the sequence of DNA has changed, the amino acid sequence of the transcribed protein has not. Alternatively, the alteration of a single codon can have a profound effect. For example, a mutation of T to A in the middle base of the sixth codon for the b chain of hemoglobin changes the amino acid from glutamine to valine. The result is a change in the shape of hemoglobin from smooth and rounded to dis- torted, and ultimately a disease known as sickle-cell anemia.

Mutations also can occur within introns, the noncoding regions of DNA, which would be expected to have little effect on the protein product of that gene. Recent evidence suggests that even mutations in introns can some- times have a profound effect on the protein product of a gene. The most serious, and easiest to recognize, mutations are ones in which a mutation in a nucleic acid produces an inadvertent “stop” codon that signals the end of transcription before the full-length gene product can be transcribed. The result is a protein that is shorter than normal, or truncated, with a cor- responding effect on its function or integrity.

Alterations can be much larger in scale than a single base pair. Recent studies show that alterations may be as large as one thousand to several

exhibit 14–1 continued

continues

51589_CH14_Printer.indd 604 09/02/13 2:38 PM

d e f i n i t i o n s a n d d i s t i n C t i o n s 605

thousand base pairs.3,4 The concept of “copy number variations” (CNVs) also has been established recently. The concept of CNV refers to a situa- tion in which the number of copies of a gene differs between individuals.1

Interestingly, many of these CNVs have been identified in human genes that reflect senses (smell, hearing, taste, and sight) and disease suscep- tibility. It has been hypothesized that several thousand of these CNVs occur; their presence raises important questions about their public health significance.

Review of Genetic Terminology Having finished this brief review of DNA and genetic variation, we present a few more definitions. The basic unit of heredity is a gene, the particular segment of a DNA molecule on a chromosome that determines the nature of an inherited trait. An allele is one of two or more alternative forms of a gene that occurs at the same locus. Of course, we have not yet defined a locus, either. It is the site or location on a chromosome occupied by a gene (i.e., a particular set of alleles). The genotype of an individual refers to his or her genetic constitution, often stated in reference to a specific trait or at a particular locus. The phenotype is the realized expression of the genotype, or the observable physical appearance or functional expression of a gene. An important situation in which genetics intersects with epidemiology happens when a genotype is modified (or interacts with the environment) to affect a phenotype (disease). Another important term is Mendelian inher- itance (named for its discoverer, the 19th century Austrian monk, Gregor Mendel), which denotes the transmission of a disease or trait from parents to offspring according to simple laws of inheritance. n

exhibit 14–1 continued

illustrative published examples and concludes with an overview of how the field of genetics has and will continue to influence the practice of public health.

Definit ions and Dist inct ions: Molecular Versus Genetic Epidemiology

It is now quite commonplace to be a hyphen-epidemiologist. That is, rarely is it suffi- cient to describe oneself as a “simple country epidemiologist” anymore. Many in the field add some sort of modifier to their title, for example, pharmaco-epidemiologist,

51589_CH14_Printer.indd 605 09/02/13 2:38 PM

606 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

behavioral-epidemiologist, or neuro-epidemiologist. To this partial list we must add molecular-epidemiologist and genetic-epidemiologist, terms that many people use interchangeably. We describe in which respects these two fields overlap, how they differ, and how technological advances in high throughput genotyping are reunit- ing molecular and genetic epidemiology. (Throughput refers to the amount of work that can be performed in a given period of time.)

Genetic Epidemiology The field of genetic epidemiology is devoted to the identification of inherited fac- tors that influence disease, and how variation in the genetic material interacts with environmental factors to increase (or decrease) risk of disease. The first textbook on the subject defines it as a “discipline that seeks to unravel the role of genetic factors and their interactions with environmental factors in the etiol- ogy of diseases, using family and population study approaches.”5 An important premise is that a better understanding of the genetic etiology of disease can facili- tate early detection in high-risk subjects and the design of more effective inter- vention strategies.6 The unifying theme of genetic epidemiology is the focus on genes and evidence for genetic influences.

Note that to answer questions two through four, families (or at least pairs of relatives) were historically required. The approach of using related persons is

1. Does the disease of interest cluster in families? 2. Is the clustering a reflection of shared lifestyle, common environ-

ment, or similar risk factor profiles? 3. Is the pattern of disease (or risk factor for a disease) within families

consistent with the expectations under Mendelian transmission of a major gene? Described in more detail later in the chapter, Mendelian transmission refers to the inheritance of characteristics in accord with Mendel’s laws of inheritance.

4. Where is the chromosomal location of the putative gene? n

Genetic Epidemiology Can Be Thought of as a Collection of Methodologies Designed to Answer

Four Questions:

51589_CH14_Printer.indd 606 09/02/13 2:38 PM

d e f i n i t i o n s a n d d i s t i n C t i o n s 607

quite different from traditional epidemiology, which assumes that the subjects under study are independent. When study subjects are biologically related, by definition they are no longer independent. Lack of independence necessitates special rules in selection of subjects and analytic approaches. The epidemiologic approach to identify genetic factors that influence disease does not require prior knowledge about the pathophysiologic process that underlies the inherited sus- ceptibility. Rather, given that the clustering of a disease in families is not due to shared environment, and is consistent with Mendelian transmission of a major gene, the goal is to identify regions of DNA that cosegregate (are inherited in the same pattern) with the disease of interest. Once the chromosomal region is nar- rowly defined, the torch is passed to molecular geneticists to identify the appro- priate gene using highly specialized techniques. This approach, called positional cloning or physical mapping, contrasts with a more traditional laboratory approach called functional cloning or functional mapping. The latter does not require epi- demiology or families and is based instead on identification of proteins that are involved in a disease process. Once a protein has been identified, scientists can determine its amino acid sequence. Working backward, the researcher is able to decipher the DNA code for the sequence of amino acids. Finally, the investigator finds where this DNA sequence occurs in the human genome. Note that physi- cal mapping has historically only been applied when the genetic influence on disease is great enough that there will be a Mendelian pattern of disease in the family. These contrasts in approaches are depicted in Figure 14–2, adapted from

Figure 14–2 Strategies to identify human genes. Source: Adapted from FS Collins, Positional Cloning Moves from Perditional to Traditional, as published in Nature Genetics, Vol 9, pp. 347–350, 1995.

Disease

Gene

FunctionMap

Disease

Gene

FunctionMap

Functional cloning Positional cloning

51589_CH14_Printer.indd 607 09/02/13 2:38 PM

608 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

a review on the subject by Dr. Francis Collins.7 Collins led the Human Genome Project, an ambitious public/private venture to determine the full sequence of the roughly 3 billion nucleotides in our DNA, and served as Director of the National Human Genome Research Institute. GWAS have clearly demonstrated that they can identify susceptibility loci with modest effect sizes. Drilling down to find the actual gene in a region of interest has traditionally been done in families. However, with advances in genotyping abilities, the ability to carefully refine the causal genetic variant is now possible through the study of unrelated individuals. Thus, the once clear lines between molecular and genetic epidemiol- ogy are beginning to blur somewhat.

Molecular Epidemiology Basically, a greater precision in estimating exposure–disease associations can be made by using molecular biology to improve the measurement of exposures and disease. The term molecular epidemiology has been attributed to researchers Perera and Weinstein. Molecular epidemiology has the possibility of providing early warnings for disease by flagging preclinical effects of exposure.8 The field is much broader than genetic epidemiology and includes a wide variety of biologic mea- sures of exposure and disease. As it relates to her research on the causes of cancer, Perera noted that molecular epidemiology combines “advances in the molecular biology and molecular genetics of cancer with epidemiology to understand the molecular dose of specific agents, their preclinical effects, and the biologic fac- tors that modulate susceptibility to their exposure.”9(p 233) Many definitions of molecular epidemiology include the concept of biomarkers.10 Consider the fol- lowing examples:

●● Rather than rely on individual recall of a usual diet to classify individuals according to intake of fruits and vegetables, assess serum levels of micro- nutrients to obtain more precise measurements of intake of fruits and vegetables.

●● Rather than conduct a clinical trial with colon cancer as the end point, use an intermediate marker (an accepted precursor lesion: the adenomatous polyp).

●● Rather than treat all cases of breast cancer as the same disease, use tumor markers to identify potentially more heterogeneous subsets.

●● In trying to identify whether clusters of cases of infectious disease are from a common source, characterize the agents according to their DNA fingerprint.

51589_CH14_Printer.indd 608 09/02/13 2:38 PM

e p i d e M i o l o G i C e v i d e n C e f o r G e n e t i C f a C t o r s 609

From these examples, the reader should notice that the markers (exposures) are based on biologic specimens (e.g., blood, tissue, urine, and sputum) rather than questionnaire or medical records data. As stated earlier, the terms molecu- lar epidemiology and genetic epidemiology are often used interchangeably. One reason is that molecular epidemiology commonly measures inherited variation in DNA (as opposed to acquired variation—somatic mutations—in our DNA) to classify subjects. Thus, when genes are involved, there is an overlap between molecular and genetic epidemiology. One distinction between the two is that molecular epidemiology does not involve studies of biologically related indi- viduals. Another distinction is that most molecular epidemiologic studies are conducted to evaluate the significance of variation in genes that would not neces- sarily manifest as Mendelian patterns of disease in a family. As molecular biolo- gists and molecular geneticists work to unravel disease processes, they discover various proteins that are involved. Genes determine these proteins; if individuals differ from one another in the genetic sequence of a protein that is functionally involved in the disease process, then evaluation of this genetic variation in epide- miologic studies could yield important insights into disease etiology.

Epidemiologic Evidence for Genetic Factors

If a disease has a genetic component, since close relatives of a case have a certain probability of sharing the same gene that influences risk of disease, there should be an excess occurrence of disease in that family. From an etiologic perspective, measurement and evaluation of family history as a risk factor may shed light on the contribution of “familial” factors on the pathogenesis of the outcome of interest. A simple definition of a positive family history is the occurrence of the same disease or trait within a family. A more precise definition would include the

Genetic epidemiology: concerned with inherited factors that influence risk of disease Molecular epidemiology: uses molecular markers (in addition to genes) to establish exposure–disease associations n

Molecular Versus Genetic Epidemiology

51589_CH14_Printer.indd 609 09/02/13 2:38 PM

610 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

specific types of relatives that will be considered, for example, first-degree rela- tives (parents, siblings, and offspring) or second-degree relatives (grandparents, aunts and uncles, nieces and nephews, grandchildren) plus specifics on the disease (such as the type of cancer) and their age at onset (since familial disease is gener- ally held to have an earlier age at onset than disease unrelated to genetic factors).

Several epidemiologic designs might be employed to evaluate the associa- tion of family history with disease. A cross-sectional survey of a representative sample of the population could assess the frequency of respondents with a posi- tive history. However, if the frequency of the disease of interest was rare, this method would not be very efficient. A more common strategy is to conduct a case-control study and compare the frequency of family history in both groups. If there was a genetic component to the disease, one would expect the odds of a positive family history of the disease to be greater among the cases than among the controls.

As with any case-control study, recall bias is always a potential problem with family studies. Family recall bias is the special situation where cases are more likely to be informed about their family history than are controls.11 It is not dif- ficult to imagine that, as a consequence of having the disease, one learns much greater detail about affected family members than had been known prior to dis- ease onset. An approach to overcome family recall bias is to perform a cohort study in which assessment of family history occurs at baseline, prior to the onset of disease. The cohort is then followed prospectively for the development of the outcome under investigation. A disadvantage, of course, is that the length of the follow-up period could be extensive before sufficient cases accrue for meaningful analysis. In addition, family history is not a static characteristic but a dynamic risk factor that can change with time as unaffected relatives develop the outcome under investigation. Assessment of family history at a single point in time cannot capture such changes.

Causes of Famil ial Aggregation

Although demonstration that a disease or trait clusters in families is certainly acceptable evidence that genetics may be important, several alternative explana- tions must be considered. The explanations include the operation of chance and the influence of environmental factors. Zhao and colleagues have presented the case that the series of inquiries that characterize genetic epidemiology can be considered as part of a sequence of studies, built upon a common epidemiologic framework.12

51589_CH14_Printer.indd 610 09/02/13 2:38 PM

C a u s e s o f f a M i l i a l a G G r e G a t i o n 611

Bad Luck The first of the explanations for the familial aggregation of disease is simply chance. Given that there is a finite probability for the development of a particu- lar disease or adverse health phenomenon, even in the absence of any genetic contribution at all, the disease may afflict several members of the same family. This occurrence is especially prevalent for the common diseases of major public health importance, such as obesity, mental illness, heart disease, and cancer. For example, an early study of the aggregation of cancer used cancer mortality rates for the adult British population to show that half of all families with more than five adults would have at least one case of cancer by chance alone.13 This is con- ceptually similar to the concept of clustering, but with the grouping based on family rather than space or time.

Two other factors—not necessarily a simple reflection of “bad luck”—would affect the likelihood that someone reports a positive family history: age and fam- ily size. Although every person has at least two first-degree biologic relatives (our parents), the person’s numbers of siblings, aunts, uncles, cousins, and children are clearly random variables. For example, someone who has 15 close relatives at risk for a common chronic disease is more likely to have a relative with the disease than a person with only two relatives at risk. Similarly, because age is the single most important risk factor for many diseases of public health importance, an older person (with older relatives) is more likely than a younger person (with corresponding younger relatives) to have a family history for nongenetic reasons. Consequently, adjustment for age and family size in the analysis is encouraged when one is trying to assess the association of family history with risk of a par- ticular disease.

Bad Environment Epidemiologists historically have devoted their energies and attention to the identification of nongenetic risk factors for disease. As illustrated throughout the other chapters of this book, history is replete with many success stories. This information on exposure–disease relationships must be considered as explana- tion for any observed clustering of disease in families. For example, although roughly 95% of all lung cancer cases are current or former smokers, fewer than 20% of heavy smokers ever develop the disease.14 This observation has caused some to hypothesize that host factors might influence response to environmen- tal agents (tobacco). A complicating factor in the study of a disease with such a strong environmental risk factor is the extent to which family members also

51589_CH14_Printer.indd 611 09/02/13 2:38 PM

612 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

smoke cigarettes. Twin studies conducted in the 1930s provided strong evidence that smoking habits clustered in families.15 Therefore, if family members of cases were more likely to smoke than family members of controls, a greater propor- tion of cases would be expected to report a positive family history of lung cancer. In this situation, however, clustering of cases in families could be due to shared lifestyle, rather than shared genes.

Studies of migrants from low-risk to high-risk countries clearly support the notion that aspects of diet are associated with coronary heart disease and can- cer. Diets low in complex carbohydrates and fruits and vegetables are associated with diabetes and cancer. Several studies suggest the familial nature of dietary intake patterns.16,17 To the extent that family members share dietary habits that increase risk for a given disease, familial clustering of disease may occur.

Similarly, health-related behaviors, such as exercise practices, alcohol intake, and use of sunscreen, may be learned within a family and indirectly relate to clustering of disease. Exercise levels are related to avoidance of overweight, pro- motion of bone strength, and cardiovascular health. Moderate alcohol consump- tion, especially intake of red wine, is thought to be protective for the heart; however, overconsumption is associated with adverse health outcomes and car crashes among drivers who are driving under the influence. Finally, avoidance of excessive sun exposure reduces the occurrence of many forms of skin cancer. The foregoing are examples of some of the many health-related practices that may be transmitted within the family environment.

Many risk factors shared by family members are a reflection of shared envi- ronment, such as water supply, radon from the soil, air quality, pesticides, lead paints, and even occupation. A case report of a family with four members with mesothelioma, a rare cancer of the lining of the peritoneal cavity, was traced back to a common occupational exposure to asbestos.18 Infectious diseases also may be included in this category, and there are a number of published examples in which familial clustering of hepatitis or tuberculosis occurs from common exposures.

Shared Family Environment and Famil ial Aggregation

A difficulty in the interpretation of “family history” data is the inability to deter- mine the influence of nongenetic risk factors on any observed familial clustering. From an etiologic perspective, measurement and evaluation of risk factors are

51589_CH14_Printer.indd 612 09/02/13 2:38 PM

s h a r e d f a M i l y e n v i r o n M e n t 613

Proband—the individual in a family who brings a disease of interest to the attention of the investigator.

necessary in order to answer the question, “Is the observed clustering due to (shared) environmental factors?” A family study can be defined as one in which data on phenotype (observable physical appearance associated with a gene) and risk factors are measured on individual family members.

Design of Case-Control Family Studies As with all epidemiologic studies, several issues must be considered regard- ing selection of case and control families. In genetic jargon, a proband is the individual in a family who brings a disease of interest to the attention of the investigator.

●● Step 1: Ascertainment of cases (probands) and controls ●● Step 2: Enumeration of the relatives of the probands and controls n

As Opposed to Most Types of Epidemiologic Study Designs, Two Steps (Rather Than Only One) Are Required to Establish the Sampling Frame for a

Case-Control Family Study:

In a case family, the proband is likely to be a person affected with the dis- ease (although one can select families on the basis of an unusual family history among healthy study participants). The proband in a control family is the sub- ject matched to the case. The definition of family also must be clearly stated: Three-generation families (or more) are most valuable for elucidation of genetic mechanisms, but this comes at the expense of less complete data and greater difficulty validating medical histories and collecting biological specimens. For diseases with late ages at onset, parents might be deceased and offspring will not be informative, owing to their youth. Therefore, some study designs include only siblings of probands as families.

51589_CH14_Printer.indd 613 09/02/13 2:38 PM

614 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

The ascertainment considerations might include:

1. From where will the probands be drawn? A random sample of the population might be considered for a common condition (high preva- lence) but is inefficient for rare diseases. If the researcher is fortunate, there might be a disease registry that contains all incident cases in a defined geographic region. In many situations, one is restricted to ascer- tainment of probands from hospitals and clinics, which raises questions about the representativeness of the probands if only specialized facilities are canvassed. Probands may be identified also through death certifi- cates,19 although this selection method would complicate the collection of biologic samples from the index case (unless there were stored samples collected previously).

2. Does the disease or trait require medical attention? It is not difficult to envision a number of important public health problems that may have a significant genetic component (e.g., alcohol or drug abuse), but the con- dition either does not require medical attention or required medical care is not sought. In those situations, it is virtually impossible to determine whether all eligible probands have been identified.

3. Is the prevalence of the disease known? When the incidence or prevalence of the condition under study is known, one has a much greater likelihood of evaluating whether or not the sampling frame of probands includes all eligible cases with the disease or condition.

The purpose of control families derives from the need to be able to evaluate whether familial clustering among cases is greater than can be expected. This consideration is especially important when there are no population-based rates to calculate the expected number of disease events in case families. Control families also are needed to be able to rule out common familial (measured) exposure as an explanation for any observed familial aggregation. Control families must be as similar as possible to the case families for all other (unmeasured) environmental exposures in order to determine properly whether or not a disease or trait truly has a genetic component (Exhibit 14–2).

Control families can be identified from several possible sources. One would be the same disease registry, clinic, or hospital from which the probands were drawn, but with sampling based on a different disease. Another choice is ran- dom selection from the general population (e.g., neighborhood controls). Relatives of the proband’s spouse have been used as controls since the beginning of the 20th century. This clever approach “matches” families on unmeasured

51589_CH14_Printer.indd 614 09/02/13 2:38 PM

s h a r e d f a M i l y e n v i r o n M e n t 615

Family Study of Lung cancer risk in Southern Louisiana

Carcinoma of the lung has been traditionally cited as an example of a malignancy solely determined by the environment.20 Indeed, the risks associated with cigarette smoking are well documented.

Numerous studies also have found that lung cancer cases are sig- nificantly more likely than controls to report a positive family his- tory of the disease.16 Before concluding that lung cancer clusters in

families as a result of a genetic predisposition, it is important to control for the fact that smoking itself “runs in families.”21 Southern Louisiana has some of the highest lung cancer mortality rates in the country.22 Epi- demiologic studies designed to identify the basis for these high rates were unable to attribute the excess rates to high prevalence of tobacco expo- sure or industrial exposures, such as those that occur in shipbuilding,23 oil refining,24 and sugarcane farming.25

To investigate the hypothesis that a genetic predisposition was opera- tive, Ooi and colleagues conducted a case-control family study.19

Probands were deceased lung cancer patients identified from a listing of all deaths attributed to lung cancer in a 10-parish (county) area over a four-year period (1976–1979). A total of 440 case probands were identi- fied. Control probands were identified as the spouse of the case (if he or she had ever been married), as listed on the death certificate. Telephone interviews were conducted to construct pedigrees of parents, siblings, offspring, and half-siblings of the cases and controls. Interviews also col- lected data on cancer history, current age (or age at death), age at onset for affected persons, smoking histories, and occupational exposures. Par- ticipation rates were equivalent for case and control families (76%). The study reported that cases were 2.4 times more likely than controls to have a first-degree relative with lung cancer. After excluding the probands and spouses, statistical models were fitted to predict the risk of lung cancer in relatives. The models adjusted for age, sex, pack-years of cigarette smoking, occupational/industrial exposures, and a variable to reflect whether or not a study participant shared genes with the lung cancer proband (i.e., was a case relative or a control relative). After adjusting for the established risk

e x

h iB

it 1

4 –2

continues

51589_CH14_Printer.indd 615 09/02/13 2:38 PM

616 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

lifestyle, economic status, education, and even religious influences, because likes tend to marry likes.27

Although the primary focus of this section has been on the selection of fami- lies through a proband with a disease of interest, other strategies may be con- sidered. For example, a random sample of families from a cross-sectional study might be considered if one is interested in the genetic epidemiology of a com- mon biologic trait or process, such as steroid hormone levels or eating behaviors.

Gene Mapping: Segregation and Linkage Analysis

Historically, when the number of genetic markers available was few and expen- sive to type, compelling evidence of a possible genetic link to a disease or condi- tion was required to justify the efforts to map complex diseases. In particular, the disease or trait had to be shown to cluster in families and not be accounted for by shared nongenetic risk factors before a hypothesis of familial aggregation of an underlying genetic influence could be made. As we explain later in the chapter, the current ease and cost efficiency of genotyping is revolutionizing the field. Nevertheless, in this section we present the traditional approach that is still commonly used for rare monogenic (single gene) traits or syndromes for which nongenetic influences are less evident.

For studies of genetic transmission, families or pairs of related individuals are required. Ideally, one would want a population-based sample of families ascer- tained through probands diagnosed in a defined geographic area over a specific time period because this sampling frame would improve the precision of the esti- mate of the frequency of the risk allele. If one had conducted a case-control fam- ily study to rule out shared environment as the cause of familial aggregation, then

factors for lung cancer, relationship to a lung cancer proband remained a significant predictor of risk (odds ratio [OR] = 2.4). Risk was especially elevated for parents (OR = 4.4) and nonsmoking sisters of the probands (OR = 4.6). Further analysis of the data suggested that susceptibility to cancer included sites besides the lung, notably other smoking-associated cancers.26 n

exhibit 14–2 continued

51589_CH14_Printer.indd 616 09/02/13 2:38 PM

G e n e M a p p i n G : s e G r e G a t i o n a n d l i n k a G e a n a l y s i s 617

the case families from that investigation could be subjected to further analysis. Segregation analysis is an approach to determine, from a sample of families, whether a particular disease or trait is inherited in Mendelian fashion. By Mende- lian, we mean the situation in which the mode of transmission of a disease or trait from parents to offspring is consistent with simple laws of inheritance. (Refer to Figure 14–3 for an image of Gregor Mendel, who is credited with specifying the laws of inheritance.) Mendelian inheritance is defined in the next section.

Modes of Inheritance The mode of inheritance refers to the association between the number of mutated alleles at a locus and a given phenotype (disease or trait) and whether or not the locus is situated on one of the 22 autosomes or on a sex chromosome. Consider the simple situation in which there are only two possible alleles at a locus, B and b. Furthermore, suppose that allele B is associated with disease. Because humans have two alleles at a locus, the three possible genotypes are therefore BB, Bb, or bb. An individual’s genotype is determined by the genotypes of his or her parents and the particular two alleles that were passed on at meiosis. According to the principles of Mendelian inheritance, the alleles transmitted from parents to offspring are ran- domly determined. Consequently, a parent whose genotype is Bb will transmit, on average, the B allele 50% of the time and the b allele the other 50%. Parents who are either BB or bb are capable of transmitting only the B and b alleles, respectively.

Figure 14–3 Gregor Mendel. Source: From National Institutes of Health, Deciphering the Genetic Code. Available at http://history.nih.gov/exhibits/ nierenberg/popup_htm/01_mendel.htm. Accessed November 14, 2007.

51589_CH14_Printer.indd 617 09/02/13 2:38 PM

618 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

Another way of stating Mendelian transmission is in terms of the probability that the deleterious gene is passed from a parent to child. The expectations that BB, Bb, and bb parents transmit a B allele are 1.0, 0.5, and 0.0, respectively. Figure 14–4 illustrates a Punnett square, named after the English geneticist Reginald C. Punnett, who devised the approach to determine the probability of an offspring's having a particular genotype. The Punnett square is a tabular summary of every possible combination of one maternal allele with one paternal allele for each gene being studied and thus is a visual representation of Mendelian inheritance.

Some diseases are referred to as autosomal dominant or recessive. The term autosomal dominant refers to the situation in which only a single copy of an altered gene located on a non-sex chromosome is sufficient to cause an increased risk of disease. In these situations, one typically expects to see an affected par- ent and, on average, half of his or her offspring affected with the same disease or condition. Autosomal recessive diseases denote those for which two copies of an altered gene are required to increase risk of disease. Carriers, individuals who have only one copy of the altered gene (heterozygotes), are typically not thought to be at increased risk of the disease. Matings of two carriers will produce, on average, affected children one-fourth of the time.

Figure 14–4 Punnett square showing possible outcome of offspring when both parents have Bb genotypes. Since bB and Bb are identical, the ratio of the genotypes is 1:2:1.

Genetic contribution of the 1st parent

BB Bb

bB bbb

Possible genotypes of the offspring

Genetic contribution of the 2nd parent

51589_CH14_Printer.indd 618 09/02/13 2:38 PM

G e n e M a p p i n G : s e G r e G a t i o n a n d l i n k a G e a n a l y s i s 619

Traits are said to be genetically additive or codominant when heterozygotes have a phenotype that is distinguishable from the two homozygous states. Codominance may occur for both quantitative and qualitative traits. Codomi- nance is easy to imagine for a quantitative trait—such as blood pressure or enzyme activity—that facilitates categorizing individuals into one of three levels (say high, medium, or low enzyme activity). For a qualitative trait (e.g., presence or absence of a type of cancer), the effect may be on the age at onset of disease. Homozygous carriers would have an earlier mean age at onset than heterozygotes, who in turn have an earlier mean age at onset than noncarriers of an altered allele. Although disease-predisposing genes may be carried on the X or the Y chromo- some, there are very few known examples of the latter. X-linked traits are passed from women to sons or daughters, but from men only to daughters (because fathers pass their Y chromosome to sons). X-linked recessive traits affect men much more frequently than women, because men need only have the gene on their one X chromosome mutated to be in the recessive state at that locus.

Determining the Mode of Inheritance Elucidation of the potential mode of inheritance of susceptibility to a com- plex disease is determined by comparing how well different hypotheses (genetic and nongenetic) fit the observed pattern of disease in a collection of families.28 Known as segregation analysis, the goal of the analysis is to determine whether a particular disease or trait is inherited in Mendelian fashion. For simple Mende- lian traits one can inspect the family pedigree and calculate a Punnett square to determine whether inheritance is dominant or recessive (See Figure 14–5).

Complex traits are characterized by pedigrees in which one cannot determine the pattern of inheritance by visual inspection. This inability can arise from sev- eral factors, such as the trait being influenced by age, environmental factors, or a modest influence on risk (often described as reduced penetrance). A variety of computer software programs are available to perform this analysis (e.g., some acronyms are SAGE, PAP, and POINTER). Coverage of these various options is beyond the scope of this chapter, as is the mathematical underpinnings showing how the various hypotheses are constructed. The concept, however, is basically as follows. In the traditional realm of science, hypotheses are generated and data are collected and analyzed to accept or refute the hypothesis. With segregation anal- ysis, the outcome of the experiment (random mating and assortment of genes from parents to offspring) has already occurred (individuals who inherit the dele- terious gene have an increased risk for disease). Thus, one is essentially asking the question, “Given the pattern of disease in these families, what is the likelihood

51589_CH14_Printer.indd 619 09/02/13 2:38 PM

620 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

that the underlying cause was transmission of an altered allele in a Mendelian fashion?” Models representing possible modes of transmission (genetic and non- genetic) are tested for goodness-of-fit (how well they fit the data). The goodness- of-fit is compared for the various models to determine which one is the most likely explanation for the observed pattern of disease. Exhibit 14–3 describes the findings of a segregation analysis of breast cancer.

Logarithm of the Odds (LOD) Score Linkage Analysis The results of the Claus et al.29 analysis (Exhibit 14–3) provided strong evi- dence, but not definitive proof, for the existence of a gene that influenced risk of breast cancer. Linkage analysis is an attempt to identify a DNA marker that co-segregates with the disease of interest and is considered strong evidence for the existence of a gene.28 The basis for linkage analysis is the simple concept that there are exceptions to Mendel’s law of independent assortment of traits. Specifically, genes that are in close physical proximity to one another on the same chromosome tend to be linked (i.e., inherited together). Two genes on different chromosomes are unlinked and will be inherited together roughly 50% of the time. It is important to emphasize that two genes on the same chromosome will

Figure 14–5 Illustration of Mendelian inheritance as evidenced by disease patterns in families.

AFFECTED FATHER

AFFECTED female

AFFECTED male

Autosomal Dominant • 50% chance of being affected

Autosomal Recessive • 25% chance of being affected • 50% chance of being a carrier

D n

D nn n n n N N d dN dN d

n n N d N d

NORMAL MOTHER

NORMAL male

CARRIER male

CARRIER female

NORMAL female

CARRIER FATHER

CARRIER MOTHER

51589_CH14_Printer.indd 620 09/02/13 2:38 PM

G e n e M a p p i n G : s e G r e G a t i o n a n d l i n k a G e a n a l y s i s 621

be linked only if they are close together on the chromosome. That is because dur- ing meiosis (the formation of gametes), homologous pairs of chromosomes can undergo recombination through “crossing over” of the DNA strands. Although earlier in the chapter we said that parents transmit half of their chromosomes to their offspring, the shifting of genetic material across homologous chromosomes often modifies the chromosomes. If genes are far apart on the same chromosome, there is a high probability that a crossover event will occur, and the genes will end up on different chromosomes. Genes that are close together on the same chromosome have a much lower probability of being separated during meiosis by a recombination (crossover) event. Therefore, the probability of a recombina- tion event is a function of the distance between two loci. In a given number of meioses, the proportion that yields a recombinant chromosome is defined as the recombination fraction (q). If the number of recombinants between two loci is estimated through analysis to be quite small (or zero), then the two loci are “linked” and the physical location of the disease-causing gene possibly has been identified. In linkage analysis between two loci, the question of whether they

Genetic epidemiology of early-Onset Breast cancer in the caSh Study

The Cancer and Steroid Hormone (CASH) study is a multicenter, population-based case-control study conducted by the Centers for Disease Control and Prevention. Researchers studied a total of 4,730 histologically confirmed cases of breast cancer among patients (aged 20–54 years) and 4,688 controls. Initial analyses confirmed that cases were significantly more likely than controls to

have a family history of the disease, especially the earlier the age at onset.30

Segregation analysis of the case families provided evidence that the pat- tern of breast cancer was consistent with Mendelian dominant transmis- sion of a rare allele (carrier frequency of 3 women per 1,000) associated with increased breast cancer risk.29 The proportion of cases predicted to carry the allele was highest (36%) among women aged 20 to 29 years but decreased to only 1% of cases aged 80 or older. The cumulative lifetime risk among carriers was predicted by the model parameters to be high: approximately 92%. For noncarriers the predicted lifetime risk was roughly 10%, essentially the same as the lifetime risk for the general population. n

e x

h iB

it 1

4 –3

51589_CH14_Printer.indd 621 09/02/13 2:38 PM

622 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

are linked is answered by comparing the value of the observed recombination fraction to the expected recombination fraction (under no linkage) of 0.5. Evi- dence for linkage is expressed as a ratio of the likelihood of the data under link- age at some specified recombination fraction to the likelihood of the data under a recombination fraction of 0.5. The results are typically expressed as an LOD (logarithm of the odds) score, which stands for the logarithm to the base 10 of the odds of linkage. A logarithm to the base 10 is the number to which the base 10 can be raised in order to produce a given number. For example, 10 may be written as 101, 100 as 102, 1000 as 103, and so forth. An LOD score of 3.0 is equivalent to a P value (probability value or significance level) of 103, or 1,000 to 1; this evidence is commonly accepted for significant linkage.

Earlier in the chapter we described ways in which specific genetic alterations can occur. Note that all of these were presented in reference to a specific gene. However, variation in DNA also occurs throughout the genome. Regardless of whether or not they occur in coding regions of DNA, base pair sequences that are highly polymorphic (i.e., that vary greatly from individual to individual) can be used to help characterize individuals as unique at a particular locus and thus can be tracked within families. Highly polymorphic, closely spaced mark- ers that cover each of the human chromosomes thus facilitate linkage analysis. Exhibit 14–4 presents an example of a linkage study.

As noted, the ultimate goal of linkage analysis is to identify the chromosomal location of a disease locus. Samples of families selected for linkage analysis are usually highly biased. Although they may be a subset of case families initially ascertained as part of a population-based case-control study, not all of the case families will be informative for linkage analysis. That is, only families with mul- tiply affected individuals are useful for attempts to identify genetic markers that co-segregate with disease.

For many diseases, evidence suggests that if two forms of the disease exist, one genetic and the other not, the genetic form of the disease will tend to have an earlier age at onset. Accordingly, researchers often prefer to study families with multiple cases of disease, especially when those affected have an age at onset ear- lier than the general population average. Finally, it is imperative that the affected individuals be living, available, and interested in participating in the research study. Linkage analysis is particularly susceptible to misclassification errors, so it is critical to validate all disease end points of interest and check the pedigrees for nonpaternity. If one is trying to track markers from parents to offspring, then it makes no sense to compare the transmission of alleles when one of the parents (usually the father) is not really biologically related.

51589_CH14_Printer.indd 622 09/02/13 2:38 PM

G e n e M a p p i n G : s e G r e G a t i o n a n d l i n k a G e a n a l y s i s 623

Linkage analysis of early-Onset Breast cancer

One of the most exciting success stories in genetic epidemiologic research was the identification of two genes—BRCA1 and BRCA2— through linkage analysis and positional cloning. Mary-Claire King and colleagues were the first to report that autosomal domi- nant susceptibility to breast cancer was due to a gene located on a region of the long arm of chromosome 17.31 This finding

was based on 23 families: 7 with an average age at onset younger than 45 years, and 16 families in which average age at onset was older than 45 years. Two of the early-onset families also included women with ovar- ian cancer. Thus, they also were considered to be part of the phenotype in the analysis. Further analysis with more markers in the 17q12-17q21 region revealed that significant evidence for linkage was limited to the seven early-onset families.32 The markers used were dinucleotide repeat polymorphisms: anonymous sequences of variable length repeats of DNA (CA)n. The most informative marker (D17S579) included 12 alleles of size 111 to 133 base pairs. The initial report was soon confirmed,33 and an international consortium was formed to pool results and clone the gene. After a furious competition, others reported the successful identification of the BRCA1 gene on chromosome 17.34 Because this locus appeared to account for roughly half of the families included in the linkage consortium, work continued to identify a second locus. Roughly a year later, a second gene (BRCA2) was identified on chromosome 13 by restricting analysis to families that included male and female breast cancer.35

Identification of these two genes has only been the beginning for genetic epidemiologic investigations. Considerable work has been done since their discovery to characterize the sites and types of mutations, the complete phenotype (cancers other than breast and ovarian), ethnic variation in allele frequency, and ethical and psychosocial issues related to genetic test- ing. Subsequent research has implicated these two genes in repair of DNA damage.36 Data have begun to emerge on the effectiveness of surgical and other prevention strategies.37 n

e x

h iB

it 1

4 –4

51589_CH14_Printer.indd 623 09/02/13 2:38 PM

624 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

It is important to understand that the phenomenon of genetic linkage is a function of the distance between the disease locus and the marker locus. Genetic linkage is not necessarily an association of a specific allele at the marker locus with disease. For example, in Exhibit 14–4, suppose that the marker locus (Mfd188) that was found to be most closely linked to the BRCA1 locus had four alleles: a, b, c, and d. The premise is that Mfd188 is close to the disease locus but is not the disease locus. This premise means that the breast cancer locus will co-segregate with allele a in some families. However, in other families, the relevant marker may be allele b, or c, or d. As a result, within a family there will be co-segregation of a particular allele at the marker locus with the disease locus. Across families, any one of the four alleles at the marker locus may co-segregate with the BRCA1 mutation. The only time one would expect to see the same marker associated with disease risk across families is when the marker is within the disease-causing gene itself and is closely linked to the actual mutation in the gene. This situa- tion, in which a specific allele at the marker locus is strongly associated with the mutant allele, is known as linkage disequilibrium. As shown in Figure 14–6, sup- pose a new mutation is generated during meisosis on a particular chromosome. Over time, recombination events create new combinations of the ancestral allele. However, note that combinations of alleles that are close together can still iden- tify the ancestral chromosome that first harbored the mutation. The importance of this phenomenon will become increasingly clear in the later section on GWAS.

Figure 14–6 Illustration of the population genetics concept of linkage disequilibrium.

Initial mutation (+) occurs on a chromosomal background (shaded)

Areas retaining LD

with + mutation

+ + +

+ +

+ + + +

+ +

(Many generations)

51589_CH14_Printer.indd 624 09/02/13 2:38 PM

G e n e M a p p i n G : s e G r e G a t i o n a n d l i n k a G e a n a l y s i s 625

Nonparametric Linkage Analysis Historically, linkage analysis has required knowledge about the mode of inheritance of the disease. Thus, a formal segregation analysis would be done to see if the pattern of disease in the family (or collection of families) was consistent with Mendelian transmission. In addition, segregation analysis provided esti- mates of important parameters that are essential for linkage analysis, such as the frequency of the disease-associated allele in the population and the age-specific risks (penetrance) of the gene. Penetrance refers to the probability that a gene or genetic trait is expressed. For example, “complete” penetrance means the gene or genes for a trait are expressed in all the population who have the genes. “Incom- plete” penetrance means the genetic trait is expressed in only part of the popula- tion. Although all of these parameters can theoretically be estimated through formal segregation analysis, it can be extremely difficult to ascertain the true genetic model for a common disease with both genetic and environmental influ- ences (described above as a “complex trait” in genetic terms). Consequently, methods have been developed that do not require specification of the parameters of the underlying genetic model (hence the term nonparametric). Initially, they were based on sibling pairs38,39 but now have been extended to include addi- tional (and more distant) relatives.40

One of the great attractions of sib-pair approaches is their simplicity. Although siblings will share an average of 50% of their genes identical by descent (IBD), at any given locus 2 siblings may share 0, 1, or 2 alleles in common. IBD means that the allele comes from the same parent. If an allele is very common in the population, siblings might share the same allele but have the allele transmitted from different parents. Such alleles that are the same but come from different parents are said to be identical by state. This complexity underscores the impor- tance of highly polymorphic markers. For a qualitative trait (presence or absence of a given disease), under the hypothesis of a genetic influence, one would expect affected sibs to share more alleles IBD than under random Mendelian segrega- tion. Similarly, for a quantitative trait, one would expect that siblings with similar trait values would share more alleles IBD than siblings with dissimilar trait values.

A limitation of sib-pair analysis occurs in the following situations: One can- not unambiguously determine the number of alleles shared IBD. This ambiguity is especially an issue when the marker locus is not highly polymorphic. When there are few alleles at a marker locus, the probability increases that alleles are identical by state rather than IBD. The uncertainty can be resolved if parents are available for genotyping at the locus, but only if the parent is heterozygous at that locus and the non-shared alleles are different. Analysis methods have been

51589_CH14_Printer.indd 625 09/02/13 2:38 PM

626 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

developed for sibling pairs that utilize multiple loci, thus improving the power and efficiency of gene mapping efforts.41 The method works for both qualitative (yes/no) and quantitative (continuous) traits and has been implemented in a computer package known as MAPMAKER/SIBS, but several others are avail- able, such as SIBLINK, SPLAT, GASP, and SOLAR.

Genome-Wide Associat ion Studies (GWAS)

As mentioned earlier, the distinction between molecular epidemiology (which does not require related individuals) and genetic epidemiology (which does) has become irrevocably blurred because of two important developments. The first is the International HapMap Project, which is described in more detail below. The second pertains to radical advances in genotyping capability. Such advances have enabled the cost per genotyping on some chip-based platforms to fall to frac- tions of a cent per single nucleotide polymorphism (SNP). SNPs denote minor variations in our genome; although many SNPs do not cause any alteration in cell functions, other SNPs increase vulnerability to disease (Figure 14–7). These developments have had a profound impact on the epidemiologic community, creating greater pressure for team science (large consortia to pool data and biologic samples across studies) and stretching the capabilities to manage, store, and ana- lyze data. (Refer to Exhibit 14–5 for information on GWAS on breast cancer.)

Genome-Wide association Studies of Breast cancer

Throughout this chapter, we have used examples from the breast cancer literature to illustrate the various genetic epidemiologic study designs. The traditional approach resulted in the identification of several major genes (i.e., BRCA1, BRCA2, and p53), but there was a general appreciation that more genes remained to be discovered.

Recently, investigators reported three GWAS that offer proof-of- principle of the merit of the segregation/linkage approach. Easton and col- leagues49 used a three-stage design. They began by genotyping 408 breast cancer cases with a strong family history of the disease and 400 cancer-free controls on a panel of 227,876 SNPs. The second stage involved 12,711 SNPs

e x

h iB

it 1

4 –5

continues

51589_CH14_Printer.indd 626 09/02/13 2:38 PM

G e n o M e - W i d e a s s o C i a t i o n s t u d i e s ( G W a s ) 627

typed on 3,990 cases and 3,916 controls. After statistical analysis, the top 30 SNPs were then typed in 22 additional case-control studies representing 21,860 cases of invasive breast cancer, 988 cases with carcinoma in situ, and 22,578 controls. Six SNPs were significant at P ≤ .00001 in the same direction as the first two stages: 5 of the 6 SNPs were in genes or LD (linkage disequilib- rium) blocks containing genes: FGFR2, TNRC9 (2 SNPs), MAP3K1, and LSP1. None of these loci had been reported previously as breast cancer risk factors. The risk SNP in the most strongly associated gene, FGFR2, is common in the population (estimated frequency of 0.38), and although associated with a modest increase in risk for breast cancer (OR = 1.63; 96% confidence interval [CI] = 1.53-1.72), the population-attributable risks are significant.

The findings implicating FGFR2 were replicated in a report published by Hunter and colleagues50 one month after Easton et al.’s study.49 The study by Hunter et al. began with 1,183 postmenopausal breast cancer cases and 1,185 individually matched controls from the Nurses, Health Study typed for 528,173 SNPs. The researchers then genotyped an additional 1,776 cases and 2,072 controls nested within three other prospective cohort studies: NHS-2, PLCO, and the American Cancer Society Cancer Prevention Study II.

The SNP in FGFR2 most strongly associated with risk in the initial scan maintained a highly significant result when pooled across all four studies (P = 4.2 × 10−10). Interestingly, the result was obtained using a different SNP than was typed in the study by Easton et al.

The third GWAS study was reported by Stacey et al.51 They genotyped approximately 300,000 SNPs among 1,600 breast cancer cases and 11,563 controls from Iceland. Selected SNPs were then typed in five replication sample sets from Sweden, Spain, and Holland as well as European Ameri- cans from the Multi-Ethnic Cohort Study. The strongest findings identified a region of chromosome 2q35, but the SNP was not within a gene and no genes were located within the identified haplotype block.

However, the results did implicate TNRC9, which was also identified in the study by Easton et al.

These recent reports have sent scientists in new directions to identify the relevant genes and alterations. As exciting as these developments have proven to be, many researchers are predicting that the lifespan of GWAS is probably going to be short, owing to the imminent promise of the ability to sequence the entire genome. n

exhibit 14–5 continued

51589_CH14_Printer.indd 627 09/02/13 2:38 PM

628 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

Figure 14–7 Illustration of single nucleotide polymorphisms—small variations in the genome. Source: Reproduced from U.S. Department of Energy, Office of Science, Joint Genome Institute. Available at: http://www.jgi.doe.gov/ education/genomics_3.html. Accessed July 18, 2007. ©2006 The Regents of the University of California.

GCA

Ala

AGA

Arg

GAT

Asp

AAT

Asn

TGT

AGA GAT AAT TGT

AAA GAT AAT TGT

Cys

GCG

Ala

Arg

Asp

Asn

Cys

Ala Lys Asp Asn Cys

GCA

1 2 3 4 5

Gene A from Person 1

Gene A from Person 2

Codon change made no difference

in amino acid sequence.

Codon change resulted in a different amino

acid at position 2.

Protein Products

...

Gene A from Person 3

Linkage Disequil ibrium Revisi ted: Haplotypes

In the section on genetic linkage, we emphasized that DNA markers that are close together are often inherited together, because crossover events between homolo- gous chromosomes are rare. This phenomenon is used to localize (map) disease susceptibility genes to chromosomal regions. Increasingly, scientists have recog- nized that the human genome is highly variable in the frequency with which such crossover events are observed within and among loci and populations. (For more detail, refer to several reviews.42–44) Moreover, when a new mutation arises, this happens on the background of a particular chromosome, and several DNA mark- ers nearby will continue to be inherited with it for many generations. This com- bination of DNA markers along the chromosome is referred to as a haplotype. A landmark study by Gabriel and colleagues45 examined genetic markers across 51 autosomal regions in DNA samples from Africa, Europe, and Asia. Their analysis of these markers provided compelling evidence of sizable regions of the genome where there is little historical evidence for crossover events during meiosis (recom- bination). They termed these regions “haplotype blocks” and noted the particular

51589_CH14_Printer.indd 628 09/02/13 2:38 PM

l i n k a G e d i s e q u i l i b r i u M r e v i s i t e d : h a p l o t y p e s 629

effect the regions could have on studies to identify susceptibility genes across the genome. In particular, the identification of SNPs that “tag” a region means, essentially, that you would be capturing the genotypes at other loci within that block that were co-inherited with the block. Thus, one could scan across the entire genome with far fewer genetic markers than would be required in the absence of haplotype blocks (which has been estimated to be around 10 million SNPs).

Figure 14–8 shows results from a recent report (refer to Permuth-Wey et al., 2011) by one of the authors (TS) that tested the hypothesis that SNPs within molecules that regulate DNA translation influence risk of ovarian cancer.46 This particular figure shows a haplotype block on chromosome 1 that is the location of LIN28. The boxes indicate the correlation between the SNPs, and three dis- tinct DNA blocks are distinguished.

The International HapMap Project This ambitious project began officially at a meeting held from October 27 to 29 in 2002. A collaboration among scientists and funding agencies in Japan, the United Kingdom, Canada, China, Nigeria, and the United States, their goal was to develop a haplotype map of the human genome and generate 200,000 to per- haps a million SNPs that would “tag” (represent) the most common haplotypes in all human populations. When completed, this public catalog would allow researchers to determine chromosomal similarities and differences worldwide as well as to seek genetic connections between human diseases and responses to pharmaceuticals.

DNA samples came from 270 individuals from the Yoruba people in Ibadan, Nigeria (30 trios of 2 parents and a child), Japanese (45 unrelated individuals), Han Chinese from Beijing (45 unrelated individuals) as well as 30 parent–child trios collected by the Centre d’Etude du Polymorphisme Humain (CEPH). The CEPH families are residents of the United States with Northern and Western European ancestry. The goal was to generate a map of 600,000 SNPs evenly spaced across the genome, such that there is one SNP for every 5,000 DNA bases across the 3 billion bases that comprise the entire genome. The data are publicly available, and updates are released in regular intervals (see the project website at http://www.hapmap.org). The website is logically organized into three main sec- tions: 1) an overview of the project (including more details on the background, ethical issues, protocols, and publications arising from the project); 2) a data section that includes downloads of the data and interactive access; and (3) useful Internet links. The reader can download and read a tutorial that provides very helpful instructions on using this resource.

51589_CH14_Printer.indd 629 09/02/13 2:38 PM

630 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

Block 2 (5 kb)

26600 k 26610 k

LIN28

26620 k

26700 k

26500 k

LIN28

Block 3 (4 kb)Block 1 (12 kb)

rs 11

24 79

rs 12

72 28

rs 94

38 62

rs 75

52 06

rs 17

16 39

rs 11

24 79

rs 11

58 17

rs 65

98 96

rs 12

72 89

rs 35

01 55

rs 94

38 62

rs 66

97 41

rs 17

26 19

83 90 73 99 96 91 80 62 37

98 93 92 92 89 35 3 99

97 12 93 84 88 45 0 74

3 72 88 40 58 88 56 80

70 87 54 38 90 89

80 28 27 31 8 88

89 62 63 21 58

56 36 57 54

39 25 26

58 99 95 99 96 74 36

3 4 5 6 7 8 9 10 11 12 13

26600 k

chr1 (p36.11)

1a12p31.1 43a41

Figure 14–8 Illustration of the concept of haplotypes, or sets of DNA polymorphisms that tend to be inherited together. Source: Permuth-Wey et al. Cancer Res 2011; 71(11):3896–3903.

Application and Implications of the HapMap Project The epidemiologic approach to the application of genetics to human disease was initially based on candidate genes. This approach requires some understanding of the biology of the disease (at least enough to identify proteins that are likely to be

51589_CH14_Printer.indd 630 09/02/13 2:38 PM

a p p l i C a t i o n o f G e n e s i n e p i d e M i o l o G i C d e s i G n s 631

involved) and information that the genes encoding the proteins are polymorphic. The latter requirement is no longer an issue because the National Center for Biotechnology Information (NCBI; available at http://www.ncbi.nlm.nih.gov/) maintains an up-to-date database of SNPs that have been identified across 43 organisms (including Homo sapiens). As of May 2012, there are over 14 million reported SNPs in humans, of which over 12.5 million have been validated. The downside, of course, is incomplete knowledge of a disease and, thus, important genes may be entirely missed. Genome-wide approaches can be considered unbi- ased to any prior information about the disease or trait, making these approaches complementary to the candidate gene (or candidate pathway) approach.

There are several important issues to consider in genome-wide studies. The efficiency of this design is informed by power (true positives), sample size, false positives, and cost. Results from GWAS to date have led to the conclusion that most variants that contribute to common diseases confer small relative risks; con- sequently, large sample sizes are required. Although the cost per genotype may be low, the cost per chip (and therefore per study subject) is still hundreds of dollars. This high expense and need for large samples have led to methodologic research on multistage designs (e.g., Boddeker and Ziegler47), whereby at the first phase a subset of subjects are genotyped for the full genome-wide panel of SNPs, and a subsequent phase is conducted in which fewer SNPs (including the most interest- ing markers identified in the first phase) are typed in a larger sample. This larger sample may be an independent sample set or one in which subjects from the first phase are carried forward and included. Given the large number of SNPs (any- where from 300,000 to 5,000,000 per subject), very stringent levels of statistical significance have been set to rule out false positive associations.48 Regardless, it has become increasingly evident that GWAS have contributed to the identification of hundreds of susceptibility loci for many diseases and traits. Indeed, a catalog of published findings is maintained by the National Human Genome Research Insti- tute and can be found online at http://www.genome.gov/gwastudies/.

Applicat ion of Genes in Epidemiologic Designs

The previous sections of this chapter outlined how genetic epidemiologists deter- mine whether a disease has a genetic component, and how they contribute to efforts (primarily by geneticists) to map genes through segregation and linkage analysis. The primary utility of epidemiology will be realized after the etiologi- cally important genes are identified. According to Friend, “As we enter this post- genomic era, we should expect more and more emphasis on something that could

51589_CH14_Printer.indd 631 09/02/13 2:38 PM

632 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

not previously be studied: the gene–gene and gene–environmental interactions that we have always suspected would end up being important.”52(p 17) The next section begins with a brief review of the recent accomplishments of the Human Genome Project and concludes with a series of examples of the types of epide- miologic studies that can be conducted with knowledge of genetics. Figure 14–9 symbolizes the far-reaching impact of the Human Genome Project.

The mapping of the human genome53,54 has been heralded as the signal accomplishment that will transform our understanding of biology and the prac- tice of medicine. Collins has enumerated several of the potential medical impli- cations this mapping may enable, including improved diagnosis, prognosis, and treatment of disease.55 Without question, increased knowledge of genetics and the incredible advances in technology also will have a profound impact on the field of epidemiology. First, the complete sequence of the 3.3 billion nucleotides comprising the genome is available over the Internet (http://genome.ucsc.edu/). This includes the location and nearly complete sequence of the 26,000–31,000 protein-encoding genes.56 As of March 2006, the reference sequence was con- sidered to be “finished,” a technical term indicating that the sequence is highly accurate (with fewer than one error per 10,000 bases) and highly contiguous (with the only remaining gaps corresponding to regions whose sequence cannot

Figure 14–9 Influence of the Human Genome Project. Source: Reproduced from the U.S. Deparment of Energy Human Genome Progam. Available at: http://genomics.energy.gov/gallery/basic_genomics/detail.np/ detail=30.html. Accessed January 24, 2008.

51589_CH14_Printer.indd 632 09/02/13 2:38 PM

a p p l i C a t i o n o f G e n e s i n e p i d e M i o l o G i C d e s i G n s 633

be reliably resolved with current technology). Although the full benefit of this milestone will not be realized for years, the initial discovery revealed some startling facts:

●● Less than 5% of the genome appears to code for actual genes, with the vast majority emerging as regions that regulate DNA.

●● The size of the human genome is only slightly larger than that of some plants.

●● Certain sequences of DNA appear to be highly conserved (highly invari- ant) across species (yeast, worms, fruit flies, mice, etc.). These sequences are thought to be highly important.

●● Human DNA is highly polymorphic, meaning that it has many forms. Aside from monozygotic twins, each person has a virtually unique DNA sequence.

Applications of Molecular Epidemiology to the Study of Infectious Diseases Use of molecular techniques in infectious disease epidemiology is continuing to grow. Maslow et al.57 noted the types of questions that infectious disease epide- miologists might want to answer, questions that can best be answered through applications of molecular epidemiology rather than traditional epidemiology:

●● Several patients on a surgical ward develop postoperative pneumonia that is caused by the same organism, Klebsiella pneumoniae. Does this occur- rence represent an outbreak?

●● Does a patient who returns to a clinic with a second infection for Esch- erichia coli suffer from a new infection or a relapse caused by the original organism?

●● A patient who has a prosthetic cardiac device is found to have blood cul- tures that are infected with Staphylococcus epidermidis. Is the infection caused by a single strain or by multiple contaminants?

By using methods derived from immunology, biochemistry, and genetics, it is possible to determine whether a bacterial disease (from the same organism) is caused by multiple bacterial isolates or by identical isolates. (This issue relates to all of the foregoing questions.) To illustrate, Hlady et al. described the use of molecular epidemiologic techniques to identify the source of an outbreak of Legionnaires’ disease.58 They conducted a case-control study of five infected patients who attended conventions at a hotel in Orlando, Florida. Nearly all

51589_CH14_Printer.indd 633 09/02/13 2:38 PM

634 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

conventioneers were probably exposed at least briefly to a decorative fountain in the hotel, ultimately determined to be the source of the infections. Statistical techniques could not identify an association between the disease and the foun- tain due to the small number of cases and the fact that many control subjects also had been near the fountain. However, it was possible to confirm an associa- tion between exposure to the fountain and Legionella by using two techniques, monoclonal antibody subtyping and pulsed-field gel electrophoresis, which are molecular biologic strategies that can assess if the cases were affected by identical bacterial strains.

Blanc et al. illustrated how the use of molecular markers helped to control an epidemic of Pseudomonas aeruginosa.59 An outbreak of infections due to this organism was linked to contaminated bronchoscopes. Two clones of the organ- ism were apparently identified in the outbreak, an epidemic and a non-epidemic form. When manual disinfection of the endoscopes was reintroduced, the inci- dence of the epidemic form declined but the incidence of the non-epidemic form did not. Investigators used a form of molecular typing known as ribotyping to distinguish between the sporadic and epidemic cases.

Applications of Molecular Epidemiology in Occupational and Environmental Epidemiologic Research Use of biomarkers helps enlarge our understanding of the etiology and preven- tion of occupationally related health problems.60 Of particular interest would be longitudinal studies in which the predictive value of biomarkers for disease is established.9 Cohorts of workers who have been exposed to hazardous materials, such as benzene or asbestos, could be followed prospectively. Other groups that have had environmental exposures to hazardous wastes or toxic chemicals could be studied also in a similar prospective manner. Clearly, the marriage of detailed exposure records with stored biological specimens (e.g., DNA) can be used for historical cohort designs, too.

The application of techniques from molecular epidemiology to the study of occupational and environmental health problems could yield several important dividends61:

●● improvements in the classification of exposures ●● more accurate definition of risk groups through the use of susceptibility

markers ●● increased specificity in the classification of disease ●● greater understanding of etiologic mechanisms for disease

51589_CH14_Printer.indd 634 09/02/13 2:38 PM

a p p l i C a t i o n o f G e n e s i n e p i d e M i o l o G i C d e s i G n s 635

Groopman et al.10 pointed out that epidemiologists are concerned with the potential associations between exposure variables and health effects. Before the advent of molecular epidemiology, it was not feasible to quantify accurately the relationships among exposure, dose, and health outcomes. Molecular biomarkers may be used to denote exposures, outcomes, or vulnerability to disease because they are more sensitive indicators than other measures of exposure.

An example of the use of a biomarker in environmental health studies is the p53 gene in liver tumors.62 The p53 gene is thought to suppress tumor forma- tion. Researchers in China examined the association between mutations in the gene and exposure to the hepatitis B virus and aflatoxin. They found that half of a small sample of liver cancer patients indeed had acquired mutations in this gene in their tumor. This research and related studies suggest that hazardous exposures may induce genetic changes that in turn are linked to cancer or other diseases.

Applications of Molecular Epidemiology to the Study of Cancer The complex nature of cancer makes it an ideal target for the use of molecular biology and molecular genetics to improve our understanding of disease etiology. Consider this brief list of possibilities:

●● Rather than rely on diet recall to assess fat intake, use biopsy of human tis- sue to determine the fatty acid composition.

●● Rather than consider cancer a simple dichotomy, use understanding of the pathologic changes from normal to malignant cells to study early stages of disease.

●● Rather than use cancer recurrence (or incidence) in studies of prevention, utilize precursor lesions, such as adenomatous polyps of the colon.

●● Rather than consider cancer at a particular site a single disease, use expres- sion of tumor markers (e.g., like Ki67 that depict the rate at which cells are dividing, RAS oncogenes, or cellular receptors for hormones like estrogen or progesterone) as evidence of distinct etiologic pathways.

As a brief example of molecular epidemiologic approaches to the study of cancer, consider the work by Taylor and colleagues at the National Institute of Environmental Health Sciences.63 Previous studies of leukemia had found only weak associations of occupational and chemical exposures with risk of the disease. Taylor et al. hypothesized that, because acute myeloid leukemia (AML) appears to be a heterogeneous disease at the molecular and cytogenetic level, certain

51589_CH14_Printer.indd 635 09/02/13 2:38 PM

636 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

environmental exposures might be linked to a specific molecular subtype. The net effect would be a strong association between an exposure with a particular subtype and a weak association with the remainder. Animal studies suggested that mutations in the ras family of proto-oncogenes are a frequent early event in chemically induced tumors. (The term ras protein refers to a protein that is associated with cell multiplication and differentiation; proto-oncogenes promote normal cell growth and division.) Moreover, ras mutations were known to occur in 15–30% of patients with AML.

Taylor et al. conducted a case-control study of 62 cases and 630 controls.63 This study revealed that cases with mutations in H-ras, K-ras-2, or N-ras were 6.8 times more likely to have worked at least 5 years in high-risk occupations than ras-negative cases, and 9 times more likely to have breathed chemical vapors on the job. When compared with the control group, the associations with occu- pational exposures were observed only among the subset of cases with ras muta- tions in their bone marrow cells. This landmark study was one of the first to demonstrate that disease etiology may be better understood if epidemiologic measures of exposure are integrated with molecular assays of the genetic defects responsible for cancer initiation and promotion.

Molecular and Genetic Epidemiology of Alzheimer’s Disease Alzheimer’s disease (AD) is a progressive, degenerative disorder that attacks the brain's nerve cells, or neurons, resulting in loss of memory, thinking and lan- guage skills, and behavioral changes. In approximately 50% of families, the disease is inherited in an autosomal dominant manner.64 The Mendelian forms of AD are caused by rare and usually highly penetrant mutations in three genes (APP, PSEN1, and PSEN2), all of which alter production of the amyloid-b pep- tide (Ab), the principal component of b-amyloid in senile plaques.65 Although probably several additional disease-causing genes remain to be identified for this type of AD, early-onset familial AD accounts for only < 5% of all AD cases. Early linkage studies implicated a region on chromosome 19.66 Evidence suggests that the relevant gene on chromosome 19 is the one that codes for apolipoprotein E (Apo E). Apo E is a constituent of plasma lipoproteins that participates in the transport of cholesterol and specific lipids. It is involved in degeneration and regeneration of nervous tissue and is found in high concen- tration in brain and cerebrospinal fluid. The fact that it is genetically highly polymorphic makes it a suitable candidate gene for AD. The three common alleles have been examined in a number of case-control studies. As one exam- ple, consider the work of Tsai et al.67 These researchers identified 77 patients

51589_CH14_Printer.indd 636 09/02/13 2:38 PM

a p p l i C a t i o n o f G e n e s i n e p i d e M i o l o G i C d e s i G n s 637

with late-onset AD from the Mayo Clinic Alzheimer’s Disease Patient Registry (operated by the Division of Community Internal Medicine). For each case, the next age- and sex-matched, cognitively normal person presenting to the same Division was recruited as a control. All subjects received a general medical examination, comprehensive neurologic examination, and rigorous neuropsy- chologic evaluation. A small sample of blood was drawn as a source of DNA. Consistent with earlier studies, the investigators found that the frequency of the e4 allele was significantly higher among cases than controls (0.351 vs. 0.130, P, 0.0001). Cases were 4.6 times more likely to carry at least one copy of the e4 allele than controls (CI: 1.9–12.3). Moreover, carriers were found to have significantly earlier age at onset than noncarriers. Thus, the Apo E-e4 allele appears to be an important genetic susceptibility risk factor for AD in the gen- eral population. Importantly, with the advent of GWAS, additional common susceptibility alleles that influence risk of nonfamilial forms of AD are being identified, shedding new insights on the biology of the disease and raising hopes for effective interventions.68

Molecular and Genetic Epidemiology of Psychiatric Disorders We noted earlier in this chapter that genetic epidemiology (in comparison with molecular epidemiology) is more concerned with the inherited basis for health outcomes. A field known as psychiatric genetic epidemiology takes several approaches to uncover the role of inheritance and environmental factors in dis- orders. The identification of familial aggregations of psychiatric disorders among close family members provides evidence for a possible genetic component.69 Additional support comes from studies that find a greater concordance of psy- chiatric illness among monozygotic (identical) than dizygotic (fraternal) twins. A more recent development has been to identify genetic markers for disorders through studies of pedigrees and sibling pairs.

Psychiatric disorders appear to result from complex mechanisms that are influenced by genes, but few present in simple Mendelian fashion. Therefore, elucidation of the genetic component is complicated by the fact that such dis- orders are likely heterogeneous (different genetic basis in different families), polygenic (multiple genes may influence the trait), and multifactorial (genes and environment matter). In recent years, a number of studies have reported linkage in psychiatric disorders, but with few findings replicated. For example, a type of affective disorder, bipolar disorder (BPAD; also known as manic depres- sion), appears to have an inherited basis, but the mode of inheritance is not well defined. Susceptibility to manic-depressive illness has been linked to the

51589_CH14_Printer.indd 637 09/02/13 2:38 PM

638 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

tyrosine hydroxylase locus within the chromosome 11p15 region using both parametric and nonparametric approaches.70 BPAD has been linked to chromo- some 18q and chromosome 21, but the regions remain poorly defined.71 The disease has been associated also with the dopamine transporter gene locus on chromosome 5.72 A Genetic Analysis Workshop (GAW10) convened 25 groups to analyze 5 distributed published data sets containing chromosome 5 and chro- mosome 18 markers. Results were suggestive, but not definitive, for linkage to a bipolar susceptibility locus on chromosome 18; evidence for chromosome 5 was minimal.73

By further clarifying the genetic component for these conditions, it may be possible to further specify high-risk groups and to clarify the environmental risk factors for BPAD and other affective disorders. Nevertheless, progress in this field of research is impeded by the fact that the brain is a very complex organ and that several genes may be implicated in the inheritance of mental disorders.

Genetics and Public Health

Although few will question the importance of new discoveries in the basic sci- ences, their true value lies in their ultimate translational potential in improved health. To a certain extent, the benefit of knowledge gleaned from the revolution in molecular genetics and the mapping of the human genome has not been real- ized. However, the potential has been recognized and is considerable given the role of genes in the etiology of many common diseases.74 Khoury commented that it is up to public health professionals to harness the utility of genetic infor- mation and technology for disease prevention.75 Khoury grouped the range of activities into three categories:

1. Assessment of the impact of genes and their interactions with modifiable disease risk factors on the health status of the population.

2. Development of policies as to when and how genetic tests are to be applied in disease prevention programs.

3. Assurance that the public health genetic programs developed are effective and of the highest quality.

Potential applications include, at a minimum, the following areas: screening for genetic susceptibility, early detection of disease, more appro- priate targeting of high-risk population subgroups for interventions, more

51589_CH14_Printer.indd 638 09/02/13 2:38 PM

G e n e t i C s a n d p u b l i C h e a l t h 639

tailored prevention approaches that recognize underlying differences in host susceptibility, and new therapies and treatments for disease. In an effort to demonstrate this potential, we will give examples related to a common theme: colorectal cancer.

Colorectal cancer is the second most common cause of cancer death in the United States; men statistically are more vulnerable than women after age 50.76 There were an estimated 143,460 cases of colorectal cancer in the United States in 2012, with 51,690 deaths.77 Colon cancer is a part of several well-defined cancer-predisposing genetic syndromes: familial adenomatous polyposis (FAP) and Lynch syndrome.78 FAP is caused by inherited mutations in the APC gene. The cancer family syndrome defined by Lynch is caused by mutations in one of four other genes (MLH1, MSH2, MSH6, and PMS2) and is associated with elevated risks of colon, rectal, stomach, small intestine, liver, gallbladder duct, upper urinary tract, brain, skin, and prostate cancer. Women with this disorder also have a high risk of cancer of the endometrium (lining of the uterus) and ovaries. Researchers estimate that roughly 2–7% of all colorectal cancers in the United States are due to mutations in 1 of these 5 genes associated with autoso- mal dominant susceptibility to cancer.

Screening for Colorectal Cancer Colon cancer is a preventable disease because it has an easily identifiable precur- sor lesion, the adenomatous polyp. Through regular sigmoidoscopy and colo- noscopy, these early lesions can be detected and removed before they become cancerous. Sigmoidoscopy is effective only for polyps that arise in the distal colon and is therefore not a perfect screening tool. Although colonoscopy enables inspection of the entire colon, it is too expensive to be used as a routine screen- ing modality and presents some inherent risks (e.g., perforation of the colon occurs in rare instances). However, the expense and the risk are not excessive for high-risk populations, where the prevalence of disease justifies the procedure. Guidelines define appropriate screening modalities for colorectal cancer based on age and family history of the disease. Because susceptibility is transmitted in an autosomal dominant fashion, on average only 50% of the offspring of affected individuals will inherit the susceptibility gene. If genetic analysis for mutations in the known colorectal cancer genes were included as part of the risk assess- ment, then the intense screening regimen would be required to be applied only to gene carriers within a given high-risk family. Family members without the inherited mutation would require only the screening intervals suggested for the

51589_CH14_Printer.indd 639 09/02/13 2:38 PM

640 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

general population. Thus, resources are targeted to those truly at greatest risk, and unnecessary screening (and risk) can be avoided.

An alternative approach for screening makes use of the following observation: Genes that influence inherited susceptibility in a minority of the population are the same genes targeted for acquired mutations in the general population. Colonic epithelial cells that acquire these mutations are occasionally shed into the lumen of the colon and excreted in the feces. Ahlquist and colleagues at the Mayo Clinic developed an assay to extract DNA from stool. Using a panel of markers, they were able to identify correctly 91% of the samples as being derived from cancer cases, 82% of the samples as being derived from patients with pre- cursor lesions (adenomatous polyps), and 93% of the samples as being derived from cancer-free controls.79 Thus, knowledge of genetics also can benefit the screening of average-risk populations.

Interventions for High-Risk Populations The public health approach posits that shifts in the population distribution of risk factors will lead to substantial changes in disease frequency. Although these shifts should continue to be the goal to which we in public health aspire, the current reality is that there are too many risk factors, too many worthy interventions, not enough intervention professionals, and not enough financial resources to apply this model routinely. A useful adjunct to the population approach is to focus energies on particular high-risk subgroups for interven- tions. Consider our example of colorectal cancer screening. Once carriers of a deleterious mutation are identified, several options for disease prevention can be considered. One would simply be to perform closer surveillance and remove any polyps as they occur. For patients with mutations in the APC gene, the colon can literally be carpeted with thousands of polyps. The solution for these patients is removal of the colon before cancer can occur. For subjects who inherit mutations in one of the other susceptibility genes, a viable option may be chemoprevention. Sulindac has been found to reduce the occurrence of polyps in high-risk individuals.80 Note the case of persons who have a family history of colon cancer and who do not have a genetic mutation in one of the syndromic genes. They may have a family history because of either undesirable lifestyle (physical inactivity, diets low in fruits and vegetables or high in red meat) or a polymorphism in a gene with a more moderate effect on risk. Such individuals would still be candidates for interventions, particularly for diet and exercise changes.

51589_CH14_Printer.indd 640 09/02/13 2:38 PM

G e n e t i C s a n d p u b l i C h e a l t h 641

Tailored Interventions Many epidemiologists traditionally have ignored the possibility that individuals respond differently to the challenges in the environment. Genetic epidemiology is predicated upon the notion first stated by Galen in 200 ad: “But remember throughout that no cause is efficient without a predisposition of the body itself. Otherwise, external causes which affect one would affect all.” Indications that individuals differ in response to an intervention have been around for decades. For example, Garrod, the founder of human biochemical genetics, and Haldane, the great British geneticist, had both suggested that biochemical individuality might explain unusual reactions to drugs and foods. We now know that the enzymes involved in these responses are genetically determined, and many are polymorphic. Indeed, this area (known as pharmacogenetics) is seeing a resur- gence as pharmaceutical companies race to determine ways to tailor prescrip- tions and doses to patients for therapeutic (and ultimately chemopreventive) purposes.

Continuing with our example of colorectal cancer, the epidemiology sug- gests several dietary strategies for risk reduction. For example, diets rich in fruits, vegetables, fiber, and calcium are associated with lower risk for the disease.81,82 Conversely, diets high in meat and fat are positively related to risk.81,83 It is tempting to speculate that dietary associations observed at the population level would translate to effective risk-reduction strategies for indi- viduals at high risk for familial or genetic reasons. However, if the biology is different for hereditary and nonhereditary colorectal cancer, then such an approach may not be prudent.84 For example, a clinical trial of calcium among 30 members of families with hereditary nonpolyposis colorectal cancer found no significant effect of calcium on cell proliferation in the colon.85 Sellers et al., utilizing the Iowa Women’s Health Study cohort, observed that high intakes of calcium were associated with 50% decreased risk of colon cancer (95% CI, 0.3–0.7) among women without, but not with (relative risk, 1.2; CI, 0.6–2.2), a family history of the disease.86 In addition, high total intakes of vitamin E had been previously associated with lower risk in this cohort.87 High total vitamin E intakes were observed to decrease risk of colon cancer by 33% among women without a family history (upper vs. lower tertile), but only a non-statistically significant 10% among the family history-positive subset. These results suggest that care should be exercised before implementing a risk reduction intervention among subjects at elevated risk for disease based on their family history.

51589_CH14_Printer.indd 641 09/02/13 2:38 PM

642 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

Conclusion

This chapter covered the topics of genetic and molecular epidemiology. These fields are truly on the cutting edge of epidemiology as a result of their promise to provide insights into the “black box” of disease etiology. We reviewed con- cepts and applications of the field in diverse areas from infectious diseases to occupational health to the epidemiology of cancer and other chronic diseases. Progress in the field will require training a new generation of scientists with req- uisite skills, as well as greater collaboration and interdisciplinary work among scientists with laboratory skills and those versed in genetics who have training in analytic epidemiology.60,61 Scientific discoveries utilizing these methods will not only improve our understanding of disease etiology, but may lead to better and more tailored approaches to screening for disease and primary and secondary prevention.

Study Questions and Exercises 1. Consider a disease with two alleles, B and b. List all of the mating types

that could produce a heterozygous child. 2. For the situation described in problem 1, which mating type gives the

highest proportion of heterozygous offspring? 3. It is impossible for you to have received a sex chromosome from one of

your four grandparents. Which grandparent could not have transmitted, via your parents, a sex chromosome to you? Answer as if you were (a) male and (b) female.

4. A case-control study of multiple sclerosis (MS) was conducted in which family history of MS was collected on all first- and second-degree rela- tives. Among the 500 cases, 16 reported an affected relative. Among the 500 age- and sex-matched controls, 8 reported an affected relative. Do these data suggest a familial component to MS?

5. For a disease with an adult age at onset, what is the rationale for match- ing cases and controls on age when one is most interested in family his- tory of the disease?

6. You are interested in determining whether or not there is a genetic pre- disposition to lung cancer. Provide at least five reasons why lung cancer might cluster in a family for nongenetic reasons.

51589_CH14_Printer.indd 642 09/02/13 2:38 PM

r e f e r e n C e s 643

7. A published segregation analysis of asthma shows that all Mendelian patterns of inheritance do not provide a good fit to the data compared with the general model. Does this rule out the possibility that genes influence risk of asthma?

References 1. McMichael AJ. Invited commentary—molecular epidemiology: new pathway or

new traveling companion? Am J Epidemiol. 1994;140:1–11. 2. Wong KK, deLeeuw RJ, Dosanjh NS, et al. A comprehensive analysis of com-

mon copy-number variations in the human genome. Am J Hum Genet. 2007;80:91–104.

3. Conrad DE, Andrews TD, Carter NP, Hurles ME, Pritchard JK. A high- resolution survey of deletion polymorphism in the human genome. Nat Genet. 2006;38: 75–81.

4. Iafrate AJ, Feuk L, Rivera MN, et al. Detection of large-scale variation in the human genome. Nat Genet. 2004;36:949–951.

5. Khoury MJ, Beaty TH, Cohen BH. Fundamentals of Genetic Epidemiology. New York, NY: Oxford University Press; 1993.

6. Ellsworth DL, Hallman DM, Boerwinkle E. Impact of the human genome project on epidemiologic research. Epidemiol Rev. 1997;19:3–13.

7. Collins FS. Positional cloning moves from perditional to traditional. Nat Genet. 1995;9:347–350.

8. Perera FP. Molecular epidemiology: insights into cancer susceptibility, risk assess- ment, and prevention. J Natl Cancer Inst. 1996;88:496–509.

9. Perera FP. Molecular epidemiology and prevention of cancer. Environ Health Perspect. 1995;103(suppl 8):233–236.

10. Groopman JD, Kensler TW, Links J. Molecular epidemiology and human risk moni- toring. Toxicol Lett. 1995;82–83:763–769.

11. Love RR, Evans AM, Josten DM. The accuracy of patient reports of a family history of cancer. J Chronic Dis. 1985;38:289–293.

12. Zhao LP, Hsu L, Davidov O, Potter J, Elston RC, Prentice RL. Population-based family study designs: an interdisciplinary research framework for genetic epidemiol- ogy. Genet Epidemiol. 1997;14:365–388.

13. Schneider NR. Familial aggregation of cancer. In: Chaganti RSK, German J, eds. Genetics in Clinical Oncology. New York, NY: Oxford University Press; 1985: 133–145.

14. Mattson ME, Pollack ES, Cullen JW. What are the odds that smoking will kill you? Am J Public Health. 1987;77:425–431.

15. Fisher RA. Lung cancer and cigarettes? Nature. 1958;182:108. 16. Sellers TA. Familial predisposition to lung cancer. In: Eeles R, Easton D, Ponder B,

Horwich A, eds. Genetic Predisposition to Cancer. London, UK: Chapman & Hall Medical; 1996:344–354.

17. Vachon CM, Sellers TA, Kushi LH, Folsom AR. Familial correlation of dietary intakes among postmenopausal women. Genet Epidemiol. 1998;15:553–563.

51589_CH14_Printer.indd 643 09/02/13 2:38 PM

644 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

18. Krousel T, Garcas N, Rothschild H. Familial clustering of mesothelioma: a report on three affected persons in one family. Am J Prev Med. 1986;2:186–188.

19. Ooi WL, Elston RC, Chen VW, et al. Increased familial risk for lung cancer. J Natl Cancer Inst. 1986;76:217–222.

20. Doll R, Peto R. The Causes of Cancer. New York, NY: Oxford University Press; 1981. 21. Tokuhata GE. Familial factors in human lung cancer and smoking. Am J Public

Health. 1964;54:24–32. 22. Rothschild HR, Voors AW, Weed SG, et al. Trends in respiratory system cancer

mortality in Louisiana: geographic distribution in 1950–1969 and 1967–1976 com- pared. Am J Public Health. 1979;69:380–381.

23. Blot WH, Harrington JM, Toledo A, et al. Lung cancer after employment in ship- yards during World War II. N Engl J Med. 1978;229:620–624.

24. Gottlieb MS, Pickle LW, Blot WJ, et al. Lung cancer in Louisiana: death certificate analysis. J Natl Cancer Inst. 1976;57:3–7.

25. Mulvey JJ, Rothschild H. Sugarcane farming—is there a link with cancer? Ecol Dis. 1983;2:267–270.

26. Sellers TA, Ooi WL, Elston RC, et al. Familial risk of non-lung cancer among rela- tives of lung cancer patients. Am J Epidemiol. 1987;126:237–246.

27. Vandenberg SG. Assortative mating, or who marries whom? Behav Genet. 1972;2:127–157.

28. Elston RC. Segregation analysis. Adv Hum Genet. 1981;11:63–120. 29. Claus EB, Risch N, Thompson WD. Genetic analysis of breast cancer in the cancer

and steroid hormone study. Am J Hum Genet. 1991;48:232–242. 30. Claus EB, Risch N, Thompson WD. Age of onset as an indicator of familial risk of

breast cancer. Am J Epidemiol. 1990;131:961–972. 31. Hall JM, Lee MK, Morrow J, et al. Linkage of early-onset familial breast cancer to

chromosome 17q21. Science. 1990;250:1684–1689. 32. Hall JM, Friedman L, Guenther C, et al. Closing in on a breast cancer gene on chro-

mosome 17q. Am J Hum Genet. 1992;50:1235–1242. 33. Narod SA, Feunteun J, Lynch HT, et al. Familial breast-ovarian cancer locus on

chromosome 17q12-q23. Lancet. 1991;338:82–83. 34. Miki Y, Swensen J, Shattuck-Eidens D, et al. A strong candidate for the breast and

ovarian cancer susceptibility gene BRCA1. Science. 1994;266:66–71. 35. Wooster R, Bignell G, Lancaster J, et al. Identification of the breast cancer suscepti-

bility gene BRCA2. Nature. 1995;378:789–792. 36. Reliene R, Bishop AJ, Schiestl RH. Involvement of homologous recombination in

carcinogenesis. Adv Genet. 2007;58:67–87. 37. Bermejo-Pérez MJ, Márquez-Calderón S, Llanos-Méndez A. Effectiveness of pre-

ventive interventions in BRCA1/2 gene mutation carriers: a systematic review. Int J Cancer. 2007;121:225–231.

38. Penrose GS. Genetic linkage in graded human characters. Ann Eugenics. 1938:8: 223–237.

39. Haseman JK, Elston RC. The investigation of linkage between a quantitative trait and a marker locus. Behav Genet. 1972;2:3–19.

40. Weeks DE, Lange K. The affected-pedigree-member method of linkage analysis. Am J Hum Genet. 1988;42:315–326.

41. Kruglyak L, Lander ES. Complete multipoint sib-pair analysis of qualitative and quantitative traits. Am J Hum Genet. 1995;57:439–454.

51589_CH14_Printer.indd 644 09/02/13 2:38 PM

r e f e r e n C e s 645

42. Pritchard JK, Przeworski M. Linkage disequilibrium in humans: models and data. Am J Hum Genet. 2001;69:1–14.

43. Jorde LB. Linkage disequilibrium and the search for complex disease genes. Genome Res. 2000;10:1435–1444.

44. Boehnke M. A look at linkage disequilibrium. Nat Genet. 2000;25:246–247. 45. Gabriel SB, Schaffner SF, Nguyen H, et al. The structure of haplotype blocks in the

human genome. Science. 2002;296:2225–2229. 46. Permuth-Wey J, Kim D, Tsai Y-Y, et al. on behalf of the Ovarian Cancer Association

Consortium (OCAC). LIN28B polymorphisms influence susceptibility to epithelial ovarian cancer. Cancer Res. 2011;71(11):3896–3903.

47. Boddeker IR, Ziegler A. Sequential designs for genetic epidemiologic linkage or asso- ciation studies. Biomed J. 2001;43:501–525.

48. Lander E, Kruglyak L. Genetic dissection of complex traits: guidelines for interpret- ing and reporting linkage results. Nat Genet 1995;11:241–247.

49. Easton DF, Pooley KA, Dunning AM, et al. Genome-wide association study identi- fies novel breast cancer susceptibility loci. Nature. 2007;447:1087–1093.

50. Hunter DJ, Kraft P, Jacobs KB, et al. A genome-wide association study identifies alleles in FGFR2 associated with risk of sporadic postmenopausal breast cancer. Nat Genet. 2007;39:870–874.

51. Stacey SN, Manolescu A, Sulem P, et al. Common variants on chromosomes 2q35 and 16q12 confer susceptibility to estrogen receptor-positive breast cancer. Nat Genet. 2007;39:865–869.

52. Friend SH. Breast cancer susceptibility testing: realities in the post-genomic era. Nat Genet. 1996;13:16–17.

53. The International Human Genome Mapping Consortium. A physical map of the human genome. Nature. 2001;409:934–941.

54. Venter JC, Adams MD, Myers EW, et al. The sequence of the human genome. Science. 2001;291:1304–1351.

55. Collins FS. Genetics: an explosion of knowledge is transforming clinical practice. Geriatrics. 1999;54:41–47.

56. Baltimore D. Our genome unveiled. Nature. 2001;409:814–816. 57. Maslow JN, Mulligan ME, Arbeit R. Molecular epidemiology: application of con-

temporary techniques to the typing of microorganisms. Clin Infect Dis. 1993;17: 153–164.

58. Hlady WG, Mullen R, Hopkins R, Mintz C, Shelton B, Daikos G. Outbreak of Legionnaires’ disease linked to a decorative fountain by molecular epidemiology. Am J Epidemiol. 1993;138:555–562.

59. Blanc DS, Parret T, Janin B, Raselli P, Francioli P. Nosocomial infections and pseu- doinfections from contaminated bronchoscopes: two-year follow-up using molecular markers. Infec Control Hosp Epidemiol. 1997;18:134–136.

60. Schulte PA. Use of biological markers in occupational health research and practice. J Toxicol Environ Health. 1993;40:359–366.

61. Soderkvist P, Axelson O. On the use of molecular biology data in occupational and environmental epidemiology. J Occup Environ Med. 1995;37:84–90.

62. Bang KM. Applications of occupational epidemiology. Occup Med. 1996;11: 381–391.

63. Taylor JA, Sandler DP, Bloomfield CD, et al. Ras oncogene activation and occupa- tional exposures in acute myeloid leukemia. J Natl Cancer Inst. 1992;84:1626–1632.

51589_CH14_Printer.indd 645 09/02/13 2:38 PM

646 C h a p t e r 1 4 M o l e C u l a r a n d G e n e t i C e p i d e M i o l o G y

64. Van Duijn CM, Clayton D, Chandra V, et al. Familial aggregation of Alzheimer’s disease and related disorders: a collaborative re-analysis of case-control studies. Int J Epidemiol. 1991;20(suppl 2):S13–S20.

65. Tanzi RE, Bertram L. 20 years of the Alzheimer's disease amyloid hypothesis: a genetic perspective. Cell. 2005;120:545–555.

66. Pericak-Vance MA, Bebout JL, Gaskell PC, et al. Linkage studies in famil- ial Alzheimer disease: evidence for chromosome 19 linkage. Am J Hum Genet. 1991;48:1034–1050.

67. Tsai M-S, Tangalos EG, Petersen RC, et al. Apolipoprotein E: risk factor for Alzheimer disease. Am J Hum Genet. 1994;54:643–649.

68. Tanzi RE, Bertram L. Genome wide association studies in Alzheimer’s disease. Hum Mol Genet. 2009;18(R2):R137-R145.

69. Kendler KS. The genetic epidemiology of psychiatric disorders: a current perspective. Soc Psychiatry Psychiatr Epidemiol. 1997;32:5–11.

70. Malafosse A, Leboyer M, d’Amato T, et al. Manic depressive illness and tyrosine hydroxylase gene: linkage heterogeneity and association. Neurobiol Dis. 1997;4: 337–349.

71. McMahon FJ, Hopkins PJ, Xu J, et al. Linkage analysis of bipolar affective dis- order to chromosome 18 markers in a new pedigree series. Am J Hum Genet. 1997;61:1397–1404.

72. Waldman ID, Robinson BF, Feigon SA. Linkage disequilibrium between the dopa- mine transporter gene (DAT1) and bipolar disorder: extending the transmission disequilibrium test (TDT) to examine genetic heterogeneity. Genet Epidemiol. 1997;14:699–704.

73. Rice J. Genetic analysis of bipolar disorder: summary of GAW10. Genet Epidemiol. 1997;14:549–561.

74. King RA, Rotter JI, Motulsky AG. The Genetic Basis of Common Diseases. New York, NY: Oxford University Press; 1992.

75. Khoury MJ. From genes to public health: the applications of genetic technology in disease prevention. Am J Public Health. 1996;86:1717–1722.

76. Schottenfeld D, Winawer SJ. Cancer of the large intestine. In: Schottenfeld D, Fraumeni JF, eds. Cancer Epidemiology and Prevention. 2nd ed. New York, NY: Oxford University Press; 1996:813–840.

77. National Cancer Institute. Colon and Rectal Cancer. http://www.cancer.gov/ cancertopics/types/colon-and-rectal. Accessed May 28, 2012.

78. Potter JD, Sellers TA, Rich SS. Colorectal cancer. In: Ponder BAJ, Waring MJ, eds. Cancer Biology and Medicine (Vol 4): The Genetics of Cancer. Amsterdam, Netherlands: Kluwer Academic Publishers; 1995:45–65.

79. Ahlquist DA, Skoletsky JE, Boynton KA, et al. Colorectal cancer screening by detection of altered human DNA in stool: feasibility of a multitarget assay panel. Gastroenterology. 2000;119:1219–1227.

80. Waddell WR, Loughry RW. Sulindac for polyposis of the colon. J Surg Oncol. 1983;24:83–87.

81. Potter JD, Slattery ML, Bostick RM, Gapstur SM. Colon cancer: a review of the epidemiology. Epidemiol Rev. 1993;15:499–545.

82. Bostick RM, Potter JD, Sellers TA, et al. Relation of calcium, vitamin D, and dairy food intake to incidence of colon cancer among older women: the Iowa Women’s Health Study. Am J Epidemiol. 1993;127:1302–1317.

51589_CH14_Printer.indd 646 09/02/13 2:38 PM

r e f e r e n C e s 647

83. Willett WC, Stampfer MJ, Colditz GA, et al. Relation of meat, fat, and fiber intake to the risk of colon cancer in a prospective study among women. N Engl J Med. 1990;323:1664–1672.

84. Alberts DS, Lipkin M, Levin B. Genetic screening for colorectal cancer and interven- tion. Int J Cancer. 1996;69:62–63.

85. Cats A, Kleibeuker JH, van der Meer R, et al. Randomized, double-blinded, placebo- controlled intervention study with supplemental calcium in families with hereditary nonpolyposis colorectal cancer. J Natl Cancer Inst. 1995;87:598–603.

86. Sellers TA, Bazyk AE, Bostick RM, et al. Diet and risk of colon cancer in a large pro- spective study of older women: an analysis stratified on family history. Cancer Causes Control. 1998;9:357–367.

87. Bostick RM, Potter JD, McKenzie DR, et al. Reduced risk of colon cancer with high intake of vitamin E: the Iowa Women’s Health Study. Cancer Res. 1993;53: 4230–4237.

51589_CH14_Printer.indd 647 09/02/13 2:38 PM

51589_CH14_Printer.indd 648 09/02/13 2:38 PM

3chapte r

649

15chapte r

Social, Behavioral, and Psychosocial Epidemiology

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● define the terms social epidemiology, behavioral epidemiology, and psychosocial epidemiology

●● state the role of psychological, behavioral, and social factors in health and disease

●● discuss the stress concept as a hypothesized determinant of disease ●● define social incongruity, person-environment fit, and stressful life

events ●● discuss moderators of the stress–illness relationship ●● state outcomes of exposure to stress

CHAPTER OUTLINE

I. Introduction II. Research Designs Used in Psychosocial, Behavioral, and Social

Epidemiology III. The Social Context of Health IV. Independent Variables V. Moderating Variables

VI. Dependent (Outcome) Variables: Physical and Mental Health VII. Conclusion

VIII. Study Questions and Exercises

51589_CH15_Printer.indd 649 09/02/13 4:47 PM

650 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Introduction

In his discussion of grief (Exhibit 15–1), the late George Engel1 questioned the adequacy of the biomedical model in explaining certain types of health problems. As noted in Exhibit 15–1, he suggested a new category of variables— psychological factors—that are not usually considered in the agent, host, and environment model. Engel was a famed psychiatrist who worked for most of his career at the University of Rochester. Engel argued for the inclusion of psy- chological, social, and behavioral factors in models for the etiology of disease. The set of factors that he identified constituted the biopsychosocial model, which was a broadened framework for the causality of health and illness. Such a broadened framework is especially relevant, given the concerns of federal gov- ernmental agencies and public health practitioners about the adverse effects of poverty, social and income inequalities, discrimination, and the structure of society. The need to address these issues has spurred the development of a growing body of research in the fields of social, behavioral, and psychosocial epidemiology.

As described previously, epidemiology (and the public health and medi- cal fields) has borrowed with increasing frequency from the theoretical and conceptual bases of the behavioral sciences for etiologic models of disease. The behavioral sciences (e.g., sociology and psychology) hold potential for the development of explanatory frameworks and for expanding knowledge of conditions of unknown etiology. Sociology contributes an interweaving of social conditions and levels as they affect disease processes. Psychology is concerned with the study of personal behavior, which is an important aspect of health outcomes. In comparison with the first part of the 20th century, the latter part of that century and the beginning of the 21st century have seen the elaboration of psychological and social models as the etiologic bases of the chronic, noninfectious diseases as well as some of the infectious diseases.

This chapter explores a rich tradition of epidemiologic theory and research: the role of psychological, behavioral, and social determinants of health and illness. These determinants include aspects of the individual’s personality, social factors enmeshed in the fabric of society, and cultural influences. They are multifacto- rial and involve a complex interaction of the person and environment. Differ- ing from some of the causal determinants discussed elsewhere in this text, they are neither single agents (e.g., specific microbes that cause illnesses) nor demo- graphic classificatory variables (e.g., age or sex).

51589_CH15_Printer.indd 650 09/02/13 4:47 PM

i n t r o d u C t i o n 651

Definitions of Important Terms Three main specializations encompass psychological, behavioral, and social factors: social epidemiology, behavioral epidemiology, and psychosocial epidemiology.

●● Social epidemiology: concerned with the influence of a person’s posi- tion in the social structure upon the development of disease.2 A broader

When Is Grief a Disease?

“To enhance our understanding of how it is that ‘problems of living’ are experienced as illness by some and not by others, it might be help- ful to consider grief as a paradigm of such a borderline condition. For while grief has never been considered in a medical framework, a significant number of grieving people do consult doctors because of disturbing symptoms, which they do not necessarily relate to grief. Fifteen years ago I addressed this question in a paper entitled, ‘Is grief

a disease? A challenge for medical research.’ Its aim too was to raise ques- tions about the adequacy of the biomedical model. A better title might have been, ‘When is grief a disease?’, just as one might ask when schizophrenia or when diabetes is a disease. For while there are some obvious analogies between grief and disease, there are also some important differences. . . . Grief clearly exemplifies a situation in which psychological factors are pri- mary; no preexisting chemical or physiological defects or agents need be invoked. Yet as with classic diseases, ordinary grief constitutes a discrete syndrome with a relatively predictable symptomatology which includes, inci- dentally, both bodily and psychological disturbances. It displays the auton- omy typical of disease; that is, it runs its course despite the sufferer’s efforts or wish to bring it to a close. A consistent etiologic factor can be identified, namely, a significant loss. On the other hand, neither the sufferer nor society has ever dealt with ordinary grief as an illness even though such expressions as ‘sick with grief’ would indicate some connection in people’s minds. And while every culture makes provisions for the mourner, these have generally been regarded more as the responsibility of religion than of medicine.” n

e x

h Ib

It 1

5 –1

51589_CH15_Printer.indd 651 09/02/13 4:47 PM

652 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

conception of the field defines it as “. . . the branch of epidemiology that studies the social distribution and social determinants of states of health.” The field “. . . aim[s] to identify socioenvironmental exposures that may be related to a broad range of physical and mental health outcomes.” The types of exposures include “. . . social phenomena such as socioeconomic strati- fication, social networks and support, discrimination, work demands, and control . . .”3(p 6) Social epidemiology “. . . uses epidemiological principles, reasoning, and methods . . .” It is “[a]n interface between epidemiology in the social sciences.”4

●● Behavioral epidemiology: studies the role of behavioral factors in health. Examples of behavioral factors are substance use (e.g., consumption of tobacco, illicit drugs, and alcohol), physical activity, risky sexual behav- ior, and consumption of unhealthful foods. Closely related to the field of behavioral epidemiology is behavioral medicine, the application of behav- ioral factors to specific clinical interventions, as in the case of biobehav- ioral approaches to the management of high blood pressure. Biobehavioral approaches include nonpharmacologic treatment methods (e.g., exercise, maintenance of desirable weight, changes in diet, and meditation).5

●● Psychosocial epidemiology: an “umbrella term” that covers a broad scope of research studies. The term psychosocial is defined by the Oxford College Dictionary as “of or relating to the interrelation of social factors and indi- vidual thought and behavior.” More broadly conceptualized than either social epidemiology or behavioral epidemiology, the field examines the influence of psychosocial influences on health. Psychosocial epidemiology often is used as a synonym for social epidemiology.

Topics Covered in the Field The topics covered include stress and stressful life events; personality factors, culture, personal behavior, and social support; and mental and physical health status (linked to psychological, social, and behavioral factors). Writings about psychological and social processes in disease often include the variables of socio- economic status, ethnicity, religion, and familial characteristics. For the sake of clarity and to highlight some of the traditional concerns of psychosocial epide- miology, the foregoing psychological and social dimensions are treated in this chapter as discrete categories. In reality, they are overlapping dimensions; for example, personal behavior is a function of sociocultural influences that also are related to stress.

51589_CH15_Printer.indd 652 09/02/13 4:47 PM

i n t r o d u C t i o n 653

Psychological, behavioral, and social factors are relevant to mental health states, including grief and depression; to physical health states, such as the chronic diseases; and to the etiology of infectious diseases, such as increased susceptibility to the common cold virus, herpes virus, and other agents. The conditions that presently compel the attention of psychosocial epidemiologic researchers, however, are the chronic, degenerative diseases: hypertension, coro- nary heart disease (CHD), arthritis, certain varieties of cancer, and diabetes, to name a few examples. Heart disease, cancer, and stroke are the leading causes of mortality in developed countries (and also to a significant extent in less developed countries) and, accordingly, the psychosocial aspects of these con- ditions should receive high priority within public health agencies and among researchers.

Given that the field of psychosocial epidemiology covers a vast body of litera- ture, this chapter cannot be exhaustive. Rather the authors survey some of the major issues and applications in this area of epidemiology.

A Model of Psychosocial Factors in Health We have developed a theoretical model in order to synthesize and integrate some of the findings of the psychosocial literature. (See Figure 15–1.) The model groups some of the components of this literature under a theoretical

Figure 15–1 Guide to psychosocial epidemiology—psychologic, behavioral, and social (examples of variables studied).

Stress

Social incongruity

Person-environment fit

Life events

Independent variables (Risks and exposures)

Moderating variables (Intervening factors)

Adverse physical and social environmental conditions (e.g., poor sanitation/water, pollution, poverty/declining economy, overcrowding)

Dependent variables (Health outcomes)

Personality factors Type A behavior Culture Sanctioned behaviors Illness behavior

Social support Lifestyle and behavior Risk taking Diet Substance use Exercise levels

Life and job dissatisfaction Mental health Anxiety Depressive symptoms PTSD Physical health Chronic diseases Infectious diseases Illness susceptibility Injuries

51589_CH15_Printer.indd 653 09/02/13 4:47 PM

654 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

framework that consists of the major categories of independent, moderating, and dependent variables. We have organized the remainder of this chapter according to these categories. The objective of the present chapter is to consider those psy- chosocial variables that are compatible with a theoretical framework for explain- ing illness etiology, that is, independent or intervening variables in the causality of disease.

●● Independent variables (also referred to as exposure or risk factor variables): defined as hypothesized causal factors in the theoretical model.

●● Moderating (intervening) variables: shown as intermediate variables in the causal process between an independent variable (risk or exposure) and outcome.

●● Dependent variables: outcome variables in the theoretical model; inde- pendent variables affect or influence dependent variables (via the pathway of moderating variables).

Multiple Causation The concept of multiple causation means “. . . a given health state or health-related process may have more than one cause. A combination of causes or alternative combinations of diseases is often required to produce the health outcome.”4 Typically, psychosocial epidemiologic studies involve diseases or conditions that have multiple independent risk factors or that involve interaction of risk factors (i.e., social, psychological, and biochemical). For example, in the case of CHD the risk factors hypertension, blood lipids, smoking, diet, and lifestyle operate jointly to increase the risk of disease. This statement is an oversimplification, however, because behavioral and psychological factors may be related indirectly to CHD by activating certain biochemical processes that, in turn, may be related directly to elevated status on risk factors for CHD.

Another characteristic of current etiologic models using multiple causation is that they do not provide a complete explanation of the phenomenon being inves- tigated; in “epi-speak” it is said that the set of known risk factors do not explain 100% of the variance in the phenomenon under study. Consider the example of CHD (and many other chronic diseases). The known risk factor variables do not form a complete explanation for CHD; many unexplained causes for CHD other than those known to science remain. In addition, a small percentage of individuals who are at high risk on all the known causal factors never develop an overt case of the disease and may go on to outlive the low-risk individuals.

51589_CH15_Printer.indd 654 09/02/13 4:47 PM

r e S e a r C h d e S i g n S 655

Research Designs Used in Psychosocial , Behavioral , and Social Epidemiology

This field of investigation has employed many of the standard epidemiologic designs (e.g., case-control, cohort, cross-sectional, and experimental designs). For example, the role of stress in health outcomes is one of the major concerns of psychosocial epidemiology. Some of the earlier research studies on this topic used cross-sectional investigations of captive and readily available populations (e.g., work-related groups of white male professionals). At present the research community acknowledges the importance of including diverse study populations in epidemiologic investigations. Cross-sectional studies of psychosocial factors in health are attractive because of the relative ease of such research. Cross-sectional designs, however, may not be adequate to detect the subtle effects associated with psychosocial factors, especially the issues of temporality of cause and effect and the influence of confounding variables.

Also needed are more longitudinal, prospective studies of the role of psy- chosocial factors in health as well as studies of women and minority groups, although much progress has been made in developing these types of studies in recent years. Among the many excellent initiatives in this area is the work of investigators in Germany.6 Researchers, conducting the Early Develop- mental Stages of Psychopathology study (EDSP), began data collection in 1994–1995. The EDSP was a prospective study of adolescents and young adults who resided in the city of Munich, Germany. A total of 3,021 adoles- cents and young adults aged 14–24 years were included at baseline. Subjects were followed up on average after a 42-month interval. This project has made possible the study of the incidence of various mental disorders, including the general anxiety syndrome and depression.7,8 See Figure 15–2 for information on the data collection intervals for the study and the types of information collected.

A unique challenge of psychosocial studies is to obtain valid and reliable operationalization of measures. The process of operationalization refers to the methods used to translate some of the concepts employed in psychosocial epi- demiology into actual measurements. For example, much controversy surrounds the development of measures of stress, epidemiologically useful measures of mental health, and measures of social support. In addition, as part of the process of improving measures used, epidemiologists need to develop well-delineated conceptual models of psychosocial processes.

51589_CH15_Printer.indd 655 09/02/13 4:47 PM

656 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Community-based participatory research (CBPR) is a methodology with much relevance to psychological and social epidemiologic studies. CBPR can be thought of as an alliance between community organizations and research units (e.g., universities or research centers) in order to investigate health- related issues of interest to the community. Many of the personnel affiliated with research units have had formal training in research methodology. The community organizations have an investment in health issues that affect the members of the community. The organizations can facilitate access to mem- bers of the community in order to collect research data. These complemen- tary skill sets help to improve the quality of information that is gathered in research projects. CBPR results in the eventual training and empowerment of community organizations to conduct their own independent research proj- ects that address their unique needs. Figure 15–3 shows members of a com- munity meeting organized by the Agency for Toxic Substances and Disease Registry

An example of CBPR is an epidemiologic study of tobacco use among the Cambodian community of Long Beach, California. A research team organized by Robert Friis collaborated with a community organization in Long Beach in order to study sociocultural factors associated with Cambodian Americans’ use of tobacco. Cambodian American men have a higher prevalence of cigarette smoking than other groups in California.9

Figure 15–2 Early Developmental Stages of Psychopathology Study. Source: Courtesy of Roselind Lieb, Basel, Switzerland, and Hans-Ulrich Wittchen, Dresden, Germany.

Birth

Lifetime

Interval

T0 T1 T2 T3

Interval Interval

1995 1997

14–24 years (N = 3021)

1999 2005

Factors in the individual:

Environmental factors:

Sex, age Genotype (T3) Substance use Early childhood conditions Natal complications DSM-IV mental disorders

Life events, traumatic events Parenting style Family environment Psychopathology in family

DSM-IV mental disorders – affective disorders – anxiety disorders – substance use disorders – Somatoform/eating disorders

51589_CH15_Printer.indd 656 09/02/13 4:47 PM

t h e S o C i a l C o n t e x t o f h e a l t h 657

The Social Context of Health

Interest in the role of the social environment in health has flourished in recent decades, perhaps because epidemiologists have recognized that the social environ- ment may contribute to the regulation of psychosocial influences upon health. As noted previously, the social environment is the totality of the behavioral, per- sonality, attitudinal, and cultural characteristics of a group of people. The social environment provides the context in which psychosocial factors operate.

An example of the potential social influence of the social environment on health comes from Scotland, which has higher mortality than England and other Euro- pean countries.10 A possible explanation is that this outcome is due to higher levels of deprivation in Scotland. However, these mortality differences persist when the level of deprivation (e.g., lack of car ownership, overcrowding, male unemployment, and social class) is controlled. The term Scottish effect signifies excess mortality in Scotland after controlling for the effects of deprivation. Excess mortality in the West of Scotland (Glasgow) after controlling for deprivation is called the Glasgow effect. When deprivation levels are controlled in Glasgow and the relatively close cities of Liverpool and Manchester, excess mortality in Glasgow can still be observed.

Figure 15–3 Town hall meeting held on behalf of the Agency for Toxic Substances and Disease Registry. Source: Reproduced from Centers for Disease Control and Prevention. Public Health Image Library. Image number 11619. Available at http://phil.cdc. gov/phil/. Accessed May 2, 2012.

51589_CH15_Printer.indd 657 09/02/13 4:47 PM

658 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

The Social Context and International Comparisons From the population perspective, we find that the social environment impacts the health of residents of both the less-developed and more-developed worlds.11 In less-developed regions, life expectancy is reduced in comparison with the developed world due to the impact of poverty, with its attendant malnutri- tion and infectious diseases. Overcrowding, poor living conditions, and lack of preventive health care foster the spread of infectious diseases. Figure 15–4 was taken during the 1970s in Bangladesh during the World Health Organization’s smallpox eradication program. The image shows an impoverished man begging for money in order to provide food for his unclothed child.

The social environment is also an influential component of the major causes of morbidity and mortality through the impact of noninfectious conditions, which are the leading causes of death in the developed world. Lifestyle factors— smoking, diet, insufficient exercise, and use of illegal substances—undoubtedly play a role in the etiology of many of these conditions. A noteworthy example is the reduced life expectancy in the countries of central and Eastern Europe and the former Soviet Union in comparison to the European Union countries. Some authorities have suggested that one of the contributing factors to this reduction

Figure 15–4 A village in Bangladesh in the 1970s during a campaign to eradicate smallpox from the country. Source: Reproduced from Centers for Disease Control and Prevention. Public Health Image Library. Image number 7405. Available at http://phil.cdc. gov/phil/. Accessed May 2, 2012.

51589_CH15_Printer.indd 658 09/02/13 4:47 PM

t h e S o C i a l C o n t e x t o f h e a l t h 659

in life expectancy is excessive alcohol consumption, although the influence of alcoholism on life expectancy is by no means clear-cut.11

In order to compare accurately among nations the contributions of disease to morbidity and mortality, work has proceeded on the development of measures that could be used in all countries. The Global Burden of Disease Study attempts to quantify and provide an epidemiologic assessment of the worldwide consequences of disease by using a measure known as the disability-adjusted life year (DALY) to assist in comparisons across countries.12 The DALY is a statistical measure applied to populations that combines information on mortality with information on mor- bidity for specific causes. The advantage of DALYs is that they provide a standard epidemiologic unit for comparative purposes. According to Murray and Lopez, “The 10 leading specific causes of global DALYs are, in descending order, lower respiratory infections, diarrhoeal diseases, perinatal disorders, unipolar major depression, ischaemic heart disease, cerebrovascular disease, tuberculosis, mea- sles, road-traffic crashes, and congenital anomalies. [A total of ] 15.9% of DALYs worldwide are attributable to childhood malnutrition and 6.8% to poor water and sanitation and personal and domestic hygiene.”12(p 1436) The social environment contributes to many of these DALYs via a range of pathways including unsanitary

First Whitehall study–initiated in 1967, the first Whitehall study covered 18,000 men in the British Civil Service. Among its findings were a higher likelihood of premature death among the lowest employment grades; these socioeconomic inequalities were not explained by standard risk factors (e.g., smoking).

Whitehall II study–noted social epidemiologist, Sir Michael Marmot, began the study in 1985. Participants were 10,308 nonindustrial civil ser- vants from 35 to 55 years of age; approximately one-third of the sample was composed of women. “Whitehall II data have been used to build one of the most detailed pictures of the determinants of health in mid-life and late-life.” One of the important contributions of the study has been to highlight “. . . the importance of psychosocial factors such as work stress, unfairness, and work-family conflict to socio-economic inequalities.” n

Source: Whitehall II History. http://www.ucl.ac.uk/whitehallII/history. Accessed August 8, 2012.

The Whitehall Study

51589_CH15_Printer.indd 659 09/02/13 4:47 PM

660 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

living conditions associated with crowding; discrimination and social exclusion; low status on the social hierarchy; social isolation; and adverse lifestyle choices.

A Landmark Study on the Effects of Socioeconomic Inequalities A major research program on socioeconomic inequalities in health is the White- hall study conducted in Britain. (Refer to the text box titled The Whitehall Study).

Independent Variables

In addition to the social environment, we will now consider how the major categories of psychosocial variables listed in Figure 15-1 affect health outcomes. Examples of independent variables covered in psychosocial epidemiologic research include general concepts of stress, social incongruity theory, person- environment fit, and stressful life events.

General Concepts of Stress According to a classic definition, psychological stress is “. . . a particular relation- ship between the person and the environment that is appraised by the person as taxing or exceeding his or her resources and endangering his or her well-being.”13 Societal structures, interpersonal processes, and the individual’s physiological, cognitive, and other responses relate to the distribution of stress. Stress as an inde- pendent, antecedent variable to health and illness represents an intriguing notion because it seems to support common sense explanations for the cause of some mental disorders, sudden death due to heart attacks, and other chronic condi- tions. Also, researchers have argued that stress induces physiological changes such as alterations to the immune system. When environmental demands challenge an organism (e.g., experimental animal or person), homeostatic regulatory mecha- nisms help the organism to maintain balance among essential biological functions. Homeostasis refers to a tendency toward a stable equilibrium among physiological processes. Allostasis denotes “. . . how the organism achieves stability (or homeo- stasis) through continual change.” 14(p 36) The term allostatic load “. . . refers to the consequences of sustained activation of primary regulatory mechanisms serving allostasis over time . . .” 14(p 36) One’s allstatic load is hypothesized to be associated with disorders and adverse health outcomes.

Many aspects of contemporary life are stressful as a result of economic forces, the changing nature of employment, and other environmental and social factors. Economic stresses due to unemployment, banking collapses, and the increased

51589_CH15_Printer.indd 660 09/02/13 4:47 PM

i n d e p e n d e n t v a r i a B l e S 661

cost of living are prominent features of modern existence worldwide. In the United States the violent consequences of stresses experienced in the workplace (e.g., “going postal”) are a common topic of media reports. Figure 15–5 shows a poster designed for an employee assistance program operated by an occupational medical service. The program targeted workers under stress due to their self- perceived inability to meet the demands of their job.

The scientific evidence for stress as an etiologic agent of disease is both con- troversial and contradictory. A review of the numerous writings on stress leads one to conclude that the concept has more than one meaning and that some of the meanings tend to be vague or inconsistent. The con- cept of stress has a venerable his- torical background in the field of medicine and in other disciplines, but it is often regarded with sci- entific skepticism. However, many research findings have tended to support the stress concept. For instance, Walter Canon’s classic research studied changes in gastro- intestinal function that accompa- nied stressful events, such as pain, hunger, and major emotion.15 (Refer to Figure 15–6 for Canon’s image.) The late Hans Selye,16 shown in Figure 15-7, specified in detail the stages of reaction to stress through the concept of the general adaptation syndrome. Selye conceived of stress as a change in the environment of the organism and proposed that the organism’s response consisted of three stages: alarm reaction, stage of resis- tance, and stage of exhaustion. Activation of the general adapta- tion syndrome, associated with corticoid secretion, may produce

Figure 15–5 A poster designed for an employee assistance program operated by an occupational medical service. Source: Reproduced from National Library of Medicine. Images from the History of Medicine. Order account number: C01097. Available at: http:// ihm.nlm.nih.gov/images/C01097. Accessed September 1, 2012.

51589_CH15_Printer.indd 661 09/02/13 4:47 PM

662 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Figure 15–6 Walter B. Cannon, 1871–1945. Source: Reproduced from National Library of Medicine. Images from the History of Medicine. Order account number: C01097. Available at: http://ihm.nlm.nih.gov/images/B04183. Accessed September 1, 2012.

Figure 15–7 Hans Selye, 1907–1982. Source: Courtesy of the Hans Selye Foundation.

51589_CH15_Printer.indd 662 09/02/13 4:47 PM

i n d e p e n d e n t v a r i a B l e S 663

somatic disease (e.g., mineralocorticoid hypertension [hypertension caused by excessive activation of mineralocorticoid receptors] and cardiac necrosis [death of cardiac tissue]).17

1. Alarm reaction: physiologic responses associated with preparation to deal with stress that lead the animal or person to fight or escape from the stressor.

2. Stage of resistance: return of physiologic responses to normal and resistance to further stressful stimuli.

3. Stage of exhaustion: failure of the organism to adapt to overwhelm- ing stresses. “Adaptation energy” becomes exhausted, and, in the case of humans, severe bodily disease and death may result. n

Selye’s Concept of the General Adaptation Syndrome

According to some experts, stress research refers to a broad area that explores how aversive environmental events control multiple response systems (verbal, physiological, and behavioral).18 Clinical findings suggest that these aversive events produce negative health outcomes. Three examples of aversive events that may produce stress responses are presentation of noxious or biologically damaging stimuli (either by actual presentation or threat of presentation), removal of reinforcements (either actual or threatened), and conflict situations. An example of a presentation of damaging stimulus is electric shock experimen- tation. The early executive monkey experiments demonstrated that physiologic arousal linked to behavioral responses to remove the threat of electric shock was associated with gastric ulcer in monkeys.18 Removal of a positively reinforc- ing stimulus includes removal of rewards, such as those associated with good behavior, and environmental supports. Removal of positive reinforcements may be associated with impaired mental health and other illnesses. Finally, a conflict situation is one that generates two or more incompatible responses in the same individual. Examples are attitude conflicts, role conflicts, and conformity con- flicts, the last of which was associated with changes in lipid metabolism in one experiment.18

51589_CH15_Printer.indd 663 09/02/13 4:47 PM

664 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Social Incongruity Theory (Status Discrepancy Models) Social incongruity is defined as a situation in which the individual is not in har- mony with or is incompatible with other persons; it can also denote lack of harmony between the individual and the larger society. Social incongruity can occur among the following: friends and significant others, a person and place of residence, members of different generations, and a foreign-born person with native-born residents. Status discrepancy refers to disharmony that arises from dif- ferences among the statuses of individuals (e.g., among higher- and lower-ranked persons within society).

Investigators have hypothesized that both social incongruity and status dis- crepancy are associated with adverse physical and mental health outcomes. General themes of research have included changes in residence from one country or culture to another, changes in residence from a rural to an urban area, upward intragenerational mobility, and marital status stress (i.e., discrepancy between husband and wife in social and educational status).

The late Sidney Cobb and associates compared women who were afflicted with rheumatoid arthritis (RA) with women who were not so afflicted. Women with RA were more likely to originate from homes with high parental status stress, defined as discrepancy between parents with respect to indicators of social status. An example was a mother of high social status married to a father from a background of lower social status. Interestingly, married women who had RA also tended to be immersed in marriages that had high status stress.19 In another study related to status stress, Shekelle and colleagues20 noted that risk of new coronary disease among men in a prospective study was associated with discrep- ancy between their social class at the time of the study and either their own or their wives’ social class in childhood.

Syme and coworkers,21 in research using urban male subjects from the California Health Survey, reported that cultural mobility was associated with CHD. Cultural mobility was defined as moving from one social setting to another or remaining stable within a given social setting while the setting itself undergoes change. In a case-control study, Syme et al. operationalized cultural mobility as cultural discontinuity and occupational mobility. Among male prog- eny of foreign-born fathers, sons who had college-level education had observed- to-expected (O/E) CHD ratios that were five times higher than those of sons who had completed only grade school or high school. Among college-educated men, sons of foreign-born fathers had O/E CHD ratios that were more than two times higher than those of sons of native-born fathers. Regarding occupational mobility,

51589_CH15_Printer.indd 664 09/02/13 4:47 PM

i n d e p e n d e n t v a r i a B l e S 665

men who held three or more different jobs for brief time periods during their lifetimes had O/E CHD ratios that were four times as high as the ratios of those who held only one or two jobs. Thus, in an urban setting, Syme et al. replicated findings regarding cultural mobility and increased CHD risk observed previously in rural areas.21

The Person-Environment Fit Model The late John French and associates defined person-environment fit “as the goodness of fit between the characteristics of the person and the properties of [his or her] environment.”23(p 316) Originally formulated to conceptualize various aspects of mental health, such as adjustment and coping, the person-environment fit model is applicable also to the etiology of physiologic illness.22 As person- environment fit decreases, adjustment to the environment decreases, and the individual’s plight becomes more and more stressful.

The model further distinguishes between the subjective environment (the environment that is perceived by the individual) and the objective environment, which exists independent of the person’s perceptions. Corresponding distinc- tions are made for the person: the objective person and the subjective person (self-concept). The model specifies two dimensions of the person and, similarly, two dimensions of the environment that are incorporated in the quantification of stress. Person characteristics are needs and abilities, and environmental charac- teristics are supplies and demands.

Person-environment fit is one aspect of a larger system of variables that relate to health outcomes. The person-environment fit model portrays a web of vari- ables that incorporates precursors of illness (e.g., lack of adjustment), mediating factors, and specific illnesses, such as CHD, arthritis, or some of the infectious diseases. Some of the variables are hypothesized to be directly related to a given outcome, others are considered intervening variables, and still others operate in concert in predicting illness. No single factor is a sufficient cause of a particular disease. The model suggests interconnections among mental health factors, phys- ical health status, and psychological factors that predispose to illness, precipitate illness, and determine recovery rates.22

Work overload provides one example of lack of person-environment fit. A defi- nition of work overload is an inability of the individual to meet demands emanat- ing from the environment (i.e., it is the result of a discrepancy between demands from the work environment and the capacity of an individual to meet those demands). Work overload contributes to a stressful state (lack of adjustment to

51589_CH15_Printer.indd 665 09/02/13 4:47 PM

666 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

the work environment) that may culminate in adverse physical and mental health outcomes. A representative case is the piece worker who is required to produce 50 widgets per hour but has the ability to produce only 25 widgets. Another example would be an overloaded executive who may have to field more work responsibili- ties than the executive feels capable of handling; the theory of person-environment fit posits that, as a result, the person is more prone than a well-adjusted executive to heart attacks, other chronic diseases, or other health problems.

Empirical studies of the person-fit model were conducted in occupational environments in which the major outcome was job dissatisfaction; findings dem- onstrated that poor person-environment fit correlated significantly with job dis- satisfaction.22 An experimental study of work overload found increases in serum cholesterol (a risk factor for CHD) among subjects who were faced with work overload.24

A second hypothetical example of poor person-environment fit is a lack of agreement between the needs of the person and supplies in the environment. For example, one may consider person-environment fit in the area of affilia- tion (the need to be with other people). If a person who has a high affiliation need has to work in isolation (e.g., by becoming a security guard at night), the person will experience stress with respect to the need for affiliation. Other workers may feel dissatisfied with their job setting because they are required to perform duties for which they are overqualified or for which their skills are underutilized.

With respect to this second example, the model theorizes that the relation- ship between stress and lack of person-environment fit is curvilinear. When a person’s needs are exactly supplied, stress is at a minimum. However, stress may result under either of the following two conditions: oversupply or undersupply of gratifications. Let us consider a hypothetical example by returning to the need for affiliation. Suppose an individual (X) desires to spend a certain time with a friend, partner, or significant other (Y)—say about two hours per day. X will experience stress if Y wants to spend less than 2 hours with X. At the other extreme, if Y wants to spend more than 2 hours with X, then X also will be under stress. A similar concept of oversupply of gratifications is called affluenza— malaise caused by too much wealth.

Stressful Life Events Stressful [life] events are defined as “. . . occurrences that [are] likely to bring about readjustment-requiring changes in people's usual activities.”25(p 477) This research field postulates a relationship between the happenings in one’s life and

51589_CH15_Printer.indd 666 09/02/13 4:47 PM

i n d e p e n d e n t v a r i a B l e S 667

the development of illness. Two crucial issues of life events research are, first, to determine which attributes distinguish more stressful from less stressful life events and, second, to refine the knowledge base regarding the pathologic effects of stressful life events.

Holmes and Rahe26 developed the Social Readjustment Rating Scale, which comprised 43 life event items. Research suggested that the items should be rank ordered in terms of importance to the individual; for example, the death of a spouse was found to be most stressful and was given the highest weight, 100 points; pregnancy was given a weight of 40; and minor violations of the law were given a weight of 11. The more severe the life change event and the higher the frequency of the event, the greater the chance that severe disease will occur. Holmes and Rahe formulated the following 10 leading life change events during the late 1960s26:

1. death of a spouse 2. divorce 3. marital separation 4. jail term 5. death of a close family member 6. personal injury or illness 7. marriage 8. being fired from a job 9. marital reconciliation

10. retirement

Holmes and Masuda27 reported that the greater the magnitude of life event as measured by the scale, the greater the probability that it would be associated with disease. In addition, there was an association between the magnitude of life change and the seriousness of illness. These investigators suggested that life stresses lower the resistance to disease and that the greater the stress or stresses in the person’s life, the more severe the illness that may develop.

Langner and Michael28 studied the association of life stresses with risk of men- tal disorders among a representative population of 1,660 residents of midtown Manhattan in New York City. Subjects were administered a carefully designed interview that contained items about childhood and adult stress factors, psychi- atric status, and other hypothesized risk factors for mental disorders. Examples of items were poor physical health as an adult or quarrels between parents during childhood. It was found that the greater the number of negative life factors, the greater the mental health risk.

51589_CH15_Printer.indd 667 09/02/13 4:47 PM

668 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Hinkle29 reported on the frequency of disabling illness for a 20-year period that occurred among a group of career telephone operators who had been employed steadily in semiskilled occupations and among a group of blue-collar workers who worked for a similar division of the same company. Some employ- ees had a much greater risk than others of becoming disabled, and, among those who had the greatest risk, there was also a greater likelihood of recurrent illnesses and more severe disability. Susceptibility to illness seemed to exacerbate the effect of stresses in the life of the individual. Among those who were not susceptible to illness, stressful life events did not seem to produce a decrement in health. There was also a tendency for the workers who were most frequently ill to be those who had occupations that seemed to be out of line with their educational and social backgrounds. For example, female telephone operators from blue-collar back- grounds were likely to be well adjusted and healthy, but college-educated women with the same job title expressed job dissatisfaction and were also more likely to experience frequent illness.

Characteristics of life events that are most salient as stressors include desirabil- ity of an event, control, and required readjustment.30 Methodologic problems in life events research include subjects’ recall ability, memory biases, reliability of measurement, and possible interconnectedness among life events.

The Stress Process Model The stress process model defines stress as a process that occurs over a course of time, with stressful events chaining from one to another and interconnectedness among these stressful events. Sociologist Leonard I. Pearlin proposed the stress process model as an organizing and orienting framework for the diverse themes of stress research.31 The model would be heuristic by guiding researchers toward potentially useful lines of research, highlighting needed data for research, and aiding in the interpretation of results.

One of the important features of the model is that it emphasizes the inter- relatedness of factors that play a role in the individual’s health and well-being. Some of the factors include the contexts of persons’ lives, their social statuses, their exposure to stress, the resources that they have available to deal with stressors, and the outcomes of stress exposure, both somatic and mental. For example, consider the role of social and economic status in the stress process. Those who are at the lower end of the economic hierarchy may face a life of fear and uncertainty or may lack economic resources. These circumstances, in turn, may be exacerbated by the individual’s neighborhood context should it be unsafe, deteriorated, or unstable. Some of the problems that arise from a stressful

51589_CH15_Printer.indd 668 09/02/13 4:47 PM

m o d e r a t i n g v a r i a B l e S 669

neighborhood context, such as lack of access to needed services, are known as daily hassles.

In addition to the individual’s social status and neighborhood context, the stress process model incorporates life events (defined earlier) and chronic or repeated strains. Examples of strains include life-course issues, for example, the need for spouses or children to assume long-term caregiving roles for elderly loved ones or parents. Finally, the model specifies moderating resources, such as coping and social support (covered later in this chapter), and stress outcomes, such as mental disorders (also covered later in the chapter). The epidemiologic features of depression and depressive symptoms indicate that these disorders are unequally distributed according to socioeconomic status, gender, marital sta- tus, and age. Turner and Lloyd provided empirical support for the stress process model with respect to the association of socioeconomic status with depressive symptoms.32

The timing and sequencing of life course transitions are salient for the stress process. Pearlin writes that “. . . it may be more difficult for 50-or 60- year-old workers who have lost their jobs to reenter the labor force than others half that age. Similarly, if unemployed persons also have family who depend on their earn- ings, they, too, may experience more hardship than out-of-work people who have not yet taken on family roles.”33

Moderating Variables

Previously, we defined moderating (intervening) variables as those that are intermediate in the causal process between risk factors and outcomes. Examples of moderating variables are the type A behavior pattern, personal behaviors and lifestyle, and supportive interpersonal relationships. Personality variables also may have a moderating effect upon health outcomes by affecting how individu- als respond to and cope with stress. For example, personality hardiness has been posited as a resistance resource that moderates the relationship between stressful life events and illness outcomes. Kobasa et al. wrote that “hardiness is consid- ered a personality style consisting of the interrelated orientations of commitment (vs. alienation), control (vs. powerlessness), and challenge (vs. threat). Persons high in commitment find it easy to involve themselves actively in whatever they are doing, being generally curious about and interested in activities, things, and people. . . . Persons high in control believe and act as if they can influence the events taking place around them through what they imagine, say, and do. . . . Challenge involves the expectation that life will change and that the changes

51589_CH15_Printer.indd 669 09/02/13 4:47 PM

670 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

will be a stimulus to personal development.”34(p 525) Kobasa et al. reported that hardiness was the most important of resistance resources studied—including social support and exercise—in predicting the probability of illness among male business executives.

Type A (Coronary-Prone) Behavior Pattern The type A behavior pattern is defined as a behavioral syndrome that includes the traits of aggressiveness, ambition, drive, competitiveness, and time urgency. The behavior pattern has been found to be associated with CHD in prospec- tive, cross-sectional, and retrospective studies. Researchers have focused upon clinical assessment of the behavior pattern through interview techniques and the development of a self-administered questionnaire measure. Rosenman and colleagues35 pioneered work with the concept; Jenkins subsequently enlarged research with the coronary-prone behavior pattern, which he conceptualized as an “overt behavioral syndrome or style of living . . . to include restlessness, hyperalertness, and explosiveness of speech.”36(p 255)

Interview measure of Type A Rosenman et al.37 measured the type A syndrome by means of a structured clini- cal interview that was administered to a sample of 3,524 men, aged 39–59 years. Known as the Western Collaborative Group Study, the research prospectively followed the incidence of CHD among men in various occupations, beginning in 1960. Interview questions were designed to measure the several dimensions of the type A personality outlined above: drive, ambition, competitiveness, aggres- siveness, hostility, and a sense of time urgency; motor and speech characteris- tics also were noted. The investigators, sometimes in collaboration with other researchers, have published numerous reports that contain data supportive of a significant and positive relationship between the type A personality and increased frequency of CHD.

Self-administered measure of Type A To measure the type A behavior pattern objectively, several researchers reported the development of a self-administered checklist. Purported advantages of a checklist measure over a structured interview are greater standardization of research procedures in replication studies, ease of administration, and elimi- nation of possible interviewer biases. Bortner38 reported the development of a short rating scale (14 items) for the behavior pattern. The measure discriminated

51589_CH15_Printer.indd 670 09/02/13 4:47 PM

m o d e r a t i n g v a r i a B l e S 671

significantly between two groups of male workers who previously had been classified as behavior pattern A or B by the interview measure. (Type B refers to those people who do not have the type A pattern.) Jenkins36 developed a self-administered, machine-scored test known as the Jenkins Activity Scale to measure the type A personality.

Social Support Social support is defined as supportive relationships that arise from friends, family members, and others. Cobb39 suggested that social support may moderate the effects of stress. Although the term social support refers to perceived emotional support that one receives from social relationships, the term social network ties is a quantitative concept that refers to the number (and, in some cases, the pattern) of ties that one has with other people or organizations. Social support systems are hypothesized to operate as mediators that serve as buffers against stress.40 Research has focused on the stress-buffering effects of social network ties and social support. For example, it has been hypothesized that social support may enhance immune status. Cancer patients’ spouses, who would be presumed to have severe, chronic life stresses, demonstrated an association between social sup- port and immune status; those who had higher levels of perceived social support tended to have better indexes of immune function.41

Social networks refer to the structure of people’s social attachments42 and are considered mediators between stressors and health outcomes because social networks help to explain differences in the individual’s ability to respond to stressors. At one time, social networks were primarily a topic of interest to researchers. Presently, social network websites captivate popular enthusiasm as indicated by the success of Facebook and other influential social networking websites. Figure 15–8 gives a diagram of a social network. The implication of the figure is that every individual is part of a social network and has many ties to other individuals who form the individual's network.

The buffering model of social support hypothesizes that one of the functions of social network ties is to lessen the adverse psychological consequences of stress.43 The underlying assumption is that support and resources provided by social network ties may act as a buffer against the potentially harmful effects (e.g., depression) of stressful life events.44 Conversely, lack of family social sup- port and close affectional ties contributes to vulnerability, onset, and severity of psychological stress.45 Variations in the effects of support are thought to be related to the source of support.46 For example, support from spouses or friends may be more important than support derived from other network ties.

51589_CH15_Printer.indd 671 09/02/13 4:47 PM

672 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Supportive relationships may be deduced from social networks, including ties with family and friends and memberships in formal and informal organiza- tions. Marital status has been found to be salient in the social support process; married older adults have more contact with family members than with friends and receive more emotional support than unmarried older adults.47 In addition, social contact, received emotional support, and anticipated support are inter- related. Increased social contact is associated with increased emotional support and perceptions of support availability.46 The specificity hypothesis postulates that interpersonal relationships provide a stress-buffering effect when there is concordance between coping requirements demanded by a particular stressor and specific types of support provided.48

Personal Behavior, Lifestyle, and Health In this section, we include characteristics such as personal risk taking, smok- ing, alcohol consumption, choice of diet, and exercise levels. Breslow49 sum- marized the results from the Human Population Laboratory in Alameda County, California, where investigators observed a positive association between seven healthful habits and physical health status and longevity. The seven habits were

Figure 15–8 A diagram of a social network. Source: Courtesy of d3images.

51589_CH15_Printer.indd 672 09/02/13 4:47 PM

m o d e r a t i n g v a r i a B l e S 673

moderate food intake, eating regularly, eating breakfast, not smoking cigarettes, moderate or no use of alcohol, moderate exercise, and 7 to 8 hours of sleep daily. There was a direct correlation between the number of healthful habits followed and good health status and reduced mortality.

Behavior and lifestyle are related to a number of diseases that are of major importance to modern Western civilization. Burkitt50 argued that these include noninfectious diseases of the large bowel, venous disorders, and obesity. Appendicitis, diverticular diseases, cancer of the large bowel, pulmonary embo- lism, gallbladder disease, ischemic heart disease, and diabetes are relatively com- mon diseases in the United States and Britain. Burkitt stated that these same conditions are uncommon in developing countries and were rare a century ago in Western nations. Suchman51 referred to the chronic, degenerative diseases as way-of-life diseases because of their closer relationship to human behavior than to any bacteriologic or infectious agent.

Personal behavior is related to unintentional injuries (accidents), the fifth major cause of death in 2007 in the United States.52 Mortality and morbid- ity from unintentional injuries are largely, if not completely, preventable. Consequently, the use of the term accidents is not encouraged. Unintentional injuries and deaths are not randomly distributed although they might appear to be unpredicted, unexpected events. Rather, they tend to be more common among individuals with certain identifiable host characteristics, such as sex, age, choice of occupation, and safety practices, which are influenced by attitudinal and behavioral variables in relation to risk taking.51

Personal behavior is intimately connected with personal health status and is a function of both psychological and contemporary sociocultural influences. For example, there is great variety in the amount of personal risk that one may want to assume in the conduct of daily existence, from minimal risk to such high-risk activities as motorcycle riding, sky diving, and hang gliding, which place one at direct risk of injury and death. Sexual behavior, dietary practices, smoking, alcohol consumption, method of infant feeding, and choice of occu- pation are all components of personal behavior that affect health and are gov- erned by personality constitution, cultural influences, and the prevailing social climate. Personal behavior is of such importance that the U.S. National Center for Chronic Disease Prevention and Health Promotion established a behavioral risk factor surveillance program. The purpose of this program was to modify behavioral risk factors to accomplish national health objectives. The objective areas for the behavioral risk factor surveillance program included obesity, lack of physical activity, smoking, safety belt use, and medical screening for breast and

51589_CH15_Printer.indd 673 09/02/13 4:47 PM

674 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

cervical cancer and elevated blood cholesterol. The surveillance program was a state-based, random-digit dialing telephone survey.53

Healthy People 2000 established a framework in the United States for national health promotion and disease prevention objectives. Several of these objectives relate to health promotion through encouragement of a desirable lifestyle. The most recent (2012) document, Healthy People 2020, builds on earlier initiatives such as Healthy People 2000 and Healthy People 2010. This latest document articulates four major overarching goals shown in Figure 15–9. Topic areas of Healthy People 2020 include nutrition, physical activity, obesity, substance abuse, and tobacco use. The topics acknowledge the awareness of the U.S. Department of Health and Human Services regarding the significant function that lifestyle plays in health (see Exhibit 15–2).

Smoking and Health Three major reports by the U.S. Surgeon General—the 1964 report,54 the 1979 report,55 and the 2004 report56—summarized conclusions regarding the rela- tionship between smoking and various adverse health consequences. Epidemio- logic research suggests that smoking is a significant cause of excess mortality and morbidity. For example:

●● Current cigarette smokers in comparison with nonsmokers have an overall 70% excess mortality regardless of amount of smoking.

●● Mortality from smoking increases with the quantity of cigarettes smoked; mortality is increased by duration of smoking, starting at earlier ages, and amount of smoke inhaled.

Figure 15–9 Graphic model of Healthy People 2020. Source: Reproduced from U.S. Department of Health and Human Services, Office of Disease Prevention and Health Promotion. Healthy People 2020 Framework, p. 3.

Overarching goals:

• Attain high quality, longer lives free of preventable disease, disability, injury, and premature death. • Achieve health equity, eliminate disparities, and improve the health of all groups.

• Create social and physical environments that promote good health for all. • Promote quality of life, healthy development, and healthy behaviors across all life stages.

Health outcomes

Determinants

Physical environment

Social environment

Individual behavior

Biology & Genetics

Health services

51589_CH15_Printer.indd 674 09/02/13 4:47 PM

m o d e r a t i n g v a r i a B l e S 675

What Is Healthy People ?

Healthy People is a set of goals and objectives with 10-year targets designed to guide national health promotion and disease prevention efforts to improve the health of all people in the United States. Healthy People is used as a tool for strategic management by the federal government, states, communities, and many other public- and private-sector partners. Its comprehensive set of objec- tives and targets is used to measure progress for health issues in

specific populations, and serves as (1) a foundation for prevention and wellness activities across various sectors and within the federal government and (2) a model for measurement at the state and local levels.

New to Healthy People 2020: Leading Health Indicators Healthy People 2020 includes a small set of high-priority health issues

that represent significant threats to the public’s health. Selected from the Healthy People 2020 objectives, the 26 Leading Health Indicators (LHIs), organized under 12 topic areas, address determinants of health that pro- mote quality of life, healthy behaviors, and healthy development across all life stages. The LHIs provide a way to assess the health of the nation for key areas, facilitate collaboration across diverse sectors, and motivate action at the national, state, and local levels. n

Table 15–1 Topic Areas and Leading Health Indicators

12 Topic Areas 26 Leading Health Indicators

Access to Health Services ●● Persons with medical insurance ●● Persons with a usual primary care provider

Clinical Preventive Services ●● Adults who receive a colorectal cancer screening based on the most recent guidelines

●● Adults with hypertension whose blood pressure is under control

●● Adult diabetic population with an A1c value greater than 9 percent

●● Children aged 19–35 months who receive the recom- mended doses of diphtheria, tetanus, and pertussis (DTaP); polio; measles, mumps, and rubella (MMR); Haemophilus influenza type b (Hib); hepatitis B; vari- cella; and pneumococcal conjugate (PCV) vaccines

Environmental Quality ●● Air Quality Index (AQI) exceeding 100 ●● Children aged 3–11 years exposed to secondhand

smoke

e x

h Ib

It 1

5 –2

continues

51589_CH15_Printer.indd 675 09/02/13 4:47 PM

676 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

●● Associations have been observed between smoking and morbidity from cardiovascular diseases; cancer of the lung, larynx, mouth, bladder, and pancreas; and non neoplastic bronchopulmonary diseases.

●● There are interactive and synergistic effects of smoking and occupational exposures to asbestos, chromium, nickel, and other potentially toxic or car- cinogenic materials.

12 Topic Areas 26 Leading Health Indicators

Injury and Violence ●● Fatal injuries ●● Homicides

Maternal, Infant, and Child Health

●● Infant deaths ●● Preterm births

Mental Health ●● Suicides ●● Adolescents who experience major depressive

episodes (MDEs)

Nutrition, Physical Activity, and Obesity

●● Adults who meet current Federal physical activity guidelines for aerobic physical activity and muscle-strengthening activity

●● Adults who are obese ●● Children and adolescents who are considered obese ●● Total vegetable intake for persons aged 2 years and

older

Oral Health ●● Persons aged 2 years and older who used the oral healthcare system in the past 12 months

Reproductive and Sexual Health

●● Sexually active females aged 15–44 years who received reproductive health services in the past 12 months

●● Persons living with HIV who know their serostatus

Social Determinants ●● Students who graduate with a regular diploma 4 years after starting ninth grade

Substance Abuse ●● Adolescents using alcohol or any illicit drugs during the past 30 days

●● Adults engaging in binge drinking during the past 30 days

Tobacco ●● Adults who are current cigarette smokers ●● Adolescents who smoked cigarettes in the past 30 days

Source: Modified from Healthy People 2020, Office of Disease Prevention and Health Promotion, U.S. Department of Health and Human Services. ODPHP Publication No. B0132, November 2010.

exhibit 15–2 continued

table 15–1 continued

51589_CH15_Printer.indd 676 09/02/13 4:47 PM

m o d e r a t i n g v a r i a B l e S 677

●● Smoking is a direct cause of reduction in birth weight and increased prenatal mortality; nicotine is found in the breast milk of mothers who smoke.

With respect to the behavioral aspects of smoking, the 1979 Surgeon General’s report concluded that the reason why the smoking habit is so widespread and difficult to break is largely unknown: “It is no exaggeration to say that smoking is the prototypical substance abuse dependency and that improved knowledge of this process holds great promise for the prevention of risk. Establishment and maintenance of the smoking habit are, obviously, prerequisite to the risk, and cessation of smoking can eliminate or greatly reduce the health threat.”55(pp 1–32) Since the 1979 report, much additional information has been acquired regarding the nature of nicotine addiction. The Surgeon General’s 2004 report identified “. . . a substantial number of diseases found to be caused by smoking that were not previously causally associated with smoking: cancers of the stomach, uterine cervix, pancreas, and kidney; acute myeloid leukemia; pneumonia; abdominal aortic aneurysm; cataract; and periodontitis.”56(p 1)

Alcohol Consumption Excessive alcohol consumption is a risk factor for specific diseases (e.g., various cancers, liver cirrhosis, and gastric disorders); increases the likelihood of involve- ment in motor vehicle crashes and other unintentional injuries; and is associated with deterioration of the social environment (e.g., interpersonal violence, family strife, and lessened job performance). In 2004, the World Health Organization (WHO) reported that approximately 2 billion people in the world partake of alcoholic beverages and that almost 80 million individuals have diagnosable alcohol use disorders. WHO highlighted the negative implications of excessive alcohol consumption for people’s health and the functioning of society.57

In the United States, over consumption of alcohol was recorded as the third leading preventable cause of death in 2001 and accounted for more than 75,000 deaths and 2.3 million years of potential life lost (YPLL).58 That number is about half the total YPLL from smoking in 1999, the most recent year for which comparative data were available. Slightly more than 40,000 deaths were from alcohol-related injuries attributed to binge drinking. (Refer to Table 15–1.)

For the population subgroup aged 12–20 years, excessive alcohol consump- tion was a factor in the three leading causes of death—unintentional injury, homicide, and suicide.59 Promotion of alcohol use by alcoholic beverage man- ufacturers in advertisements placed in magazines that have a 15–30% youth readership has remained a common practice.

51589_CH15_Printer.indd 677 09/02/13 4:47 PM

678 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Pregnant women’s alcohol consumption is related to fetal alcohol syndrome (FAS), which is characterized by postnatal growth deficiency, mental retarda- tion, and various physical abnormalities.60 According to Streissguth,60 FAS presents as a set of specific characteristics: small stature, small head, small eyes, flattened nasal bridge, and a thin, narrow upper lip. In addition to FAS, there is an association between low to moderate intake of alcohol during pregnancy and low birth weight. The sequelae of FAS last a lifetime, even though the effects may range from mild to severe.61 As a result of pervasive alcohol consumption among women of childbearing age (more than 50%), there is a high likelihood of exposure to alcohol during pregnancy. Additional health education programs are needed to inform women of childbearing age about FAS and other hazards of alcohol consumption during pregnancy.

Dietary Practices Dietary practices refer to one's usual consumption of foods including saturated fats, sugar, sodium, fresh fruits, and fresh vegetables. Our dietary preferences are related to whether we will become overweight or develop chronic diseases such as CHD, diabetes, and cancer. Figure 15–10 depicts a woman who is making a healthy dietary choice by eating an apple.

Dietary choices are an aspect of personal behavior that is related to sociocul- tural influences governed by level of economic development. In affluent societies, people tend to consume highly refined and processed foods and large amounts of protein from animal sources. Burkitt50 noted the association between the con- sumption of refined carbohydrate foods and obesity and diabetes. He suggested that lack of fiber in the food of Western diets is related to diseases of the bowel, such as colon cancer and diverticular disease.

The diet–heart hypothesis suggests that a diet high in saturated fats and cholesterol is linked to high blood lipids, which are in turn associated with arte- riosclerosis and heart disease. Low levels of high-density lipoproteins and high levels of low-density lipoproteins in the blood are associated with heart disease. Diets that maintain a high ratio of polyunsaturated to saturated fat lower risk of CHD, a phenomenon supported by findings from studies of California Seventh- Day Adventists. About half the membership consumes a nonmeat, lacto-ovo vegetarian diet. Adult members have CHD mortality rates that are about 30% of the rate for an equivalent age group within the total California population.

Dietary practices such as high consumption of fats and low consumption of fruits and vegetables are hypothesized to be related to cancer incidence. Consumption of animal fats, a high-meat diet, and vegetables has been studied

51589_CH15_Printer.indd 678 09/02/13 4:47 PM

m o d e r a t i n g v a r i a B l e S 679

in relationship to colon cancer.62 The linkage between low blood cholesterol and risk of developing cancer also has been investigated. Dietary fat may play a role in mediating the relationship between blood cholesterol and cancer.63 Among residents of Shanghai, China, a case-control study determined that con- sumption of fruits, certain dark green/yellow vegetables, and garlic was a protec- tive factor for laryngeal cancer, whereas intake of salt-preserved meat and fish increased risk.64 Another factor studied in relation to cancer is b-carotene, for which investigators hypothesize a protective role. Low serum b-carotene may be associated with cancers of the lung, stomach, cervix, esophagus, small intestine, and uterus.65

Coffee and tea consumption as factors in morbidity and mortality from CHD and other chronic diseases has been studied extensively. Although the findings

Figure 15–10 Healthy dietary choice. Source: Reproduced from Centers for Disease Control and Prevention. Public Health Image Library. Image number 13637. Available at http://phil.cdc. gov/phil/. Accessed March 21, 2012.

51589_CH15_Printer.indd 679 09/02/13 4:47 PM

680 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

are conflicting for the association of coffee drinking with cardiovascular disease, there does appear to be an indirect connection through association with adverse lifestyle factors. An Austrian study reported that coffee drinking was related to lifestyle factors—smoking, drinking, eating, and lack of physical activity—that could increase risk of cardiovascular disease.66 The reverse of this association was found between tea drinking and lifestyle factors.

One of the crucial diet-related issues in the United States is obesity (often referred to as a modern epidemic), given the association of overweight and obesity with diabetes, several types of cancer, and CHD risk factors. The rising obesity rates in the United States have become a dominant concern of public health practitioners,

Increased consumption of fruits and vegetables helps to reduce occurrence of obesity and is a desirable health practice in general. Currently, only about one-third of the population consumes fruit (e.g., apples) two or more times per day and vegetables three or more times per day. Although school cafeterias, fast food restaurants, and grocery chains are increasing the availability of low-fat and healthful food choices, many individuals prefer high-fat foods and large por- tion sizes. Often, foods sold by fast food restaurants contain unhealthy trans fats, which are made by hydrogenating oils and are known to increase the risk of CHD. In 2007 New York City banned the use of trans fats in foods sold in restaurants. As a result of the ban, fast food restaurant patrons in the city have reduced their intake of trans fats.67 The American Medical Association advocates for legislation that would implement nationwide bans on the use of artificial trans fats in restaurants and bakeries.68

Sedentary Lifestyle Sedentary Western existence, with its use of labor-saving devices and reduced level of physical activity, is identified as a risk factor for obesity, CHD, type 2 diabetes, osteoporosis, cancer, and many other conditions. Major research studies have documented the positive effects of physical activity among diverse groups, e.g., workers, college alumni, and the elderly. Figure 15–11 shows a fitness cen- ter operated by one of the CDC’s campuses.

Morris et al.69 examined the leisure time activities of 17,000 male executive- grade British civil servant office workers aged 40–64 years. Workers who partici- pated in vigorous active recreations, such as swimming and heavy gardening, had about one-third the incidence of CHD as the less active workers. Light exercise that did not have a training effect on the cardiovascular system did not reduce

51589_CH15_Printer.indd 680 09/02/13 4:47 PM

m o d e r a t i n g v a r i a B l e S 681

the incidence of heart disease. The results were interpreted as demonstrating that vigorous exercise promotes cardiovascular health.

Paffenbarger et al.70 corroborated these results in a study of about 17,000 Harvard male alumni aged 35–74. Risk of first heart attack was inversely related to involve- ment in vigorous physical exercise (e.g., stair climbing, walking, and strenuous ath- letics). Those who were college athletes and discontinued exercise during later life were at greater risk of heart attack than adults who began exercising at a later age.

Streja and Mymin71 designed an exercise program to determine whether there is a relationship between exercise and cholesterol level among sedentary persons. They observed that low levels of high-density lipoprotein cholesterol (HDL-C, the so-called good cholesterol) have been shown to precede arteriosclerosis and that athletes in comparison with control subjects have high levels of HDL-C.

Figure 15–11 People exercising at a fitness center. Source: Reproduced from Centers for Disease Control and Prevention. Public Health Image Library. Image number 13982. Available at http://phil.cdc. gov/phil/. Accessed May 2, 2012.

51589_CH15_Printer.indd 681 09/02/13 4:47 PM

682 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

In an exercise program that consisted of walking and slow jogging, a small sample of middle-aged men with coronary artery disease had increased levels of HDL-C. The results of the study suggest that exercise programs may retard the develop- ment of arteriosclerosis.

Many epidemiologic investigations have followed the early work of Morris et al. on the role of physical activity in health. These later studies have incor- porated diverse populations in terms of sex, ethnicity, age, social composition, and geographic location. In their follow-up study, Paffenbarger et al. concluded that, “These studies have extended and amplified those by the Morris group, thereby helping to solidify the cause-and-effect evidence that exercise protects against heart disease and averts premature mortality.”72(p 1184) Not only does exercise benefit the general population, but it also promotes physical functioning in subgroups of the population, for example, persons with disabilities73 and the elderly.74

Sociocultural Influences on Health One of the concerns of contemporary epidemiologic research is the role of social and cultural factors in health. Epidemiologic research has produced a volumi- nous literature regarding the effects of social, cultural, and psychosocial factors in the etiology of disease.75 Culture may be defined as the set of values to which a group of people subscribes, as the way of life of a group of people, or as the totality of what is learned and shared through interaction of the members of a society. Specific behaviors associated with a particular culture have implications for the health of the individual. In support of this notion, Susser et al. wrote, “Habits that affect health, in childbearing and midwifery, in nutrition and in daily living, are not merely the negative result of ignorance among people who know no better. They are often an intrinsic element in a way of life, customs that have positive value and symbolic significance.”76(pp 152–153)

One explanation for the role that cultural factors play in health is that they may mediate the amount of stress to which the individual is exposed. Matsumoto77 observed that Japan historically has had one of the lowest rates of CHD in the world and that the United States has had one of the highest. He hypothesized:

The etiology of coronary heart disease is multiple and complex, but in urban-industrial Japan, the in-group work community of the individual, with its institutional stress-reducing strate- gies, plays an important role in decreasing the frequency of the disease. . . . Deleterious circumstances of life need not be expressed

51589_CH15_Printer.indd 682 09/02/13 4:47 PM

m o d e r a t i n g v a r i a B l e S 683

in malfunctioning of the physiologic or psychological systems if a meaningful social group is available through which the individual can derive emotional support and understanding.77(p 14)

Marmot and colleagues78 studied a large population of men of Japanese ancestry: 2,141 men who were being followed by the Atomic Bomb Casualty Commission in Hiroshima and Nagasaki, Japan; 8,006 men in Honolulu; and 1,844 men in the San Francisco Bay area. Japanese who lived in California had a higher prevalence of CHD and its manifestations than those who resided in Hawaii or Japan. The Hawaiian Japanese tended to have higher CHD rates than residents of Japan. The investigators speculated that differences in prevalence of CHD may have been due to differences in the way of life in Japan and the United States. For example, there are major variations between the two countries in diet, occupation, and the social and cultural milieu.

Studies have pursued the role of sociocultural factors in mental retardation. Mercer79 suggested that excessively large numbers of individuals from minor- ity backgrounds were being labeled by the public schools as mentally retarded because available standardized tests of intelligence did not adequately take into account the background of the students. The dimensions that are measured on IQ and other tests are taken from the white, middle-class society and do not constitute a culture-free measure.

Both utilization of health services and, in fact, the very definition of illness are related to cultural background and show variation from person to person. According to Mechanic, illness refers to objective symptoms, whereas illness behavior “refers to the varying perceptions, thoughts, feelings, and acts affect- ing the personal and social meaning of symptoms, illness, disabilities and their consequences. . . . [Some people will] make light of symptoms and impairments. Others magnify even minimal problems, allowing them to affect their life adjust- ments substantially.”80(p 79) It is possible that those who seek medical care readily may, over the long run, experience increased life expectancy through early iden- tification of potentially life-threatening illness.

In citing differences in preference for type of medical services by cultural group membership, Mechanic81 stated that some Mexican Americans might pre- fer folk medicine and family care. Non-Hispanic whites, however, would show a predilection for modern medical services in a technologically advanced medi- cal center. Elsewhere, Mechanic wrote, “Illness perception and response may be socially learned patterns developed early in life as a result of exposure to particular cultural styles, ethnic values, or sex role socialization.”80(p 79)

51589_CH15_Printer.indd 683 09/02/13 4:47 PM

684 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Social and cultural factors are related to the successful control of communicable diseases; a notable example is control of tuberculosis among immigrants to the United States from Mexico. People of Mexican descent with tuberculosis showed delayed response in seeking medical attention. These delays were attributed to diagnosis by a layperson of the symptoms of the folk illness susto, a condition not considered susceptible to the ministrations of physicians. Among undocu- mented Mexican workers residing in Orange County, California, the average delay between acknowledgment of tuberculosis symptoms and the presentation of a complaint to a physician was 8.5 months.82

Dependent (Outcome) Variables: Physical and Mental Health

Psychosocial epidemiology covers both physical and mental health outcomes, examples of the former being specific chronic diseases such as CHD. One set of mental health outcomes that has been researched extensively embraces affective states (having to do with feelings and emotion), e.g., life and job dissatisfaction and depression. This section covers the following topics: the effects of dissat- isfaction with one’s life and career; mental health and stressors; psychological factors and cancer; personality and smoking; and psychosocial aspects of employ- ment. We also point out that physical and mental dimensions of health have been shown to overlap: Psychiatric and physical disorders are risk factors for each other.83 An illustration is the person with a chronic disease who also experiences concurrent adverse mental health effects. Moreover, one’s habitual mental out- look tends to be associated with one’s physical health status and longevity.

Life and Job Dissatisfaction According to Jenkins, “The hypothesis that life dissatisfaction is a risk factor for coronary disease is a promising one and deserves careful examination in prospec- tive studies.”36(p 254) One aspect of life dissatisfaction that is increasingly shown to be related to coronary disease is job dissatisfaction. Empirical findings suggest that life and job dissatisfaction are directly related to morbidity and mortality from CHD. An ecologic analysis by Sales and House84 reported strong negative correlations between job satisfaction and coronary disease death rates for white- collar and blue-collar workers when the effects of social class were controlled. In a study of identical twins, Liljefors and Rahe85 similarly reported a strong asso- ciation between various life dissatisfactions, including job dissatisfactions, and

51589_CH15_Printer.indd 684 09/02/13 4:47 PM

d e p e n d e n t ( o u t C o m e ) v a r i a B l e S 685

heart disease. Other studies have implicated various themes of job dissatisfaction in coronary disease. Tedious work, feeling ill at ease at work, lack of recognition, difficulties with coworkers, demotion, and prolonged emotional strain associated with work overload have all been shown to be related to coronary disease.

Several investigators have focused upon extrinsic–intrinsic motivation (moti- vation for power and money as opposed to motivation for the internal qualities of an occupation) as a job-related motivational dimension that may be associated with CHD. Using a sample from the University of Michigan Tecumseh Study, House86 found that the association between extrinsic–intrinsic motivation and risk of CHD was conditioned by occupational status. Among white-collar work- ers, extrinsic motivation was positively related to risk of CHD; the relationship between intrinsic motivation and risk of CHD was negative. The reverse associa- tions were found for blue-collar workers.

Mental Health and Stressors Epidemiologic research has examined various mental health outcomes, such as psychological disorders (e.g., posttraumatic stress disorder [PTSD] and major depression) and affective states, as outcomes of the stress–illness paradigm. Prevalence studies have documented that depression and major depressive dis- order are extremely common adverse mental health outcomes; almost one-fifth of U.S. adults are afflicted with major depressive disorder during their lifetimes. Mental disorders (anxiety, stress, and neurotic disorders) are a substantial burden for U.S. workers in private industry, although the occurrence of disorders varies by industry sector and rates have declined over time. Between 1992 and 2001, one sector (finance, insurance, and real estate) had higher rates of anxiety, stress, and neurotic disorders than other industry sectors. (See Figure 15–12.)

PTSD is an example of a mental disorder that is associated with exposure to extremely traumatic events. PTSD victims tend to reexperience the traumatic event—whether a natural disaster, war, rape, or other trauma—and may progress in some instances to persisting psychopathology. PTSD is a potentially notewor- thy disorder for epidemiologic research because it often stems from a massive trauma that has been experienced by an entire population.87 Examples of trau- matic events include the 1994 Northridge earthquake in southern California, the 2001 terrorist attack on the World Trade Center in New York, the 2005 flooding of New Orleans caused by Hurricanes Katrina and Rita, and the 2011 tsunami in northern Japan. Still another example is the frequent occurrence of PTSD among returning war veterans, most recently those stationed in Iraq and Afghanistan. Not only do individuals who experience a natural disaster, such as

51589_CH15_Printer.indd 685 09/02/13 4:47 PM

686 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

an earthquake, vary in the severity of their responses, but there also appears to be a dose–response relationship between proximity to the traumatic event and degree of mental impairment.

Let us now turn to a unique problem for mental health epidemiology. Of particular interest to epidemiologists has been the development of an easily administered instrument to assess the prevalence of mental disorders in popula- tion surveys. Prior to the development of such a measure, some epidemiologic investigations relied on individual clinical ratings, hospital admission rates, or other utilization data to determine rates of conditions such as depression. Langner’s 22-item Index of Psychophysiologic Disorder represented one of the first efforts to design an epidemiologic measure of psychiatric impairment.88 A later instrument is the Center for Epidemiologic Studies’ depression (CES-D) scale, a brief, 20-item self-report depression symptom scale.89,90 Possible scores on the CES-D scale range from a minimum of 0 to a maximum of 60. Research studies have found it to be as reliable, sensitive, and valid a measure of depres- sive symptoms and change in depressive symptoms as clinical interview ratings. The instrument permits differentiation between acute depressives and recov- ered depressives as well as between depressives and other diagnostic groups.91 The measure has been validated in predominantly urban populations92–94 and more recently in rural populations.95 Sample items are self-reported feelings of

Figure 15–12 Annual rates of anxiety, stress, and neurotic disorder cases involving days away from work by private industry sector, 1992–2001. Source: Reproduced from National Institute for Occupational Safety and Health. Worker Health Chartbook, 2004. DHHS (NIOSH) Publication NO. 2004-146, September 2004, p. 39.

3.5

3.0

2.5

2.0

1.5

R at

e pe

r 10

,0 00

w or

ke rs

1.0

0.5

0.0 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

Finance, insurance, and real estate Transportation and public utilities Services Wholesale trade Manufacturing Retail trade Construction Mining Agriculture, forestry, and fishing

51589_CH15_Printer.indd 686 09/02/13 4:47 PM

d e p e n d e n t ( o u t C o m e ) v a r i a B l e S 687

depression, fearfulness, loneliness, and sadness. Subjects indicate the frequency with which these symptoms have occurred during the past week (range, 0 to 5–7 days; range of item scores, 0–3). A total CES-D score of 16 or greater has been defined in literature reports as the criterion for a case of depression.91

Several major surveys have examined the prevalence of self-reported symp- toms of depression as assessed by the CES-D scale. The prevalence of depres- sion in a representative sample of adults in Los Angeles County was 19%.96 Rates of depression were higher among women than men (23.5% vs. 12.9%). Depressed persons reported more physical illnesses than the nondepressed. Using data from the Hispanic Health and Nutrition Examination Survey, other inves- tigators reported a caseness (i.e., CES-D score of 16 or greater) rate for high lev- els of depressive symptoms of 13.3%; female sex, low educational achievement, low income, birth in the United States, and white-oriented acculturation of the Hispanic sample were associated with depressive symptoms.97

The National Institute of Mental Health Epidemiological Catchment Area Program was a comprehensive collaborative effort by scientists to gather data on the prevalence of mental disorders in the United States. The disorders studied were the major psychiatric illnesses classified in the third edition of the Diagnostic and Statistical Manual (a manual for the classification of mental disorders). The study was unprecedented in its scope, covering 17,000 residents of 5 community sites across the United States.98

Each year the Substance Abuse and Mental Health Services Administration (SAMHSA) collects data on the national prevalence of depression (major depres- sive disorder) as part of the National Survey on Drug Use and Health. Data on the prevalence of depression among adults and by sex for 2004–2008 are pro- vided in Figure 15–13. During this time period, the prevalence of depression among U.S. adults (part A) ranged from 7.9% in 2004 to 6.4% in 2008. Part B reveals higher rates of major depressive disorder among females in comparison with males (almost two times as high in 2008).

Premorbid Psychological Factors and Cancer Fox99 compiled a comprehensive literature review and evaluation of studies of premorbid psychological and personality factors associated with cancer. Possible deficiencies of the prospective and retrospective studies done in this field through 1977 include small sample sizes, inappropriate use of statistical tests, methodologic flaws, and possible alternative interpretations. One group of prospective studies suggested that cancer patients show lack of warm relationships with their parents and pathologic responses to the Rorschach test (a personality test that uses ink blot designs to evoke associations). Other studies mentioned in Fox’s review suggested

51589_CH15_Printer.indd 687 09/02/13 4:47 PM

688 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

Figure 15–13 Prevalence of depression among U.S. adults, 2004–2008: (Part A) By year; (Part B) By gender. Source: (A) Reproduced from National Institute of Mental Health. Statistics. Major depressive disorder among adults. http:/www.nimh.nih .gov/statistics/1MDD_ADULT.shtml. Accessed May 3, 2012. (B) Reproduced from National Institute of Mental Health. Statistics. Major depressive disorder among adults. http:/www .nimh.nih.gov/statistics/1MDD_ADULT.shtml. Accessed May 3, 2012.

(a)

7.9 7.3 7.2 7.5

6.4

P er

ce nt

o f a

du lts

0 2004 2005 2006

Year

2007 2008

MaleFemale (b)

P er

ce nt

2005

2006

2007

2008

9.3 9.0 9.5

8.1

5.2 5.3 5.3 4.6

51589_CH15_Printer.indd 688 09/02/13 4:47 PM

d e p e n d e n t ( o u t C o m e ) v a r i a B l e S 689

that women who were later found to have breast cancer deliberately repress and fail to express anger. Still another investigation reported that lung cancer patients had higher-than-average scores on the lie scale of the Minnesota Multiphasic Personal- ity Inventory, but the interpretation of this finding has been challenged.

Deficiencies of research notwithstanding, Fox99 summarized two major per- sonality types at increased risk of cancer. He portrayed the first type as yielding, compliant, and eager to please. Among the first personality type, activation of hormonal mechanisms might be associated with repression of feelings. Repressed emotion might alter immune system responses to carcinogenic agents, thereby increasing incidence of cancer in individuals with this personality type. The sec- ond type consisted of extroverted, non-neurotic individuals who tend toward heaviness. He predicted that male or female extroverts, as a result of physical lifestyle (sometimes involving excessive eating, drinking, and smoking), would have higher rates than others of colorectal, breast, lung, prostate, esophageal, and cervical cancer. The primary etiologic mechanism would be the indirect linkage between personality and cancer through lifestyle factors.

A subsequent review of the associations among psychological variables (e.g., stress, bereavement, depressed mood, mental illness, suppressed emotions, helplessness and hopelessness, and social support) and various cancer outcomes, including mortality or course of the disease, appeared approximately two decades later.100 It was concluded that the literature remains contradictory, marked by both positive findings and the absence of associations; however, the evidence against the relationship between psychological factors and cancer outcomes is most notable for stress, depressed mood, psychosis, and bereavement.

Effects of Major Diseases on Personality Not only might one conceive of personality characteristics as a cause of diseases, but one might look also at the reverse side of the coin and examine the effect of disease upon personality. Affliction with a chronic illness may become a substantial stress factor for a person and for members of his or her immediate social environment. A severe drinking problem, substance abuse, or a heart attack may induce person- ality changes in the afflicted individual and may affect other persons, including one’s children, spouse, and coworkers. A number of personality effects accompany severe illness; for example, wives of heart attack victims experience depression, fear, anxiety, and guilt.101 Wives have increasing anxiety about the future and guilty feelings about being a possible cause of the attack. Additionally, the vic- tim may experience increased feelings of depression, anxiety, guilt, and hopeless- ness. One implication for epidemiologic research is that studies should carefully

51589_CH15_Printer.indd 689 09/02/13 4:47 PM

690 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

separate out the temporality of causality. Did the personality characteristic cause the disease, or did the disease cause the personality characteristic?

Personality and Smoking An issue that has commanded the attention of epidemiologic researchers and others concerns the extent to which smoking behavior is determined by person- ality factors. The Surgeon General’s Report of 197956 indicated that personality factors that may be related to smoking behavior are extroversion, neuroticism, antisocial tendencies, and the belief that one is externally controlled (i.e., that fate, luck, or other factors beyond one’s control will bring one rewards). Smokers show a greater willingness than nonsmokers to take risks and are more impulsive, more likely to divorce and change jobs, more interested in sex, and more likely to consume tea, coffee, and alcohol.

Research conducted subsequently to the Surgeon General’s Report also sug- gested that cigarette smokers may possess distinctive personality characteristics in comparison with nonsmokers, for example, with respect to risk behaviors.102 A major review of the epidemiology of tobacco use concluded that some studies have linked cigarette smoking to several types of psychiatric disorders.102 Associa- tions between depressive states and smoking, anxiety disorders and smoking, and schizophrenia and smoking have been reported by several investigators. For exam- ple, Acton et al.103 examined smoking status and diagnosis among patients hospi- talized for psychiatric disorders. Among never smokers, rates of currently diagnosed major depressive disorder were lower than among patients who were ever smokers.

Habitual Mental Outlook and Health Status One’s prevailing attitudes toward life and one’s mental health status have been probed with respect to their association with physical health status and longevity. A major study of male mental health followed up 204 men biennially over a period of 4 decades, beginning at adolescence.104 Information regarding the mental health status as well as the physical health status of the subjects was routinely collected during the study. Among the 59 men with the best mental health between the ages of 21 and 46, only 2 developed chronic illness or died by age 53. Among 46 men with the worst mental health levels, 18 developed chronic illnesses or died. The association between mental and physical health remained statistically significant when the variables of alcohol and tobacco con- sumption, obesity, and longevity of relatives were controlled.

The foregoing research addressed the possible role of mental health and adult adjustment in men’s physical health. Aspects of mental health

51589_CH15_Printer.indd 690 09/02/13 4:47 PM

C o n C l u S i o n 691

potentially associated with physical health include habitual mental outlook, such as optimism. The self-reported health of midlife women has been shown to be positively related to optimism.105 However, the positive association between cheerfulness—one aspect of habitual mental outlook—and health (specifically longevity) has been contradicted by a study of subjects from Terman’s seven- decade longitudinal investigation of highly intelligent children.106 While the findings suggested that conscientiousness in childhood was associated with sur- vival in middle to old age, cheerfulness—characterized by optimism and sense of humor—was inversely related to longevity.

In conclusion, while research into the association between characteristics of habitual mental outlook (mental health, adult adjustment, cheerfulness, opti- mism, and sense of humor) and health status (self-reported physical health, chronic disease, and longevity) has evolved and produced some intriguing hypotheses and findings, further work is needed because of inconsistent results. For example, there is a possible need to reconceptualize the health relevance of variables such as cheerfulness.106 The role of habitual mental outlook in health may be elucidated further by prospective studies that refine measurement tech- niques and clarify the theoretical pathways through which this factor may operate.

Psychosocial Aspects of Employment and Health Stresses and other psychosocial aspects of the work environment represent an important area of investigation for occupational health epidemiologists. Researchers have probed the psychosocial aspects of the job environment in great depth. In addition to work overload described previously, examples of topics examined in research include CHD, job stresses and absenteeism due to infec- tious and chronic diseases, shift work and physical and mental health, health effects of physical activity at work, and variations in chronic disease mortality according to occupational status. A representative study reported that stress- related characteristics of work (e.g., overwork, time pressure, and high levels of mental demand) were associated with periodontal health status.107

Conclusion

The field of social epidemiology is concerned with the influence of a person’s position in the social structure upon the development of disease. Behavioral epi- demiology studies the role in health of factors such as tobacco use, physical activ- ity, and risky sexual behavior. The term psychosocial epidemiology has been more broadly conceptualized to include social, behavioral, and psychological factors.

51589_CH15_Printer.indd 691 09/02/13 4:47 PM

692 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

This field of investigation has used case-control, cohort, cross-sectional, and experimental designs to research a wide variety of health outcomes.

This chapter first examined the independent variables of stress, status incon- gruity, person-environment fit, and stressful life events. Stress as an independent, antecedent variable to health and illness represents an intriguing notion because it seems to agree with commonsense explanations for the cause of some men- tal disorders and chronic physical illnesses, yet documentation of stress as an etiologic agent of disease is inconsistent. Investigators have hypothesized that either social mobility or status incongruity may be associated with morbidity and mortality. Person-environment fit, originally formulated to conceptualize various aspects of mental health such as adjustment and coping, is another example of a social and psychological precursor to the etiology of physiologic illness. A fourth example of an independent variable is the concept of stressful life events. The central postulate of life events research is that there is a relationship between the happenings in one’s life and the development of illness.

Examples of moderating (intervening) variables in psychosocial epidemiologic research are the type A behavior pattern, personal behaviors and lifestyle, sup- portive interpersonal relationships, and social and cultural influences. Outcome variables include affective states, life and job dissatisfaction, chronic disease, and depressive symptoms. The increasing number of psychosocial epidemiologic research studies point to the need for well-delineated conceptual models, which at present have not been implemented satisfactorily.108 Also needed is the devel- opment of strong, research-based intervention programs.109

Study Questions and Exercises

1. Define the following: a. stress b. general adaptation syndrome c. social incongruity d. person-environment fit e. life events f. type A behavior pattern g. social support h. lifestyle i. depressive symptoms j. stress process model k. allostatic load

51589_CH15_Printer.indd 692 09/02/13 4:47 PM

S t u d y Q u e S t i o n S a n d e x e r C i S e S 693

2. Propose a model for the relationship between stress and illness. Be sure to choose a specific outcome and include moderating factors. How will your model operationalize stress? Draw a diagram of all major relation- ships among the variables.

3. How do one’s culture and environment relate to health? In answering this question, consider the cultural dynamics of a Western and non-Western country. Identify an immigrant group in your community and discuss how this group’s health-related practices differ from those of the larger community.

4. Give examples of two personality traits that may modify the relationship between stress and disease. Can you give your own examples of personal- ity traits that are not mentioned in the chapter?

5. What is the role of stressful life events as an influence upon disease? To what extent are each of the 10 leading life events identified during the 1960s relevant to the 21st century? Can you identify other life events that might be more salient to contemporary society?

6. Describe the association between social incongruity and chronic disease. What types of study populations would be appropriate to an epidemio- logic investigation of the health effects of social incongruity?

7. How is person-environment fit relevant to studies of occupational health and job-related stress? What type of person would adjust well to an authoritarian work environment, and what type to an unstructured environment? What is meant by a curvilinear relationship between person-environment fit and stress?

8. Give examples of how the following lifestyle variables affect health: a. alcohol and smoking b. exercise c. risk taking

9. Stress has been hypothesized to be associated with human illness. Apply Sir Austin Bradford Hill’s causal criteria to an argument for and against a causal association between stress and CHD.

10. To what extent are health outcome variables distinct or overlapping? For example, discuss the possibility that impaired mental health may be a risk factor for impaired physical health and vice versa. What types of epide- miologic study designs might be able to disentangle the time sequencing of overlapping mental disorders and impaired physical health?

11. Capstone exercise: This exercise is included here because it requires skills developed in the previous chapters of the text. Select a data-based

51589_CH15_Printer.indd 693 09/02/13 4:47 PM

694 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

article from the American Journal of Public Health, the American Journal of Epidemiology, or other public health journal of your choice. Refer to Appendix A (Guide to the Critical Appraisal of an Epidemiologic/ Public Health Research Article). Using the criteria shown in the Appendix, write a brief critique of the article. In addition to discussing the criteria sug- gested in the appendix, mention what improvements, if any, you would make in the article.

References

1. Engel G. The need for a new medical model: a challenge for biomedicine. Science. 1977;196:129–136.

2. Syme SL. Behavioral factors associated with the etiology of physical disease: a social epidemiological approach. Am J Public Health. 1974;64:1043–1045.

3. Berkman LF, Kawachi I. A historical framework for social epidemiology. In: Berkman LF, Kawachi I, eds. Social Epidemiology. New York, NY: Oxford University Press; 2000.

4. Porta M. (ed.) A Dictionary of Epidemiology, 5th ed. New York, NY: Oxford University Press; 2008.

5. Blumenthal JA, Sherwood A, Gullette EC, et al. Biobehavioral approaches to the treatment of essential hypertension. J Consult Clin Psychol. 2002;70:569–589.

6. Wittchen HU, Perkonigg A, Lachner G, Nelson CB. Early developmental stages of psychopathology study (EDSP): objectives and design. Eur Addict Res. 1998;4:18–27.

7. Lieb R, Zimmermann P, Friis RH, et al. The natural course of DSM-IV somato- form disorders and syndromes among adolescents and young adults: a prospective- longitudinal community study. Eur Psychiatry. 2002;17:321–331.

8. Friis RH, Wittchen HU, Pfister H, Lieb R. Life events and changes in the course of depression in young adults. Eur Psychiatry. 2002;17:241–253.

9. Friis RH, Garrido-Ortega C, Safer AM, et al. Socioepidemiology of cigarette smoking among Cambodian Americans in Long Beach, California. J Immigr Minor Health. 2012; 14(2):272–280.

10. McCartney G, Collins C, Walsh D, Batty GD. Why the Scots die younger: synthesiz- ing the evidence. Public Health. 2012;126:459–470.

11. Marmot M. Introduction. In: Marmot M, Wilkinson RG, eds. Social Determinants of Health. New York, NY: Oxford University Press; 1999:1–16.

12. Murray CJ, Lopez AD. Global mortality, disability, and the contribution of risk fac- tors: global burden of disease study. Lancet. 1997;349:1436–1442.

13. Lazarus RS, Folkman S. Stress, Appraisal, and Coping. New York: Springer Publishing Company, 1984.

14. Monroe SM. Modern approaches to conceptualizing and measuring human life stress. Annu Rev Clin Psychol. 2008; 4:33–52.

15. Wolf S. Psychosocial influences in gastrointestinal function. In: Levi L, ed. Society, Stress, and Disease. London, UK: Oxford University Press; 1971:362–366.

16. Selye H. The Stress of Life. New York, NY: McGraw Hill; 1956.

51589_CH15_Printer.indd 694 09/02/13 4:47 PM

r e f e r e n C e S 695

17. Selye H. The evolution of the stress concept—stress and cardiovascular disease. In: Levi L, ed. Society, Stress, and Disease. London, UK: Oxford University Press; 1971:299–311.

18. Crider A. Experimental studies of conflict-produced stress. In: Levine S, Scotch NA, eds. Social Stress. Chicago, IL: Aldine; 1970:156–188.

19. Cobb S, Schull WJ, Harburg F, et al. The intrafamilial transmission of rheumatoid arthritis—VIII: summary of findings. J Chronic Dis. 1969;22:295–296.

20. Shekelle RB, Ostfeld AM, Paul O. Social status and incidence of coronary heart dis- ease. J Chronic Dis. 1969;22:381–394.

21. Syme SL, Borhani NO, Buechley RW. Cultural mobility and coronary heart disease in an urban area. Am J Epidemiol. 1966;82:334–346.

22. Caplan RD, Cobb S, French JRP Jr, et al. Job Demands and Worker Health. Department of Health, Education and Welfare Publication (NIOSH) 75–160. Washington, DC: Department of Health, Education and Welfare; 1975.

23. French JRP Jr, Rodgers W, Cobb S. Adjustment as person-environment fit. In: Coelho GV, Hamburg DA, Adams JE, eds. Coping and Adaptation. New York, NY: Basic Books; 1974:316–333.

24. Sales SM. Differences Among Individuals in Affective, Behavioral, Biochemical, and Psychologicalal Responses to Variations in Workload. Thesis. Ann Arbor, MI: University of Michigan; 1969.

25. Dohrenwend BP. Inventorying stressful life events as risk factors for psychopathol- ogy: toward resolution of the problem of intracategory variability. Psychol Bull. 2006;132(3):477–495.

26. Holmes T, Rahe R. The social readjustment rating scale. J Psychosom Res. 1967;11:213–218.

27. Holmes TH, Masuda M. Life change and illness susceptibility. In: Dohrenwend BS, Dohrenwend BP, eds. Stressful Life Events: Their Nature and Effects. New York, NY: Wiley: 1974:45–72.

28. Langner TS, Michael ST. Life Stress and Mental Health. New York, NY: Free Press; 1963.

29. Hinkle LE Jr. The effect of exposure to culture change, social change, and changes in interpersonal relationships on health. In: Dohrenwend BS, Dohrenwend BP, eds. Stressful Life Events: Their Nature and Effects. New York, NY: Wiley; 1974:9–44.

30. Pilkonis PA, Imber SD, Rubinshy P. Dimensions of life stress in psychiatric patients. J Hum Stress. 1985;11:5–10.

31. Pearlin LI. The stress process revisited: reflections on concepts and their interrelation- ships. In: Aneshensel CA, Phelan JC, eds. Handbook of the Sociology of Mental Health. New York, NY: Plenum Publishing Corp; 1999.

32. Turner RJ, Lloyd DA. The stress process and the social distribution of depression. J Health Soc Behav. 1999;40:374–404.

33. Pearlin LI. The life course and the stress process: some conceptual comparisons. J Gerontol B Psychol Sci Soc Sci. 2010;65B(2):207–215.

34. Kobasa SCO, Maddi SR, Puccetti MC, Zola MA. Effectiveness of hardiness, exercise and social support as resources against illness. J Psychosom Res. 1985;29:525–533.

35. Rosenman RH, Friedman M, Straus R, et al. Coronary heart disease in the Western Collaborative Group Study: a follow-up experience of two years. JAMA. 1966;195:86–92.

51589_CH15_Printer.indd 695 09/02/13 4:47 PM

696 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

36. Jenkins CD. Psychological and social precursors of coronary disease. N Engl J Med. 1971;284:244–255;307–317.

37. Rosenman RH, Friedman M, Straus R, et al. A predictive study of coronary heart disease: the Western Collaborative Group Study. JAMA. 1964;189:113–120.

38. Bortner RW. A short rating scale as a potential measure of pattern A behavior. J Chronic Dis. 1969;22:87–91.

39. Cobb S. Social support as a moderator of stress. Psychosom Med. 1976;38:300–314. 40. Rabkin JG, Struening EL. Life events, stress, and illness. Science. 1976;194:

1013–1020. 41. Baron RS, Cutrona CE, Hicklin D, et al. Social support and immune function among

spouses of cancer patients. J Pers Soc Psychol. 1990;59:344–352. 42. Pearlin LI. The sociological study of stress. J Health Soc Behav. 1989;30:241–256. 43. Aneshensel CS, Stone JD. Stress and depression: a test of the buffering model of

social support. Arch Gen Psychiatry. 1982;39:1392–1396. 44. Lin N, Woefel MW, Light SC. The buffering effect of social support subsequent to an

important life event. J Health Soc Behav. 1985;26:247–263. 45. Mitchell RE, Cronkite RC, Moos RH. Stress, coping, and depression among married

couples. J Abnorm Psychol. 1983;92:433–447. 46. Dean A, Kolody B, Wood P. Effects of social support from various sources on depres-

sion in elderly persons. J Health Soc Behav. 1990;31:148–161. 47. Krause N, Liang J, Keith V. Personality, social support, and psychologicalal distress in

later life. Psychol Aging. 1990;5:315–326. 48. Tetzloff CE, Barrera M Jr. Divorcing mothers and social support: testing the specific-

ity of buffering effects. Am J Community Psychol. 1987;15:419–434. 49. Breslow L. Prospects for improving health through reducing risk factors. Prev Med.

1978;7:449–458. 50. Burkitt DP. Some diseases characteristic of modern Western civilization. In: Logan

MH, Hunt EE Jr, eds. Health and the Human Condition: Perspectives on Medical Anthropology. North Scituate, MA: Duxbury; 1978:137–147.

51. Suchman EA. Health attitudes and behavior. Arch Environ Health. 1970;20:105–110. 52. Heron MP. Deaths: Leading Causes for 2007. National Vital Statistics Reports; 59(8).

Hyattsville, MD: National Center for Health Statistics; 2011. 53. Siegel PZ, Frazier EL, Mariolis P, et al. Behavioral risk factor surveillance, 1991:

monitoring progress toward the nation’s year 2000 health objectives. MMWR. 1993;42(SS-4):1–21.

54. U.S. Department of Health, Education and Welfare, Public Health Service. Smoking and Health, Report of the Advisory Committee to the Surgeon General of the Public Health Service. Public Health Service publication 1103. Washington, DC: Government Printing Office; 1964.

55. U.S. Department of Health, Education and Welfare, Public Health Service. Smoking and Health, a Report of the Surgeon General. Department of Health, Education and Welfare publication 79-50066. Washington, DC: Department of Health, Education and Welfare; 1979.

56. U.S. Department of Health and Human Services. The Health Consequences of Smoking: A Report of the Surgeon General. Atlanta, GA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion. Office on Smoking and Health; 2004.

51589_CH15_Printer.indd 696 09/02/13 4:47 PM

r e f e r e n C e S 697

57. World Health Organization. Global Status Report on Alcohol 2004. Geneva, Switzerland: World Health Organization, Department of Mental Health and Substance Abuse; 2004.

58. Centers for Disease Control and Prevention. Alcohol-attributable deaths and years of potential life lost—United States, 2001. MMWR. 2004;53:866–870.

59. Centers for Disease Control and Prevention. Youth exposure to alcohol advertising in magazines—United States, 2001–2005. MMWR. 2007;56:763–767.

60. Streissguth AP. Fetal alcohol syndrome: an epidemiologic perspective. Am J Epidemiol. 1978;107:467–478.

61. Centers for Disease Control and Prevention. Guidelines for identifying and referring persons with fetal alcohol syndrome. MMWR. 2005;54:1–10.

62. Graham S, Mettlin C. Diet and colon cancer. Am J Epidemiol. 1979;109:1–20. 63. Kritchevsky SB. Dietary lipids and the low blood cholesterol-cancer association.

Am J Epidemiol. 1992;135:509–520. 64. Zheng W, Blot W, Shu X, et al. Diet and other risk factors for laryngeal cancer in

Shanghai, China. Am J Epidemiol. 1992;136:178–191. 65. Smith AH, Waller KD. Serum b-carotene in persons with cancer and their immediate

families. Am J Epidemiol. 1991;133:661–671. 66. Schwartz B, Bischof H-P, Kunze M. Coffee, tea, and lifestyle. Prev Med. 1994;23:

377–384. 67. CNN.com. NYC’s fat ban paying off. http://www.cnn.com/2012/07/16/health/

nyc-fat-ban-paying-off/index.html. Accessed August 19, 2012. 68. Reuters. AMA supports trans-fat bans. http://www.reuters.com/article/2008/11/11/

us-ama-trans-fat-idUSTRE4AA6C720081111. Accessed August 19, 2012. 69. Morris JN, Chave SPW, Adam C, et al. Vigorous exercise in leisure-time and the

incidence of coronary heart-disease. Lancet. 1973;1:333–339. 70. Paffenbarger RS Jr, Wing AL, Hyde RT. Physical activity as an index of heart attack

risk in college alumni. Am J Epidemiol. 1978;108:161–175. 71. Streja D, Mymin D. Moderate exercise and high-density lipoprotein-cholesterol.

JAMA. 1979;242:2190–2192. 72. Paffenbarger RS, Blair SN, Lee I-M. A history of physical activity, cardiovascular

health, and longevity: the scientific contributions of Jeremy N. Morris, DSc, DPH, FRCP. Int J Epidemiol. 2001;30:1184–1192.

73. Cooper RA, Quatrano LA. Research on physical activity and health among people with disabilities: a consensus statement. J Rehabil Res Dev. 1999;36:142–154.

74. Friis RH, Nomura WL, Ma CX, et al. Socioepidemiologic and health-related cor- relates of walking for exercise among the elderly: results from the Longitudinal Study of Aging. J Aging Phys Act. 2003;11:54–65.

75. Fabrega H, Van Egeren L. A behavioral framework for the study of human disease. Ann Intern Med. 1976;84:200–208.

76. Susser MW, Watson W, Hopper K. Sociology in Medicine, 3rd ed. New York, NY: Oxford University Press; 1985.

77. Matsumoto YS. Social stress and coronary heart disease in Japan: a hypothesis. Milbank Mem Fund Q. 1970;48:9–36.

78. Marmot MG, Syme SL, Kagan A, et al. Epidemiologic studies of coronary heart dis- ease and stroke in Japanese men living in Japan, Hawaii and California: prevalence of coronary and hypertensive heart disease and associated risk factors. Am J Epidemiol. 1975;102:514–525.

51589_CH15_Printer.indd 697 09/02/13 4:47 PM

698 Chapter 15 SoCial, Behavioral, and pSyChoSoCial epidemiology

79. Mercer JR. Sociocultural factors in educational labeling. In: Begab MJ, Richardson SA, eds. The Mentally Retarded and Society: A Social Science Perspective. Baltimore, MD: University Park Press; 1975:141–157.

80. Mechanic D. Illness behavior, social adaptation, and the management of illness. J Nerv Ment Dis. 1977;165:79–87.

81. Mechanic D. Social psychological factors affecting the presentation of bodily com- plaints. N Engl J Med. 1972;286:1132–1139.

82. Rubel AJ, Garro LC. Social and cultural factors in the successful control of tuberculo- sis. Public Health Rep. 1992;107:626–634.

83. Dohrenwend BP. A psychosocial perspective on the past and future of psychiatric epidemiology. Am J Epidemiol. 1998;147:222–231.

84. Sales SM, House J. Job dissatisfaction as a possible risk factor in coronary heart dis- ease. J Chronic Dis. 1971;23:861–873.

85. Liljefors I, Rahe RH. An identical twin study of psychosocial factors in coronary heart disease in Sweden. Psychosom Med. 1970;32:523–542.

86. House JS. The Relationship of Intrinsic and Extrinsic Work Motivations to Occupational Stress and Coronary Heart Disease Risk. Thesis. Ann Arbor, MI: University of Michigan; 1972.

87. Friis RH, Lee JA. Natural disasters and posttraumatic stress disorder: an epidemio- logic approach. Paper presented at the Seventh International Conference on Social Stress Research, Budapest, Hungary, May 26–29, 1998.

88. Langner TS. A twenty-two item screening score of psychiatric symptoms indicating impairment. J Health Hum Behav. 1962;3:267–269.

89. Markush RE, Favero RV. Epidemiologic assessment of stressful life events, depressed mood, and psychophysiological symptoms: a preliminary report. In: Dohrenwend BS, Dohrenwend BP, eds. Stressful Life Events: Their Nature and Effects. New York, NY: Wiley; 1974:171–190.

90. Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psychol Meas. 1977;1:385–401.

91. Weissman MM, Sholomskas D, Pottenger M, et al. Assessing depressive symptoms in five psychiatric populations: a validation study. Am J Epidemiol. 1977;106:203–214.

92. Comstock GW, Helsing KJ. Symptoms of depression in two communities. Psychol Med. 1976;6:551–563.

93. Craig TJ, Van Natta PA. Presence and persistence of depression symptoms in patient and community populations. Am J Psychiatry. 1976;133:1426–1429.

94. Frerichs RR, Aneshensel CS, Clark VA. Prevalence of depression in Los Angeles County. Am J Epidemiol. 1981;113:691–699.

95. Husaini BA, Neff JA, Harrington JB, et al. Depression in rural communities: validat- ing the CES-D scale. J Community Psychol. 1980;8:20–27.

96. Frerichs RR, Aneshensel CS, Yokopenic PN, Clark VA. Public health and depression: an epidemiologic survey. Prev Med. 1982;11:639–646.

97. Moscicki EK, Locke BZ, Rae DS, Boyd JH. Depressive symptoms among Mexican Americans: the Hispanic Health and Nutrition Examination Survey. Am J Epidemiol. 1989;130:348–360.

98. Freedman DX. Psychiatric epidemiology counts. Arch Gen Psychiatry. 1984;41: 931–933.

99. Fox BH. Premorbid psychologicalal factors as related to cancer incidence. J Behav Med. 1978;1:45–133.

51589_CH15_Printer.indd 698 09/02/13 4:47 PM

r e f e r e n C e S 699

100. Fox BH. The role of psychologicalal factors in cancer incidence and prognosis. Oncology. 1995;9:245–253.

101. Croog SH, Fitzgerald EF. Subjective stress and serious illness of a spouse: wives of heart patients. J Health Soc Behav. 1978;19:166–178.

102. Giovino GA, Henningfield JE, Tornar SL, et al. Epidemiology of tobacco use and dependence. Epidemiol Rev. 1995;17:48–65.

103. Acton GS, Prochaska JJ, Kaplan AS, et al. Depression and stages of change for smok- ing in psychiatric outpatients. Addict Behav. 2001;26:621–631.

104. Vaillant GE. Natural history of male psychological health: effects of mental health on physical health. N Engl J Med. 1979;301:1249–1254.

105. Thomas SP. Psychosocial correlates of women’s health in middle adulthood. Issues Mental Health Nurs. 1995;16:285–314.

106. Friedman HS, Tucker JS, Tomlinson-Keasey C, et al. Does childhood personality predict longevity? J Pers Soc Psychol. 1993;65:176–185.

107. Marcenes WS, Sheiham A. The relationship between work stress and oral health status. Soc Sci Med. 1992;35:1511–1520.

108. Martikainen P, Bartley M, Lahelma E. Psychosocial determinants of health in social epidemiology. Int J Epidemiol. 2002;31:1091–1093.

109. Syme SL. Historical perspective: the social determinants of disease—some roots of the movement. Epidemiol Perspect Innov. 2005;2(1):2.

51589_CH15_Printer.indd 699 09/02/13 4:47 PM

51589_CH15_Printer.indd 700 09/02/13 4:47 PM

3chapte r

701

16chapte r

Epidemiology as a Profession

LEARNING OBJECTIVES

By the end of this chapter the reader will be able to:

●● describe five areas of specialization within epidemiology ●● describe four career roles for epidemiologists ●● name three resources for education and employment in epidemiology ●● name three epidemiology associations and three journals that

publish articles on epidemiology ●● list four competencies required in the field of epidemiology ●● state five ethical issues that pertain to the practice of epidemiology

CHAPTER OUTLINE

I. Introduction II. Specializations within Epidemiology

III. Career Roles for Epidemiologists IV. Epidemiology Associations and Journals V. Competencies Required of Epidemiologists

VI. Resources for Education and Employment VII. Professional Ethics in Epidemiology

VIII. Conclusion IX. Study Questions and Exercises

51589_CH16_Printer.indd 701 11/02/13 6:02 AM

702 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

Introduction

This chapter responds to the many questions received from students about career opportunities in epidemiology. As noted previously in the introduction to this text, epidemiology is an exciting and rewarding field. It is unusual for a day to pass without media reports about health-related studies based on epidemiologic methods. To illustrate, headlines have announced findings about obesity and consumption of diet sodas, obesity and choice of friends, the association between contaminants in imported foods and potential adverse health effects, and hazards from lead in toys. Moreover, the authors have provided many examples of epi- demiology applied to public health practice throughout this book. We hope that our enthusiasm for epidemiology expressed in this text will be “contagious” and that our readers will consider a career in this growing field.

What types of work does an epidemiologist perform? What are sources of career information about the field? What are the professional and ethical obligations of epidemiologists? For those interested in embarking on a career in epidemiology, this chapter will provide resources such as examples of specializations, possible career roles, contacts for employment, epidemiology journals and organizations, and pro- fessional and ethical issues. For a quick introduction to ongoing issues and profes- sional concerns of the field, we recommend The Epidemiology Monitor (http://www .epimonitor.net; accessed November 23, 2012), a monthly newsletter that is always an interesting read. A typical issue of The Epidemiology Monitor lists more than 150 employment opportunities in national and international locations (Figure 16–1).

Figure 16–1 The Epidemiology Monitor. Source: With permission from the Epidemiology Monitor.

51589_CH16_Printer.indd 702 11/02/13 6:02 AM

s p e C i a l i z a t i o n s w i t h i n e p i d e m i o l o g y 703

Special izat ions within Epidemiology

Infectious disease epidemiology is one of the oldest and most familiar specializations within the discipline. Investigations of outbreaks of foodborne diseases, communicable diseases such as tuberculosis, and nosocomial infections illustrate this specialization in operation. Among the many other fields within epidemiology are environmental epidemiology, chronic disease epidemiology, social epidemiology, and molecular and genetic epidemiology. Table 16–1 lists more than 30 specializations in epidemiology, ranging from the epidemiology of aging to public health practice epidemiology, occupational epidemiology, and women’s health epidemiology. Each of these specific applications is a potential career path for future epidemiologists who aspire to stimulating and challenging employment. Given the expanding number of specializations within the field, the information provided in the table is not exhaustive. Following are some more detailed statements regarding several of the examples shown in the table:

●● Reproductive epidemiology: “. . . Covers a wide range of topics from the development, adult capacity, and senescence of the reproductive systems, to conception and pregnancy, to delivery and health of the offspring.”1(p 585)

Table 16–1 Specializations within Epidemiology

Epidemiology of Aging Assessment of Health Care Epidemiology Behavioral Epidemiology Biostatistics and Epidemiologic Methods

●● Disease Informatics ●● Meta-Analysis

Birth Defects Epidemiology Chronic Disease Epidemiology

●● Coronary Heart Disease ●● Diabetes ●● Obesity

Clinical Epidemiology Primary Care Epidemiology Environmental Health Epidemiology Epidemiology of Cancer Epidemiology of Urban Health Field Epidemiology/Public Health Practice Epidemiology Genetic and Molecular Epidemiology Health and Policy Administration Epidemiology Health Services Research Epidemiology Life Course Epidemiology Infectious Disease Epidemiology

●● Infection Control and Hospital Epidemiology

●● Parasitology ●● Surveillance ●● Vector-borne ●● Virology ●● Zoonoses

Injury Epidemiology Neuroepidemiology Nutrition Epidemiology Occupational Epidemiology Oral/dental Epidemiology Pediatric Epidemiology Pharmacoepidemiology Psychiatric Epidemiology Psychologic Epidemiology Renal Epidemiology Reproductive Epidemiology Screening Epidemiology Social Epidemiology Spatial Epidemiology Sport Epidemiology Substance Abuse Epidemiology Veterinary Epidemiology Women’s Health Epidemiology

●● Perinatal ●● Pregnancy

51589_CH16_Printer.indd 703 11/02/13 6:02 AM

704 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

●● Nutrition epidemiology: “. . . Uses epidemiologic approaches to determine relations between dietary factors and the occurrence of specific diseases.”2(p 623) An illustration of a field project in nutrition is shown in Figure 16–2.

●● Health services research epidemiology: Applies the methods of epidemiology to the assessment of needs for health services and the evaluation of the quality of health services.

●● Neuroepidemiology: Epidemiologic approaches are applied to investigate the prevalence and risk factors for neurological diseases such as Alzheimer’s disease, Parkinson’s disease, and multiple sclerosis.

●● Injury epidemiology: The methods of epidemiology are used to investigate the incidence, prevalence, and risk factors for unintentional injuries and deaths; the broad topic of injuries includes motor vehicle crashes, work- place injuries, and sports-related injuries.

●● Coronary heart disease epidemiology: Focuses on “. . . the determinants, distribu- tion, and sequelae of coronary heart disease (CHD) in populations . . .”3 (p 7)

●● Pharmacoepidemiology: This field examines drug utilization and associated effects among the population. Examples are the appropriate and inappro- priate use of drugs as well as clinical outcomes linked to drugs.

Figure 16–2 A “weight and height measurement team” in place to assess the nutritional status of Biafran war refugees. Source: Reproduced from CDC Public Health Image Library, ID# 7115. Available at: http://phil.cdc .gov/phil/details.asp. Accessed December 25, 2012.

51589_CH16_Printer.indd 704 11/02/13 6:02 AM

C a r e e r r o l e s f o r e p i d e m i o l o g i s t s 705

●● Oral epidemiology/dental epidemiology: Studies population-based outcomes related to dental health. Concerns of the field are effects of fluoride consumption, oral cancer, dental caries, and tooth loss (partial and full edontulism).

●● Renal epidemiology: Examines the distribution and determinants of renal (kidney) diseases and abnormality in populations. Renal diseases include end-stage kidney failure, kidney stones, and hematuria.

●● Veterinary epidemiology: Studies the occurrence of diseases in animal popu- lations. Information acquired from veterinary epidemiology studies can be used to promote the productivity and welfare of animal populations.

Career Roles for Epidemiologists

The field of epidemiology provides numerous potential career roles in teach- ing, research, and applied settings. Private industry, government, universities, research organizations, hospitals, and nonprofit organizations are among the employers of epidemiologists. The following list (not exhaustive) contains a few of the many occupations found within epidemiology:

●● academic workers, including college professors ●● research workers ●● public health nurse epidemiologists ●● biostatisticians who focus on epidemiology ●● healthcare planners ●● pharmaceutical and biotech industry researchers ●● consultants in epidemiology (e.g., for community-based participatory

research) ●● program evaluation/community needs assessment specialists ●● epidemiology surveillance workers (Figure 16–3)

Epidemiologists who work in local health departments investigate local out- breaks of disease such as foodborne illnesses, vector-borne diseases, disease clusters, and communicable diseases in schools. They provide assistance to local hospitals and other healthcare providers regarding control of infectious diseases. They also participate in local programs for prevention and control of chronic diseases.

In the United States, depending on the size of the state, epidemiologists contribute to statewide initiatives to monitor and control infectious and chronic diseases. Also, they aid in implementing health-related mandates issued by the executive and legislative branches of state government. In larger states, epide- miologists may hold staff and research positions in specialized health-related

51589_CH16_Printer.indd 705 11/02/13 6:02 AM

706 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

programs such as those devoted to air pollution control and environmental health impacts of toxic wastes.

Nationally, various U.S. government agencies employ epidemiologists in administrative and research positions. Two examples are the Centers for Dis- ease Control and Prevention (CDC) and the National Institute of Occupational Safety and Health (NIOSH).

With the growth of initiatives that require programs to be evidence based, the skills of epidemiologists are crucial for design of protocols, reliable and valid data collection, and analysis of data. Epidemiologists provide helpful input into the quantitative aspects of community and academic teams that perform community-based participatory research.

Centers for Disease Control and Prevention At the CDC in Atlanta, Georgia, epidemiologists are involved in program administration, basic research, and “shoe leather” data collection activities. Numerous examples of reports from the CDC have been cited throughout this

Figure 16–3 A CDC field researcher is shown detaching a collection bag from a CDC light trap, which has captured numerous mosquitoes. Source: Reproduced from CDC Public Health Image Library, ID# 5600. Available at: http://phil.cdc.gov/phil/ home.asp. Accessed December 25, 2012.

51589_CH16_Printer.indd 706 11/02/13 6:02 AM

C a r e e r r o l e s f o r e p i d e m i o l o g i s t s 707

text. An important educational and public health practice arm of the CDC is the Epidemic Intelligence Service, established in 1951 to warn against the use of biological warfare (Figure 16–4).

National Institute of Occupational Safety and Health (NIOSH) Headquartered in Washington, DC, NIOSH is a branch of the CDC located in the U.S. Department of Health and Human Services. NIOSH was “. . . estab- lished to help assure ‘safe and healthful working conditions for working men and women by providing research, information, education, and training in the field of occupational safety and health.’”4 (p1) NIOSH positions are primarily located in Cincinnati, OH; Morgantown, WV; Pittsburgh, PA; and Spokane, WA, with fewer employment opportunities in Atlanta, GA, and Washington, DC. As stated on the agency’s website,

NIOSH is seeking career professionals to conduct research or field investigations, serve as expert consultants, design research studies, and develop recommended standards for occupational safety and health. Applicants should be trained in chemistry, physics, engineering, industrial hygiene, safety, biology, toxicology, medicine, public health, epidemiology, environmental health, occupational health, statistics, or computer science.5

Figure 16–4 The Epidemic Intelligence Service (EIS) graduating class of 2004. Source: Reproduced from CDC Public Health Image Library, ID# 7284. Available at: http://phil.cdc.gov/phil/home.asp. Accessed December 25, 2012.

51589_CH16_Printer.indd 707 11/02/13 6:02 AM

708 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

Epidemiologists employed by NIOSH assess, research, and monitor workplace-related injuries and diseases. Examples of topics addressed by previous epidemiologic investigations are the association of use of video display terminals with women’s reproductive hazards, utility linemen’s electrocution hazards, and construction workers’ fall hazards. Often NIOSH epidemiologists collaborate with international colleagues in studies of occupational hazards such as asbestos, benzene, and formaldehyde. For more information refer to an interesting article by Halperin and Howard, “Occupational Epidemiology in the National Institute for Occupational Safety and Health.”6

Epidemiology Associat ions and Journals

A list of professional associations related to epidemiology is presented in Table 16–2. Also provided in the text is a list of journals that publish epide- miologic studies. Contact information for both associations and journals is shown. The lists are not exhaustive but provide a launching point for your fur- ther investigation. For example, The Society for Epidemiologic Research, one of the lead organizations for epidemiologists, publishes the American Journal of Epidemiology.

Professional Associations The following paragraphs describe in more detail some of the associations listed in Table 16–2. The quoted material is reproduced verbatim from the organiza- tions’ websites.

American Academy of Pediatrics, Section on Epidemiology: “. . . Founded in 1988, [the AAP] is dedicated to educating the general pediatrician and pediatric subspecialist on the basic principles of epidemiology and its relation to pediatrics and its various disciplines. The section consults to the AAP Board of Directors and works with various AAP committees, sections, and task forces to provide methodological support and expertise in epidemiology and evidence-based med- icine.” (website: http://www.aap.org/sections/epidemiology/default.cfm)

American College of Epidemiology (ACE): “. . . Incorporated in 1979 to develop criteria for professional recognition of epidemiologists and to address their professional concerns . . . ACE serves the interests of its members through sponsorship of scientific meetings, publications, and educational activities, recognizing outstanding contributions to the field and advocating for issues per- tinent to epidemiology.” (website: http://www.acepidemiology2.org/)

51589_CH16_Printer.indd 708 11/02/13 6:02 AM

e p i d e m i o l o g y a s s o C i a t i o n s a n d J o u r n a l s 709

American Public Health Association: “The mission of the Epidemiology section is to foster epidemiologic research and science-based public health prac- tice and serve as a conduit between the epidemiologic research community and users of scientific information for the development, implementation, and evalu- ation of policies affecting the public’s health.” (website: http://www.apha.org)

International Epidemiological Association: “The International Corresponding Club, as the IEA was first called, was started in 1954 by John Pemberton of Great Britain and Harold N. Willard of the United States with the

Table 16–2 Domestic and International Organizations in Epidemiology

American Academy of Pediatrics, Section on Epidemiology http://www.aap.org/sections/epidemiology/default.cfm

International Genetic Epidemiology Society http://www.genepi.org/

American College of Epidemiology http://www.acepidemiology.org/

International Society for Pharmacoepidemiology http://www.pharmacoepi.org/

American Public Health Association, Epidemiology Section http://www.apha.org/membergroups/sections/aphasections/epidemiology/

International Society for Environmental Epidemiology http://www.iseepi.org/

American Statistical Association, Section on Statistics on Epidemiology http://www.amstat.org/sections/epi/SIE_Home.htm

Society for Epidemiologic Research http://www.epiresearch.org/

Association for Professionals in Infection Control and Epidemiology http://www.apic.org/

Society for Pediatric and Perinatal Epidemiologic Research http://www.sper.org/

Association of Public Health Epidemiologists in Ontario http://www.apheo.on.ca/

The Canadian Society for Epidemiology and Biostatistics http://www.cseb.ca/

Australasian Epidemiological Association http://www.aea.asn.au/

The Danish Epidemiologic Society http://www.dansk-epidemiologisk-selskab.dk/

Council of State and Territorial Epidemiologists http://www.cste.org/

The German Society for Epidemiology http://dgepi.de/

European Epidemiology Federation http://www.iea-europe.org/

The Society for Clinical Trials http://www.sctweb.org/

International Epidemiological Association http://www.ieaweb.org/

The Society for Healthcare Epidemiology in America http://www.shea-online.org/

Accessed August 18, 2012.

51589_CH16_Printer.indd 709 11/02/13 6:02 AM

710 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

advice and help of the late Robert Cruickshank. They had found, as traveling Research Fellows each in the other’s country, that they were handicapped by not being sufficiently well informed about the research and teaching in the field of social and preventive medicine in the various medical schools and research insti- tutes.” (website: http://www.dundee.ac.uk/iea/)

Society for Epidemiologic Research: “. . . established in 1968 as a forum for sharing the latest in epidemiologic research. The SER is committed to keep- ing epidemiologists at the vanguard of scientific developments.” (website: http:// www.epiresearch.org/)

Epidemiology Journals Numerous scientific journals publish epidemiologic research. However, the following journals feature predominantly epidemiologic studies. The quoted descriptions are reproduced verbatim from the journals’ websites.

American Journal of Epidemiology: “. . . the premier epidemiological journal devoted to the publication of empirical research findings, opinion pieces and methodological developments in the field of epidemiological research. It is aimed at both fellow epidemiologists and those who use epidemiological data, including public health workers and clinicians.” (website: http://aje.oxfordjournals.org/)

American Journal of Public Health: “. . . is dedicated to original work in research, research methods, and program evaluation in the field of public health. This prestigious journal also regularly publishes authoritative editorials and com- mentaries and serves as a forum for the analysis of health policy. The stated mis- sion of the Journal is ‘to advance public health research, policy, practice, and education.’” (website: http://www.ajph.org/)

Epidemiology: “. . . is a peer-reviewed scientific journal that publishes original research on the full spectrum of epidemiologic topics. Journal content ranges from cancer, heart disease and other chronic illnesses to reproductive, environ- mental, psychosocial, infectious-disease and genetic epidemiology. The journal places special emphasis on theory and methodology, and welcomes commentar- ies that explore fundamental assumptions or offer provocative dissent.” (website: http://www.jstor.org/journals/10443983.html)

Epidemiology and Infection: “. . . publishes original reports and reviews on all aspects of infection in humans and animals. Particular emphasis is given to the epidemiology, prevention, and control of infectious diseases. The field covered is broad and includes the zoonoses, tropical infections, food hygiene, vaccine studies, statistics and the clinical, social and public-health aspects of infectious disease.” (website: http://www.jstor.org/journals/09502688.html)

51589_CH16_Printer.indd 710 11/02/13 6:02 AM

C o m p e t e n C i e s r e q u i r e d o f e p i d e m i o l o g i s t s 711

International Journal of Epidemiology: “. . . an essential requirement for any- one who needs to keep up to date with epidemiological advances and new developments throughout the world. It encourages communication among those engaged in the research, teaching, and application of epidemiology of both communicable and noncommunicable disease, including research into health services and medical care.” (website: http://ije.oxfordjournals.org/ archive/)

Cancer Epidemiology, Biomarkers & Prevention: Published by the American Association for Cancer Research, “. . . publishes original, peer-reviewed research on cancer causation, mechanisms of carcinogenesis, prevention, and survivorship. Topics include descriptive, analytical, biochemical, and molecular epidemiology; the use of biomarkers to study the neoplastic and preneoplastic processes in humans; chemoprevention and other types of prevention trials; and the role of behavioral factors in cancer etiology and prevention.” Many of these areas are based on epidemiologic methods. (website: http://cebp .aacrjournals.org/)

Competencies Required of Epidemiologists

Competencies refer to skills that a professional should acquire in order to perform effectively. Competency levels vary according to the degree acquired—bachelor’s, master’s, doctoral, or postdoctoral—with more advanced degree holders trained to perform more complex tasks and assume greater independence and responsi- bility. For example, candidates with a doctoral degree should be able to perform advanced epidemiologic research. Figure 16–5 shows a research worker engaged in laboratory research.

Master of Public Health (MPH) degree candidates are capable of performing a wide range of tasks in the work environment. The Association of Schools of Public Health has developed the following list of competencies in epidemiology to be achieved by MPH degree candidates in public health7:

1. Identify key sources of data for epidemiologic purposes. 2. Identify the principles and limitations of public health screening

programs. 3. Describe a public health problem in terms of magnitude, person, time,

and place. 4. Explain the importance of epidemiology for informing scientific, ethical,

economic, and political discussion of health issues.

51589_CH16_Printer.indd 711 11/02/13 6:02 AM

712 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

5. Comprehend basic ethical and legal principles pertaining to the collection, maintenance, use, and dissemination of epidemiologic data.

6. Apply the basic terminology and definitions of epidemiology. 7. Calculate basic epidemiology measures. 8. Communicate epidemiologic information to lay and professional

audiences. 9. Draw appropriate inferences from epidemiologic data.

10. Evaluate the strengths and limitations of epidemiologic reports.

Resources for Education and Employment

Continuing Education Degree Programs Education in epidemiology can be obtained via special summer session programs and online programs, such as those offered by Johns Hopkins University and the University of Michigan, Ann Arbor. Examples of summer programs are:

●● Graduate Summer Session in Epidemiology: “The summer program offers instruction in the principles, methods, and applications of epidemiology.

Figure 16–5 Here a laboratory technician is using a dissecting type of microscope to view a Legionella pneumophilia culture specimen. Source: Reproduced rom CDC Public Health Image Library, ID# 4105. Available at: http://phil.cdc.gov/phil/details.asp. Accessed December 25, 2012.

51589_CH16_Printer.indd 712 11/02/13 6:02 AM

r e s o u r C e s f o r e d u C a t i o n a n d e m p l o y m e n t 713

Distinguished faculty from academic centers and governmental agencies throughout the United States and other countries will be responsible for introductory and advanced courses in epidemiology, biostatistics, and data management.” (website: http://www.sph.umich.edu/epid/GSS/. Accessed November 26, 2012.)

●● Graduate Summer Institute of Epidemiology and Biostatistics—Johns Hopkins Bloomberg School of Public Health: “The Program is intended to develop an understanding of basic and advanced principles of epidemiological research, and will offer courses that present epidemiologic methods and their application to the study of the natural history and etiology of disease.” (website: http://www.jhsph.edu/academics/con- tinuing-education/institutes/summer-institutes/summerepi/. Accessed November 26, 2012.)

●● Summer Session for Public Health Studies–Harvard University, School of Public Health: “. . . is intended for health professionals in training or those who are considering a mid-career change into public health and feel the need to strengthen their skills. Participants include public health profes- sionals, primary care practitioners, physicians engaged in the evaluation of healthcare delivery and management, physicians in training (includ- ing preventive medicine residents and medical students in an MD/MPH joint degree program), and candidates for a part-time MPH program . . .” (website: http://www.hsph.harvard.edu/academics/public-health-studies/. Accessed November 26, 2012.)

Colleges and universities in the United States and abroad offer Master’s and doctoral degree programs in epidemiology. Consult the website of the Council on Education for Public Health (CEPH) for a current list of accredited schools of public health and graduate public health programs in the United States (web- site: http://www.ceph.org).

Employment According to the United States Department of Labor, Bureau of Labor Statistics (BLS), there were 5,000 epidemiologist positions in 2010. Through 2020 the BLS projects the need for an additional 1,200 epidemiologists. The 24% increase is faster than average. Exhibit 16–1 provides median annual wages in the industries employing the largest numbers of epidemiologists in May 2010. The median annual wage for all industries was $63,010 in 2010. For informa- tion on positions in your local area or at the state level, research the Internet

51589_CH16_Printer.indd 713 11/02/13 6:02 AM

714 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

for health departments, hospitals, pharmaceutical firms, and biotech companies. Plan to attend the annual meetings of the Society for Epidemiologic Research and the American Public Health Association for on-site job listings and contacts. Refer to the following websites to obtain national listings of employment oppor- tunities for epidemiologists:

●● APHA Public Health Career Mart: http://www.apha.org/about/careers/ ●● APIC Career Center: http://www.apic.org/Resources/Career-Center ●● Centers for Disease Control and Prevention: http://www.cdc.gov/

employment/ ●● Council of State and Territorial Epidemiologists Employment Listings:

http://www.cste.org/dnn/Employment/EmploymentOpenings/tabid/138/ Default.aspx

●● Epimonitor.net Job Bank: http://www.epimonitor.net/JobBank.htm ●● National Institutes of Health: http://www.jobs.nih.gov/ ●● World Health Organization: http://www.who.int/employment/en/

Professional Ethics in Epidemiology

The ACE has developed a framework for ethical principles in epidemiology. Some of these guidelines are reprinted in Exhibit 16–2. Examples of ethical guidelines include the obligation to submit research proposals to ethics commit- tees for review, avoidance of conflicts of interest and partiality, and respecting cultural diversity.

Median remuneration for epidemiologists

Employment category Median wage in 2010

Pharmaceutical and medicine manufacturing

$92,920

Hospitals; state, local, and private $72,990 Scientific research and development services

$67,160

Colleges, universities, and professional schools

$61,870

Federal, state, and local government $57,390

Source: United States Department of Labor, Bureau of Labor Statistics. http:// www.bls.gov/ooh/life-physical-and-social-science/epidemiologists.htm. Accessed November 26, 2012. n

e x

h ib

it 1

6 –1

51589_CH16_Printer.indd 714 11/02/13 6:02 AM

p r o f e s s i o n a l e t h i C s i n e p i d e m i o l o g y 715

ethics Guidelines for epidemiologists

Part I. Core Values, Duties, and Virtues in Epidemiology 1.1. Definition and Discussion of Core Values Like other scientists, epidemiologists uphold values of free inquiry and the pursuit of knowledge. The goal of science, after all, is to explain and to predict natural phenomena. Epidemiologists not only pursue knowledge about the distribution and determinants of health and disease in populations, but also uphold the value

of improving the public’s health through the application of scientific knowledge.

These core values underlie the mission and purpose of epidemiology. Here we are concerned with core values that are internal to the profession of epidemiology. As such, they are more restricted in scope than gen- eral ethical principles such as beneficence (which relates to the balancing of risks and benefits and the promotion of the common welfare). On the other hand, core values in epidemiology are more general (and more basic) than ethical rules and norms within the profession such as the need to obtain the informed consent of research participants. [Here and elsewhere in this document the term research participants is used instead of human subjects, which is sometimes regarded as paternalistic; never- theless, the term participants may incorrectly imply that there has been valid consent to participate, which is not always feasible in epidemiologic studies.] This section provides a concise set of ethics guidelines for epide- miologists. [Part III, which provides further detail, is not included in this Exhibit.]

2.1. The Professional Role of Epidemiologists The profession of epidemiology has as its primary roles the design and conduct of scientific research and the public health application of sci- entific knowledge. This includes the reporting of results to the scientific community, to research participants, and to society; and the mainte- nance, enhancement, and promotion of health in communities. Other professional roles in epidemiology include teaching, consulting, and administration.

e x

h ib

it 1

6 –2

continues

51589_CH16_Printer.indd 715 11/02/13 6:02 AM

716 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

2.2. Minimizing Risks and Protecting the Welfare of Research Participants Epidemiologists have ethical and professional obligations to minimize risks and to avoid causing harm to research participants and to society. The risks of nonresearch public health practice activities also should be minimized.

2.3. Providing Benefits Epidemiologists should ensure that the potential benefits of studies to research participants and to society are maximized by, for example, com- municating results in a timely fashion. Steps also should be taken to maxi- mize the potential benefits of public health practice activities.

2.4. Ensuring an Equitable Distribution of Risks and Benefits Epidemiologists should ensure that the potential benefits and burdens of epidemiologic research and public health practice activities are distributed in an equitable fashion.

2.5. Protecting Confidentiality and Privacy Epidemiologists should take appropriate measures to protect the privacy of individuals and to keep confidential all information about individual research participants during and after a study. This duty also applies to personal information about individuals in public health practice activities.

2.6. Obtaining the Informed Consent of Participants Epidemiologists should obtain the prior informed consent of research par- ticipants (with exceptions noted below in Section 2.6.3), in part by dis- closing those facts and any information that patients or other individuals usually consider important in deciding whether or not to participate in the research.

2.6.1. Elements of informed consent Information should be provided about the purposes of the study, the sponsors, the investigators, the scientific methods and procedures, any anticipated risks and benefits, any anticipated inconveniences or discom- fort, and the individual’s right to refuse participation or to withdraw from the research at any time without repercussions.

exhibit 16–2 continued

continues

51589_CH16_Printer.indd 716 11/02/13 6:02 AM

p r o f e s s i o n a l e t h i C s i n e p i d e m i o l o g y 717

2.6.2. Avoidance of manipulation or coercion Research participants must voluntarily consent to the research without coercion, manipulation, or undue incentives for participation.

2.6.3. Conditions under which informed consent requirements may be waived Requirements to obtain the informed consent of research participants may be waived in certain circumstances, such as when it is not feasible to obtain the informed consent of research participants, in some studies involving the linkage of large databases routinely collected for other purposes, and in studies involving only minimal risks. In such circumstances, research participants generally need protection in other ways, such as through con- fidentiality safeguards and appropriate review by an independent research ethics committee (often referred to as institutional review boards in the United States or as ethics review boards in Canada). Informed consent requirements may also be waived when epidemiologists investigate disease outbreaks, evaluate programs, and conduct routine disease surveillance as part of public health practice activities.

2.7. Submitting Proposed Studies for Ethical Review Epidemiologists should submit research protocols for review by an inde- pendent ethics committee. An exception may be justified when epidemi- ologists investigate outbreaks of acute communicable diseases, evaluate programs, and conduct routine disease surveillance as part of public health practice activities.

2.8. Maintaining Public Trust To promote and preserve public trust, epidemiologists should adhere to the highest ethical and scientific standards and follow relevant laws and regulations concerning the conduct of these activities, including the pro- tection of human research participants and confidentiality protections.

2.8.1. Adhering to the highest scientific standards Adhering to the highest scientific standards includes choosing an appropri- ate study design for the scientific hypothesis or question to be answered; writing a clear and complete protocol for the study; using proper

exhibit 16–2 continued

continues

51589_CH16_Printer.indd 717 11/02/13 6:02 AM

718 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

procedures for the collection, transmission, storage, and analysis of data; making appropriate interpretations from the data analyses; and writing up and disseminating the results of the study in a manner consistent with accepted procedures for scientific publication.

2.8.2. Involving community representatives in research To the extent possible and whenever appropriate, epidemiologists also should involve community representatives in the planning and conduct of the research such as through community advisory boards.

2.9. Avoiding Conflicts of Interest and Partiality Epidemiologists should avoid conflicts of interest and be objective. They should maintain honesty and impartiality in the design, conduct, interpre- tation, and reporting of research.

2.10. Communicating Ethical Requirements to Colleagues, Employers, and Sponsors and Confronting Unacceptable Conduct

Epidemiologists, as professionals, should communicate to their students, peers, employers, and sponsors the ethical requirements of scientific research and its application in professional practice.

2.10.1. Communicating ethical requirements Epidemiologists should provide training and education in ethics to stu- dents of the discipline as well as to practicing scientists. They should demonstrate appropriate ethical conduct to colleagues and students by example.

2.10.2. Confronting unacceptable conduct Epidemiologists should confront unacceptable conduct such as scientific misconduct, even though confronting it can be difficult in practice. Steps should be taken to provide protections for persons who confront or allege unacceptable conduct. The rights of the accused to due process should also be respected.

2.11. Obligations to Communities Epidemiologists should meet their obligations to communities by under- taking public health research and practice activities that address health

exhibit 16–2 continued

continues

51589_CH16_Printer.indd 718 11/02/13 6:02 AM

C o n C l u s i o n 719

problems including questions concerning the utilization of health care resources, and by reporting results in an appropriate fashion.

2.11.1. Reporting results All research findings and other information important to public health should be communicated in a timely, understandable, and responsible manner so that the widest possible community stands to benefit. [In addition, studies that report null results should be published in order to help the scientific community discontinue investigations that do not have merit.]

2.11.2. Public health advocacy In confronting public health problems, epidemiologists sometimes act as advocates on behalf of members of affected communities. Advocacy should not impair scientific objectivity.

2.11.3. Respecting cultural diversity Epidemiologists should respect cultural diversity in carrying out research and practice activities and in communicating with community members. n

Source: Adapted from American College of Epidemiology, Ethics Guidelines. This article was published in Annals of Epidemiology, Vol 10, No 8, 2000, pp. 487–497, “Ethics guidelines”, Copyright Elsevier, 2000. Available at: http://www.acepidemiology.org/statement/ethics_ guidelines. Accessed December 25, 2012.

exhibit 16–2 continued

Conclusion

This chapter provided an introduction to the profession of epidemiology. Because epidemiology is an interdisciplinary field, personnel come from a wide variety of backgrounds. Some epidemiologists may have augmented their pre- vious education and experience in a different field with specialized training in epidemiology; others may have received beginning and/or advanced degrees exclusively in epidemiology. Among the societal trends that reinforce the demand for well-trained epidemiologists are the impact of bioterrorism and the emer- gence and re emergence of infectious diseases. Another trend relates to the impacts of demographic changes in the population due to immigration, population growth, and maturation of the “Baby Boom” generation. These demographic changes have resulted in increases in the occurrence of chronic conditions such

51589_CH16_Printer.indd 719 11/02/13 6:02 AM

720 C h a p t e r 1 6 e p i d e m i o l o g y a s a p r o f e s s i o n

as diabetes and obesity. Epidemiologists will be needed to conduct research on these and other chronic conditions. Adverse environmental conditions such as global warming will add to the demand for epidemiologic researchers. Finally, the increasing financial pressures associated with increasing healthcare costs demand the use of epidemiologic methods to identify cost-effective allocation of resources. In view of these trends, epidemiology will continue to be an exciting field well into the foreseeable future.

Study Questions and Exercises

1. Define the following terms: a. Pharmacoepidemiology b. Neuroepidemiology c. Oral/dental epidemiology d. Renal epidemiology e. Veterinary epidemiology

2. Conduct a search of the websites of your local and state health depart- ments regarding activities in epidemiology. Then log on to the websites for the following national and international agencies. Describe how an epidemiologist might contribute to the following employment settings: a. A local health department b. A state government health agency, such as a state department of health

services c. A federal health agency; examples include:

i. Centers for Disease Control and Prevention ii. Environmental Protection Agency

iii. National Institute of Occupational Safety and Health iv. The Agency for Toxic Substances and Disease Registry v. Food and Drug Administration

d. An international agency; examples include: i. World Health Organization

ii. Pan American Health Organization iii. The European Union Health Organization

3. An epidemiologist has been awarded a large grant to conduct a cohort study on the health of a minority population. What ethical issues might arise in carrying out this research project?

4. Describe how the employment responsibilities might differ among individuals trained at the bachelor’s, master’s, and doctoral levels.

51589_CH16_Printer.indd 720 11/02/13 6:02 AM

r e f e r e n C e s 721

5. How do the employment roles of epidemiologists differ from those of other public health professionals? An example of an area in which an epidemiologist might specialize is in the design of research studies. How would the epidemiologist’s contributions to a research study be different from those of other public health professionals?

6. Refer to the resources provided for finding employment opportuni- ties in epidemiology. Access an active website and identify five current employment openings in the field. Describe what these opportunities involve.

7. Professional networking is a key aspect of obtaining employment in a professional field. Using your own ideas, describe how you might estab- lish a personal network of epidemiologic professional who could facili- tate your search for employment opportunities.

8. Invite an epidemiologist (perhaps a graduate of your program) to your class and request that the individual describe his or her job responsibilities.

9. Arrange a field visit to a health-related agency that is located in your community. An example might be a public health department, nonprofit health-related organization, or research institute. Find out about how epidemiologists contribute to these settings.

References

1. Weinberg CR, Wilcox AJ. Reproductive epidemiology. In: Rothman KJ, Greenland S, eds. Modern Epidemiology, 2nd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 1998:585–608.

2. Willett WC. Nutritional epidemiology. In: Rothman KJ, Greenland S, eds. Modern Epidemiology, 2nd ed. Philadelphia, PA: Lippincott Williams & Wilkins; 1998: 623–642.

3. Tyroler HA. Coronary heart disease epidemiology in the 21st century. Epidemiol Rev. 2000;22:7–13.

4. Centers for Disease Control and Prevention. National Institute for Occupational Safety and Health (NIOSH). Fact Sheet. http://www.cdc.gov/niosh/pdfs/2003-116 .pdf. Accessed August 18, 2012.

5. Centers for Disease Control and Prevention. National Institute for Occupational Safety and Health (NIOSH). http://www.cdc.gov/niosh/employ/nioshcar.html. Accessed August 18, 2012.

6. Halperin W, Howard J. Occupational epidemiology and the National Institute for Occupational Safety and Health. MMWR. 2011;60 (supplement):97–103.

7. Association of Schools of Public Health, Education Committee. Master’s Degree in Public Health Core Competency Development Project, Version 2.3. Washington, DC: Association of Schools of Public Health; August 11, 2006. http://www.asph .org/. Accessed March 4, 2012.

51589_CH16_Printer.indd 721 11/02/13 6:02 AM

51589_CH16_Printer.indd 722 11/02/13 6:02 AM

723

Appendix

A Guide to the Crit ical Appraisal of an Epidemiologic/Public Health Research Article

For a review of the terms used in this guide, refer to the relevant chapters in the text. The ability to read an empirical research report and critique its strengths and

weaknesses is a desirable skill for epidemiology students to master. When you analyze a published article carefully, you will observe that all research reports contain strengths and weaknesses. The perfect epidemiologic research study has not yet been designed! Despite this assertion, a well-written article effectively communicates the purpose of the research study. It contains sufficient informa- tion so that the reader may assess the internal and external validity of the study design, that is, the degree of confidence that the reader can place in the findings that are reported. Many articles, such as those published in the American Journal of Public Health, are divided into sections (e.g., introduction, methods, results, and conclusion). Here is a list of criteria to consider when reading an empirical journal article (organized according to the sections contained in a typical empiri- cal research report):

1. Introduction The introduction provides a review of the relevant literature and sets the stage for the research study. From the introduction, the reader should be able to infer the:

●● problem studied ●● specific aims of the study ●● research question(s) and/or hypotheses ●● variables included in the study

51589_APPA_Printer.indd 723 09/02/13 1:00 PM

724 A p p e n d i x A

The report should communicate clearly the problem being investigated. If the investigator has not stated the objectives of the investigation explic- itly, the reader is left uncertain about the purpose of the study. The intro- duction also may provide an opportunity for the author to introduce variables measured as part of a theoretical framework. In addition to ascertaining the purpose of the research, the astute reader should try to determine the study’s principal outcome variable or variables (dependent variable or variables). The next step is to identify exposure, risk factor, or independent variables. Further, it should be possible to assess whether the researchers used intervening (moderating) and control variables.

2. Methods The methods section introduces the type of study design, that is, case- control, cohort, intervention, ecologic, or other design. Based upon the type of design specified, it should be apparent to the reader whether a cross-sectional, prospective, or retrospective time frame was used and whether the study was observational or experimental. In addition, the investigator should specify the unit of analysis: individual or group. Depending upon the study design, the investigator should state the number of observations made, that is, a single observation or multiple observations. a. Study Sample

This section presents the method for selecting the subjects. Possible sample designs include, but are not limited to:

●● probability-based sample (simple random sample; stratified ran- dom sample)

●● systematic sample of available clinic patients (known as a grab-bag sample)

●● sample of medical records from a healthcare setting The choice of a study sample forms a crucial aspect of external valid-

ity (generalizability) of the study’s findings; the reader needs to know the group or groups to which the study’s findings apply. These groups com- prise the target population to whom the study’s results can be generalized. Most researchers intend that the results of their research can be applied to populations other than the specific group from which their sample was selected. Sometimes the target population of the research may not be specified explicitly, leaving to the reader the task of divining the exact nature of the groups to which the findings might apply.

Information should be provided regarding the nature of control or comparison groups; the method of assignment of subjects to study

51589_APPA_Printer.indd 724 09/02/13 1:00 PM

G u i d e t o t h e C r i t i C A l A p p r A i s A l 725

conditions affects the internal validity of the study (as in the control of possible systematic group differences among study conditions). The reader should note whether single or double blinding has been used, espe- cially in clinical trials. How many subjects are chosen is especially crucial to population research; the number of subjects affects statistical power (ability to detect a statistically significant difference) and also must be suf- ficient for diseases and conditions that have low prevalence in the popula- tion. Inappropriate sampling designs may introduce errors and bias into the study. One type of bias is selection bias, which arises from nonrandom assignment of subjects into the study. Other sampling issues include sub- ject attrition (an important feature of sampling that could affect cohort studies and longitudinal studies adversely), which is measured by:

●● Refusal rates (the number of selected subjects who refused to participate)

●● Follow-up rates ●● Drop-out rates

b. Measures (Instruments) This section provides information on how the study’s investigators operationalized the variables used in the study. The term operational- ization refers to the schema for creation of actual numerical measure- ments that correspond to the concepts used in the study. For example, the article may discuss how outcome, exposure, and moderating vari- ables were measured. Studies that investigate diseases (e.g., diabetes, hypertension, or depression) must provide very clear “rule-in” and “rule-out” criteria for making these diagnoses. Other examples of measures that one might find in a study include:

●● Physiological measures (e.g., blood pressure, height, weight) ●● Blood chemistry (e.g., serum cholesterol, hemoglobin, glucose) ●● Control (demographic) variables (e.g., age, sex, race, socioeconomic

status) ●● Attitudinal surveys/questionnaire measures/clinical assessments

(e.g., measures of depressive symptoms, anxiety, type A personality, perceived social support)

●● Outcome measures (e.g., death, recovery, response to a medication) ●● Exposure variables (e.g., medications used, use of toxic chemicals,

type of employment) The author should present information on the reliability and validity

of measures used, particularly in the case of newly developed measures and measures of psychosocial variables. This information might come

51589_APPA_Printer.indd 725 09/02/13 1:00 PM

726 A p p e n d i x A

from previous reports of research with the measures and from the author’s own reliability data. c. Data Analysis

In this section, the investigator presents information regarding how the association between independent (exposure) and outcome vari- ables was assessed, along with methods to control for potential con- founding factors. Some authors cite the statistical software package used (e.g., SAS, SPSS, Epi Info). The report also should mention statistics used (e.g., chi square, t-test, or other appropriate statistic), depending upon the problem investigated. An important issue for data analysis is the use of appropriate statistical techniques. Some data analysis methods (e.g., multiple regression analysis) demand a large number of cases; yet, occasionally reports of research with small sam- ple sizes will show the inappropriate use of multiple regression analysis and other multivariate techniques. Some studies may not test specific hypotheses related to exposure–outcome associations, but instead present descriptive analyses of results (e.g., counts and percents of the number of cases in various categories). While the issue of data analysis can be quite complex, the reader should try to develop a familiarity with some of the common statistical procedures used by referring to a statistics text.

3. Results This section should refer back to the original issue addressed in the intro- duction (e.g., research question or hypothesis) and then discuss how the findings either do or do not answer the research question(s) posed in the introduction.

4. Discussion The discussion places the findings in the context of previous research and an existing theoretical framework. In this section, the study’s author may present weaknesses and limitations of the research. At this point, the reader should form an opinion regarding the extent to which the findings and conclusions follow from the design of the research study.

5. Conclusions The conclusions section should provide a concise review of the main points covered in the paper. Public health or clinical implications of the findings and directions for future research may be disclosed. After review- ing the entire article, the reader should reflect on how the research and its presentation might be improved. The reader also should consider the strengths and weaknesses of the article.

51589_APPA_Printer.indd 726 09/02/13 1:00 PM

Appendix

B Answers to Selected Study Questions

Chapter 2

Question 3. Professor Morris’s list of uses of epidemiology is probably exhaustive.

Question 9. a. The death rate dropped to 793.7 in 2009. The percentage change in

death rate was (1,719.1 - 793.7)/1,719.1 = 925.4/1,719.1 = (0.538305 × 100) = 53.8%

Chapter 3

Question 2. Age-specific death rates for malignant neoplasms of trachea, bronchus, and lung:

Ages 25–34 154/39,872,598 × 100,000 = 0.4 Ages 35–44 2,478/44,370,594 × 100,000 = 5.6 Ages 45–54 12,374/40,804,599 × 100,000 = 30.3 Ages 55–64 30,956/27,899,736 × 100,000 = 111.0 Ages 65–74 49,386/18,337,044 × 100,000 = 269.3 Inferences: Rate increases with age. Question 3. Age-specific death rates: Ages 20–24 19,973/20,727,694 × 100,000 = 96.4 Ages 25–34 41,300/39,872,598 × 100,000 = 103.6 Ages 35–44 89,461/44,370,594 × 100,000 = 201.6 Age- and sex-specific death rates:

Males aged 20–24 14,964/10,663,922 × 100,000 = 140.3

727

51589_APPB_Printer.indd 727 09/02/13 1:00 PM

728 A p p e n d i x B

Males aged 25–34 28,602/20,222,486 × 100,000 = 141.4 Males aged 35–44 56,435/22,133,659 × 100,000 = 255.0 Females aged 20–24 5,009/10,063,772 × 100,000 = 49.8 Females aged 25–34 12,698/19,650,112 × 100,000 = 64.6 Females aged 35–44 33,026/22,236,935 × 100,000 = 148.5

Question 4. a. Calculate the crude death rates:

Crude death rate 2,448,288/290,810,789 × 100,000 = 841.9 Death rate, males 1,201,964/143,037,290 × 100,000 = 840.3 Death rate, females 1,246,324/147,773,499 × 100,000 = 843.4 Due to accidents:

Total 109,277/290,810,789 × 100,000 = 37.6 Males 70,532/143,037,290 × 100,000 = 49.3 Females 38,745/147,773,499 × 100,000 = 26.2

Due to malignant neoplasms: Total 556,902/290,810,789 × 100,000 = 191.5 Males 287,990/143,037,290 × 100,000 = 201.3 Females 268,912/147,773,499 × 100,000 = 182.0

Due to Alzheimer’s disease: Total 63,457/290,810,789 × 100,000 = 21.8 Males 18,335/143,037,290 × 100,000 = 12.8 Females 45,122/147,773,499 × 100,000 = 30.5

b. Proportional mortality ratios (PMRs): PMR for accidents:

Total 109,277/2,448,288 × 100 = 4.5 Males 70,532/1,201,964 × 100 = 5.9 Females 38,745/1,246,324 × 100 = 3.1

PMR for malignant neoplasms: Total 556,902/2,448,288 × 100 = 22.7 Males 287,990/1,201,964 × 100 = 24.0 Females 268,912/1,246,324 × 100 = 21.6

PMR for Alzheimer’s disease: Total 63,457/2,448,288 × 100 = 2.6 Males 18,335/1,201,964 × 100 = 1.5 Females 45,122/1,246,324 × 100 = 3.6

c. Maternal mortality rate 495/4,089,950 × 100,000 = 12.1 per 100,000 live births

51589_APPB_Printer.indd 728 09/02/13 1:00 PM

A n s w e r s t o s e l e c t e d s t u d y Q u e s t i o n s 729

d. Infant mortality rate 28,025/4,089,950 × 1,000 = 6.9 per 1,000 live births

e. Crude birth rate 4,089,950/290,810,789 × 1,000 = 14.1 per 1,000 population

f. General fertility rate 4,089,950/61,910,608 × 1,000 = 66.1 per 1,000 women aged 15–44 (Refer to Table 3A-2 and obtain the total number of women aged 15-44 by summing the number of females in each age group [e.g., 15-19, 20-24, etc.]. The total is 61,910,608.)

Question 5. The prevalence of HIV is 137.0 per 100,000 population. The incidence of HIV is 24.1 per 100,000 population.

Question 8. a. 1.3 to 1.0 (sex ratio, male to female regular drinkers) b. 42.8% (proportion of regular drinkers who are women) c. 565.2 per 1,000 (men only); 393.2 per 1,000 (women only); 476.1 per

1,000 (total population) Question 10. These data are prevalence data from which it is not possible to

infer risk. Question 11. 20%

Chapter 5

Question 8. SURVEY Advantages

-More control over the quality of the data. - More in-depth data can be collected on each case than is usually possible with surveillance.

- Can identify spectrum of childhood injuries, including those that do not warrant medical care.

-More accurate assessment of true incidence and prevalence. Disadvantages

- More costly to perform since survey requires development of de novo data collection system and hiring of interviewers who require training and supervision.

- Represents only single point in time (“snapshot”); may miss seasonal trends; misses rare diseases; misses rapidly fatal diseases.

51589_APPB_Printer.indd 729 09/02/13 1:00 PM

730 A p p e n d i x B

- Tells little if anything about changes over time in incidence or prevalence of a behavior or outcome.

- Recall bias more likely to affect results because data are collected retrospectively (surveillance is usually prospective).

SURVEILLANCE Advantages

-Cheaper (for the health department). -Can often use existing systems and health personnel for data collection. -Allows monitoring of trends over time. - Ongoing data collection may allow collection of an adequate number of cases to study those at risk. With surveys, an event may be too infrequent to gather enough cases for study; with surveillance, the observation period can be extended until sufficient numbers of cases are collected.

Disadvantages - May not provide a representative picture of the incidence or prevalence unless care is taken in selecting reporting sites and ensuring complete reporting.

- Data that can be collected are limited by the skill, time, and good will of the data collectors, who usually have other responsibilities.

-Quality control may be a major problem in data collection. -The quality of data may vary between collection sites.

Question 9. 1. Change in surveillance system/policy of reporting 2. Change in case definition 3. Improved diagnosis

-New laboratory test - Increased physician awareness of the syndrome, new physician in town, etc.

-Increase in publicity/public awareness may have prompted individuals or parents to seek medical attention for compatible illness

4. Increase in reporting (i.e., improved awareness of requirement to report) 5. Batch reporting (unlikely in this scenario) 6. True increase in incidence

Question 10. No right answer, but one sequence might be as follows:

Table 1: Number of reported cases this week, disease by county

51589_APPB_Printer.indd 730 09/02/13 1:00 PM

A n s w e r s t o s e l e c t e d s t u d y Q u e s t i o n s 731

Table 2: Number of reported cases, disease by week (going back 6–8 weeks for comparison)

Table 3: Number of reported cases for past 4 weeks, disease by year (going back 5 years for comparison)

Table 1 addresses disease occurrence by place. Tables 2 and 3 address dis- ease occurrence by time. Together, these tables should give an indication of whether an unusual cluster or pattern of disease is occurring. If such a pattern is detected, person characteristics may then be explored. Question 11. Many state health department newsletters do not go to “all who need to know.” Even among those who receive the newsletter, some do not read it at all; many others skim the articles and ignore the tables altogether. In addition, depending on the timing of the laboratory report and publication deadlines, the information may be delayed by up to several weeks. This information is important for all who may be affected, and for all who

may be able to take preventive measures, including: - Other public health agencies (e.g., neighboring local health departments, animal control staff, etc.)

-Healthcare providers -Veterinarians -The public (inform by issuing press release to the media)

Source: (Answers to Questions 8–11): Reprinted from Centers for Disease Con- trol and Prevention. Principles of Epidemiology, 2nd ed. Atlanta, GA: CDC; 1998:337–338.

Chapter 6

Question 7. (37)(121) OR = ————— = 2.74 (68)(24)

Question 8. (105)(137) OR = —————— = 0.37 (a protective effect) (463)(84)

Note: A = 189 – 105; B = 600 – 137

51589_APPB_Printer.indd 731 09/02/13 1:00 PM

732 A p p e n d i x B

Question 9. (37)(752) OR = —————— = 1.60 (48)(362)

Note: C = 399 – 37; D = 800 – 48

Chapter 7

Question 5. Relative risk of anxiety: (500/10,000)/(200/20,000) = 0.05/0.01 = 5 Question 8. A cohort study would probably not be necessary given the strong association.

Chapter 8

Question 8. d Question 9. a. (case-control); b. (prospective cohort); c. (clinical trial); d. (retrospective cohort); e. (prospective cohort); f. (clinical trial); g. (cross-sectional)

Chapter 9

Question 1. (1.2 – 1)/1.2 = 0.167 (1.8 – 1)/1.8 = 0.444 (3.0 – 1)/3.0 = 0.667 (15.0 – 1)/15.0 = 0.933 Question 2. c Question 3. RR (lung cancer) = (71/100,000)/(7/100,000) = 10.1 Question 4. RR (coronary thrombosis) = (599/100,000)/(422/100,000) = 1.4 Question 5. Etiologic fraction (lung cancer) = [(71/100,000 – (7/100,000)]/

(71/100,000) = 0.90

51589_APPB_Printer.indd 732 09/02/13 1:00 PM

A n s w e r s t o s e l e c t e d s t u d y Q u e s t i o n s 733

Question 6. Etiologic fraction (coronary) = [(599/100,000) – (422/100,000)]/ (599/100,000) = 0.30

Question 7. Population etiologic fraction (lung cancer) = [0.55(10.1 – 1)]/ [0.55(10.1 – 1) + 1] = (5.005/6.005) × 100 = 83%

Question 8. Population etiologic fraction (coronary) = [0.55 (1.4 – 1)]/ [0.55(1.4 – 1) + 1] = (0.22/1.22) × 100 = 18.0%

Question 9. a. False b. True. RR for lung cancer is 10.1 vs. 1.4 for coronary thrombosis; and EF

for lung cancer is 90% vs. 30% for coronary thrombosis. c. False d. False e. False Question 10. a. RR = (500/10,000)/(200/20,000) = 0.05/0.01 = 5 b. Risk difference = 0.05 – 0.01 = 0.04 c. Etiologic fraction = [(0.05 – 0.01)/(0.05)] × 100 = 80% d. Population etiologic fraction = 0.33(5 – 1)/0.33(5 – 1) + 1 = 1.32/2.32 ×

100 = 56.9%

Chapter 10

Question 1. Yes. Because one-third of the subjects were lost to follow-up, a selec- tion bias is suspected. We do not know why the 20 people died. If they died of the disease in question, the incidence rate would prob- ably be higher than 8%.

Question 2. Stratify the cases and controls. Question 3. Device B gives greater validity because the two measurements are

closer to the actual number of cells to be counted. Device A has greater reliability because its two measurements are closer together than those of Device B.

Question 4. Yes. Smoking is strongly associated with disease (lung cancer) and is likely to be associated with diet. Therefore, smoking should be controlled for in either the study design or data analysis.

Question 5. Yes. The rate among the self-referred group was much higher than the rate the investigator found among members, leading to a poten- tially elevated RR. The self-referred group may represent a subset of

51589_APPB_Printer.indd 733 09/02/13 1:00 PM

734 A p p e n d i x B

people with special circumstances, because they were not included in the first military group (i.e., discharged, different duty assign- ment, etc.). They volunteered for the study, so it is possible they suspected their exposure or had a particular interest in the outcome of the study.

Question 6. There is no validity to the scale, except for subjects who actually weigh 30 kg! However, the scale has high reliability because it gives consistent readings.

Question 7. b Question 8. e Question 9. d Question 10. c

Chapter 11

Question 6. Sensitivity = 50%; Specificity = 25%; Predictive value (+) = 25%; Predictive value (–) = 50%

Question 7. Sensitivity = 69% Specificity = 97%

Question 8. Predictive value (+) = 90%; Predictive value (–) = 90%

Question 9. Accuracy = 83% Question 10. a. Predictive value of a positive test = 63.0%

Predictive value of a negative test = 98.8%

Chapter 12

Question 2. Secondary attack rate = [(14 – 2)/(20 – 2)] × 100 = 12/18 × 100 = 66.7%

Question 10. a. Based on the difference in attack rates between those persons who ate and

did not eat the food items served, the rare beef was the food most likely to be responsible for the outbreak. The attack rate among those who ate rare beef was 17/23 × 100 = 74%. The attack rate among those who did not eat rare beef was 0/1 × 100 = 0%. The difference in attack rates was

51589_APPB_Printer.indd 734 09/02/13 1:00 PM

A n s w e r s t o s e l e c t e d s t u d y Q u e s t i o n s 735

74% – 0% = 74%. When all of the differences in attack rates between those persons who ate and did not eat the food items served are calculated by using procedures similar to those for the rare beef calculations, one can determine that 74% is the greatest difference.

Chapter 14

Question 1. Bb × BB; Bb × Bb; Bb × bb; BB × bb. Question 2. BB × bb = 100% heterozygotes Question 3. A female receives no sex chromosome from her father’s father,

whereas a male receives no sex chromosome from his father’s mother. Question 4. This result corresponds to an odds ratio of about 2. Therefore, the

data provide minimal evidence for a familial component. Question 5. This procedure provides an indirect match on the age of the relatives. Question 6. Smoking, occupation, radon exposure, secondhand smoke, diet, and

age Question 7. No. Aside from possible biases in selection of the families, and errors

in the analysis, another possibility may be that any genetic influence on the disease is insufficient to manifest in a Mendelian pattern of disease (low penetrance).

51589_APPB_Printer.indd 735 09/02/13 1:00 PM

51589_APPB_Printer.indd 736 09/02/13 1:00 PM

Glossary

Acculturation Modifications that individuals or groups undergo when they come into contact with another culture.

Acculturation hypothesis Proposes that as immigrants become acculturated to a host country, their health profiles tend to converge with that of the native- born population.

Adjusted rate Rate of morbidity or mortality in a population in which statisti- cal procedures have been applied to permit fair comparisons across populations by removing the effect of differences in the composition of various populations; an example is age adjustment.

Agent In the epidemiologic triangle, the cause of a disease; in infectious dis- eases, often the agent is a microbe such as a virus or bacterium.

Age-specific rate Frequency of a disease in a particular age stratum divided by the total number of persons within that age stratum during a time period.

Allele One of two or more alternative forms of a gene that occurs at the same locus (site or location on a chromosome occupied by a gene, which comprises a particular set of alleles).

Allostasis Refers to how the organism achieves stability (homeostatsis) through continual change.

Allostatic load Refers to the consequences of sustained activation of primary regulatory mechanisms serving allostasis over time.

Alternative hypothesis A part of significance testing that signifies that the null hypothesis is false.

Analytic study A type of research design concerned with the determinants of disease and the reasons for relatively high or low frequency of disease in specific population subgroups. Analytic studies identify causes of the problem, test spe- cific etiologic hypotheses, generate new etiologic hypotheses, and suggest mecha- nisms of causation; they also may include case-control studies, cohort studies, and some types of ecologic studies.

737

51589_GLOS_Printer.indd 737 11/02/13 2:45 PM

Antigenicity Ability of an agent to induce antibody production in the host.

Association A linkage, connection, or statistical dependence between (among) two or more factors.

Attack rate An alternative form of the incidence rate that is used when the nature of a disease or condition is such that a population is observed for a short time period. The attack rate is calculated by the formula ill/(ill  well) 3 100 (during a time period). The attack rate is not a true rate because the time dimen- sion is often uncertain.

Attributable risk A measure of risk difference. In a cohort study, refers to the difference between the incidence rate of a disease in the exposed group and the incidence rate in the nonexposed group.

Autosomal dominant A situation in which only a single copy of an altered gene located on a nonsex chromosome is sufficient to cause an increased risk of disease.

Autosomal recessive Denotes those diseases for which two copies of an altered gene are required to increase risk of disease.

Availability of the data Refers to the investigator’s access to data (e.g., patient records and databases in which personally identifying information has been removed).

Basic reproductive rate (R0) (during an epidemic) A measure of the number of infections produced on average by an infected individual in the early stages of an epidemic when virtually all contacts are susceptible.

Behavioral epidemiology Field of epidemiology that studies the role of behavior factors in health.

Behavioral medicine The application of behavioral factors to specific clinical interventions.

Bias (also, systematic errors) Refers to deviations of results, or inferences, from the truth.

Biostatistics A field of statistics that is applied to living organisms.

Blinding (also, masking) An aspect of study design wherein the subject is not aware of his/her group assignment of placebo or treatment; seeks to alleviate bias in study results.

Buffering model of social support Hypothesizes that one of the functions of social network ties is to lessen the adverse psychological consequences of stress.

738 G l o s s a r y

51589_GLOS_Printer.indd 738 11/02/13 2:45 PM

Case clustering An unusual aggregation of health events grouped together in space or time.

Case-control study A study that compares individuals who have a disease with individuals who do not have the disease in order to examine differences in expo- sures or risk factors for the disease.

Case fatality rate Number of deaths caused by a disease among those who have the disease during a time period.

Causality Referring to the relationship between cause and effect.

Cause Act, event, or state of nature that initiates/permits, alone or in conjunc- tion with other causes, a sequence of events resulting in an effect.

Cause-specific rate Measure that refers to mortality (or frequency of a given disease) divided by the population size at the midpoint of a time period times a multiplier.

Choropleth map A map that represents disease rates (or other numerical data) for a group of regions by different degrees of shading.

Chronic disease A long-lasting illness that is difficult to eradicate.

Classification Method of arranging disease entities into categories that share similar features, allowing for statistical compilation.

Clinical trial A carefully designed and executed investigation of the effects of a treatment or technology that uses randomization, blinding of subjects to study conditions, and manipulation of the study factor.

Codon A set of three of four amino acids that can combine in different orders to code for 64 different amino acids, which when strung together into a long sequence comprise the genetic code of DNA or RNA.

Cohort A group of individuals who share an exposure in common and who are followed over time; an example is an age cohort.

Cohort effect Consequence of long-term secular trends in exposure within a specific cohort.

Cohort life table A table that presents mortality statistics of all persons born during a particular year.

Cohort study (also, prospective or longitudinal study) A type of study that collects data and follows a group of subjects who have received a specific expo- sure. The incidence of a specific disease or other outcome of interest is tracked

G l o s s a r y 739

51589_GLOS_Printer.indd 739 11/02/13 2:45 PM

over time. The incidence in the exposed group is compared with the incidence in groups that are not exposed, have different levels of exposure, or have different types of exposures.

Colonization Situation wherein an infectious agent may multiply on the surface of the body without invoking a tissue or immune response.

Common source epidemic A disease outbreak caused by common exposure of a group of individuals to a disease agent.

Communicable disease A illness caused by an infectious agent that can be transmitted from one person to another.

Community-based participatory research (CBPR) An alliance between community organizations and research units in order to investigate health- related issues of interest to the community.

Community intervention (also community trial) An intervention designed for the purpose of educational and behavioral changes at the population level.

Completeness of population coverage Encompasses representativeness of the data and thoroughness of case identification.

Concurrent validity A type of measurement obtained by correlating a mea- sure with an alternative measure of the same phenomenon taken at the same point in time (see Validity).

Confidence interval A computed interval of values that, with a given prob- ability, is said to contain the true value of the population parameter; a measure of uncertainty about a parameter estimate. An example is the confidence interval about a relative risk measure.

Confounding Masking of an association between an exposure and an out- come because of the influence of a third variable that was not considered in the study design or analysis.

Construct validity Degree to which a measurement agrees with the theoretical concept being investigated.

Content validity (also, rational or logical validity) Degree to which a mea- sure covers the domains or range of meanings included within a concept.

Continuous common source epidemic An outbreak that lasts longer than the time span of a single incubation period and is caused by a common source of exposure.

740 G l o s s a r y

51589_GLOS_Printer.indd 740 11/02/13 2:45 PM

Continuous variable A type of variable that can have an infinite number of values within a specified range (e.g., blood pressure measurements).

Count Total number of cases of a disease or other health phenomenon being studied.

Criterion-referenced validity Type of validity found by correlating a measure with an external criterion of the entity being assessed to determine the accuracy of the data (can include predictive and concurrent validity).

Crossover design (also, treatment crossover) Any change of treatment for a patient in a clinical trial that involves a switch of study treatments.

Cross-sectional study (also, prevalence study) A type of descriptive study (e.g., a population survey) designed to estimate the prevalence of a disease or exposure.

Crude birth rate Number of live births during a specified period of time per the resident population during the midpoint of the time period (expressed as rate per 1,000).

Crude rate A summary rate based on the actual number of events in a popula- tion over a given time period. An example is the crude death rate, which approxi- mates the proportion of the population that dies during a time period of interest.

Cumulative incidence (also, cumulative incidence rate) Number or pro- portion of a population (or group of people) who become diseased or develop a condition being studied during a stated period of time; used to calculate risk.

Cyclic fluctuation An increase or decrease in the frequency of a disease or health condition in a population over a period of years or within each year.

Demographic transition Historical shift from high birth and death rates found in agrarian societies to much lower birth and death rates found in devel- oped countries.

Demography The study of data (e.g., births, deaths, and socioeconomic sta- tus) related to the structure of human populations.

Dependent variable A factor in a theoretical model that is affected or influ- enced by an independent variable.

Descriptive epidemiology Epidemiologic studies that are concerned with characterizing the amount and distribution of health and disease within a population.

G l o s s a r y 741

51589_GLOS_Printer.indd 741 11/02/13 2:45 PM

Descriptive study A type of study designed to portray the health characteristics of a population with respect to person, place, and time. Such studies are uti- lized to estimate disease frequency and time trends, and include case reports, case series, and cross-sectional surveys.

Determinant A factor or event that is capable of bringing about a change in the health status of a population.

Direct transmission Spread of infection through person-to-person contact.

Disability-adjusted life years (DALY) A measure that adds the time a person has a disability to the time lost to early death; thus, one DALY indicates one year of life lost to the combination of disability and early mortality.

Disappearing disorder Type of disease or illness that was formerly a common source of morbidity and mortality and that has nearly disappeared in epidemic form.

Distribution Differential frequency in the occurrence of disease and mortality among population groups (or among subgroups of a population).

Dose–response curve Graphical representation of the relationship between changes in the size of a dose or exposure and changes in response. This curve generally has an “S” shape.

Double-blind design Feature of a clinical trial in which neither the subject nor the experimenter is aware of the subject’s group assignment in relation to control or treatment status.

Dynamic population A population that adds new members through immi- gration and births or loses members through emigration and death.

Ecologic comparison study Type of research design that assesses the cor- relation (association) between exposure rates and disease rates among differ- ent groups or populations over the same time period. The unit of analysis is the group.

Ecologic fallacy A misleading conclusion about the relationship between a factor and an outcome that occurs when the observed association obtained between study variables at the group level does not necessarily hold true at the individual level.

Ecologic trend study Type of study that examines the correlation of changes in exposure and changes in disease over time within the same community, country, or other aggregate unit.

742 G l o s s a r y

51589_GLOS_Printer.indd 742 11/02/13 2:45 PM

Effect measure A quantity that measures the effect of a factor on the frequency or risk of a health outcome.

Emerging infection An abrupt increase in the incidence or geographic scope of a seemingly new infectious disease (e.g., hantaviral pulmonary syndrome found in the Southwestern United States).

Endemic A disease or infectious agent that is habitually present in a commu- nity, geographic area, or population group. Often an endemic disease maintains a low but continuous incidence.

Environment Domain in which the disease-causing agent may exist, survive, or originate.

Environmental epidemiology The study of diseases and conditions (occur- ring in the population) that are linked to environmental factors.

Epidemic Occurrence of a disease clearly in excess of normal expectancy.

Epidemiologic transition A shift in the pattern of morbidity and mortality from causes related primarily to infectious and communicable diseases to causes associated with chronic, degenerative diseases; is accompanied by demographic transition.

Epidemiology Study of the distribution and determinants of health and dis- ease, morbidity, injuries, disability, and mortality in populations. Epidemiologic studies are applied to the control of health problems in populations.

Exposure A causal factor or variable; to be in contact with a causal factor such as a disease agent or toxic chemical.

Etiologic fraction Proportion of the rate of disease in an exposed group that is due to the exposure.

Etiology Having to do with causes or the mode in which diseases are caused.

Experimental study Research design in which the investigator manipulates the study factor and randomly assigns subjects to exposed and nonexposed conditions.

External validity Measure of the generalizability of the findings from the study population to the target population.

Family recall bias A type of bias that occurs when cases are more likely to remember the details of their family history than are controls (see Bias).

G l o s s a r y 743

51589_GLOS_Printer.indd 743 11/02/13 2:45 PM

Family study A type of research design in which data on phenotype and risk factors are measured in individual family members.

Fertility rate (see General fertility rate).

Fetal death rate Number of fetal deaths after 20 weeks or more gesta- tion divided by the number of live births plus fetal deaths after 20 weeks or more gestation during a year (expressed as rate per 1,000 live births plus fetal deaths).

Fetal death ratio Number of fetal deaths after a gestation of 20 weeks or more divided by the number of live births during a year (expressed as rate per 1,000 live births).

Fixed population A population distinguished by a specific happening and addings no new members; population decreases in size as a result of deaths only.

Functional cloning (also, functional mapping) After the proteins involved with a disease process are identified, a geneticist can determine their amino acid sequence and decipher the DNA code for that sequence; then the geneticist can search for where the sequence occurs in the human genome.

Gene A particular segment of a DNA molecule on a chromosome that deter- mines the nature of an inherited trait in an individual.

General adaptation syndrome A term used to describe the body’s short-term and long-term reactions to stress; made up of three stages: alarm reaction, stage of resistance, and stage of exhaustion.

General fertility rate Number of live births reported in an area during a given time interval divided by the number of women aged 15 to 44 years in that area (expressed as rate per 1,000 women aged 15 to 44).

Generation time An interval of time between lodgment of an infectious agent in a host and the maximal communicability of the host.

Genetic epidemiology Field of epidemiology concerned with inherited factors that influence risk of disease.

Genotype In an individual, refers to his or her genetic constitution, often stated in reference to a specific trait or at a particular locus.

GINI Index A common measure of income inequality expressed as a number that ranges from 0 to 1. The closer the index is to one, the greater is the level of inequality.

744 G l o s s a r y

51589_GLOS_Printer.indd 744 11/02/13 2:45 PM

Glasgow effect Excess mortality in the West of Scotland (Glasgow) after controlling for deprivation.

Halo effect Refers to the influence on an observer’s perception of an individu- al’s characteristics (other than the characteristic that is under study).

Health disparities Differences in health that occur among population groups, especially differences related to gender, race or ethnicity, education, or income.

Health policy An action statement or principle that an organization has adapted with respect to health.

Healthy migrant effect In studies of migration and health, a bias that results from the migration of younger, healthier persons in comparison with those who remain at home (see Bias).

Healthy worker effect Error linked to the observation that employed persons tend to have lower mortality rates than the general population; stems from the fact that good health is necessary for obtaining and maintaining employment (see Bias).

Herd immunity Resistance of an entire community to an infectious disease due to the immunity of a large proportion of individuals in that community to the disease.

Homeostasis Refers to a tendency toward a stable equilibrium among physi- ological processes.

Host Person (or animal) who (that) has a lodgment of an infectious disease agent under natural conditions.

Hypothesis Supposition tested by collecting facts that lead to its acceptance or rejection.

Identical by descent (IBD) In siblings, an allele that comes from the same parent.

Identical by state Alleles that are the same in siblings but come from different parents.

Infectious disease Synonym for communicable disease.

Infestation Presence of a living infectious agent on the body’s exterior surface, causing a local tissue or immune response.

Immunogenicity The ability of an infection to produce specific immunity (protection against a disease).

G l o s s a r y 745

51589_GLOS_Printer.indd 745 11/02/13 2:45 PM

Inapparent infection A type of infection that shows no clinical or obvious symptoms.

Incidence density Number of new cases of disease during a time period divided by the total person-time of observation; used to calculate incidence when subjects have been observed for varying periods of time.

Incidence rate Number of new cases of a disease—or other condition—in a population divided by the total population at risk over a time period times a multiplier (e.g., 100,000).

Incubation period Time interval between exposure to an infectious agent and the appearance of the first signs or symptoms of disease.

Independent variable (also, exposure or risk factor variable) Hypothesized causal factor in a theoretical model.

Index case In an epidemiologic investigation of a disease outbreak, the first case of disease to come to the attention of authorities (e.g., the initial case of Ebola virus).

Infant mortality rate Number of infant deaths among infants aged 0 to 365 days during a year divided by the number of live births during the same year (expressed as the rate per 1,000 live births).

Infectivity Capacity of an agent to enter and multiply in a susceptible host and thus produce infection or disease.

Information bias Measurement error in assessment of exposure and/or disease; types include recall bias and interviewer bias (see Bias).

Internal validity Measures the extent to which differences in an outcome between or among groups in a study can be attributed to the hypothesized effects of an exposure, an intervention, or other causal factor being investigated. A study is said to have internal validity when there have been proper selection of study groups and a lack of error in measurement (see Validity).

Intervention study A type of research design that tests the efficacy of a preven- tive or therapeutic measure. Intervention studies include controlled clinical trials and community interventions.

Late fetal death rate Number of fetal deaths after 28 weeks or more gestation divided by the number of live births plus fetal deaths after 28 weeks or more ges- tation during a year (expressed as rate per 1,000 live births plus late fetal deaths).

746 G l o s s a r y

51589_GLOS_Printer.indd 746 11/02/13 2:45 PM

Latency Time period between initial exposure to an agent and development of a measurable response. The latency period can range from a few seconds (in the case of acutely toxic agents) to several decades (in the case of some forms of cancer).

Lead time bias Erroneous perception that a screening-detected case of dis- ease has a longer survival than an unscreened case of the same disease simply because the screened case was identified earlier in its natural history than was the unscreened case (see Bias).

Length bias Error resulting from the fact that some cases of disease (particu- larly tumors) detected by screening programs tend to progress more slowly than those detected by clinical manifestations (see Bias).

Levels Denotes the hierarchy of tasks that epidemiology studies seek to accomplish.

Life expectancy Number of years that a person is expected to live, at any particular year.

Lifestyle Habits, behaviors (smoking, drinking, and exercise levels), or dietary practices that influence health.

Linkage disequilibrium When a specific allele at a marker locus is strongly associated with a mutant allele.

Linked Genes that are in close physical proximity to each other on the same chromosome.

Manipulation of the study factor Issues that an investigator can determine such as timing of exposure, intensity, and duration.

Mass screening (also, population screening) A type of screening that col- lects data on total population groups, regardless of any a priori information as to whether the individuals are members of a high-risk subset of the population.

Maternal mortality rate Number of maternal deaths ascribed to childbirth divided by the number of live births times 100,000 live births during a year.

Measurement bias Refers to the constant errors that are introduced by a faulty measuring device.

Measures of natality Statistics associated with births.

Mendelian inheritance Named for the 19th century Austrian monk, Gregor Mendel, denotes the transmission of a disease or trait from parents to offspring according to simple laws of inheritance.

G l o s s a r y 747

51589_GLOS_Printer.indd 747 11/02/13 2:45 PM

Metropolitan statistical areas (MSAs) Formerly known as standard metropol- itan statistical areas (SMSAs), geographic areas of the United States established by the Bureau of the Census to provide a distinction between metropolitan and nonmetropolitan areas by type of residence, industrial concentration, and popu- lation concentration.

Moderating variable (also, intervening variable) Third variable (effect modifier) that is intermediate in the causal process between exposure factor (independent variable) and outcome (dependent variable); modifies the effect of the exposure upon disease status.

Molecular epidemiology Field of epidemiology that uses biomarkers to estab- lish exposure–disease associations. Examples of biomarkers are serum levels of micronutrients and DNA fingerprints.

Morbidity Occurrence of an illness or illnesses in a population.

Mortality Occurrence of death in a population.

Mortality difference Measure of the difference between an exposed and non- exposed population in the frequency of death (see Risk difference).

Multiphasic screening Use of two or more screening tests simultaneously among large groups of people.

Multiple causality (also, multicausality or multifactorial etiology) A por- trayal of causality wherein several individual, community, and environmental factors may interact to cause a particular disease or condition.

Mutation A change in DNA that may adversely affect an organism.

Nativity Place of origin (e.g., native-born or foreign-born) of the individual or his or her relatives.

Natural experiment A type of research design in which the experimenter does not control the manipulation of a study factor(s). The manipulation of the study factor occurs as a result of natural phenomena or policies that impact health, an example being laws that control smoking in public places.

Nature of the data Refers to the source of the data (e.g., vital statistics, physi- cian’s records, case registries, etc.).

Neonatal mortality rate Number of infant deaths under 28 days of age divided by the number of live births during a year.

748 G l o s s a r y

51589_GLOS_Printer.indd 748 11/02/13 2:45 PM

Nested case-control study A type of research design wherein both cases and controls come from the population of a cohort study (see Case-control study and Cohort study).

Nomenclature A highly specific set of terms for describing and recording clin- ical or pathologic diagnoses to classify ill persons into groups.

Nonprobability sample Type of sample in which the population does not have a nonzero probability of being included in the sample (e.g., quota samples and judgmental samples).

Null hypothesis A hypothesis of no difference in a population parameter among the groups being compared.

Observational study A type of research design in which the investigator does not manipulate the study factor or use random assignment of subjects. There is careful measurement of the patterns of exposure and disease in a popula- tion in order to draw inferences about the distribution and etiology of diseases. Observational studies include cross-sectional, case-control, and cohort studies.

Odds ratio Measure of association between frequency of exposure and fre- quency of outcome used in case-control studies. The formula is (AD)/(BC), where A is the number of subjects who have the disease and have been exposed, B is the number who do not have the disease and have been exposed, C is the number who have the disease and have not been exposed, and D is the number who do not have the disease and have not been exposed.

Operationalization Methods used to translate concepts used in research into actual measurements.

Operations research A type of study of the placement of health services in a community and the optimum utilization of such services.

Outbreak A localized disease epidemic (e.g., in a town or healthcare facility).

P value An assessment that indicates the probability that the observed findings of a study could have occurred by chance alone.

Pandemic An epidemic that spans a wide geographic area. A worldwide influ- enza outbreak is an example of a pandemic.

Passive smoking (also, secondhand or sidestream exposure to cigarette smoke) Refers to the involuntary breathing of cigarette smoke by nonsmokers in an environment where cigarette smokers are present.

G l o s s a r y 749

51589_GLOS_Printer.indd 749 11/02/13 2:45 PM

Pathogenesis Process and mechanism of interaction of disease agent(s) with a host in causing disease.

Pathogenicity Capacity of an agent to cause overt disease in an infected host.

Penetrance Refers to the probability that a gene or genetic trait is expressed.

Perinatal mortality rate Number of late fetal deaths after 28 weeks or more gestation plus infant deaths within 7 days of birth divided by the number of live births plus the number of late fetal deaths during a year (expressed as rate per 1,000 live births and fetal deaths).

Perinatal mortality ratio Number of late fetal deaths after 28 weeks or more gestation plus infant deaths within 7 days of birth divided by the number of live births during a year (expressed as rate per 1,000 live births).

Period prevalence Number of cases of illness during a time period divided by the average size of the population.

Periodic life table A type of statistical table that provides an overview of the present mortality experience of a population and shows projections of future mortality experience.

Persisting disorder A type of illness or disease that remains common because an effective method of prevention or cure has not yet been discovered.

Person–environment fit The fit between the characteristics of a person and the properties of his or her environment.

Point epidemic Response of a group of people circumscribed in place to a common source of infection, contamination, or other etiologic factor to which they were exposed almost simultaneously.

Point prevalence Number of cases of illness in a group or population at a point in time divided by the total number of persons in that group or population.

Polymorphic traits Traits that vary greatly from individual to individual.

Population-based cohort study A type of cohort study that includes either an entire population or a representative sample of the population (see Cohort study).

Population etiologic fraction (also, attributable fraction in the popula- tion) Proportion of the rate of disease in the population that is due to an expo- sure. It is calculated as the population risk difference (Ip – Ine) divided by the rate of disease in the population (Ip).

750 G l o s s a r y

51589_GLOS_Printer.indd 750 11/02/13 2:45 PM

Population risk difference Difference between the incidence rate (risk) of disease in the nonexposed segment of the population (Ine) and the overall inci- dence rate (IP): (IP – Ine). It measures the benefit to the population derived by modifying a risk factor.

Positional cloning (also, physical mapping) Method for geneticists to iden- tify a gene responsible for a disease from a chromosomal region of DNA.

Postneonatal mortality rate Number of infant deaths from 28 days to 365 days after birth divided by the number of live births minus neonatal deaths during a year (expressed as rate per 1,000 live births).

Predictive validity Ability of a measure to forecast some attribute or charac- teristic in the future.

Predictive value (positive and negative) () Proportion of individuals who are screened positive by a test and actually have the disease. (2) Proportion of indi- viduals who are screened negative by a test and actually do not have the disease.

Prepathogenesis Period of time that precedes the interaction between an agent of disease and the host.

Prevalence Number of existing cases of a disease or health condition in a pop- ulation at some designated time.

Prevalence difference Measure that computes the difference in prevalence between an exposed and nonexposed population (see Risk difference).

Primary prevention Activities that are designed to reduce the occurrence of disease and that occur during the period of prepathogenesis (i.e., before an agent interacts with a host).

Primordial prevention Actions and measures that inhibit the emergence and establishment of factors such as environmental, economic, social and behavioral conditions, and cultural patterns of living known to increase the risk of disease. Seeks to minimize health hazards in general.

Probability sample Type of sample in which every element in the population has a nonzero probability of being included in the sample (e.g., simple random sample).

Proband Individual in a family who brings a disease to the attention of the investigator. In a case family, the proband is likely to be the person who is affected with the disease of interest.

G l o s s a r y 751

51589_GLOS_Printer.indd 751 11/02/13 2:45 PM

Prophylactic trial A type of clinical trial designed to evaluate the effectiveness of a treatment or substance used to prevent disease. Examples are clinical trials to test vaccines and vitamin supplements.

Proportion Fraction in which the numerator is a part of the denominator.

Proportional mortality ratio (PMR) Number of deaths within a population due to a specific disease or cause divided by the total number of deaths in the population during a time period such as a year.

Prospective cohort study A type of cohort study design that collects data on exposure at the initiation (baseline) of a study and follows the population in order to observe the occurrence of health outcomes at some time in the future.

Protective factor A circumstance or substance that provides a beneficial envi- ronment and makes a positive contribution to health.

Psychosocial epidemiology Field of epidemiology that examines the role of psychological, behavioral, and social factors in health.

Quasi-experimental study Type of research design in which the investiga- tor manipulates the study factor but does not assign subjects randomly to the exposed and nonexposed groups.

Random errors Errors that reflect fluctuations around a true value of a param- eter (such as a rate or a relative risk) because of sampling variability.

Randomization A process whereby chance determines the subjects’ likelihood of assignment to either an intervention group or a control group. Each subject has an equal probability of being assigned to either group.

Rate A ratio that consists of a numerator and denominator in which time forms part of the denominator. Example: The crude death rate refers to the num- ber of deaths in a given year divided by the size of the reference population (dur- ing the midpoint of the year) (expressed as rate per 100,000).

Rate difference Measure of the difference between two rates (for example, inci- dence rates) between exposed and nonexposed populations (see Risk difference).

Rate ratio A more precise term when relative risk is calculated with incidence rates or incidence density.

Ratio A fraction in which there is not necessarily any specified relationship between the numerator and denominator.

752 G l o s s a r y

51589_GLOS_Printer.indd 752 11/02/13 2:45 PM

Record linkage System of joining data from two or more sources.

Reemerging infection Recurrence of certain “old” diseases, possibly as a result of changes in the components of the epidemiologic triangle. Examples of such infections are tuberculosis and syphilis.

Reference population Group from which cases of a disease (or health-related phenomenon under study) have been taken; also refers to the group to which the results of a study may be generalized.

Registry Centralized database for collection of information about a disease.

Relative risk Ratio of the risk of disease or death among the exposed to the risk among the unexposed. The formula used is Relative risk 5 Incidence rate in the exposed/Incidence rate in the nonexposed.

Reliability (also, precision) Ability of a measuring instrument to give consis- tent results on repeated trials.

Reportable disease statistics Statistics derived from diseases that physicians and other healthcare providers must report to government agencies according to legal statute. Such diseases are called reportable diseases.

Representativeness (also, external validity) Refers to the degree to which a sample resembles the parent population; affects the generalizability of the find- ings of an epidemiologic study to the population.

Residual disorder An illness or disease for which the key contributing factors are known but specific methods of control have not been effectively implemented.

Resistance Ability of an agent to survive adverse environmental conditions.

Retrospective cohort study Type of cohort study that uses historical data to determine exposure level at some time in the past; subsequently, follow-up mea- surements of occurrence(s) of disease between baseline and the present are taken.

Risk difference (also, attributable risk) Difference between the incidence rate of disease in the exposed group (Ie) and the incidence rate of disease in the nonexposed group (Ine): risk difference 5 Ie 2 Ine.

Risk factor An exposure that is associated with a disease, morbidity, mortality, or adverse health outcome.

Sampling error As a result of sampling methods, the misrepresentation of the sample selected for a study in relation to the target population.

G l o s s a r y 753

51589_GLOS_Printer.indd 753 11/02/13 2:45 PM

Scottish effect Signifies excess mortality in Scotland after controlling for the effects of deprivation.

Screening Presumptive identification of unrecognized disease or defects by the application of tests, examinations, or other procedures that can be applied rapidly.

Screening survey Investigation of a particular group of persons in order to identify individuals who have unrecognized health conditions.

Secondary attack rate Measure of the spread of a disease within a household or similar circumscribed unit. It is calculated by the formula [number of new cases in a group – initial case(s)] / [number of susceptible persons in the group – initial case(s)] 3 100.

Secondary prevention Intervention designed to reduce the progress of a disease after the agent interacts with the host; occurs during the period of pathogenesis.

Secular trends Gradual changes in disease frequency over long time periods.

Segregation analysis When the outcome of an experiment (random mating and assortment of genes from parents to offspring) is known, models represent- ing possible modes of transmission (genetic and nongenetic) are fit to the data.

Selection bias Error that occurs when the relationship between exposure and disease is different for those who participate in a study versus those who would be theoretically eligible for the study but do not participate (see Bias).

Selective factor A circumstance that results in the choice of persons for a group because of their health status or other characteristic.

Selective screening (also, targeted screening) A type of presumptive identifi- cation of unrecognized disease or defects applied to subsets of the population at high risk for disease or certain conditions as the result of family history, age, or environmental exposures (see Screening).

Sensitivity Ability of a test to identify correctly all screened individuals who actually have the disease being screened for.

Significance level Chance of rejecting the null hypothesis when, in fact, it is true.

Social desirability effects Introduced when a respondent answers questions in a manner that corresponds to the prevailing socially acceptable norms instead of giving a true answer.

754 G l o s s a r y

51589_GLOS_Printer.indd 754 11/02/13 2:45 PM

Social determinants Causes (of diseases) that originate from society and its configuration.

Social epidemiology The specialization within epidemiology that examines the social distribution and social determinants of health outcomes.

Social incongruity Defined as a situation in which the individual is not in harmony with or is incompatible with other persons; can also denote lack of harmony between the individual and larger society.

Social network Structure of people’s social attachments; helps to explain dif- ferences in the individual’s ability to respond to stressors.

Social support Perceived emotional support that one receives from family members, friends, and others; mediates against stress.

Spatial clustering Concentration of cases of a disease in a particular geo- graphic area.

Specific rate Statistic that refers to a particular subgroup of the population defined in terms of race, age, or sex; also may refer to the entire population but is specific for some single cause of death or illness, as in a cause-specific rate.

Specificity Ability of a test to identify nondiseased individuals who actually do not have a disease.

Standard metropolitan statistical areas (SMSAs) Standard areas of the United States established by the U.S. Bureau of the Census to make regional comparisons in disease rates and also to make urban/rural comparisons (see also Metropolitan statistical areas (MSAs)).

Standardized mortality ratio (SMR) Number of observed deaths divided by the number of expected deaths during a time period. The analogous term for morbidity is the standardized morbidity ratio.

Status discrepancy Refers to the disharmony that arises from differences among the statuses of individuals.

Stratum A homogeneous population subgroup, such as that characterized by a narrow age range (e.g., a five-year age group).

Strengths versus limitations Pertains to the usefulness of data for various types of epidemiologic research.

G l o s s a r y 755

51589_GLOS_Printer.indd 755 11/02/13 2:45 PM

Stressful [life] events Occurrences that are likely to bring readjustment or changes in people’s usual activities.

Stress-process model Defines stress as a process that occurs over time, with stressful events chaining from one to another and the interconnectedness among these stressful events.

Surveillance (also, public health surveillance) Systematic collection, analysis, interpretation, dissemination, and consolidation of data pertaining to the occur- rence of a specific disease.

Synergism Situation in which the combined effect of several exposures is greater than the sum of the individual effects.

Team science Uses large consortia of scientists to pool data and biological samples across studies.

Temporal clustering Association between common exposure to an etiologic agent at the same time and the development of morbidity or mortality in a group or population.

Temporality Timing of information about cause and effect; whether the infor- mation about cause and effect was assembled at the same time point or whether information about the cause was garnered before or after the information about the effect.

Tertiary prevention Intervention that takes place during late pathogenesis and is designed to reduce the limitations of disability from disease.

Therapeutic trial A type of study designed to evaluate the effectiveness of a treatment in bringing about an improvement in the patient’s health. An example is a trial that evaluates new curative drugs or a new surgical procedure.

Thoroughness A sub-criterion related to the care that has been taken to iden- tify all cases of a given disease including subclinical cases.

Threshold Lowest dose (often of a toxic substance) at which a particular response may occur.

Toxigenicity Capacity of an agent to produce a toxin or poison.

Type A behavior pattern A behavioral syndrome that includes the traits of agressiveness, ambition, drive, competitiveness, and time urgency.

756 G l o s s a r y

51589_GLOS_Printer.indd 756 11/02/13 2:45 PM

Vaccination Procedure in which a vaccine (a preparation that contains a killed or weakened pathogen) is introduced into the body to invoke an immune response against a disease-causing microbe such as a virus or bacterium. Also called inoculation, immunization.

Validity (also, accuracy) Ability of a measuring instrument to give a true measure (how well the instrument measures what it purports to measure).

Virulence An agent’s capacity to induce overt disease in the host; sometimes used as a synonym for pathogenicity.

Vital statistics Mortality and birth statistics maintained by government agencies.

Work overload An inability of the individual to meet demands emanating from the environment.

Years of potential life lost (YPLL) A type of statistical measure that takes into account the effect of premature death caused by diseases; computed by subtract- ing the actual age of death of an individual from the average age of death in a population (arriving at the YPLL for the individual), and then summing the YPLL for each individual in the population (i.e., the United States) being studied for a specific cause of mortality.

G l o s s a r y 757

51589_GLOS_Printer.indd 757 11/02/13 2:45 PM

51589_GLOS_Printer.indd 758 11/02/13 2:45 PM

759

A AADR. See Age-adjusted death rate Absenteeism data, 249t, 271–272 Absolute effects, 410–414, 430 Acculturation

defined, 187 hypothesis, 188

Accuracy, in screening, 473, 479 ACE. See American College of Epidemiology Acquired immune deficiency syndrome.

See AIDS Active follow-up, for cohort studies, 346 Active immunity, 498 Active prevention, 100 Activities of daily living, elderly population

and, 167 Activity limitations, percentage of adults

with, by age group and type of limitation, 169

Acute myeloid leukemia, 635–636 Acute myocardial infarction and passive

smoking, case-control study on, 314–315

Adaptive randomization, 387 Adenine, 602 Adjusted rates, 144–151

defined, 144 methods for, 144

ADLs. See Activities of daily living Adults, causes of mortality for, 167 Adventist Health Study 1, 196 Adventist Health Study 2, 196 Adventist Mortality Study, 196 Affluenza, 666 African-American

BHP diagnosis in men, 83 African-Americans

age-adjusted female breast cancer death rate, 185

category in 2010 Census, 180

differential mortality study and, 184 female life expectancy rate, 185 HIV diagnoses and, 184 hypertension and, 186 life expectancy at birth, U.S.,

1970–2007, 185 male deaths, calculation of among

African-American and white boys aged 5 to 14 years, 111t

male life expectancy rates, 185, 185 mortality rates of, 200–201 obesity in women, 173 sickle cell anemia and, 69

African sleeping sickness, place variation for, 217

Age, 163–164, 166–170 rate adjustment and, 144 sensory impairments and, 299, 300 trends in mortality for specific age groups,

164, 166–170 trends in mortality from leading causes of

death, 164 Age-adjusted death rates, 147, 149

group comparison of crude death rates and, 145t

for selected leading causes of death, 61 United States, 1960–2007, 146

Age adjustment defined, 146 indirect, 149, 150, 150t

Age and sex distribution, health of the community and, 67, 68, 69

Age associations, four reasons for, 167 Agency for Toxic Substances and Disease

Registry, 656, 657 Agent, in chain of infection, 500 Agents of infectious disease, 493–495

characteristics of, 496–497 types of, 493–494

Age-specific death rates, calculation of, 140t

Index Note: Italicized page locators indicate figures/photos; tables are noted with t.

51589_IDXx_Printer.indd 759 11/02/13 10:20 PM

760 I n d e x

Age-specific disease rates, 163 Age-specific incidence rates, 168 Age-specific rates, 143t

calculating, 138–139 defined, 138 formula for, 140

Agricultural workers, zoonotic diseases and, 572

AIDS, 63, 206, 519, 536, 539 African Americans and, 184 descriptive data on epidemiology of, 159 estimated number of diagnoses and

deaths and estimated number of persons living with AIDS and diagnosed/undiagnosed HIV, U.S., 1981–2008, 520

estimated prevalence of diagnoses, at end of 2008, 520

geographic variation in rates of, 210 health disparities and, 73

AIDS prevention programs, community trials and, 397

Air bags in vehicles, 74 Air pollution, 549, 549, 575, 590

bronchitis and, confounding, 447 constituents of, 578 ecologic study of, 287–288 health effects associated with, 579–581 indoor, 578–579 method of agreement and, 160–161

Alameda County Study, 350t Alarm reaction stage, in Selye’s general

adaptation syndrome, 661 Alaska Natives

category in 2010 Census, 180 disease burden for, 186–187

Alcohol and lung cancer risk in IWHS, nested case-control analysis of, 453

Alcohol consumption, 677–678 family environment and, 612 lifestyle and, 46

Alcohol use disorders, 48 Alleles, 605, 619

identical by descent and, 625 identical by state and, 625 mode of inheritance and, 617

Allergenic agents, work-related infections and, 572

Allergens, 557, 558t, 567–568 Allostasis, 660 Allostatic load, 660 Alpha particles, 566 Altitude and CHD, confounding and, 447 Alzheimer’s disease, 167, 704

death rates for, 60 Mendelian forms of, 636 molecular and genetic epidemiology of,

636–637 Ambispective cohort study, 355 Amebiasis, 495, 519 American Academy of Pediatrics, 298

Section on Epidemiology, 708 American Cancer Society, 399

Cancer Prevention Study I, 350t Cancer Prevention Study II, 627 mammography guidelines, 463

American College of Epidemiology, 708 American Housing Survey, 70 American Indians

category in 2010 Census, 180 morbidity and mortality profiles of,

186–187 American Journal of Epidemiology, 710 American Journal of Public Health, 710 American Medical Association, 339, 680 American Public Health Association,

709, 714 Amino acids, coding, 602 AML. See Acute myeloid leukemia Amyloid-beta peptide, Alzheimer’s disease

and, 636 Analogy, 92, 424 Analytic epidemiology, descriptive

epidemiology vs., 158–159 Analytic studies, 284 Anesthetics, history behind, 375 Angell, Marcia, 84 Angiosarcoma, clustering and development

of, 222 Anopheles mosquito, 501 Anthrax, 316, 531 Anthrax attacks, 2001, 8–9, 531 Antibiotic-resistant organisms, 316 Antibiotics, 493 Anticancer drugs, phases in testing of,

385–386

51589_IDXx_Printer.indd 760 11/02/13 10:20 PM

I n d e x 761

Antigenicity, 496–497 Antipsychotic drug use, regulation of, in

nursing homes, 284 Anxiety, 685

annual rates of, involving days away from work by private industry sector, 686

coffee use and, cohort study data for, 433t APHA Public Health Career Mart, 714 APIC Career Center, 714 Apo E-e4 allele, Alzheimer’s disease

and, 637 APP, Alzheimer’s disease and, 636 Apparent case ratio, 505, 505t AR. See Attack rate Arboviral diseases, 534–535 Armed forces

morbidity in: data on active personnel and veterans, 272–273

statistics on morbidity in, 250t Arsenic

carcinogenicity of, 562 contamination of drinking water and, 586 medical applications with, 561

Arsenic poisoning, 561–562 Arthritis, 46 Arthropod-borne diseases, 494, 495,

534–536 arboviral diseases, 534–535 Lyme disease, 535–536

Arthropod vectors, 501 Artificial, active immunity, 499 Artificial, passive immunity, 499 Asbestos, 557, 558t, 560–561

smoking and occupational exposure to, 676

smoking in lung cancer and, synergism, 557

Asbestosis, 21, 550, 561 ASD. See Autism spectrum disorder Aseptic meningitis, 495 Asians

category in 2010 Census, 180 morbidity and mortality profiles of,

187–188 Aspergillosis, 495 Assault death rates, South Atlantic states,

U.S., 2003, 72

Association of Schools of Public Health, 711 Atherosclerotic disease, community trial

on, 394 Atomic Bomb Casualty Commission,

Hiroshima and Nagasaki, 683 Atomic bombs, 566 Attack rate, 109, 143t, 506–507, 539

defined, 506 for foodborne illness, 506–507, 507t formula for, 120, 506

Attributable fraction in the population, 417

Attributable fractions, 410 Attributable proportion (or attributable

fraction), 415 Attributable risk, 410 AUDs. See Alcohol use disorders Autism spectrum disorder, prevalence study

on, 297–298, 299 Autosomal dominant, 618, 620 Autosomal recessive, 618, 620 Autosomes, 617 Availability of data criterion, in

epidemiology, 239, 239–240 Avian influenza, 45

cases of, 44–45 web of causation for, 427

B Babesiosis, 495 Baby Boom generation, maturation of, 719 Bacillus anthracis, 9, 531 Bacteria, diseases caused by, 493–494 Bacteriology, 16 Bangladesh, village in 1970s, during

WHO smallpox eradication program, 658

Bar graphs, 18, 19 Basic reproductive rate (R0), 511 Behavioral epidemiologists, 606 Behavioral epidemiology, 652, 655–656 Behavioral factors in health

examples of, 652 mental health states and, 653

Behavioral medicine, 652 Behavioral Risk Factor Surveillance

System, 266 Behavioral sciences, 650

51589_IDXx_Printer.indd 761 11/02/13 10:20 PM

762 I n d e x

Benign prostatic hyperplasia, epidemiologic evaluation of surgical interventions for, 83

Benzene National Cancer Institute, Tianjin, China

study, 554–555 occupational exposure to, 574

Bereavement, 170 Beta particles, 566 BHP. See Benign prostatic hyperplasia Bias, 456

in analysis and publication, 454–455 defined, 476 in environmental and occupational

epidemiology, 553–554 in epidemiologic research, 442–447

confounding, 445, 446t, 447 information bias, 443–445 selection bias, 442–443

reducing, techniques for, 449–450 Bibliographic resources, 242 Binge drinking, teenage years through young

adulthood, 164 Biologic agents, work-related infections

and, 572 Biologic markers, 600 Biologic clock, 170 Biologic gradient, causality and

epidemiologic research and, 91 Biologic plausibility, 424 Biomarkers, 608–609, 635 Biomedical model, 650 Biomphalaria glabrata, schistosomiasis

and, 516 Biopsychosocial model, 650 Biostatistics, 15, 15, 16–17, 48 Bioterrorism, 719 Bipolar disorder, 637–638 Birth certificates, 254 Birth control pills and breast cancer, nested

case-control study on, 357 Birth defects

Chernobyl power plant accident and, 583 hazardous waste exposure and, 578 maternal surgery during first trimester

and, case-control study on, 315–316 Birth rates, 151

decline in, reasons for, 67

seasonal trends in, 218 by selected age of mother, U.S.,

1990–2009, 128 Birth records, medical data from, 248t Births, 65, 66, 66 Birth statistics, 254 Bisphenol A, 558 Black Death, 23, 25–26 Black lung disease, 573 Bladder cancer, 492, 550 Blastomycosis, 495 Blinding, clinical trials and, 381–383 Blood lead levels (BLLS), 564 Blood pressure screening, 465 BLS. See Bureau of Labor Statistics Body weight-lung cancer association,

confounding and, 447, 448 Bogalusa Heart Study, 440 Botulism, 514t Bovine spongiform encephalopathy,

international spread of, 207–208

BRCA1, 623, 624, 626 BRCA2, 623, 626 Breast cancer, 620

African American females and death rate from, 185

age-adjusted total U.S. mortality rates for, all ages, females for 1995–2004, 221, 221

birth control pills and, nested case-control study on, 357

early-onset, genetic epidemiology of, in CASH study, 621

early-onset, linkage analysis of, 623 genome-wide association studies of,

626–627 HRT and, 369 international comparison of, 205 invasive, five-year relative and period

survival, by race and sex in U.S., 2002–2008, 96

marital status and, 174 mortality and dietary fat intake, ecologic

correlation of, 290, 290 postmenopausal, calculating incidence

rate of, in Iowa Women’s Health Study, 119

51589_IDXx_Printer.indd 762 11/02/13 10:20 PM

I n d e x 763

premorbid psychological factors and, 689 radiation from atomic bombs and, 566 screening for, 465

information bias and, 444–445 by mammography, women aged 40

through 49 years and, 462–464 percentage of women up-to-date on,

U.S., 2000–2010, 466 Breast Cancer Detection Demonstration

Project, 463 Breeder hypothesis, 199 BRFSS. See Behavioral Risk Factor

Surveillance System Bronchitis and air pollution, confounding

and, 447 Bronchus, malignant neoplasms of, by age

group, U.S., 2003, 152t Brown lung disease, 550 Brucella abortus, 531 Brucellosis, 531 BSE. See Bovine spongiform

encephalopathy Bubonic plague, 25–26, 510 Buffering model of social support, 671 Bunyarviridae, 538t Bureau of Labor Statistics, 713

C California, Smokefree Bars (SFB) Law in,

policy evaluation of, 77–78 California Health Interview Survey,

266–267, 664 California Seventh-Day Adventists

studies, 678 California Tumor Registry, 260 Cambodian Americans

cigarette smoking among, cross-sectional prevalence survey of, 297–298, 299, 656

smoking rates among, 188 Camplylobacter enteritis, 515t Campylobacter jejuni, 515t Cancer, 46, 63

African Americans and, 184 age-adjusted death rates, females, U.S.,

1930–2008, 172 age-adjusted death rates, males, U.S.,

1930–2008, 172

causes of, extraordinary range in, 436–437

death rates for, 60 dietary practices and, 678–679 environmental exposures and, 548, 550 etiology, case-control studies and, 313 familial aggregation and, 611 health disparities and, 73 infectious agents and, 492 invasive, five-year relative and period

survival, by race and sex in U.S., 2002–2008, 96

molecular epidemiology applications in study of, 635–636

mortality, geographic variation in rates of, 209

mortality rate, age-specific, 139, 140 premorbid psychological factors and,

687, 689 rates of

increase in, worldwide, 205 urban areas and, 212

selected, SEER estimated 2010 U.S. mortality for, 568

smoking and, case-control study of, 313–314

socioeconomic groups in England and Wales and, 201

time sequence and, 91 trends in age-specific incidence rates

of, 141t Cancer and Steroid Hormone (CASH)

study, genetic epidemiology of early-onset breast cancer in, 621

Cancer Epidemiology, Biomarkers & Prevention, 711

Cancer screening, secondary prevention and, 100

Cancer sites, age-specific (crude) SEER incidence rates for, by race and sex, 1992–2008, 168

Cancer Surveillance Program, 452 Candidate genes, 630, 631 Candidiasis, 495 Cannon, Walter, 661, 662 Carbon monoxide, 578, 579, 580 CARDIA. See Coronary Artery Risk

Development in Young Adults

51589_IDXx_Printer.indd 763 11/02/13 10:20 PM

764 I n d e x

Cardiovascular disease health disparities and, 73 as leading cause of death worldwide, 204 socioeconomic status and, 200 women and, 171–173

Cardiovascular Health Study, 351t Career opportunities in epidemiology,

701–720 career roles for epidemiologists, 705–708 competencies required of epidemiologists,

711–712 epidemiology associations and journals,

708–711 resources for education and employment,

712–714 specializations, 703–705, 703t

Carryover effect, 388 Case-control design, 93 Case-control family studies

design of, 613–614, 616 sampling frame for, 613

Case-control studies, 269, 284, 303–317, 324, 325, 418

cohort studies vs., 358 comparable cases and controls in, guide

for selection, 307t comparative analysis of, 360t–361t defined, 303 distribution of exposures in, 311 in environmental and occupational

epidemiology, 555 limitations of, 316–317 matching and confounding in, 452–453 measure of association used in,

309–316 nested, 357–358 number of controls for, 306 observation point for, 303 research conducted with, examples

of, 312t selection of cases for, 304

illustration of, 304 selection of controls in, 305–306 selection of study groups for, 303 sources of cases for, 304–305 sources of controls in, 307–309

patients from same hospital as the cases, 308–309

population-based controls, 307–308 relatives or associates of cases, 309

summary of, 316–317 Case fatality rate, 143t, 496, 539

cause-specific mortality rate and, 510–511

Case fatality ratio, 510–511 Case finding, 468 Case reports, 162, 284 Case series, 162–163, 284 CASH study. See Cancer and Steroid

Hormone (CASH) study Casmalia Waste Disposal Facility,

California, 575 Cassel, John, 86 Categorical variables, controlling

confounding and, 454 Caucasian men, BHP diagnosis and, 83 Causal criteria, for classification of ill

persons, 486 Causal inference, statistical inference and, 93 Causality, 101

criminality and, 94 in epidemiologic research, 84–87 Hill’s criteria of, 424 modern concepts of, 87–93 suggesting, aspects of an association

tied to, 90t Causal relationships

models of, 425–430 multiple causality, 426–430

possible types of associations, 425 Causal research, 411 Causal statistical association, 425, 425 Causal Thinking in the Health Sciences

(Susser), 86 Cause-specific mortality rate, case fatality

rate and, 510–511 Cause-specific rate, 138, 139, 143t CBPR. See Community-based participatory

research CCOP. See Community Clinical Oncology

Program CDC WONDER online databases, 265 Cellular phones, radiofrequency fields

and, 566 Census data, 250t Census tracts, 212

51589_IDXx_Printer.indd 764 11/02/13 10:20 PM

I n d e x 765

Center for Epidemiologic Studies’ depression (CES-D) scale, 686–687

Centers for Disease Control and Prevention, 21, 40, 44, 73, 239, 300, 315, 468

on anthrax cases, 2001, 9, 9 CASH study, 621 employment web site, 714 Epidemic Intelligence Service, 707, 707 epidemiologists working at, 706–707 on H1N1 pandemic, 2, 4, 8 on mammograms, 444 notifiable disease surveillance

system, 254 on waterborne disease out breaks, 588

Centers for Medicare and Medicaid Services, 307

Centre d’Etude du Polymorphisme Humain, 629

CEPH. See Council on Education for Public Health

Cervical cancer, 492 case-control study of, 313 human papillomavirus and, 41, 385 international comparison of, 205 invasive, five-year relative and period

survival, by race and sex in U.S., 2002–2008, 96

screening, 465 percentage of women up-to-date on,

U.S., 2000–2010, 466 CFR. See Case fatality rate; Case fatality

ratio Chagas disease, 206 Chain of infection, 500 CHD. See Coronary heart disease Chemical agents

health effects of, 558–559 in household products, 559

Chemicals, drinking water contamination and, 586

Chemical spills, 549 Chernobyl nuclear power plant accident,

582, 583, 584 Chicken pox, vaccination for, 521 Childhood, trends in mortality during, 164 Childhood cancers and parental smoking,

case-control study on, 314 Childhood Cancer Survivor Study, 339

Childhood lead poisoning, wheel model and, 428–429

Children, passive smoking and effects on, 580, 581

Chili pepper consumption and gastric cancer, sample calculation of odds ratio, 310–311

Chimney sweeps, 90 CHIS. See California Health Interview

Survey Chlordane, 559 Chlorination, drinking water and, 587 Chloropleth map, 215 Cholera, 21, 23, 206, 515t

causes and characteristics of, 517 Haiti earthquake of 2010 and outbreak

of, 518, 518 isolation and, 502 London epidemic, 1854, 30, 48, 214,

220, 221 John Snow on, 31–32, 34

Soho epidemic, commemorative plaque, 1854, 36

Cholera Foundation, Dresden, Germany, 23, 24

Cholesterol levels, exercise and, 681–682 Chromated copper arsonate, portions of

wood darkened by, 562 Chromium, smoking and occupational

exposure to, 676 Chromosomes, transmission of, 603 Chronbach’s alpha coefficient, 472 Chronic diseases, 46–47

completing clinical picture of, epidemiology and, 57–58

gender, burden of morbidity and, 173 health outcomes and, 71 personality and, 689–690 sex differences in, 170–171 as way-of-life diseases, 673

Chronic obstructive pulmonary disease air pollution and, 580 rubber dust exposure and, 573

CHS. See Cardiovascular Health Study CI. See Confidence interval Cigarette consumption, annual adult per

capita and smoking-and-health events, U.S., 1900–2011, 327

51589_IDXx_Printer.indd 765 11/02/13 10:20 PM

766 I n d e x

Cigarette smoking. See also Smoking among Cambodian Americans in

Long Beach, California, cross-sectional prevalence survey of, 298–299, 299, 656

among U.S. adolescents, cross- sectional surveys on, 300–301, 301, 302

changes in prevalence of, among successive birth cohorts of U.S. men, 1900–1987, 328

Cirelli, Dorothy, 376 Cirrhosis of the liver, 212 Class, health outcomes and, 196–197.

See also Socioeconomic status Classification, defined, 485 Climate change, 549 Clinical data sources, 267–268, 268 Clinical disease, iceberg concept of infection

and, 504, 504 Clinical end points, 379 Clinical equipoise, 389 Clinical medicine, 15, 15, 16, 48 Clinical significance, statistical

significance vs., 422 Clinical trial evidence, efficacy vs.

effectiveness and, 392 Clinical trials, 283, 373, 374–392, 404

blinding (masking) and, 381–383 community trials vs., 403t–404t crossover designs and, 387–388, 388 defined, 377 ethical aspects of experimentation with

human subjects, 388–390 examples of, 380–381 glossary of terms for, 379t history behind, 374–376 NIH on, 368 outcomes of, 379–380 phases of, 383–386, 386t placebo control and, 402 randomization and, 386–387 reporting results of, 390 schematic diagram of, 378 strengths of, 390–391 weaknesses of, 391–392 why, what, when, and where of,

376–377

Clostridium botulinum, 514t Clostridium perfringens, food poisoning

and, 515t Clotting factor VIII, 603 Clustering, 217, 221–223

defined, 221 of disease in families, exposure-disease

relationships and, 611–612 spatial, 222–223 temporal, 222

Cluster randomized trials, 393 CNVs. See Copy number variations Coal dust, 557, 558t Coal workers’ pneumoconiosis, 573

years of potential life lost and, 573 Cobb, Sidney, 664 Coccidioides imitis, 534 Coccidioidomycosis, 495, 499, 533, 534

reported cases-U.S., 2009, 535 Codominance, for quantitative and

qualitative traits, 619 Codons, 602 Coffee consumption

anxiety and, cohort study data for, 433t lifestyle factors and, 679–680

Coherence, 91–92, 424 Cohort, defined, 325 Cohort analysis, 41, 325 Cohort designs

options on timing of data collection, 342 temporal differences in, 341–344

prospective cohort studies, 342 retrospective cohort studies, 343–344

Cohort effect defined, 325 examples of, 325–327, 330

Cohort formation options, sampling and, 335–341

Cohort life table, 330 Cohort studies, 93, 284, 285, 323–362, 335

advantages and limitations with, 358–359 anxiety and coffee use data, 433t availability of exposure data, 344 case-control studies vs., 316 comparative analysis of, 360t–361t data collection and management of,

345–346 defined, 325

51589_IDXx_Printer.indd 766 11/02/13 10:20 PM

I n d e x 767

examples of, 348–349, 353–357, 359t–352t

exposure-based, 338–340, 359 follow-up issues, 346–347 outcome measures in, 340–341, 341t population-based, 336–337, 359 prospective, 342 retrospective, 343–344 sampling and cohort formation options,

335–341 size and cost of cohort, 344–345 sufficiency of scientific justification, 347

Colds, 96, 218, 497 College graduates, as exposure-based

cohorts, 339 Collins, Francis, 608 Colon cancer

dietary practices and, 678–679 time sequence and, 91

Colonization, 504 Colonoscopic screening, efficacy of,

case-control study on, 314 Colorado Plateau Uranium Miners

Study, 350t Colorectal cancer

diet and risk reduction for, 640, 641 screening for, 465, 639–640

case-control study on, 314 percentage of men and women up-to-

date on, U.S. 2000–2010, 466 Common source epidemics, 219 Communicable diseases, 21, 39, 43, 48, 703

defined, 20 foreign-born persons in U.S. and, 193 international comparisons of, 206 sociocultural influences and, 683

Community-acquired infections, 533 Community-based participatory

research, 656 Community Clinical Oncology Program, 397 Community environmental health hazards,

575–588 air pollution, 578–581 drinking water, 586–588 Gulf War syndrome, 575, 577 hazardous waste sites, 575, 578 nuclear facilities, 581–586 sick building syndrome, 575, 576

Community infrastructure availability of health and social

services, 70 health-related outcome variables, 71–72 housing stock, quality of, 70 social stability, 70 variables related to, 68

Community interventions, 373, 399–404 evaluation of, 399–404

posttest, 399, 401 pretest/posttest, 401 pretest/posttest/control, 401 Solomon four-group assignment,

401–402 miscellaneous issues, 402–403 typical, 283

Community Intervention Trial for Smoking Cessation, 397

Community policing programs, 70 Community trials, 283, 373, 374,

392–399, 404 advantages and disadvantages of, 398–399 clinical trials vs., 403t–404t defined, 392 enrollment in, 392 schematic diagram of, 393

Competencies for epidemiologists, 711–712 Completeness of population criterion, in

epidemiology, 239, 240 Complete penetrance, 625 Complex traits, 625 Concurrent validity, 474 Confidence interval

for case-control study with three different sample sizes, 421t

defined, 312, 421 odds ratios and, 311

Confidentiality, of epidemiologic data, 244–245

Confounding, 441, 445, 456 controlling, methods for, 450–454

analysis strategies, 453–454, 456 prevention strategies, 451–453, 456

defined, 445, 553 in environmental and occupational

epidemiology, 553–554 Simpson’s paradox and, 445,

446, 446t, 447

51589_IDXx_Printer.indd 767 11/02/13 10:20 PM

768 I n d e x

Confounding variable, defined, 445 Congenital malformations, environmental

hazards and, 550 Connecticut Tumor Registry, 260 Consistency, 424 Consistency reliability, measuring, 472 Consistency upon repetition, causality and

epidemiologic research and, 90 CONSORT statement, 390 Construct validity, 474–475 Contagion (film), 492 Content validity, 472–473 Continuing education degree programs in

epidemiology, 712–713 Continuous common source epidemics, 220 Continuous variables, 337, 454 Coprimary cases, 508, 509 Copy number variations, 605 Coronary Artery Risk Development in

Young Adults, 351t Coronary bypass surgery, study of risks and

prognosis of survival, 96–97 Coronary heart disease, 46, 203

altitude and, confounding, 447 diet and, 678 exercise and, cohort studies on, 349,

355–357 Hispanics and, 189–190 in Japan vs. in United States, 188,

682–683 as leading cause of death in U.S., 205 method of difference and, 160 method of residues and, 161 multiple causation in, 654 psychosocial epidemiology and, 684 race and, 13 risk difference and, 411 screening programs for, 462 sedentary lifestyle and, 680–681 sex differences in mortality from, 171 type A behavior pattern and, 670

Coronary heart disease epidemiology, 704 Council of State and Territorial

Epidemiologists Employment Listings, 714

Council on Education for Public Health, 713

Count, 109, 151

defined, 108 examples of, 108–109

Counties, NCHS urban-rural classic classification scheme for, 212

County and City Data, 273 Cowpox, 28, 29 Coxiella burnetii, 532 CRC. See Colorectal cancer Creutzfeldt-Jakob disease, 206 Criminality, causality and, 94 Criterion-referenced validity, 474 “Crossing over,” of DNA strands, 621 Crossover designs, 404

clinical trials and, 387–388, 388 Cross-sectional ecologic studies, 289 Cross-sectional studies, 162, 163, 282,

294–303, 317, 324, 435 comparative analysis of, 360t–361t in environmental and occupational

epidemiology, 554–555 examples of, 296–303 individual level in, 294 principal weakness of, 302–303 sample designs, 295–296 subject selection in, 295

Cross-sectional surveys, 284 Crude birth rate, 127–128, 143t Crude death rate, 59, 113, 143t, 144, 146

group comparison of age-adjusted death rates and, 145t

United States, 1960–2007, 146 Crude mortality rate, 112, 151 Crude rates, 126–137, 144

birth rate, 127–128 defined, 126 examples of, 127 fertility rate, 129–130 fetal mortality, 130–131 infant mortality rate, 132, 133, 134 maternal mortality rate, 136 neonatal mortality rate, 134 perinatal mortality, 136 postneonatal mortality rate, 134–135

Cruickshank, Robert, 710 Cryptococcosis, 495 Cryptosporidiosis, 495 Cryptosporidium, 538t Cultural mobility, 664

51589_IDXx_Printer.indd 768 11/02/13 10:20 PM

I n d e x 769

Culture, defined, 682 Cumulative incidence, 121–122, 412 Cutaneous anthrax, 9 CWP. See Coal workers’ pneumoconiosis Cyclic fluctuations, 217, 218–219 Cytosine, 602

D Daily hassles, stress and, 669 DALYs. See Disability-adjusted life years Data, graphic presentation of, 18 Data and safety monitoring board, 390 Data interpretation issues, 435–456

bias in analysis and publication, 454–455 control of confounding, methods for,

450–454 reducing bias, techniques for, 449–450 sources of error in epidemiologic research,

440–447, 448 validity of study designs, 437–440

Data perturbation, 240 Data sharing, 245 Day care centers, giardiasis and, 492 DDT, health effects associated with, 559 Death certificate

description of, 247, 251 sample, 252–253

Death rates age-adjusted, for selected leading causes of

death, 61 decline in, reasons for, 67 direct adjustment of, 147, 148t for leading causes of death by age group,

U.S., 1997–2007, 166 variation in diagnosis, reporting, or case

fatality and, 61–62 Deaths, 65

leading causes of, and rates for those causes, 1900 and 2009, U.S., 103

population size and, 66, 66 Decennial Censuses of Population and

Housing, 273 Deepwater Horizon Oil Spill, 2010, 548 Defense mechanisms

disease-specific, 498–499 nonspecific, 498

Demographic transition, 67 Demographic variables, 17, 67, 68, 69

Demography, 15, 15 Dengue fever, 206, 495 Deoxyribonucleic acid. See DNA Department of Veterans Affairs Medical

Center, Minneapolis, 270 Dependent (outcome) variables, 654, 692

health-related, 68 physical and mental health, 684–691

effects of major diseases on personality, 689–690

habitual mental outlook and health status, 690–691

life and job dissatisfaction, 684–685 mental health and stressors,

685–687 personality and smoking, 690 premorbid psychological factors and

cancer, 687, 689 psychosocial aspects of employment

and health, 691 Depression, 685, 686

age-adjusted percentages of sadness, hopelessness, worthlessness, 175

marital status and, 174 postpartum, temporal clustering and, 222 prevalence of, among U.S. adults,

2004–2008, by year and by gender, 688

seasonal trends in, 218 surveys on, 686–687

DES and vaginal cancers, case-control study of, 313

Descriptive epidemiology, 223 analytic epidemiology vs., 158–159 objectives of, 159 of a selected health problem, 233–234 three approaches to, 162–163

Descriptive studies, 284 epidemiologic hypotheses and, 159–160

Determinants, 8–9, 12, 159 Developed countries, population age

distribution for, by age group and sex-worldwide, 1950, 1990, and 2030, 64, 65

Developing countries, population age distribution for, by age group and sex-worldwide, 1950, 1990, and 2030, 63, 64, 65

51589_IDXx_Printer.indd 769 11/02/13 10:20 PM

770 I n d e x

Developmentally disabled children, in the Bronx, improving coordination of health services to, 80–81

DHHS. See U.S. Department of Health and Human Services

Diabetes, 46, 203, 720 health disparities and, 73 prevalence among adults 20 years of age

and over, U.S., 19 type 2, 71

Diabetes mellitus, 69 American Indians/Alaska Natives

and, 186 mortality rates and, 201

Diagnosis, screening vs., 465 Diagnostic and Statistical Manual, third

edition, 687 Diagnostic and Statistical Manual of Mental

Disorders-IV-TR, 486 Diagnostic studies, 376 Dialogues, The, 348–349 Dichlorodiphenyltrichloroethane. See DDT Dictionary of Epidemiology, 439 Diesel exhaust, lung cancer and, 579 Diet, 46, 652

colon health and, 641 familial clustering of disease and, 612

Dietary practices, 678–680, 679 Diet-heart hypothesis, 678 Dight Institute of Genetics, University of

Minnesota, 343 Dioxins, 559 Diphtheria, 256, 521 Direct adjustment, of death rates, 147, 148t Direct transmission of diseases, 500 Dirty bombs, ionizing radiation and, 567 Disability-adjusted life years, 331, 659–660 Disability limitation, goals of, 100 Disappearing disorders, 62, 62–63 Disease causality, complexity of, 426 Disease etiology, epidemiologic applications

to, 83–101 Disease frequency

epidemiology and study of, over time, 59 international comparisons of, 203–208

Disease outbreak measures, 506–511 attack rate, 506–507 basic reproductive rate, 511

case fatality rate, 510–511 secondary attack rate, 507–509

Disease rates, urban/rural differences in, 211–213

Disease registries, 248t, 260–261 Diseases. See also Chronic diseases;

Communicable diseases; Infectious diseases; Screening for disease

of affluence, 203 clustering of, 221–223 cyclic fluctuations in, 218–219 enlargement of clinical picture for, 97 epidemic frequency of, 21–22 epidemiologically significant, 494 exposure and, internal validity, 438–439 geographic variation in rates of, 208–210 germ theory of, 23, 38 Koch’s postulates and, 38, 48 localized place comparisons and, 213–214 magnitude of life event and, 667 natural history of, in relation to time of

diagnosis, 484 prevention of, 97–101

primary prevention, 98–100 secondary prevention, 100 tertiary prevention, 100–101

reasons for place variation in, 215–217 searching for causes of, epidemiology

and, 58 single agent causal model vs.

multifactorial causality doctrine for, 86

Diseases of unknown origin, method of analogy and, 161–162

Disease-specific defense mechanisms, 498–499

Disinfection by-products, drinking water contamination and, 586, 588

Disorders, four trends in, 62, 62–63 Disparities in health care, American

Indians/Alaska Natives and, 187 Distribution, 13 Division of Microbiology and Infectious

Diseases, 390 DNA, 602

facts about, 603–604 genetic variation and, 604–605 helix, 600, 601

51589_IDXx_Printer.indd 770 11/02/13 10:20 PM

I n d e x 771

human, highly polymorphic, 633 physical arrangement of, 602–603 sequences, 607 typing, 47 variation of, throughout genome, 622

Dose-response, 424 Dose-response curve, 91, 555–556, 556 Double-blind design, 383 Dowland’s Illustrated Medical Dictionary, 18 Downward drift hypothesis, 199 Drinking water, 586–588

trihalomethanes in, 589 waterborne disease outbreaks associated

with, by year and etiology, U.S., 1971–2008, 587

Drug testing, clinical trials and, 376–377 DSMB. See Data and safety monitoring

board; Division of Microbiology and Infectious Diseases

Dust, 557, 558t Dynamic population, defined, 66

E Early detection

scientific factors in screening and, 470 through screening, 462

Early Development Stages of Psychopathology study, 655, 656

Eastern equine encephalitis, 534 Eastern European Jewish population,

Tay-Sachs disease and, 69, 116–217 Ebola virus, 12, 538t Ecologic comparison studies, defined, 289 Ecologic fallacy, 292

defined, 293 examples of, 292, 292–294, 293t

Ecologic studies, 284, 287–294 applications of, 290–292 breast cancer and dietary fat intake,

290, 290 childhood lead poisoning in

Massachusetts, 290–291 comparative analysis of, 360t–361t defined, 287 limitations of, 292–294 questions investigated by, examples

of, 288t sample selection for, 288

Seven Countries Study, 291 stroke mortality rates and mean systolic

blood pressure, 291 Ecologic study designs, in environmental

and occupational epidemiology, 554 Ecologic time series analysis, mortality from

homicides in São Paulo, Brazil, 291 Ecologic trend (time series) studies, 289, 484 Economic stresses, 660–661 Ectoparasites

place variation for, 217 Rickettsia and, 495

EDSP. See Early Development Stages of Psychopathology study

Education age-adjusted percent distributions of health

status among persons aged 18 years and over by, U.S., 2010, 202

in epidemiology, resources for, 712–713 socioeconomic status and, 199, 201

Effect measure, defined, 336, 410 EIS. See Epidemic Intelligence Service Elderly population, leading causes of death

in, 167 Electric fields, 564–566 Electric generating plant, fumes issuing

from, 549, 549 Electric radiation, 557, 558t Electric utility workers and suicide, nested

case-control study on, 357–358 Electromagnetic radiation, from

high-tension wires, 549 Electronic health records, defined, 269 Emerging infections, 536–539

defined, 536 Escherichia coli, 537 hantavirus pulmonary syndrome, 537 West Nile virus, 539

Emigration, 65, 66, 66 Emphysema

indoor air pollution, smoking and, 579 mortality data, ecologic fallacy and, 282,

292–293 Empirical validity, 474 Employee assistance program poster, 661 Employment, 47

in epidemiology, resources for, 713–714 psychosocial aspects of, 691

51589_IDXx_Printer.indd 771 11/02/13 10:20 PM

772 I n d e x

Endemic, defined, 21 Engel, George, 650 Entamoeba histolytica, transmission of, 519 Environment

disease causation and, 23–25 health and, 46 infectious diseases and, 499–500 types of agents in, 557–570, 558t

Environmental epidemiologic research, molecular epidemiology applied in, 634–635

Environmental epidemiology, 46 defined, 549, 588 study designs used in, 550–555

case-control studies, 555 confounding and bias in occupational

epidemiology and, 553–554 cross-sectional studies, 554–555 ecologic study designs, 554 methods for selection of research

population and collection of exposure data, 551–553

retrospective cohort studies, 551 toxicologic concepts related to, 555–557

dose-response curve, 555–556 latency, 557 synergism, 557 threshold, 556–557

Environmental exposures, cancer and, 548 Environmental hazards

health effects associated with, 550 notable human exposures to, 549

Environmental health, epidemiologists’ role in, 549

Environmental Protection Agency, 239 environmental data compiled by, 238 on environmental tobacco smoke, 580 on Hanford weapons production, 585 mercury advisories, 563 water quality standards and, 586, 587

Environmental reservoir, 499 Environmental risk factors, clustering of

disease in families and, 611–612 Environmental variables, 73 Environmental variables, health of the

community and, 68 Enzootic arboviruses, 535 Enzootic diseases, 531

EPIC database, 581 Epidemic frequency of disease, ascertaining,

methods for, 21–22 Epidemic Intelligence Service (CDC), 41,

44, 46, 707, 707 Epidemics

common source, 219 continuous common source, 220 defined, 18 point epidemics, 219

Epidemiological associations evaluating, 423–424

bias, 423 cause-and-effect relationship, 424 confounding variables, 423 observing by chance, 423 to whom does association apply?,

423–424 Epidemiologic approach, aspects in, 48 Epidemiologic data, 235–274

absenteeism data, 271–272 birth statistics, 254 Census data, 273 clinical data sources, 267–271 confidentiality, sharing of data, and record

linkage, 244–247 criteria for quality and utility of, 239,

239–240 death certificate, sample, 252–253 disease registries, 260–261 diverse sources of, 274 insurance data, 267 morbidity in armed forces, 272–273 morbidity surveys of the general

population, 262–267 mortality statistics, 247, 251, 254 notifiable infectious diseases, 257–259 online sources of, 241–244 overview of, 236–238 presentations of, 18, 19 reportable disease statistics, 254–256 school health programs, 272 screening surveys, 259–260 sources of, overview, 248t–250t

Epidemiologic designs, application of genes in, 631–638

Epidemiologic findings, popular interest in, 41

51589_IDXx_Printer.indd 772 11/02/13 10:20 PM

I n d e x 773

Epidemiologic hypotheses, descriptive studies and, 159–160

Epidemiologic measures count, 108–109 incidence rate, 118–120 overview of, 109 prevalence, 113–118 proportion, 110–111 rate, 111–113 ratio, 109–110

Epidemiologic methods, policy evaluation of Smokefree Bars (SFB) Law in California, 77–78

Epidemiologic rates, elements in, 112 Epidemiologic research

causality in, 84–87 sources of error and bias in, 440–447,

448 typology of, 281t

Epidemiologic studies application of, 48 conflicting nature of, 84 methods and procedures, 17, 18 proliferation of, reasons for, 58–59 sources of error and bias in, 448 special vocabulary, 18, 20–21

Epidemiologic surveillance, aims of, 467–468

Epidemiologic transition, 67 Epidemiologic triangle, 426, 427

disease causality and, 492–493 three-factor, 493

Epidemiologists career roles for, 705–708

CDC, 706–707 NIOSH, 707–708

competencies required of, 711–712 ethics guidelines for, 715–719 Internet addresses of interest to, 243–244 median remuneration for, 714t modifiers added to titles of, 605–606 public health policy-making and, 76 well-trained, trends reinforcing demand

for, 719 Epidemiology, 15–22, 48

aims and levels, 14–15 defined, 8 descriptive vs. analytic, 158–159

determinants, 8–9, 12 distribution, 13 domestic and international organizations

in, 709t health phenomena, 14 historical antecedents of, 23–41, 48 historical use of: study of past and

future trends in health and illness, 59–63, 65

impact of Human Genome Project on, 632

interdisciplinary foundations of, 15, 15–22, 48

journals in, 710–711 morbidity and mortality, 14 overview of study designs used in, 282,

282–287 population, 13–14 population dynamics and, 65–67 professional associations for, 708–710 professional ethics in, 714–719 recent applications of, 41–48 seven uses of, 56, 57–58 specializations within, 703–705, 703t triumphs in, 42t–43t

Epidemiology, 710 Epidemiology and Infection, 710 Epidemiology Monitor, The, 702, 702 Epi Info™, 215 Epimonitor.net Job Bank, 714 Epizootic diseases, 531 Eradication of disease, 43 Eras in Epidemiology (Susser and Stein), 86 ERIC™ database, 242 Errors

poor precision, 440 random, 440–441 sampling, 440–441 systematic, 441–445

Escherichia coli, 537, 633 Escherichia coli O104, outbreak in Germany,

2011, 7 Escherichia coli O157:H7, 5, 6, 6, 7–8,

536, 537 Escherichia coli O145 multistate outbreak

(2012), 7 Essentials of Toxicology (Casarett and

Doull), 555

51589_IDXx_Printer.indd 773 11/02/13 10:20 PM

774 I n d e x

Estradiol, 171 Estrogen, 171 Ethical factors in screening, 469, 470 Ethics guidelines for epidemiologists,

715–719 avoiding conflicts of interest and

partiality, 718 communicating ethical requirements to

colleagues, employers, sponsors and confronting unacceptable conduct, 718

definition/discussion of core values, 715 ensuring equitable distribution of risks

and benefits, 716 maintaining public trust, 717–718 minimizing risks and protecting

the welfare of research participants, 716

obligations to communities, 718–719 obtaining informed consent of

participants, 716–717 professional role of epidemiologists, 715 protecting confidentiality and

privacy, 716 providing benefits, 716 submitting proposed studies for ethical

review, 717 Ethnic composition, health of the

community and, 68, 69–70 Ethnic diversity, 176 Ethnicity, 176–177, 182–191

overall trends in mortality according to, 182–183

percent distribution of 10 leading causes of death by, U.S., 2007, 183

socioeconomic status, health and, 197 tuberculosis incidence by, U.S.,

1995–2009, 189 Etiologic fraction, 430

equation 1, 415–416 equation 2, 416 relative risk and, 415–417

Europe, heat-related deaths in, 570 Evaluation, stages of, 400 Executive monkey experiments, 663 Exercise, 46, 652

coronary heart disease and, cohort studies on, 349, 355–357

family environment and, 612 health and, 680–682, 681

Exhaustion stage, in Selye’s general adaptation syndrome, 661

Exons (or expressed sequences), 603 Experiment, causality and epidemiologic

research and, 92 Experimental approach in epidemiology,

observational approach vs., 281–282 Experimental designs, most appropriate use

for, 373 Experimental studies, 281t, 282,

282–283, 317 Experimental study designs, 367–404

clinical trials, 374–392 community trials, 392–404 intervention studies, 373–374

Exposure, internal validity and, 438 Exposure assessment, environmental

epidemiology and, 552 Exposure-based cohort studies, 338–340, 359

comparison groups for use in, 340 sources of, 339

Exposure data cohort studies and availability of, 344 environmental, collection of, 551–553 examples of, 289

Extensively drug-resistant tuberculosis, 525 actions to protect public health,

527t–528t case study of, 526

External environment, infectious diseases and, 499

External validity, 240, 402, 439–440 Extrinsic motivation, heart disease, job

dissatisfaction and, 685

F Facebook, 671 Falls, 170, 569 False negatives, 477, 480, 490t False positives, 477, 478, 490t Familial adenomatous polyposis, 639 Familial aggregation

causes of, 610–612 bad environment, 611–612 chance, 611

shared family environment and, 612–616

51589_IDXx_Printer.indd 774 11/02/13 10:20 PM

I n d e x 775

Family recall bias, 443, 610 FAP. See Familial adenomatous polyposis Farr, William, 23, 37–38, 550 Fasting blood glucose, sensitivity, specificity

and, 482, 482–483 Federal Trade Commission, 308 Female paradox, 170 Females

age-adjusted cancer death rates, U.S., 1930–2008, 172

mortality and, 170 Fertility rate, 127, 129–130, 151

general, 129 total, 130

Fetal alcohol syndrome, 678 Fetal death certificate, 254 Fetal death rate, 127, 130, 143t Fetal death ratio, 127, 130–131, 143t Fetal mortality, 130–131 Fetal mortality rates, by period of gestation,

U.S., 1990–2005, 131 FGFR2, 627 Filoviradae, 538t Firearm death rates South Atlantic states,

U.S., 2003, 72 Fish, mercury in, 563 Fixed population, defined, 65–66 Fixed randomization, 387 Fleming, Alexander, 40 Flies, 495 Fluoridation of water, 74 Fluorosis, 213 Flu pandemic, 23 Flu season, 22 FN. See False negative Folic acid, neural tube defects and, 380 Follow-up, active and passive, for cohort

studies, 346–347 Fomites, indirect disease transmission

and, 501 Food and Drug Administration, drug

approval and, 376–377 Foodborne illnesses, 21, 493, 494,

536, 703 attack rate for, 506–507, 507t college cafeteria outbreak, data from, 545 in the community, 513 incubation period and, 503

infectious agents causative of, partial list, 514t–515t

“Food deserts,” 71 Food production chain, how food gets

contaminated, 515 Formaldehyde-containing construction

materials, 559 Formative evaluation, 400 Fourfield 2  2 table, for classification of

screening test results, 477 FP. See False positive Frameshift mutations, 604 Framingham Heart Study, 41, 350t Freedman, Larry, 436 Freedom of Information Act, 245 French, John, 665 Frequency matching, 306, 452 Frost, Wade Hampton, 40, 41, 325 Fruits and vegetables, health and increased

consumption of, 680 Fukushima nuclear reactor meltdown, 2011,

548, 582, 583–584 Functional cloning, 607, 607 Functional mapping, 607, 607 Fungal diseases, 494, 495, 533–534 Fungal meningitis, outbreak in 2012, 495

G Galen, 641 Gamma rays, 566 Gardisil, 41, 385 GASP, 626 Gastric cancer and chili pepper

consumption, sample calculation of odds ratio, 310–311

Gender, 47 Gene mapping, 626

determining mode of inheritance, 619–620

LOD score linkage analysis, 620–622, 624

models of inheritance, 617–619 nonparametric linkage analysis, 625–626 segregation and linkage analysis, 616–626

General adaptation syndrome, 661, 663 General fertility rate, 129, 143t Generalizability, of data, 240 Generation time, 503–504

51589_IDXx_Printer.indd 775 11/02/13 10:20 PM

776 I n d e x

Genes, 605 candidate genes, 630, 631 defined, 602 in epidemiologic designs, application of,

631–638 linked, 620–621

Genetic code, 602 Genetic epidemiology, 47–48, 600, 703

of Alzheimer’s disease, 636–637 defined, 606 four questions related to, 606 molecular epidemiology vs.,

605–609, 626 of psychiatric disorders, 637–638

Genetic factors, epidemiologic evidence for, 609–610

Genetics and public health, 638–641 interventions for high-risk

populations, 640 screening for colorectal cancer, 639–640 tailored interventions, 641

Genetic terminology, 605 Genetic variation, 604–605 Genome-wide association studies, 600, 608,

626–628 of Alzheimer’s disease, 637 of breast cancer, 626–627 issues related to and contributions made

by, 631 Genomics, 48 Genotypes, 605, 618, 631 Genotyping, 616, 626 Geographic information systems, 214–215

defined, 214 finding out about, 215 map of infant mortality rates in

Idaho, 216 representative applications of, 215t

Germ theory of disease, 23, 38 Giardia lamblia, 209–210, 495 Giardiasis, 209, 492, 495 Gini index, 74

state-specific, of inequality in number of healthy days and average number of healthy days, U.S., 2007, 75

GIS. See Geographic information systems Glasgow effect, 657

Glaucoma, screening for, referral criteria, 483

Global Burden of Disease Study, 659 Global warming, 548, 549, 720

effects of, in northeastern U.S., by end of 21st century, 570

heat waves and, 569 Goiter, 213 Goldberger, Joseph, 40, 41 Google, 241 Graunt, John, 48, 422

on sex differences in death rates, 26 ten leading causes of mortality in time, 27

Greeks (ancient), epidemiology and, 23 Greenhouse gases, 569 Green tea consumption and lung cancer,

women who smoked vs. did not smoke and, 313

Grief, 651 Groundwater contamination, sources of,

586 Group A streptococci, 538t Group Health Incorporated, 467 Group Health of Puget Sound, 339 Guanine, 602 Gulf of Mexico, Deepwater Horizon Oil

Spill, 2010, 548 Gulf War syndrome (GWS), 575, 577 GWAS. See Genome-wide association

studies GWS. See Gulf War syndrome

H Haemophilus influenzae type B infections,

vaccination for, 521 Haiti earthquake of 2010, 236

cholera outbreak in wake of, 518, 518 Halo effect, 476 HANES. See Health and Nutrition

Examination Survey Hanford, Washington, nuclear weapons

production at, 584–585 Hansen’s disease (leprosy)

migration and, 193 place variation for, 217 reported cases, by year, U.S.,

1970–2009, 194

51589_IDXx_Printer.indd 776 11/02/13 10:20 PM

I n d e x 777

Hantaviruses, 538t Hantavirus pulmonary syndrome, 162–163,

536, 537 Haplotype blocks, 628–629 Haplotypes, 628–631, 630 HapMap Project

applications and implications of, 630–631 launch of, 629

Hardiness, 669–670 Harvard University School of Public Health,

Summer Session for Public Health Studies, 713

Hats and sunburn, hypothetical ecologic relationship between, 293–294, 293t

HAV. See Hepatitis A virus Hawthorne effect, 402 Hazardous chemicals, 16 Hazardous wastes, 549 Hazardous waste sites, 20, 575, 578 HBV. See Hepatitis B virus HCV. See Hepatitis C virus HDV. See Hepatitis D virus Health

environment and, 46 habitual mental outlook and, 69–691 model of psychosocial factors in, 653–654 personal behavior, lifestyle and, 672–674 psychological and social factors in, 47 psychosocial aspects of, 691 searching for causes of, epidemiology

and, 58 smoking and, 674, 676–677 social context of, 657–660 sociocultural influences on, 682–684

Health and Nutrition Examination Survey, 264

Healthcare services, community infrastructure and, 70

Health disparities, 73–74 Health Examination Survey, 263–265 Health fairs, 260 Health insurance coverage, minority

populations’ access to, 83 Health Insurance Plan of New York, 267 Health Insurance Portability and

Accountability Act, 245, 246–247

Health insurance statistics, 248t Health of the community, 67–77

age and sex distribution, 67, 69 demographic and social variables, 67, 69 descriptive variables for, 68 diagnosis of, epidemiology and, 57 environmental variables, 73 health disparities, 73–74 policy evaluation, 74–77 racial, ethnic, and religious composition,

69–70 socioeconomic status, 69 variables related to community

infrastructure, 70–72 Health phenomena, 14 Health promotion, 46–47 Health-related data, uses for, 236–238 Health-related outcome variables, 71–72

chronic and infectious diseases, 71 homicide rate, 72 infant mortality rate, 71 suicide rate, 71 teenage pregnancy rate/sexually

transmitted diseases, 72 Health services, sociocultural influences and

utilization of, 683 Health services research epidemiology, 704 Health surveillance, rationale for, 572 Healthy migrant effect, 195 Healthy People

description of, 675 topic areas and Leading Health Indicators,

675t–676t Healthy People 2000, 674 Healthy People 2010, 73, 674 Healthy People 2020, 73, 173, 466, 564, 674

Leading Health Indicators, 675t–676t overarching goals of, 674

Healthy worker effect, bias and, 553–554 Heart disease, 47

death rates for, 60 familial aggregation and, 611 gender and, 171–173 job dissatisfaction and, 684–685 passive smoking and, case-control study

on, 314–315 seasonal trends in, 218

51589_IDXx_Printer.indd 777 11/02/13 10:20 PM

778 I n d e x

Heart disease mortality, secular trends in, 223

Heart disease research project, person-years of observation for, 123t

Heat waves deaths attributed to, 569–570 global warming and, 569

Heavy metals, 549 Helicobacter pylori, 205 Helminth-borne disease, 495 Hematotoxicity, benzene and, 554 Hemolytic-uremic syndrome, 5, 537 Hemophilia, 603 Henle-Koch postulates, 84 Hepatic angiosarcoma, vinyl chloride

exposure and, 574 Hepatitis A virus, 495, 502–503, 525, 530t

inapparent/apparent case ratio and, 505t secondary attack rate and, 509 vaccination for, 521

Hepatitis B virus, 525, 526, 530t day care centers and, 492 inapparent/apparent case ratio and, 505t vaccination for, 521

Hepatitis C virus, 525, 529, 530t, 536, 538t Hepatitis D virus, 530t Hepatitis E virus, 530t Hepatitis virions, 525, 529 Herbicides, 559 Herd immunity, 503 Heredity, 605 Herpes simplex, 495 HES. See Health Examination Survey Heterozygotes, 619 HEV. See Hepatitis E virus H5N1 (avian influenza), 44–45, 45 HHANES. See Hispanic Health and

Nutrition Examination Survey High-density lipoproteins, exercise and,

681–682 Highly pathogenic Avian influenza, 44–45 Highly polymorphic markers, importance

of, 625 High-risk populations, genetics, public

health and, 640 Hill, Sir Austin Bradford, 89, 91

nine issues relevant to causality and epidemiologic research, 90–92

HIPAA. See Health Insurance Portability and Accountabilty Act

Hip fractures, seasonal variations in rates of, 412–413

Hippocrates, 25, 48 Hispanic Health and Nutrition Examination

Survey, 83, 189, 264, 265, 687 Hispanic mortality paradox, 190–191 Hispanic origin, race and, Census 2010,

177–181, 182t Hispanics

HIV diagnoses and, 184 morbidity and mortality profiles of,

189–191 obesity and, 191 percent distribution of 10 leading causes

of death, U.S., 2007, 192 total population growth in U.S.,

2000–2010, 181, 182t Historical Statistics of the United States,

Colonial Time to 1970, 273 History of health of populations,

epidemiology and study of, 57 Hi-tech trash disposal, 575 HIV, 206, 495, 519, 536

acuteness of epidemic, from worldwide perspective, 520

adults and children estimated to be living with, 2011, 521

African Americans and, 184 cause-specific rate due to, 138 diagnosis rates-U.S. and U.S. territories,

2009, 211 geographic variation in rates of, 210 health disparities and, 73 incidence and prevalence, U.S.,

1977–2006, 125 percentage of diagnosed cases, by race/

ethnicity, U.S., 2009, 184 HIV prevention programs, community trials

and, 397, 398 Hodgkin’s disease

age and, 169 spatial clustering and, 222

Homelessness, point vs. period prevalence and study of, 116–117

Homeostasis, 660 Homicide rates, health outcomes and, 72

51589_IDXx_Printer.indd 778 11/02/13 10:20 PM

I n d e x 779

Homologous chromosomes, shifting of genetic material across, 621

Homozygous carriers, 619 H1N1 influenza pandemic

basic reproductive rate and, 511 brief chronology of, 3–4 CDC estimates of cases in U.S., by age

group, 5 in 2009, 2–5, 7, 40, 43, 108

Honolulu Heart Program, 350t Honolulu Heart Study, 187 Hormesis, 556–557 Hormone replacement therapy

experimental studies and case of, 368–369 follow-up protocol and cohort studies

on, 347 Women’s Health Initiative and, 369–371

Hospice care patients, discharged, primary admission diagnoses of, 19

Hospital-acquired infections, 492, 533 Hospital controls, in case-control studies,

308–309 Hospitals

data from, 268–269 diseases treated in, 269–270 inpatient statistics from, 249t

Host, 497–499 in chain of infection, 500 disease-specific defense mechanisms,

498–499 nonspecific defense mechanisms, 498

Household Interview Survey, 263–264 Housing, quality of, in U.S., 70, 71 HPAI. See Highly pathogenic Avian

influenza HPS. See Hantavirus pulmonary syndrome HPV vaccine, phases in testing of, 385 “Human biologic clock” phenomenon, 170 Human genes, identifying, strategies for,

607, 607–608 Human genetics, principles of, 601–605 Human genome, mapping of, 600, 632 Human Genome Project, 608

completion of, 48 multiple impacts of, 632, 632–633

Human immunodeficiency virus. See HIV Human papillomavirus, 205

cervical cancer and, 41

inapparent/apparent case ratio and, 505t

Human Population Laboratory, California, 672

Human subjects, ethical aspects of experimentation with, 388–390

Hurricane Katrina, 66, 569 Hurricane Rita, 569 HUS. See Hemolytic-uremic syndrome Hydrocarbons, 578 Hypercholesterolemia, 419, 462 Hypertension, 203

African Americans and, 186 death rates for, 60 screening, 462

Hyperthermia, 571 Hypertrophic pyloric stenosis, 120 Hypotheses, declaring, 159–160

I IADLs. See Instrumental activities of daily

living Iatrogenic reactions, 389 IBD. See Identical by descent ICD-10, 486 Iceberg concept of infection, 504, 504 Idaho, GIS map of infant mortality rates

in, 216 Identical by descent, 625 Identical by state, 625 IEA. See International Epidemiological

Association Illness, 683. See also Diseases Illness behaviors, 683 Ill persons, classification of, criteria for, 486 Immigration, 65, 193, 719 Immune system, stress and alterations

to, 660 Immunity to disease, types of,

498–499 Immunizations

health disparities and, 73 migration and, 195

Impact evaluation, 400 Implicit question, stating hypotheses

and, 160 Inapparent case ratio, 505, 505t Inapparent infection, 502–503

51589_IDXx_Printer.indd 779 11/02/13 10:20 PM

780 I n d e x

Incidence, 138 analogy of prevalence and, 115 defined, 118 interrelationship between prevalence and,

124, 126 prevalance and, HIV, U.S., 1977–2006,

124, 125 Incidence data, applications of, 126 Incidence density, 122–124, 143t, 412

defined, 122 women at risk for postmenopausal breast

cancer in IWHS, 123 Incidence rate, 109, 118–120, 143t, 151

attack rate, 120 elements of, 118 number of new cases, 118 population at risk, 118–119 of postmenopausal breast cancer in

IWHs, calculating, 119, 122 specification of a time period, 120

Income age-adjusted percent distributions of

health status among persons aged 18 years and over by, U.S., 2010, 202

measurement of, 198–199 socioeconomic status and, 201

Income inequality, health disparities and, 74 Incomplete penetrance, 625 Incubation period, 503, 504 IND. See Investigational New Drug

Application Independent variables, 654, 660–661,

663–669, 692 general concepts of stress, 660–661, 663 person-environment fit model,

665–666 social incongruity theory, 664–665 stressful life events, 666–668 stress process model, 668–669

Index cases, 508, 509 Indirect exposure measurements, in

environmental epidemiology, 552–553

Indirect method, of age adjustment, 149, 150, 150t

Indirect transmission, 501 Individual matching, 306, 452 Individual risks, epidemiology and, 57

Indoor air pollution, 578–579 Inductive reasoning, Mill’s canons of,

160–161 Industrial chemicals, occupational exposure

to, 574 Infant mortality, 63, 73 Infant mortality rates, 74, 127, 132–134,

143t, 151 African American military personnel

and, 200 determining, 132 health outcomes and, 71 in Idaho, GIS map of, 216 international comparison of, 206

by selected countries, 2007, 134 low social class standing and, 200 by race, U.S., 1950–1995, 133 by race and ethnicity, 1995–2004, 133 by race and Hispanic origin of mother,

U.S., 2007, 133 United States, 1940–2007, 135

Infants, trends in mortality for, 164 Infectious disease epidemiology, 40, 42–44,

46, 491–545, 703 arthropod-borne diseases, 534–536

arboviral diseases, 534–535 Lyme disease, 535–536

emerging infections, 536–539 Escherichia coli, 537 examples of, 538t hantavirus pulmonary syndrome, 537 West Nile virus, 539

epidemiologic triangle, 492, 493 foodborne illness in the community,

513–514 fungal diseases, 533–534 means of transmission, 500–505

colonization and infestation, 504 direct transmission, 500 generation time, 503–504 herd immunity, 503 iceberg concept of infection, 504 inapparent/apparent case ratio, 505 inapparent infection, 502–503 incubation period, 503 indirect transmission, 501 portals of exit and entry,

501–502, 501t

51589_IDXx_Printer.indd 780 11/02/13 10:20 PM

I n d e x 781

person-to-person contact and spread of disease, 524–529

tuberculosis, 524–525 viral hepatitis, 525–526, 529

sexually transmitted diseases, 519–520 vaccine-preventable diseases, 520–524 water- and foodborne diseases, 516–519 zoonotic diseases, 531–533

Infectious disease outbreaks procedures used in investigation of,

511–512 appraise existing data, 511–512 case identification, 512 clinical observations, 512 define the problem, 511 draw conclusions and formulate

practical applications, 512 formulate a hypothesis, 512 identification of responsible agent, 512 tabulation and spot maps, 512 test the hypothesis, 512

Infectious disease research, case-control studies and, 316

Infectious diseases, 21, 38, 48 agents of, 493–495

characteristics of, 496–497 defined, 20 designated as notifiable at the national

level during 2009, 257–259 environment and, 499–500 geographic variation in rates of,

209–210 in United States, 210t

health outcomes and, 71 host, 497–499

disease-specific defense mechanisms, 498–499

nonspecific defense mechanisms, 498 international comparison of, 205 low social class standing and, 199–200 migration and, 193 molecular epidemiology and study of,

633–634 variation in severity of, 497

Infectivity, 496 Infertility, environmental hazards and, 550 Infestation, 504

Influenza, 96 as cause of death, 59 CDC surveillance system for, 255 geographic variation in rates of, 210 as leading cause of death, 492 mortality and, 96 1918 pandemic, 40 number of cases at a residential facility, by

illness onset date and severity-Ohio, 2011, 220

percentage of deaths attributable to, 22 seasonal trends in, 218 surveillance systems for, 256 vaccination for, 521

Influenza A pandemic, basic reproductive rate and, 511

Information bias, 433–445, 441 defined, 443 reducing, techniques for, 450

Informed consent clinical trials and, 390 experimentation with human subjects

and, 388–390 Inheritance

determining mode of, 619–620 Mendelian, 605, 617, 619 mode of, 617–619

Injury epidemiology, 704 Insecticides, 559 Insects, infectious diseases and, 495 Institutional review boards, 717 Institutional settings

infectious disease outbreaks in, 492 measles outbreaks in, 523

Instrumental activities of daily living, defined, 167

Insurance data, 267 Interjudge reliability, 472 Internal consistency reliability, 472 Internal validity, 437–439 International Classification of Diseases,

37, 251 International Epidemiological Association,

709–710 International HapMap Project, 626 International Journal of

Epidemiology, 711

51589_IDXx_Printer.indd 781 11/02/13 10:20 PM

782 I n d e x

International Statistical Classification of Diseases and Related Health Problems, 251, 486

International travel, infectious diseases and, 492

Internet, 238, 241 Internet addresses, for epidemiologists,

243–244 Interpersonal relationships, as buffers against

stress, 671–672 Intervening variables. See Moderating

(intervening) variables Intervention, defined, 392 Intervention studies, 373–374, 404 Interviewer/abstractor bias, 443, 450 Intestinal parasites, 495 Intrinsic motivation, heart disease, job

dissatisfaction and, 685 Introns (or intervening sequences), 603, 604 Invalidity, sources of, 476 Investigational New Drug Application, 377 Ionizing radiation, 549, 550, 557, 558t,

566–567, 590 Chernobyl power plant accident, 582 from nuclear power facilities, 575

Iowa Women’s Health Study, 113, 114, 337, 346, 351t, 354, 442, 453

calcium intake and decreased risk of colon cancer in, 641

calculating incidence rate of postmenopausal breast cancer in, 119, 121

calculating rate of ovarian cancer in, 119 information bias and, 444

Isolation, infectious disease and, 502 IWHS. See Iowa Women’s Health Study

J Japan

breast cancer and radiation from atomic bombs in, 566

Fukushima nuclear reactor meltdown, 2011, 548, 582, 583–584

Japanese, low mortality rates for, 187 Jenner, Edward, 23, 26–29, 28, 375 Job dissatisfaction, physical and mental

health and, 684–685 “Jogging female heart,” 171

Johns Hopkins Bloomberg School of Public Health, Graduate Summer Institute of Epidemiology and Biostatistics, 713

Johns Hopkins University, 41 John Snow Pub, 37 Journals in epidemiology, 710–711 Judgmental samples, 296

K Kaiser Medical Plan, 267 Kaiser Permanente, 339, 467 Kaplan-Meyer survival probability plots,

of malignant plural mesothelioma patients, 334, 334–335

Kassirer, Jerome, 84 Kidney disease, death rates for, 60 KiKK study, Germany, 584 King, Mary-Claire, 623 Klebsiela pneumoniae, 633 Koch, Robert, 38 Koch’s disease, 38 Koch’s postulates, 48, 84, 85–86 Kuder-Richardson reliability

coefficient, 472

L Labor statistics, 250t Langmuir, Alexander, 40 Langner’s 22-item Index of

Psychophysiologic Disorder, 686 Lassa virus, 538t Late fetal death rate, 143t Latency, 557 Latinas, obesity and, 173 Latinos

diabetes mellitus and, 69 morbidity and mortality profiles of,

189–191 Lead, 557, 558t

contamination of drinking water and, 586 health effects and exposure to, 563–564

Leading Health Indicators, Healthy People 2020, 675t–676t

Lead poisoning, 212 Lead time, 468 Lead time bias, 484 Legionella pneumophila, 97, 634

51589_IDXx_Printer.indd 782 11/02/13 10:20 PM

I n d e x 783

Legionella pneumophilia culture specimen, laboratory technician viewing with dissecting type of microscope, 712

Legionella spp., 587, 588 Legionnaires’ disease, 12, 16, 587

method of analogy and, 161 Philadelphia outbreak, 1976, 97

Leishmaniasis, 495 Length bias, 484–485 Leprosy (Hansen’s disease)

migration and, 193 place variation for, 217 reported cases, by year, U.S.,

1970–2009, 194 Leptospira, 532 Leptospirosis, 531, 532 Lesbian, gay, bisexual, or transgender,

eliminating health disparities among, 173

Leukemia, 550 benzene exposure and, 554 childhood, magnetic and electric fields

and, 565, 566 spatial clustering and, 222

Levels, defined, 14 LGBT. See Lesbian, gay, bisexual, or

transgender LHIs. See Leading Health Indicators Life change events, leading, 667 Life cycle phenomena, aging-associated

problems and, 170 Life dissatisfaction, physical and mental

health and, 684–685 Life expectancy

American Indians/Alaska Natives, 186 at birth, by race and sex, U.S.,

1970–2007, 185 defined, 330 in Eastern Europe and former Soviet

Union, 658–659 international comparison of, 206 in less-developed countries, 658 life tables and projections of, 330 low social class standing and, 200 ranking at birth, by sex in selected

countries and territories, 204, 204 Life insurance statistics, 249t

Lifestyle personal behavior, health and, 672–674 physical health outcomes and, 46–47

Life tables, 330–333, 359 cohort life table, 330 columns of, explanation for, 332–333 period life table, 330 for total population, U.S., 2007, 331t

Lind, James, 374, 375 Lindane, 559 Line graphs, 18, 19 Linkage analysis

basis for, 620 of early-onset breast cancer, 623 LOD score, 620–622, 624 nonparametric, 625–626 reclassification errors and, 622 ultimate goal of, 622

Linkage disequilibrium, 627 population genetics concept of, 624

Live births and rates, United States, 1920–2009, 129

Liver cancer, 205, 492 Liver transplantation, HCV infection

and, 529 Localized place comparisons, disease

outbreaks and, 213–214 Locus (loci), 605

linkage analysis between, 621–622 mode of inheritance and, 617

Logarithm of the odds (LOD) score linkage analysis, 620–622, 624

Logical validity, 474 London, 214

cholera epidemic in, 1854, 30, 48, 220, 221

deaths in neighborhood of Broad Street, 1849, 33

Golden Square district, fatal attacks and deaths, 34

cholera outbreaks in, 23 replica of Broad Street pump in, 36 visit to Broad Street pump and Sir John

Snow Public House in, 35 Long Beach Smoking Ordinance,

California, 78 Longevity, healthful habits, physical health

status and, 672–673

51589_IDXx_Printer.indd 783 11/02/13 10:20 PM

784 I n d e x

Longitudinal study, 335 Louisiana, southern, family study of lung

cancer risk in, 615–616 Love Canal, 20, 575, 578 Low birth weight, environmental hazards

and, 550 LSP1, 627 Lung cancer, 63

asbestos exposure and, 561 death rates in UK and U.S., cohort effect

and, 328, 329t, 330 diesel exhaust and, 579 Doll and Peto study on smoking and, 41 dust exposure and, 573 etiology of, pie model, 429, 429–430 family study of risk for, in southern

Louisiana, 615–616 green tea consumption and, among

women who smoked vs. did not smoke, 313

international comparison of, 205 invasive, five-year relative and period

survival, by race and sex in U.S., 2002–2008, 96

obesity and, confounding, 447 radon-associated, 567 smoking and, 580

1964 Surgeon General’s report, 88–89, 90

shift in distribution of age of onset of, 327

synergism between asbestos and smoking in, 557

Lung cancer mortality, sex differences in, 171

Lungs, malignant neoplasms of, by age group, U.S., 2003, 152t

Lyme disease, 495, 535–536 incidence per 1000,000 population of

reported confirmed cases, by county, U.S., 2009, 536

Lynch syndrome, 639

M Mad cow disease, international spread of,

207–208 Magnetic fields, 564–566 Magnetic radiation, 557, 558t

Magnitude, screening and, 469 Magnitude of effect, relative risk and, 422 Malaria, 495

Anopheles mosquito and, 501 international comparisons of, 205–206 migration and, 193 seasonal trends in, 218

Males age-adjusted cancer death rates, U.S.,

1930–2008, 172 mortality and, 170

Malignant neoplasms, of trachea, bronchus, and lung deaths, by age group, U.S., 2003, 152t

Malignant plural mesothelioma patients, Kaplan-Meyer survival probability plots of, 334, 334–335

Mammography, 465 NCI recommendations, 444 use of, in past two years among women

40 years of age and over, by age, U.S., 2000–2010, 444

women aged 40 through 49 years and, 462–464

“Mandatory Reporting of Infectious Diseases by Clinicians, and Mandatory Reporting of Occupational Diseases by Clinicians” (CDC), 255

Man-environment interactions, wheel model of, 428

Manhattan Project, 584 Manic depression, 637 Manifestational criteria, for classification of

ill persons, 486 Manipulation of the study factor, 368, 373 Mantel-Haenszel procedure, 453 MAP3K1, 627 MAPMAKER/SIBS, 626 Marburg, 538t Marginal totals, in 2  2 table, 286 Marital status, 47, 173–176

protective or selective factor and, 175 social support process and, 672

Marmot, Michael, 659 Married adults, health status of,

173–174 Masking, clinical trials and, 381–383

51589_IDXx_Printer.indd 784 11/02/13 10:20 PM

I n d e x 785

Massachusetts, childhood lead poisoning in, ecologic study of, 290–291

Mass diagnostic and screening surveys, 248t Mass health examinations, 467–468 Mass screening, 467 Master of Public Health (MPH) degree

competencies in epidemiology and, 711–712

Matching of subjects, control of confounding and, 451, 452

Maternal mortality rate, 127, 143t defined, 136 formula for, 137 leading causes of, 2007, 137

Mayo Clinic, 269–270, 343 Alzheimer’s Disease Patient Registry, 637 hip fracture study at, 412–413

MBG. See Molecular biology and genetics MDR TB. See Multidrug-resistant

tuberculosis Measles, 43, 63, 71, 497

arriving refugee and outbreak of, Los Angeles County, 2011, 523–524

incidence by year-U.S., 1974–2009, 522 incubation period for, 503 reported cases, by year-U.S.,

1950–1997, 522 resurgence of, 1989-early 1990s, in

U.S., 521 secondary attack rate for, calculation

of, 509 vaccination for, 521 variation in severity of, 497

Measles-mumps-rubella (MMR) vaccine, 523

Measurement bias, 476 Measures of effect, 409–433

absolute effects, 410–414 in case-control studies, 309 cohort studies and, interpretation and

examples, 347–349, 355–358 evaluating epidemiologic associations,

423–424 models of causal relationships, 425–430 relative effects, 414–419 statistical, 420–422

Mechanical energy, 557, 558t, 569

Medical Research Council Vitamin Study, 380

Meditation, 652 MEDLINE, 241 Meiosis, 603, 621 Men, marital status, health and, 174, 176 Mendel, Gregor, 605, 617 Mendelian inheritance, 605, 617, 619, 625

as evidenced by disease patterns in families, 607, 609, 620

visual representation of, 618 Mendel’s law of independent assortment of

traits, 620 Meningococcal disease

age and, 169 age group, 2000–2009, 164, 165 seasonal trends in, 218

Meningococcal infections, summer school outbreak, interrelationship between incidence and prevalence, 124, 125

Meningococcal meningitis, vaccination for, 521

Mental disorders classifying, 486 life stresses and, 667

Mental health factors related to, 47 stressors and, 685–687

Mental illness familial aggregation and, 611 social causation explanations of, 199

Mental outlook, habitual, health status and, 69–691

Mental retardation mild, socioeconomic status and, 201 sociocultural factors in, 683

Men who have sex with men, HIV infection among, 519

Mercury, 562–563 in certain species of ocean fish, 563 total lake acres under advisory for,

2010, 563 MESH headings, 241 Mesothelioma

asbestos exposure and, 561 family occupational exposure to asbestos

and, 612 latency period for, 557

51589_IDXx_Printer.indd 785 11/02/13 10:20 PM

786 I n d e x

Metallic compounds, 561–564 arsenic, 561–562 lead, 563–564 mercury, 562–563

Methicillin-resistant Staphylococcus aureus, 494

Methionine, 602 Metropolitan Atlanta Congenital Defects

Program, 315 Metropolitan statistical areas, 211 Mexican Americans, 189, 191 MI. See Myocardial infarction Microbial agents, 38, 493–494 Microbiology, 15, 15, 16, 48 MIDSPAN Study, Scotland, 580 Migration, nativity and, 192–193, 195 Military service data, 272–273 Mill, John Stuart, canons of inductive

reasoning, 160–161 Mill’s canons of inductive reasoning,

160–162 method of agreement, 160–161 method of analogy, 161 method of concomitant variation, 161 method of difference, 160 method of residues, 161

Minamata Bay, Japan, methyl mercury disaster, 562

Mineral and organic dusts, in work settings, 572–573

Minnesota Breast Cancer Family Study, 346 Minnesota Heart Health Program, 394 Minnesota Multiphasic Personality

Inventory, 689 Minorities, in rural areas, 213 Miscarriages, tap water consumption and,

findings on, 589–590 Mites, 495 MLH1, cancer family syndrome and, 639 MMR vaccine. See Measles-mumps-rubella

(MMR) vaccine Mode of inheritance, 617–619, 620 Mode of transmission, portal of entry, portal

of exit and, 501t Moderating (intervening) variables, 654,

669–684, 692 alcohol consumption, 677–678 defined, 669

dietary practices, 678–680 personal behavior, lifestyle, and health,

672–674 sedentary lifestyle, 680–682 smoking and health, 674, 676–677 social support, 671–672 sociocultural influences on health,

682–684 type A (coronary-prone) behavior pattern,

670–671 Mold, 557, 558t, 568–569

allergies to, 568 in wake of Hurricanes Katrina and

Rita, 569 Molecular biology and genetics, 601 Molecular epidemiology, 47, 550, 600, 703

applications of Alzheimer’s disease, 636–637 cancer studies, 635–636 infectious diseases, 633–634 occupational and environmental

epidemiologic research, 634–635 psychiatric disorders, 637–638

defined, 608 genetic epidemiology vs., 605–609, 626

Morbidity, 14 age-related changes in rates of, 170 in armed forces: data on active personnel

and veterans, 272–273 classification of, issues in, 485–486 environmental hazards and, 550 epidemiologic transition and, 67 excess, smoking and, 674 factors reducing reliability of observed

changes in, 61–62 male vs. female differences relative to,

170, 171 marital status and, 175 population dynamics and, 63 rates of, 112 screening and secondary prevention

of, 462 summary of unadjusted measures

of, 143t urban and rural residents for 2005–2007

and, 213 Morbidity and Mortality Weekly

Report, 519

51589_IDXx_Printer.indd 786 11/02/13 10:20 PM

I n d e x 787

Morbidity surveys, of general population, 248t, 262–267

Mormons mortality rates among, 196 place variation for CHD and, 216

Morris, Jerry, 56, 59 Mortality, 14, 412

age effects on, explanations for, 169–170 age-related changes in rates of, 170 classification of, issues in, 485–486 counts, use of, 26 death certificate and cause of, 251 environmental hazards and, 550 epidemiologic transition and, 67 excess, smoking and, 674 factors reducing reliability of observed

changes in, 61–62 heat-related, 569–570, 571 marital status and, 174, 175 population dynamics and, 63 race and ethnicity and overall trends in,

182–183 rates of, 112 screening and secondary prevention

of, 462 by selected age groups, males and females,

U.S., 2003, 153t sex differences in, 170–173 statistics, 247, 248t, 251 summary of unadjusted measures of, 143t ten leading causes of, 1900 and 2009, 60 total from selected causes, males and

females, U.S., 2003, 153t urban and rural residents for 2005–2007

and, 213 use of, as study end point, 551

Mortality difference, 412 Mortality figures, accurate, importance of,

236–237 Mosquitoes, 495, 501 Motorcycle fatalities, simple sex ratio

for, 109 Motorcycle helmets, 74 Motor vehicle death rates, South Atlantic

states, U.S., 2003, 72 Motor vehicle-traffic injuries, 569 MPM patients. See Malignant plural

mesothelioma patients

MRSA. See Methicillin-resistant Staphylococcus aureus

MSAs. See Metropolitan statistical areas MSH2, cancer family syndrome and, 639 MSH6, cancer family syndrome and, 639 MSN. See Men who have sex with men Multicenter AIDS Cohort Study, 351t Multicenter trials, 375–376 Multidrug-resistant tuberculosis, 526 Multifactorial etiology, 426 Multiphasic screening, 260, 467 Multiple causality, 426–430, 654

pie model, 429, 429–430 web of causation, 427, 427 wheel model, 427–429, 428

Multiple causality doctrine, 86 Multiple sclerosis, 162, 209, 704 Multivariate techniques, for controlling

confounding, 451, 454 Mumps

generation time and, 503 vaccination for, 521

Mutations, 604, 628 Mycosis, 495 Myocardial infarction, 13

N NAMCS. See National Ambulatory Medical

Care Survey National Academy of Sciences-

National Research Council, twin panel, 273

National Ambulatory Medical Care Survey, 81, 264

National Cancer Institute, 397 mammography guidelines, 444, 463 on prostate cancer, 481 Surveillance, Epidemiology, and End

Results Program, 114 National Cancer Institute, Tianjin, China,

benzene study, 554–555 National Center for Biotechnology

Information, 631 National Center for Chronic

Disease Prevention and Health Promotion, Behavioral Risk Surveillance program, 673–674

51589_IDXx_Printer.indd 787 11/02/13 10:21 PM

788 I n d e x

National Center for Health Statistics, 116, 146–147, 189, 211, 212, 245, 330, 346

National Change of Address Service, 123 National Child Development Study, 358 National Death Index, 124, 346 National Do Not Call (DNC) Registry, 308 National Fetal Mortality Survey, 264 National Health and Nutrition Examination

Survey, 116, 264, 299 on children’s blood lead levels, 564, 565 portable units for data collection, 266

National Health Interview Survey, 174, 201, 262, 263, 466

National Health Survey Act of 1956, 262 National Hospital Discharge Survey, 264 National Human Genome Research

Institute, 608, 631 National Institute for Occupational Safety

and Health, 12 National Institute of Environmental Health

Sciences, 635 National Institute of Mental Health,

Epidemiological Catchment Area Program, 687

National Institute of Occupational Safety and Health, epidemiologists working at, 706, 707–708

National Institutes of Health, 368, 376, 377, 390, 714

National Library of Medicine, 241 National Longitudinal Mortality Study, 200 National Mortality Followback Survey, 264 National Natality Surgery, 264 National Nursing Home Survey, 264 National Nutrition Surveillance Survey, 264 National Occupational Mortality

Surveillance Program, 200 National Survey-CSHCN, 298 National Survey on Drug Use and

Health, 687 Native Hawaiians, category in 2010

Census, 180 Nativity, migration and, 192–193, 195 Natural, active immunity, 498 Natural, passive immunity, 499 Natural disasters, PTSD and, 685–686

Natural experiments, use of, 29–31 Natural history of disease

levels of preventive measures in, 99 prepathogenesis and pathogenesis periods

of, 98 in relation to time of diagnosis, 484

Natural immunity, 498 Natural Selection Foods, 6 Nature of data criterion, in epidemiology,

239, 239 NCBI. See National Center for

Biotechnology Information NCHS, 251 NDI. See National Death Index Negative declaration, of stating

hypotheses, 160 Neighborhood context, stress process model

and, 668, 669 NEJM. See New England Journal of Medicine Nelmes, Sarah, 28, 29 Neonatal mortality rate, 127, 143t

formula for, 134 United States, 1940–2007, 135

Nested case-control studies, 357–358 Neural tube defects

folic acid and, 380 maternal surgery during first trimester

and, case-control study on, 315–316 Neuro-epidemiologists, 606 Neuroepidemiology, 704 Neurotic disorders, annual rates of,

involving days away from work by private industry sector, 686

New drugs, clinical trials and, double-blind design, 383

New England Journal of Medicine, 58, 84, 95 New epidemic disorders, 62, 63 New Haven, Connecticut, mental illness

survey of, 198, 199 New York City Department of Health, 10 New York State Cancer Registry, 260 New York University Women’s Health Study,

351t NHANES. See National Health and

Nutrition Examination Survey NHANES II, 264 NHIS. See National Health Interview Survey

51589_IDXx_Printer.indd 788 11/02/13 10:21 PM

I n d e x 789

Nickel, smoking and occupational exposure to, 676

Nile River, transmission of schistosomiasis along, 516

NIOSH. See National Institute for Occupational Safety and Health

Nitrogen oxides, 578 Noise, adverse effects of, 569 Nomenclature, defined, 485 Noncausal statistical association, 425, 425 Noncompliance, 403 Nonintervention influences, 399 Nonparametric linkage analysis, 625–626 Nonprobability samples, 295, 296 Nonresponse, selection bias and, 442–443 Nonspecific defense mechanisms, 498 Nonsteroidal anti-inflammatory drugs

peptic ulcer disease and etiologic fraction, equation 1

and, 416 population etiologic fraction, equation

3 and, 417–418 risk difference and, 414, 426

North Karelia Project, 394 Nosocomial infections, 44, 492, 703 Nosologist, death certificate and, 251 Notifiable diseases, 254

infectious, at national level during 2009, 257–259

NSAIDs. See Nonsteroidal anti-inflammatory drugs

NTDs. See Neural tube defects Nuclear facilities, 581–586

nuclear power plants, 582, 582–584 nuclear weapons plants, 584–585 nuclear weapons testing, 585–586

Nuclear power facilities, ionizing radiation from, 575

Nuclear power plants, 582, 582–584 Nuclear weapons plants, 584–585 Nuclear weapons testing, 585–585 Nucleic acids, 602 Null hypothesis, 420 Nurses’ Health Study, 350t, 353, 627 Nurses’ Health Study II, 353 Nursing homes, antipsychotic drug use in,

regulation of, 284

Nutrition, nonspecific defense mechanisms and, 498

Nutrition epidemiology, 704

O Oak Ridge, Tennessee, nuclear weapons

production at, 584 Obesity, 46, 47, 63, 71, 720

among children, by age, U.S., 19 familial aggregation and, 611 Hispanics and, 191 lung cancer and, confounding, 447 minority women and, 173 National Health and Examination Survey,

116, 117 prevalence of, among adults over 20 years,

by race/ethnicity and sex, U.S., 2009–2010, 117

prevalence of, among public school children, by school year and selected characteristics, NYC, 2006–2007 to 2010–2011, 193

teenage years through young adulthood, 166

Objective environment, person-environment fit model and, 665

Objective person, person-environment fit model and, 665

Observational approach in epidemiology, experimental approach vs., 281–282

Observational studies, 281t, 282, 284–285, 317, 373

Observational study designs, comparison of, 360t–361t

Occupational epidemiologic research, molecular epidemiology applied in, 634–635

Occupational epidemiology, 46 Occupational hazards

industrial chemicals, 574 mineral and organic dusts, 572–573 monitoring and surveillance of exposure

to, 571–572 pesticides, 574

Occupational mobility, 664–665 Occupational prestige, social class and, 197 Occupations, within epidemiology, 705–706

51589_IDXx_Printer.indd 789 11/02/13 10:21 PM

790 I n d e x

Odds ratios, 410, 416, 418 for case-control study with three different

sample sizes, 421, 421t, 422 defined, 310 interpreting, 311–312 sample calculation of, 310–311

Office of Management and Budget, 177 definition of Hispanic or Latino origin in

2010 Census, 178 definitions of race categories in 2010

Census, 179 Office visits, percent distribution of, by

primary expected source of payment according to patient’s age, U.S., 2003, 82

Older population chronic diseases and, 167 sensory impairments and, 299, 300

Olmsted Medical Group, Minnesota, 269, 270

OMB. See Office of Management and Budget

Omnibus Budget Reconciliation Act, 284 On Airs, Waters, and Places

(Hippocrates), 25 One-sample cohort study, sample selection

in, 336 Online sources of epidemiologic data, 241–244 Oocytes, 603 Operationalization, in psychosocial

epidemiology, 655 Operations research, 101

defined, 79 epidemiology and, 79–81 goals of, 80

Opportunistic infections, 533 Opportunistic screening, 468 Optimism, self-reported health of midlife

women and, 691 Oral Contraception Study of the

Royal College of General Practitioners, 350t

Oral epidemiology/dental epidemiology, 705 Organization for Economic Cooperation

and Development, 203 Organocarbamates, 559 Organochlorides, 559 Organochlorines, 559

Organophosphates, 559, 574 ORs. See Odds ratios Osteoporosis, 412, 412–413

among elderly persons, 412 as risk factor for fractures, 412

Outbreak, defined, 20 Outbreak investigations, case-control design

and, 303 Outcome, internal validity and, 438 Outcome evaluation, 400 Outcome variables. See Dependent

(outcome) variables Ovarian cancer, 119, 307, 629 Oversupply of gratifications, stress and, 666 Ozone, 578

P Pacific Islanders, 180, 187–188 Pancreas cancer, invasive, five-year relative

and period survival, by race and sex in U.S., 2002–2008, 96

Pandemic influenza, secondary attack rate for, calculation of, 509

Pandemics defined, 21 1918 influenza pandemic, 40

PAP, 619 Parasites, migration and, 193 Parasitic diseases

geographic variation in rates of, 209–210 in United States, 210t

Paré, Ambroise, 374 Parkinson’s disease, 704 Participation, enhancing, incentives for, 449 Passive follow-up, for cohort studies,

346–347 Passive immunity, 498 Passive prevention, 100 Passive smoking

children and, 580, 581 health effects related to, 580 research on, methodologic difficulties tied

to, 581 Pathogenecity, 496 Pathogenesis

defined, 98 of natural history of disease, 98 secondary prevention and, 100

51589_IDXx_Printer.indd 790 11/02/13 10:21 PM

I n d e x 791

Pawtucket Heart Health Program, 394, 396–397

Pearlin, Leonard I., 668 Pellagra, three Ds of, 40, 41 Pemberton, John, 709 Penetrance, 619, 625 Penicillin, discovery of, 40 Peptic ulcer disease and NSAIDs

etiologic fraction, equation 1 and, 416 population etiologic fraction, equation 3

and, 417–418 risk difference and, 414, 426

Perinatal mortality, two measures of, 136 Perinatal mortality rate, 127, 136, 143t, 151

formula for, 136 global estimates of, by geographical region

and subregion, 2000, 137 Perinatal mortality ratio, formula for, 136 Period life table, 330 Period prevalence, 109

counting cases and, 116 determining, 114 Iowa Women’s Health Study, 116 point prevalence vs., 116

Persisting disorders, 62, 63 Personal behavior, lifestyle, health and,

672–674 Personal characteristics, 163–202

age, 163–164, 166–170 findings, 199–201 marital status, 173–176 measurement, 197–199 nativity and migration, 192–193, 195 race and ethnicity, 176–177, 182–191 religion, 195–196 sex/gender, 170–171, 173 socioeconomic status, 196–197

Personality effects of major diseases on, 689–690 factors, epidemiologic studies of disease

and, 47 smoking and, 690 variables, 669

Person-environment fit model, 665–666, 692

Person-time, defined, 122 Person-to-person contact and disease

transmission, 494, 524–531

tuberculosis, 524–525, 526, 527t viral hepatitis, 525–526, 529, 529,

530t, 531 Person-years, of observation for hypothetical

heart disease research project, 123t Pertussis

incidence by year-U.S., 1980–2010, 542 vaccination for, 521

Pesticide drift, health effects associated with, 560

Pesticides, 549, 559–560 contamination of drinking water and, 586 health effects of, 559–560 occupational exposure to, 574

p53 gene, 626, 635 Pharmaco-epidemiologists, 605 Pharmacoepidemiology, 704 Pharmacogenetics, 641 Phase I trials, 383, 384, 385, 386t Phase II trials, 383, 384, 385, 386t Phase III trials, 383, 384, 385–386, 386t Phase IV trials, 386, 386t Phenotypes, 605, 617 Phenylketonuria, screening neonates

for, 465 Phipps, James, 28 Physical activity, 680–682, 681. See also

Exercise CHD and, cohort studies on, 349,

355–357 Physical disorders, psychiatric disorders

and, 684 Physical energy, 557, 558t, 569 Physical environment, infectious diseases

and, 499 Physical health, 47 Physical mapping, 607 Physicians’ practices, data from, 249t,

270–271 Pie charts, 18, 19 Pie model, 426, 429, 429–430 Pinworms, 495 PKU. See Phenylketonuria Placebo control, in clinical trials of

pharmacologic agents, 402 Place characteristics, 203–217

geographic information systems, 214–215

51589_IDXx_Printer.indd 791 11/02/13 10:21 PM

792 I n d e x

Place characteristics (Continued) international comparisons of disease

frequency, 203–208 localized place comparisons, 213–214 reasons for place variations in disease,

215–217 urban/rural differences in disease rates,

211–213 within-country geographic variation in

rates of disease, 208–210 cancer mortality, 209 infectious, vector-borne, and parasitic

diseases, 209–210 multiple sclerosis, 209

Place comparisons, types of, 203 Plague, 23, 502 Plasma carotenoids, feeding study on utility

of, 439–440 Plasmodium vivax, 501 Plausibility, causality, epidemiologic research

and, 91 PMR. See Proportional mortality ratio PMS2, cancer family syndrome and, 639 Pneumococcal disease, vaccination for, 521 Pneumoconiosis, 550

years of potential life lost and, 573 Pneumocystis carinii pneumonia, 519 Pneumonia

as leading cause of death, 59, 492 percentage of deaths attributable to, 22 seasonal trends in, 218

Point epidemics, 217, 219 POINTER, 619 Point prevalence, 109, 114, 115

period prevalence vs., 116 Poisonings, 569 Policy cycle, 76–77 Policy decision-making, factors related to, 77 Policy evaluation, epidemiology and, 74–77 Polio/poliomyelitis, 43, 62, 376

inapparent/apparent case ratio and, 505t paralytic, vaccination for, 521

Polio vaccinations, promoting public health awareness of, 384

Polychlorinated biphenyls, 559 Polymorphic, 622, 625 Population, 13–14

decreasing in size, 67

in equilibrium or a steady state, 67 increasing in size, 67

Population age distribution, for developing and developed countries, by age group and sex-worldwide, 1950, 1990, and 2030, 64

Population-based cohort studies, 336–337, 359

Population-based controls, in case-control studies, 307–308

Population dynamics, epidemiology and, 65–67

Population etiologic fraction, 430 equation 3, 417 equation 4, 418 impact of exposure on population,

418–419 Population growth, 719 Population medicine, 13 Population pyramid, 63, 67 Population risk difference, defined, 413 Population screening, 467 Population size, factors related to, 66, 66–67 Portal of entry

chain of infection and, 500 portal of exit, mode of transmission

and, 501t Portal of exit, mode of transmission, portal

of entry and, 501t Port Pirie Cohort Study, 351t Positional cloning, 607 Positive declaration, of stating hypotheses,

159–160 Positive family history, defined, 609–610 Positive reinforcements, 663 Postneonatal mortality rate, 127, 143t

defined, 134 formula for, 135 United States, 1940–2007, 135

Postpartum depression, temporal clustering and, 222

Posttest, 399, 400t, 401 Posttraumatic stress disorder, 685 Postvaccination reactions, temporal

clustering and, 222 Pott, Percival, 90 Precision, in screening tests, 472 Predictive valdidity, 474

51589_IDXx_Printer.indd 792 11/02/13 10:21 PM

I n d e x 793

Predictive value, 486 Predictive value ()

calculation of, 479 prostate cancer screening and importance

of, 480–481 screening and, 476, 478 of screening test, effects of disease

prevalence on, 480t Predictive value (-)

calculation of, 479 screening and, 476, 478 of screening test, effects of disease

prevalence on, 480t Premorbid psychological factors, cancer and,

687, 689 Prepathogenesis

defined, 98 of natural history of disease, 98

Pretest/posttest, 400t, 401 Pretest/posttest/control, 400t, 401 Prevalence, 109, 113–118, 138, 151

analogy of incidence and, 115 of disease, effects on screening test results,

479–480 expressing, examples, 113 incidence and, HIV, U.S., 1977–2006,

124, 125 interrelationship between incidence and,

124, 126 period prevalence, 114, 116 point prevalence, 114, 115

Prevalence data, uses for, 117 Prevalence difference, 412 Prevalence studies, examples of,

297–300 Prevarication (lying) bias, 443–444 Prevention of disease, 97–101

primary prevention, 98–100, 101, 464

secondary prevention, 100, 101 tertiary prevention, 100–101

Prevention studies, 376 Primary prevention, 98–100, 101, 464 Primordial prevention, 98–99 Principles of Epidemiology (CDC), 255 Prions, 207 Privacy Act of 1974, 245 Privacy Rule (HIPAA), 246–247

Probability samples defined, 295 examples of, 296

Probands in case-control family studies, 613,

614, 616 defined, 613 for family study of lung cancer risk in

southern Louisiana, 615–616 Process evaluation, 400 Professional ethics in epidemiology,

714–719 Prognosis, 95 Program evaluation, 81, 83, 101 Project RESPECT, 398 Prophylactic trials, 378 Proportion, 109, 110–111 Proportional mortality ratio, 140,

142, 143t calculating, formula for, 142 defined, 140

ProQuest Dialog, 241 Prospecitve studies, 335 Prospective cohort studies, 285, 342 Prospective studies, 282

case-control studies vs., 316 Prostate biopsy, 481 Prostate cancer

African American males and, 185 international comparison of, 205 invasive, five-year relative and period

survival, by race and sex in U.S., 2002–2008, 96

prevalence of, 480–481 screening for, positive predictive values

for, 480–481 treatment for, 481

Prostate-specific antigen, 481 Prostate-specific antigen test, 462, 481 Prostatitis, 481 Protective hypothesis, marital status

and, 175 Protein, 603 Protozoa, diseases caused by, 495 PSA test. See Prostate-specific antigen test PSEN1, Alzheimer’s disease and, 636 PSEN2, Alzheimer’s disease and, 636 Pseudomonas aeruginosa, 634

51589_IDXx_Printer.indd 793 11/02/13 10:21 PM

794 I n d e x

Psychiatric disorders molecular and genetic epidemiology of,

637–638 physical disorders and, 684

Psychiatric genetic epidemiology, 637 Psychological factors, health and, 653 Psychological stress, defined, 660 Psychological therapy interventions, rigorous

evaluations of, developing, 389–390 Psychology, 650 Psychosocial epidemiology, 652, 653

guide to, 653 multiple causation and, 654 operationalization in, 655 research designs used in, 655–656

Psychosocial factors, model of, in health, 653–654

Publication bias, 455 Public health

genetics and, 638–641 laws and ordinances related to,

examples, 76t Public health clinic data, 249t Public Health Service Act, 245 Public health surveillance, 254 PubMed, 241, 242 Puerperal fever, 35, 37 Puerperal psychoses, temporal clustering

and, 222 Puerto Ricans, coronary heart disease

and, 190 Punnett, Reginald C., 618 Punnett square, 618, 619

possible outcome of offspring when both parents have Bb genotypes, 618

Pure determinism, 85 P values, 420, 422

for case-control study with three different sample sizes, 421t

Pyrethroids, 559

Q Q fever, 495, 531, 532

acute and chronic, number of reported cases-U.S., 2010, 532

Quadrivalent vaccines, 385 Qualitative sources of information, 17 Qualitative traits, IBD and, 625

Quantification, 17, 18, 48 Quantitative traits, IBD and, 625 Quasi-experimental designs, four major

variations on, 399, 400t, 401–402 Quasi-experimental studies, 281t, 282,

283–284, 372 Quota samples, 296

R Rabies, 497, 510, 531, 533

case of soldier with, 533 inapparent/apparent case ratio and, 505t number of reported cases among wild and

domestic animals, by year, U.S. and Puerto Rico, 1975–2005, 541

variation in severity of, 497 Race, 47, 176–177, 182–191

African Americans, 184–186 American Indians/Alaska Natives,

186–187 Asians/Pacific Islanders, 187–188 children’s blood lead levels and, 564, 565 coronary heart disease and, 13 diagnosis of BHP and, 83 Hispanic origin and, 2010 Census,

177–181, 182t Hispanics/Latinos, 189–191 overall trends in mortality according to,

182–183 percent distribution of 10 leading causes

of death by, U.S., 2007, 183 socioeconomic status, health and, 197 tuberculosis incidence by, U.S.,

1995–2009, 189 Racial composition, health of the

community and, 68, 69–70 Racial diversity, 176 Radioactive iodine, nuclear weapons testing

and, 585–586 Radionuclides, drinking water

contamination and, 586 Radon, 566, 567 Railroad Retirement Board, 355 Random allocation, clinical trials and, 377 Random-digit dialing, 308 Random errors, 456

factors contributing to, 440–441 poor precision, 440

51589_IDXx_Printer.indd 794 11/02/13 10:21 PM

I n d e x 795

sampling error, 440–441 variability in measurement, 441

Randomization clinical trials and, 386–387, 404 of subjects, 368, 373

control of confounding and, 451 Randomized controlled trials, 373, 374,

390, 391–392, 484 Randomized trials, phases of, progress

through, 391 Random samples, 296 Rapid City, South Dakota, statistician in

field collecting data for asphalt milling in, 238

Rate, 109, 111–113, 151 calculation of, 112 expression of, 113 risk vs., 121

Rate difference, 410, 412 Rate ratio, 348 Ratio, 109, 151

defined, 109 demographic sex ratio, 109 sex ratio at birth, 110 simple sex ratio, 109

Rational validity, 474 RCTs. See Randomized controlled trials RDD. See Random-digit dialing Recall bias, 443, 450, 610 Recombinant chromosomes, defined, 621 Recombination events, 621, 624 Record linkage, 245–246, 247 Red spots on airline flight attendants,

11, 12, 20 Reference group, 413 Registry, defined, 260 Relative effects, 414–419, 430

etiologic fraction, 415–417 population etiologic fraction, 417–419

Relative risk, 410 calculating, 348 defined, 347, 414 magnitude of effect and, 422 sexual abuse and suicide attempt, 349

Reliability defined, 472 evaluation of screening tests and,

472–473

interrelationships between validity and, 475, 475

of set of measurements, expressing, 473 types of, 472–473

Reliability coefficient, 473 Religion, morbidity and mortality rates and,

195–196 Religious composition, health of the

community and, 68, 69–70 Renaissance, 23 Renal epidemiology, 705 Repeated measurement reliability, 472 Reportable disease statistics, 248t, 254–256

information cycle, 255 Reproductive epidemiology, 703 Research designs

manipulation of study factor (M) in, 281, 282

randomization of study subjects (R) in, 281, 282

Research population, environmental hazards and methods of selection, 551–553

Reservoir chain of infection and, 500 direct or indirect from: means of

transmission and, 500–505 infectious diseases and, 499

Residential care residents, selected characteristics of, U.S., 2010, 82

Residual disorders, 62, 63 Resistance, of agent, 496 Resistance stage, in Selye’s general adaptation

syndrome, 661 Restriction of admission criteria, control of

confounding and, 451–452 Retirement syndrome, 170 Retrospective cohort studies, 343–344, 551 Retrospective studies, 282 Retroviruses, 495 RFETS. See Rocky Flats Environmental

Technology Site Rheumatoid arthritis, in women, Cobb

study on, 664 Ribonucleic acid. See RNA Rickettsia, 495 Rickettsialpox, 495 Ringworm, 495 Risk, rate vs., 121

51589_IDXx_Printer.indd 795 11/02/13 10:21 PM

796 I n d e x

Risk difference, 410, 411–413, 412 defined, 411, 414 multivariate causality and, 426

Risk factors, requisite criteria for, 87 Risks to individuals, study of, 93, 95–97 Robison, Leslie, 222 Rochester Epidemiology Project, 269, 270 Rochester Methodist Hospital,

Minnesota, 270 Rocky Flats, Colorado, nuclear weapons

production at, 584, 585 Rocky Flats Environmental Technology

Site, 585 Rocky Mountain spotted fever, 218, 495 Rodenticides, 559 Romans, epidemiology and, 23 Rorschach test, 687 Ross, Julie, 222 Rotaviral enteritis, vaccination for, 521 Rotavirus infection, seasonal trends in, 218 Rotavirus tests, with positive results,

percentage of by surveillance week-participating laboratories, U.S., 2000–2009, 219

Roundworms, 495 RR. See Relative risk Rubber dust, COPD and, 573 Rubella, 92

inapparent/apparent case ratio and, 505t vaccination for, 521

Rural diseases, 212–213

S SAGE, 619 Saint Mary’s Hospital, Minnesota, 270 Salaries, for epidemiologists, 713, 714 Salmon bias effect, Mexican Americans

and, 191 Salmonellosis, 254, 493, 514t Salvarsan, 561 Samet, Jonathan, 245 SAMHSA. See Substance Abuse and Mental

Health Services Administration Sampling, cohort formation options and,

335–341 Sampling error, 440–441 Sampling frame, 296 Sampling unit, 296

San Antonio Heart Study, 190 San Joaquin Valley fever, 495, 499, 533, 534 San Onofre Nuclear Power Plant, California,

582, 582 São Paulo, Brazil, ecologic analysis of

mortality from homicides in, 291 SARS. See Severe acute respiratory syndrome SBS. See Sick building syndrome Schistosoma haematobium, 516 Schistosoma intercalulatum, 516 Schistosoma mansoni, 495, 516, 517 Schistosomes, 516 Schistosomiasis

international comparison of, 205 life cycle of, 516, 517

School health program data, 249t, 272 Science, 84 Scientific factors in screening, 468, 470 Scottish effect, 657 Scrapie, 207 Screening

case-control design and, 303 defined, 259, 464 diagnosis vs., 465 multiphasic, 260, 467

Screening for disease, 461–490 appropriate situations for, 468–470

ethical, 469, 470 scientific, 468, 470 social, 468, 469–470

characteristics for good screening tests, 471

classification of morbidity and mortality, issues in, 485–486

colorectal cancer, 639–640 effects of prevalence of disease on results,

479–480 evaluation of screening tests, 471–475

interrelationships between reliability and validity, 475

reliability, 472–473 validity, 473–475

mammography screening for women aged 40 through 49 years, 462–464

mass health examinations, 467–468 mass screening, 467 measures of validity in, 476, 478–479 multiphasic screening, 260, 467

51589_IDXx_Printer.indd 796 11/02/13 10:21 PM

I n d e x 797

overview, 464–466 relationship between sensitivity and

specificity, 482–483 selective screening, 467 sources of unreliability and invalidity, 476

Screening surveys, defined, 259–260 Scurvy, 374–375, 375 Search engines, 241 Seasonal trends, 218–219 Seat belt use, 74 Secondary attack rate, 507–509, 539

calculating, 508, 509 data for military cadets, 507–508, 508t defined, 507

Secondary prevention, 100 Secondhand smoke

health effects related to, 580, 581 heart disease and, case-control study on,

314–315 tobacco control policy and, 78

Secular time trends, 217, 220–221 Secular trends, 59, 223 Sedentary lifestyle, 680–682 SEER database, 268 SEER program

cancer data, 261 geographical areas in U.S. covered by, 262

Segregation analysis, 617, 619, 625 of early-onset breast cancer, 621

Selection bias, 441, 442–443, 485 environmental hazard research population

and, 552 preventing, guidelines for, 449

Selective factor, marital status and, 175 Selective screening, 467 Selye, Hans, 661, 662, 663 Semmelweis, Ignaz, 35 Sensitivity, 486

calculation of, 479 improving, 483 interrelationship between specificity and,

482, 482–483 of PSA test, 481 screening and, 476, 478

Sensory impairments, minimizing, cross-sectional prevalence data and, 299–300

Sentinel health event, 572

September 11, 2001 terrorist attacks, 8 Sequential design, 389 SER. See Society for Epidemiologic Research SES. See Socioeconomic status Seven Countries Study, 291 Seventh-Day Adventists

low rates of CHD and other chronic diseases among, 195–196

place variation for CHD and, 216 Severe acute respiratory syndrome,

17–18, 43 Sewage workers, biohazards and, 572 Sex chromosomes, 617 Sex/gender, 170–173

marital status, health and, 174, 176 Sex ratio, 110

at birth, 1940–2002, 110 demographic, 109 simple, 109

Sexual abuse and suicide attempt, relative risk and, 349

Sexually transmitted diseases, 63, 494, 519–521, 539

direct transmission of, 500 education and prevention of, 381 health outcomes and, 72

Shanghai Women’s Cohort Study, 345 Shigellosis, CDC surveillance system

for, 255 SIBLINK, 626 Sib-pair approaches, nonparametric linkage

analysis and, 625, 626 Sick building syndrome, 575, 576 Sickle-cell anemia, 69, 604 Sickle-cell gene, gene/environment

interaction and, 216 Sidestream exposure to smoking, 580 Sigmoidoscopy, 639 Significance level, 420 Significance tests, 420 Silent mutations, 604 Silica, 557, 558t Silicosis, 573 Simpson’s paradox, 445, 446, 446t, 447 Single agent causal model, of disease, 86 Single nucleotide polymorphisms, 626, 627,

628, 629, 631 Size of risk, 95

51589_IDXx_Printer.indd 797 11/02/13 10:21 PM

798 I n d e x

Skin cancer, prevention of, family environment and, 612

Smallpox, 23, 43, 62, 497, 499 vaccine, 26–29, 28, 375

Smokefree Bars (SFB) Law (California), epidemiologic methods used in policy evaluation of, 77–78

Smokeless tobacco use, cross-sectional study on, 296–297, 301

Smoking, 46. See also Cigarette smoking; Tobacco use

age of onset of lung cancer and, 327 cancer at various sites and, case-control

study of, 313–314 Doll and Peto study on lung cancer

and, 41 health and, 674, 676–677 indoor air pollution, emphysema

and, 579 lung cancer and, familial clustering of

disease, 612 lung cancer and, method of concomitant

variation, 161 parental, childhood cancers and,

case-control study on, 314 personality and, 690 Surgeon General’s reports on, 88–89, 90,

674, 677, 690 synergistic relationship between lung

cancer risk among asbestos workers and, 557

Smoking and Health, Report of the Advisory Committee to the Surgeon General of the Public Health Service, 87, 90

SMR. See Standardized mortality ratio Snail fever, 495 Snow, John, 18, 23, 36, 37, 214

London cholera epidemic and, 29, 31–32, 34, 35, 48

Snow on Cholera, 23 SNPs. See Single nucleotide polymorphisms Social and behavioral sciences, 15, 15,

16, 48 Social class

findings on health and, 199–201 measures of, 197–199

Social context of health, 657–660 international comparisons, 658–660 Whitehall Study, 659, 660

Social desirability effects, 476 Social determinants, 650 Social environment, infectious diseases

and, 499 Social epidemiology, 651–652, 655–656 Social factors

health and, 653 in screening

cost-effectiveness and, 469–470 public acceptance and, 468

Social incongruity, defined, 664 Social incongruity theory, 664–665 Social networks, 671, 672 Social Readjustment Rating Scale, 667 Social Security numbers, linkage to National

Death Index and, 347 Social Security statistics, 250t, 267 Social stability, community infrastructure

and, 70 Social status, stress process model and,

668, 669 Social status discrepancy, adverse physical

and mental health outcomes and, 664

Social support, 47 buffering model of, 671–672 defined, 671

Social variables, 67, 68, 69 Society for Epidemiologic Research,

710, 714 Sociocultural influences on health, 682–684 Socioeconomic status, 47

BHP diagnosis and, 83 health and: example of confounding, 553 health of the community and, 68, 69 health outcomes, mortality and, 196–197 mental illness survey of New Haven,

Connecticut, 198, 199 Sociology, 650 Socrates, 349 Soho cholera epidemic, commemorative

plaque, 1854, 36 SOLAR, 626 Solomon four-group, 400t, 401–402

51589_IDXx_Printer.indd 798 11/02/13 10:21 PM

I n d e x 799

Solvents, drinking water contamination, 586 Southern Community Cohort Study, 83 Spanish Flu, 40 Spatial clustering, 222–223 Special clinics, diseases treated in, 269–270 Specializations within epidemiology,

703–705, 703t Specificity, 424, 486

calculation of, 479 causality and epidemiologic research

and, 90 improving, 483 interrelationship between sensitivity and,

482, 482–483 of PSA test, 481 screening and, 476, 478

Specificity hypothesis, interpersonal relationships and, 672

Specific rates advantages/disadvantages of, 139 age-specific rates, 138–139 cause-specific rates, 138, 139 defined, 138

Speizer, Frank, 353 Sperm, 603 SPLAT, 626 St. Louis encephalitis, 534 Standard error of measurement, 473 Standardized mortality ratio, 149–151 Standard metropolitan statistical areas, 211 Stanford Five-City Project, on cardiovascular

disease risk factors, 394–395, 396 Staphylococcal food poisoning, 515t Staphylococcus aureus, 513, 515t, 633 Statistical Abstract of the United States, 273 Statistical inference, causal inference and, 93 Statistical measure of effect, 420–422

clinical vs. statistical significance, 422 confidence interval, 421–422 P value, 420 significance tests, 420

Statistical power, 402–403, 422 Statistical significance, clinical

significance vs., 422 Statistical validity, 474 Statistics

birth statistics, 254

epidemiologic data sources, 248t–250t health insurance statistics, 248t hospital inpatient statistics, 249t labor statistics, 250t life insurance statistics, 249t on morbidity in armed forces, 250t mortality statistics, 247, 251 reportable disease statistics, 248t, 254–256 Social Security statistics, 250t

Status discrepancy defined, 664 models, 664–665

STDs. See Sexually transmitted diseases Stein, Zena, 86 Sterilization, prevalence study on, 297 Stillbirths, 130, 136 Stomach cancer, international comparison

of, 205 Strains, examples of, 669 Stratification, control of confounding and,

451, 453–454 Stratified sampling, 296 Stratum, 452 Strength of association, 90, 424 Strengths vs. limitations criterion, in

epidemiology, 239, 240 Streptococcal infections, 493–494 Stress, 47, 692

annual rates of, involving days away from work by private industry sector, 686

aversive events and, 663 defined, 660 general concepts of, 660–661, 663 social support as buffer against, 671–672 workplace, 661, 661

Stressful life events, 692 characteristics most salient as stressors, 668 defined, 666 ten leading life change events, 667 timing and sequencing of, 669

Stress process model, 668–669 Stress research, areas encompassed by, 663 Stringfellow acid pits, California, 575 Stroke, 13, 203

death rates for, 60 mortality rates and mean systolic blood

pressure, ecologic study of, 291

51589_IDXx_Printer.indd 799 11/02/13 10:21 PM

800 I n d e x

Study designs, 279–317 availability of subjects for, 281 case-control studies, 303–317 cohort studies, 323–362

defined, 325 measures of effect, 347 practical considerations, 344–347 primary data and, 335 sampling and cohort formation

options, 335–341 summary of, 358–359 temporal differences in cohort designs,

341–344 cross-sectional studies, 294–303 data collection methods for, 280 differences among, 280–281 directionality of exposure in, 280 ecologic studies, 287–294 in environmental epidemiology,

550–555 experimental studies, 282–283 hierarchy of, 371–373 number of observations made in, 280 observational studies, 284–285 observational vs. experimental

approaches in epidemiology, 281–282

overview of, used in epidemiology, 282, 282–287

quasi-experimental studies, 283–284 temporality and, 324 timing of data collection for, 280 2  2 table, 285–287 unit of observation in, 281 validity for etiologic inference according

to, 372 validity of, 437–440

external validity, 439–440 internal validity, 437–439

Study of Osteoporotic Fractures, 351t, 354 Subclinical disease, iceberg concept of

infection and, 504, 504 Subjective environment,

person-environment fit model and, 665

Subjective person, person-environment fit model and, 665

Substance abuse, teenage years through young adulthood, 164, 166

Substance Abuse and Mental Health Services Administration, 687

Subthreshold phase, 556 Sudden infant death syndrome, 46 Suffocation, 569 Suicide among electric utility workers,

nested case-control study on, 357–358

Suicide attempts, sexual abuse and, relative risk for, 349

Suicide rates health outcomes and, 71 international comparison of, 206 marital status and, 174

Sulfur oxides, 578 Sunburn and wearing hats, ecologic fallacy

and, 293–294, 293t Surgical operations, epidemiologic

evaluations of, 83 Surveillance, 21 Surveillance, Epidemiology, and End

Results program. See SEER program

Survey of Pathways to Diagnosis and Services, 298

Survival curves, 359 applications for, 334–335 constructing, 333–334

Susser, Mervyn, 86 Swedish Cancer Registry, 565 Swine influenza, 3 Syndromes, identifying, epidemiology

and, 57 Synergism, 557 Systematic samples, 296

T Taco Bell restaurants, E. coli 0157:H7

outbreak, 6 Tailored interventions, genetics, public

health and, 641 Tamoxifen and Finasteride Prevention

Trials, 397 Tapeworms, 495

51589_IDXx_Printer.indd 800 11/02/13 10:21 PM

I n d e x 801

Tap water consumption, miscarriages and, findings on, 589–590

Tay-Sachs disease, 69, 216–217, 467 TB. See Tuberculosis Tea consumption, lifestyle factors and,

679, 680 Team science, 626 Tecumseh study, Michigan, as

population based cohort study, 336–337

Teenagers, trends in mortality for, 164, 166 Teen fatherhood, nested case-control study

on, 358 Teen pregnancy, 72, 164 Temporal clustering, 222 Temporality, 324, 424 Tertiary prevention, 100–101 Tetanus, vaccination for, 521 Thalidomide, 92 Thames River, London, drinking water

source, 30, 30 Therapeutic trials, 379 Third National Health and Nutrition

Examination Survey, 83 Thomas, Lewis, 84 Thoroughness, of data, 240 Three Mile Island nuclear power plant, 582 Threshold, 556–557 Throughput, 606 Thucydides, 23 “Thucydides plague,” 23–25 Thymine, 602 Thyroid cancer

Chernobyl power plant accident and, 583 nuclear weapons testing and, 585–586

Ticks, 495 Time, 158 Time characteristics, 217–223

clustering, 221–223 spatial, 222–223 temporal, 222

common source and point epidemics, 219–220

cyclic fluctuations/seasonal trends, 218–219

secular time trends, 220–221 Times Beach, Missouri, 575

Time sequence, causality, epidemiologic research and, 91

Time series study, 289 Tinea capitis, 495 TN. See True negative TNRC9, 627 Tobacco control policy, laws, local

ordinances and, 78 Tobacco use

cancer and, case-control studies of, 313–314

products used in U.S., cohort effect and, 326–328, 327

teenage years through young adulthood, 164

Total fertility rate, 130 Toxic agents, work-related infections

and, 572 Toxic chemicals, 46, 557, 558t

health effects and exposure to, 548 synergistic effects of smoking and

occupational exposure to, 676 Toxicology, 15, 15, 16, 550 Toxic shock syndrome, 13, 16, 20, 316 Toxic waste dumps, 575, 590 Toxigenicity, 496 TOXLINE, 241 TP. See True positive Trachea, malignant neoplasms of, by age

group, U.S., 2003, 152t Traumatic events, examples of, 685 Traumatic forces, 569 Treatment efficacy, case-control design

and, 303 Treponema pallidum, 205 Trichinella spiralis, 495 Trichinellosis, 495 Trichinosis, 495

associated with meat from Alaskan grizzly bear, 513–514

Trihalomethanes, 589 True negatives, 477, 480, 490t True positives, 477, 480, 490t Trypanosomiasis, 217, 495 Tryptophan, 602 TSS. See Toxic shock syndrome TTHMs. See Trihalomethanes

51589_IDXx_Printer.indd 801 11/02/13 10:21 PM

802 I n d e x

Tubal sterilization, prevalence study on, 297 Tuberculosis, 71, 212, 468, 493, 524–525,

539, 703 age and, 169 Asians and rate of, 188 expected and observed number of cases,

U.S., 1980–1992, 525 extensively drug-resistant, 525, 526,

527t–528t incidence by race/ethnicity, U.S.,

1995–2009, 189 isolation and, 502 mortality rates from Massachusetts, Frost’s

data on, 325–326, 326t number and rate of, among U.S.-born

and foreign-born persons, 1993–2011, 194

resurgence of, 524 sociocultural influences and, 683 variation in severity of, 497

Tucson Epidemiological Study of Airways Obstructive Diseases, 580

2  2 table, 285–287, 348 association between exposure and disease

status and, 285–286, 285t cohort study design, 286, 287 fourfield, for classification of screening

test results, 477 marginal totals in, 286

Type A (coronary-prone) behavior pattern defined, 670 interview measure of, 670 self-administered measure of, 670–671

Typhoid fever, 500 Typhoid Mary, 500 Typhus fever, 495

U UCLA Center for Health Policy Research,

266–267 Undersupply of gratifications, stress

and, 666 Unintentional injuries

for adults, 167 childhood to early adolescence, 164 descriptive epidemiology of, example, 569 personal behavior and, 673 seasonal trends in, 218

teenage years through young adulthood, 164

United Kingdom, lung cancer death rates in, 328, 329t, 330

United Kingdom Childhood Cancer Study, 314

United Nations, 251 United States

Gini index for, 74 high temperature and mortality in, 571 housing quality in, 71 infant, neonatal, and postneonatal

mortality rates in, 1940–2007, 135 infant mortality rate in, 73–74 life table for total population in,

2007, 331t lung cancer death rates in, 328,

329t, 330 population by HIspanic or Latino

origin and by race for, 2000 and 2010, 182t

ten leading causes of death, all races, both sexes, 2003, 139t

within-country comparisons in, 208–209

years of potential life lost before age 65, 2009, U.S., all races, both sexes, all deaths, 333

University of Michigan Graduate Summer Session in

Epidemiology, 712–713 Tecumseh Study, 336, 685

University of Minnesota hospitals, 270 University of Pittsburgh, Health Sciences

Library System at, 242 Unreliability, sources of, 476 Urban diseases, 212 Urbanization, residential water supplies

and, 591 Urban/rural differences in disease rates,

211–213 U.S. Bureau of the Census, 65, 70, 177,

197, 211, 273 U.S. Department of Health and Human

Services, 73, 376, 674 U.S. Department of Labor, 713 U.S. Preventive Services Task

Force, 481

51589_IDXx_Printer.indd 802 11/02/13 10:21 PM

I n d e x 803

U.S. Preventive Services Task Force (USPSTF), 465–466

Usual frequency, meaning of, 20

V Vaccination program, stages in development

of, 383 Vaccine-induced immunity, 499 Vaccine-preventable diseases, 494,

521–524 Vaccines/vaccinations

clinical trial phases and, 384–385 effectiveness of, case-control design

and, 303 HPV, 41 smallpox, 23, 26–29, 375

Vaginal cancers and DES, case-control study of, 313

Validity defined, 473 evaluation of screening tests and,

473–475 interrelationships between reliability and,

475, 475 of screening tests, measures of, 476,

478–479 of study designs

external validity, 439–440 internal validity, 437–439

types of, 473–475 Validity criterion, 474 Valley of the Drums, Kentucky, 575 Vampire bat rabies, 206 Variability in measurement, 441 Variant Creutzfeldt-Jakob disease,

international spread of, 207–208 Variolation, 27 Vasectomies, prevalence study on, 297, 301 vCJD. See Variant Creutzfeldt-Jakob disease Vector-borne diseases

geographic variation in rates of, 209–210 in United States, 210t

Vectors, indirect disease transmission and, 501

Vegetables and fruits, health and increased consumption of, 680

Vehicles, indirect disease transmission and, examples of, 501

Veterans data on, 272–273 as exposure-based cohorts, 339 PTSD and, 685

Veterans Administration Cooperative Study, 96

Veterinary epidemiology, 705 Vibration, adverse effects of, 569 Vibrio cholerae, 515t, 517, 538t Vinyl chloride, occupational exposure to, 574 Viral encephalitis, 495 Viral hepatitis, 525–526, 529, 529,

530t, 531 five types of, 530t incidence of, by year, U.S., 1979 to

2009, 531 Viral infections, variations in clinical

presentations of, with comparative rank in inapparent/apparent case ratios, 505t

Virology, 15, 15, 16 Virulence, 496 Viruses, diseases caused by, 494–495 Vital statistics, Farr’s use of, 23 Vocabulary, specialized, 48

W Wages, for epidemiologists, 713, 714 War veterans, PTSD among, 685 Washout period, 388 Water, fluoridation of, 74 Water-borne diseases, 494, 516–519,

586, 587 Water hardness, geographic variation in, 214 Water not intended for drinking, water-

borne disease outbreaks and, 588 Water quality

degradation of, 575 EPA and regulation of, 586

Water supplies cholera epidemic in London and, 30–32,

34, 35, 48, 214, 220, 221 degradation of, 586–588

Water use of unknown intent, water-borne disease outbreaks and, 588

Web of causation, 426 for avian influenza, 427 defined, 427

51589_IDXx_Printer.indd 803 11/02/13 10:21 PM

804 I n d e x

Weighted method, for direct rate adjustment, 148t

Weiss, Noel, 437 Western Collaborative Group Study, 670 Western equine encephalitis, 535 West Nile virus, 539 Wheel model, 426, 427–428

applications of, 428–429 of man-environment interactions, 428

Wheezing respiratory illness in children and passive smoking by parent, statistical test, 420

WHI. See Women’s Health Initiative Whitehall Study, 659, 660 White persons, category in 2010

Census, 180 WHO. See World Health Organization Widowed individuals, depression and, 174 Willard, Harold N., 709 Willett, Walter, 353 WNID. See Water not intended for drinking WNV. See West Nile virus Women

coronary heart disease and, 171–173 lung cancer mortality in, 223 marital status, health and, 174, 176 with rheumatoid arthritis, Cobb study

on, 664 Women’s health, cohort studies on, 353–354 Women’s Health Initiative, 353–354,

369–371 Women’s Health Initiative Observational

Study, 352t Work environment

industrial chemicals in, 574 mineral and organic dusts in, 572–573 smoke exposure in, health effects of, 580

Working health services, epidemiology and, 57

Work overload, 691 person-environment fit model and,

665–666 Workplace

stress and psychosocial aspects in, 691 stress in, 661, 661

Work-related infections, biologic agents and, 572

World Health Organization, 10, 45, 203, 239, 251, 254, 486, 714

alcohol consumption, 677 environmental hazards, 548 HIV estimates, 520 H1N1 pandemic, 3 perinatal mortality rate, 136 plague cases reported to, 26 sick building syndrome, 576 smallpox eradication program in

Bangladesh, 658, 658 World War I, distribution of free cigarettes

to troops during, 326 World War II, penicillin developed

during, 40 World Wide Web, 241, 242 WUI. See Water use of unknown intent

X X chromosomes, 603, 619 XDR TB. See Extensively drug-resistant

tuberculosis X-rays, 566, 567

Y Yaws

international comparison of, 205 place variation for, 217

Y chromosomes, 603, 619 Years of potential life lost, 330–331, 333

before age 65 years and mean YPLL for decedents with coal workers pneumoconiosis as cause of death, 573

overconsumption of alcohol and, 677 Yersinia pestis, 25, 26 YPLL. See Years of potential life lost

Z Zoonoses, 499, 531 Zoonotic agents, work-related infections

and, 572 Zoonotic diseases, 206, 494, 531–533

51589_IDXx_Printer.indd 804 11/02/13 10:21 PM

Title Page
Copyright Page
Contents
New to This Edition
Introduction
Preface
Acknowledgments
About the Authors
Chapter 1: History and Scope of Epidemiology

Introduction
Epidemiology Defined
Foundations of Epidemiology
Historical Antecedents of Epidemiology
Recent Applications of Epidemiology
Conclusion
Study Questions and Exercises
References

Chapter 2: Practical Applications of Epidemiology

Introduction
Applications for the Assessment of the Health Status of Populations and Delivery of Health Services
Applications Relevant to Disease Etiology
Conclusion
Study Questions and Exercises
References

Chapter 3: Measures of Morbidity and Mortality Used in Epidemiology

Introduction
Definitions of Count, Ratio, Proportion, and Rate
Risk Versus Rate; Cumulative Incidence
Interrelationship Between Prevalence and Incidence
Applications of Incidence Data
Crude Rates
Specific Rates and Proportional Mortality Ratio
Adjusted Rates
Conclusion
Study Questions and Exercises
References

Chapter 4: Descriptive Epidemiology: Person, Place, Time

Introduction
Characteristics of Persons
Characteristics of Place
Characteristics of Time
Conclusion
Study Questions and Exercises
References
Appendix 4—Project: Descriptive Epidemiology of a Selected Health Problem

Chapter 5: Sources of Data for Use in Epidemiology

Introduction
Criteria for the Quality and Utility of Epidemiologic Data
Online Sources of Epidemiologic Data
Confidentiality, Sharing of Data, and Record Linkage
Statistics Derived from the Vital Registration System
Reportable Disease Statistics
Screening Surveys
Disease Registries
Morbidity Surveys of the General Population
Insurance Data
Clinical Data Sources
Absenteeism Data
School Health Programs
Morbidity in the Armed Forces: Data on Active Personnel and Veterans
Other Sources: Census Data
Conclusion
Study Questions and Exercises
References

Chapter 6: Study Designs: Ecologic, Cross-Sectional, Case-Control

Introduction
Observational Versus Experimental Approaches in Epidemiology
Overview of Study Designs Used in Epidemiology
Ecologic Studies
Cross-Sectional Studies
Case-Control Studies
Conclusion
Study Questions and Exercises
References

Chapter 7: Study Designs: Cohort Studies

Introduction
Cohort Studies Defined
Sampling and Cohort Formation Options
Temporal Differences in Cohort Designs
Practical Considerations
Measures of Effect: Their Interpretation and Examples
Summary of Cohort Studies
Conclusion
Study Questions and Exercises
References

Chapter 8: Experimental Study Designs

Introduction
Hierarchy of Study Designs
Intervention Studies
Clinical Trials
Community Trials
Conclusion
Study Questions and Exercises
References

Chapter 9: Measures of Effect

Introduction
Absolute Effects
Relative Effects
Statistical Measures of Effect
Evaluating Epidemiologic Associations
Models of Causal Relationships
Conclusion
Study Questions and Exercises
References
Appendix 9—Cohort Study Data for Coffee Use and Anxiety

Chapter 10: Data Interpretation Issues

Introduction
Validity of Study Designs
Sources of Error in Epidemiologic Research
Techniques to Reduce Bias
Methods to Control Confounding
Bias in Analysis and Publication
Conclusion
Study Questions and Exercises
References

Chapter 11: Screening for Disease in the Community

Introduction
Screening for Disease
Appropriate Situations for Screening Tests and Programs
Characteristics of a Good Screening Test
Evaluation of Screening Tests
Sources of Unreliability and Invalidity
Measures of the Validity of Screening Tests
Effects of Prevalence of Disease on Screening Test Results
Relationship Between Sensitivity and Specificity
Evaluation of Screening Programs
Issues in the Classification of Morbidity and Mortality
Conclusion
Study Questions and Exercises
References
Appendix 11—Data for Problem 6

Chapter 12: Epidemiology of Infectious Diseases

Introduction
Agents of Infectious Disease
Characteristics of Infectious Disease Agents
Host
The Environment
Means of Transmission: Directly or Indirectly from Reservoir
Measures of Disease Outbreaks
Procedures Used in the Investigation of Infectious Disease Outbreaks
Epidemiologically Significant Infectious Diseases in the Community
Conclusion
Study Questions and Exercises
References
Appendix 12—Data from a Foodborne Illness Outbreak in a College Cafeteria

Chapter 13: Epidemiologic Aspects of Work in the Environment

Introduction
Health Effects Associated with Environmental Hazards
Study Designs Used in Environmental Epidemiology
Toxicologic Concepts Related to Environmental Epidemiology
Types of Agents
Environmental Hazards Found in the Work Setting
Noteworthy Community Environmental Health Hazards
Conclusion
Study Questions and Exercises
References

Chapter 14: Molecular and Genetic Epidemiology

Introduction
Definitions and Distinctions: Molecular Versus Genetic Epidemiology
Epidemiologic Evidence for Genetic Factors
Causes of Familial Aggregation
Shared Family Environment and Familial Aggregation
Gene Mapping: Segregation and Linkage Analysis
Genome-Wide Association Studies (GWAS)
Linkage Disequilibrium Revisited: Haplotypes
Application of Genes in Epidemiologic Designs
Genetics and Public Health
Conclusion
Study Questions and Exercises
References

Chapter 15: Social, Behavioral, and Psychosocial Epidemiology

Introduction
Research Designs Used in Psychosocial, Behavioral, and Social Epidemiology
The Social Context of Health
Independent Variables
Moderating Variables
Dependent (Outcome) Variables: Physical and Mental Health
Conclusion
Study Questions and Exercises
References

Chapter 16: Epidemiology as a Profession

Introduction
Specializations within Epidemiology
Career Roles for Epidemiologists
Epidemiology Associations and Journals
Competencies Required of Epidemiologists
Resources for Education and Employment
Professional Ethics in Epidemiology
Conclusion
Study Questions and Exercises
References

Appendix A: Guide to the Critical Appraisal of an Epidemiologic/Public Health Research Article
Appendix B: Answers to Selected Study Questions
Glossary
Index