1 Page summary

profileGsanc1
Week5_PPslides1-11_Spring20201.pps
Home>Information Systems homework help>1 Page summary

HCS 420:
RESEARCH METHODS IN HEALTHCARE

Week 5

Chapters 8 & 9

Study Design II
Case Control Studies

Chapter 8

What are case-control studies?


A case-control study identifies one group of subjects with the disease and another without it, then looks backward to find differences in predictor variables that may explain why the cases got the disease and the controls did not

Therefore, they are retrospective in nature.

Cases vs Control?

Case-Control Studies

  • Cases: Those with the disease
  • Could also be the patients who have experienced a desirable outcome (i.e. recovered from a disease)

  • Controls: The comparator
  • Generally, those without the disease

Important Notes on Case Control Studies

Case-Control Studies

  • Cross-sectional and cohort studies are not fit for certain situations
  • e.g. when studying rare diseases
  • The investigator determines the proportion of study subjects who have the disease and how many control
  • Cannot estimate incidence or prevalence
  • Estimate the strength of the association in the form of odds ratio
  • Can estimate the relative risk if the risk of the disease in both exposed and unexposed subjects is relatively low

Examples of Case-Control Studies?

A study showing the link between prone sleeping position and sudden infant death syndrome.

This study has helped save many lives!

How would such a study be designed?

Can you describe how you would go ahead conducting a case-control study?

What are some Strengths of Case-Control Studies?

Strengths

  • Yields rapid, high yield information from relatively few subjects

  • Efficient for studying rare outcomes

  • Can be used to examine a large number of predictor variables

  • Excellent for generating hypotheses

What are some Weaknesses of case-control studies?

Weaknesses

  • Compared to cohort and cross-sectional studies, only one outcome can be studied at a time

  • Since no direct estimate of the incidence or prevalence of the disease is provided, the information provided is limited

  • High susceptibility to bias
  • Separate sampling of the cases and control
  • Retrospective measurement of the predictor variables

How is sampling bias a problem for case-control studies?

Issues of Sampling Bias in Case-Control Studies

  • Typically, cases are samples from patients in whom the disease has already been diagnosed and who are available for study
  • The undiagnosed, misdiagnosed, unavailable, or dead are unlikely to be included
  • While common disease that require hospitalization (e.g. hip fracture) may be ok for study, conditions that may not come to medical attention may not be
  • Selection of cases is often limited to the accessible sources of subjects

Strategies for improving the quality of controls in case-control studies?

Strategies from Sampling Controls

  • Clinic or hospital-based controls
  • Choosing the control from the same facility where the cases come from can reduce bias

  • Using population-based samples of cases and controls
  • Very desirable when registries are available

Strategies (Contd.)

  • Using two or more control groups
  • Multiple control groups provide a way to test the validity of the result
  • Matching
  • Ensuring that cases and control are comparable on major factors related to the disease but not of interest to the investigator



The overall goal of sampling in case-control studies is to sample controls from the population who would have become a case in the study if they had developed the disease

NOTE

Other Case-Control Designs?

Other Case-Control Designs

  • Nested Case-Control

  • Incidence-Density Nested Case-Control

  • Nested case-cohort

  • Case-Crossover

Nested Case-Control

A case-control study “nested” within a defined cohort

i.e. a case-control study within a cohort study

Nested Case-Control

  • First, the investigator defines the cohort
  • The cohort may have been already defined as part of a formal cohort study
  • Or, the cohort may not be already defined

  • Next, the investigator defines the occurrence of the outcome of interest

Nested Case-Control (Contd)

  • Individuals in the cohort who develop the outcome by the end of the follow-up are the cases
  • A random sample of the subjects who were also part of the cohort but didn’t develop the outcome constitute the controls

Strengths of Nested Case-Control?

Strengths of Nested Case-Control

  • Useful for taking measurements on for expensive measurements like serum and similar specimens that have been archived

  • Shares the advantage of cohort studies in terms of collecting predictor variables before the outcomes have happened

  • Avoids the biases of a traditional case-control that cannot make measurements on fatal cases and draws cases and controls from different populations

Weaknesses?

Weaknesses

  • Like other observational designs, observed associations may be due to the effect of unmeasured or imprecisely measured confounding variables

  • Baseline measurements may be affected by silent preclinical disease

Incidence-density Nested Case-Control

  • Used when follow-up is variable or incomplete, or if the exposure varies over time

  • Controls are samples from risk sets (defined for each case)
  • Members of the cohort who were followed the same length of time as the case but had not yet become cases

  • Considers incidence-density (considers person-time at risk)

Nested case-cohort?

Nested case-cohort

  • Similar to a nested case-control design

  • However, instead of selecting controls who didn’t develop the outcome of interest, the investigator selects a random sample of all the members of the cohort, regardless of outcome

Advantages?

Nested case-cohort

  • A single random sample of the cohort can provide the controls for several case-control studies of different outcomes

  • The random sample of the cohort provides information on the overall prevalence of risk factors in the cohort

Case-Crossover?

Case-Crossover

  • A type of case-control that’s useful for studying the short-term effects of intermittent exposures
  • Begins with a group of cases (people who have had the outcome of interest)
  • Each case serves as it’s own control
  • Exposures of the cases at the time (or right after) the outcome occurred are compared with exposures of those same cases at one or more other points in time
  • Analyzed as a matched case-control study
  • Control exposures are exposures of the case at different time periods (rather than exposures of the matched controls)

Causal Inference in Observation Studies

Chapter 9

In general, observational studies are designed to suggest that an association exist between a predictor and an outcome

Not all associations are cause-effect

  • Chance: Spurious

  • Bias: Spurious

  • Effect-cause: Real

  • Confounding: Real

These four alternative explanations must be dismissed before a cause-effect can be established

“Chance”

  • “Random error”
  • E.g., recruiting more people with a disease/condition among the cases compared to control, creating a spurious association
  • Type I error occurs when an association due to random error is taken to be statistically significant.

How can you reduce random error in research?

Strategies

  • Both “design” and “analysis” strategies
  • Design
  • Increasing precision and measurements
  • Increasing the sample size
  • Analysis
  • Calculating P values and confidence intervals
  • Helps in quantifying the magnitude of the observed association in comparison with what might have occurred by chance alone

“Bias”

  • “Systematic errors”
  • E.g. differences between the original research question and the one that is actually answered by the study
  • Reflects the compromises made for the study to be feasible
  • Also reflects mistakes in the design and execution of the study

How can you reduce systematic error in research?

Strategies

  • Both “design” and “analysis” strategies
  • Design phase
  • Samples of study subjects represent the populations of interest
  • Measurements of the predictor variables represent the predictors of interest
  • Measurements of the outcome variables represent the outcomes of interest
  • Analysis phase
  • Assess likely severity of bias by analyzing data that have been collected to identify that
  • Compare to results of other studies

Effect-Cause

  • Occurs when the “outcome” causes the “predictor” (rather than the other way around; “cause-effect”)

  • Often a problem in cross-sectional, case-control studies, and case-crossover

  • Less of a problem in cohort studies

Confounding

  • Occurs when a third variable/factor is a real cause of the outcome and the predictor of interest is associated with, but not a cause of, the third factor
  • The variable must be associated with the predictor of interest and also be a cause of the outcome
  • If the predictor causes a third variable (C) which then causes the outcome, then the variable C is a mediator

Controlling for Confounding (Design)

  • Specification

  • Matching

  • Opportunistic studies

Controlling for Confounding (Analysis)

  • Stratification

  • Adjustment

  • Propensity scores