Statistical application and the interpretation of data

WhatAreStatisticsandWhyAreTheyImportanttoHealthScience.pdf

Home >Nursing homework help >Statistical application and the interpretation of data

What Are Statistics and Why Are They Important to Health Science?By June Helbig & Jayme Ambrose Essential Questions

• How are statistical concepts used in daily life? • Historically, how have statistics been used in health care? • Why is it important for a person working in health care to understand

statistical concepts? • What is the first step in beginning the data analysis process?

Introduction

• There is generally a lack of understanding of what statistics is and how much it is present in everyday life. Understanding statistics requires learning new terms and adopting a different way of observing the world. This book will provide an explanation of how statistics is applied in both the personal and professional lives of health care providers. Additionally, this book will provide the statistical tools to determine the efficacy of research, use statistics to identify patient care trends using data, and increase the ability to create and utilize evidence-based practice, all of which are crucial to improve patient outcomes and ensure quality care. Statistics is a component of mathematics used to analyze data. Learning about statistics does not require learning a foreign language, but it does require an understanding of certain terms and their meanings as well as theorems and how they are used to calculate numbers. It is a mathematical journey to getting to a set of numbers that provide information about a sample population to make decisions about patient care (Campbell, 2012). After reading this chapter, the health care professional, or baccalaureate-prepared registered nurse will have an understanding of statistical concepts and their application to the health care setting. This chapter will provide an overview of statistics, its history, and the definitions of applied concepts. How statistics is applied to quality and leadership will be discussed, as well as the statistical impact on future trends.

History of Statistics

•

• Statistics looks for correlations between entities. Do peanut butter and jelly sandwiches taste better than plain peanut butter sandwiches? If six out of ten people say yes, then statistically, more people like peanut butter and jelly sandwiches than plain peanut butter sandwiches. Statistics is not a new concept or way of looking for answers but one that has been around for hundreds of years. The first documentation of social statistical concepts originated in the 1600s as published discussions on probability and normal curve equations (Fienberg, 1992). As the study of research continued, it broadened to include the collection, summary, and analysis of data and was joined with probability. The roots of nursing practice are based in the use and statistical analysis of patient data. Florence Nightingale had “an ardent desire to remedy the scandalous neglect of sanitary precautions in the Army” (Kopf, 1916, p. 390). During the Crimean War (1853– 1856), Nightingale collected statistics on wounded soldiers, including injuries, infections, and mortalities. She observed unsanitary conditions, bad food, and bad air and began looking for ways to improve these conditions. As Nightingale reviewed the data she had collected on the soldiers’ causes of death, she found that those who had died were seven times more likely to have died from disease in a hospital than to have died in combat. Nightingale used this data analysis to formulate the knowledge of nursing (Nightingale, 1969). Her focus for proper nursing care was on ventilation, appropriate housing, and cleanliness. She stressed the need to keep the air and water pure. She identified the need for efficient drainage and light. “People lose their health in a dark house, and if they get ill they cannot get well again” (Nightingale, 1969). She also stressed the importance of cleanliness of the environment, patient, and all items coming in contact with the patient. At the time of Nightingale’s observations, the world had poor sanitary conditions in which raw sewage ran through the city streets and there was no in-home running water.

The air smelled foul from the unfiltered sewage that flowed down the river. Nightingale’s suggestion to open windows ran contrary to the belief that the bad smells caused disease. Nightingale was able to show that the bad smells and, ultimately, the risk of disease could be eradicated thorough cleaning. Nightingale was also a pioneer in producing graphic illustrations of statistics (Kopf, 1916) (see Figure 1.1). Nightingale went on to help reform the way the hospitals were collecting data. She proposed the process of routine collection and analysis of benchmarked health data. She was the leader in adding questions on health status and housing. The hospitals were then able to share their statistical findings and make changes to patient care on a larger scale. These results showed the “value of how large a field of qualified statistical inquiry had been opened” (Kopf, 1916, p. 394). Figure 1.1

Nightingale Diagram Causes of Mortality

Note. Adapted from Notes on Matters Affecting the Health, Efficiency and Hospital Administration of the British Army, by Florence Nightingale, 1858.

Nightingale was deeply influenced by Adolphe Quetelet, an astronomer and mathematician, who is considered to be the father of quantitative statistics in social sciences (Fienberg, 1992). His book focusing on the statistical methods applied to the life of man was published in 1835. Quetelet provided the advancement of statistical knowledge that allowed Nightingale to develop the methods, direction, and results of qualified inquiry. (Kopf, 1916). Statistics and its utilization has continued to evolve over time. The 18th century focused on the development of mathematics and the theories of probability. In the 19th century,

the American Statistical Association was formed in Boston, Massachusetts (American Statistical Association, n.d.). Following the development of civil registration of births and deaths, the U.S. government began expanding their use of statistics from the mere collection of the data for census knowledge to the analysis of this data. The analysis of the data began the use of epidemiology to improve population health.

•

• Epidemiology is the study of health-related states or events, including disease. Epidemiology is also used to control diseases and other health problems. Epidemiologic investigations include observation or surveillance and descriptive or experimental studies to examine the distribution of the health determinants. Epidemiology is considered a methodology for studying health conditions and developing the body of knowledge of the health condition. This approach enhances both the practice of nursing and other health care professions. In the 21st century, the need for a basic understanding of statistics has increased, making statistical literacy a necessity. Statistics is widely used in government, business, the sciences, and computer technology. The advancement of computer technology has changed the way statistics are used and simplified data analysis for researchers. The appearance of statistical software packages that could analyze data first originated in the 1960s. (Griffith, 2010). In the 1980s, statistical software became user-friendly tools available on personal computers. Statistical software, the most common known of which is Statistical Package for the Social Sciences (SPSS). Software tools can perform calculations and summarize results; however, it is still necessary to interpret the output. As with all software, good data in is good data out.

Data Analysis and Application

The enormous volume of health care data collected on a daily basis from the electronic medical record (EMR), including information such as biostatistics, satisfaction surveys, birth weights, mortality rates, diagnoses, and testing results, can help to answer questions through data analysis. For example, analyzing health care data could help to identify quality outcomes related to infection rates and fall risks, changes to procedures based on data findings, and changes in patient interactions based on satisfaction survey results. Data analysis can also measure intake and output and look for trends.

Leaders’ Role

• As leaders in health care, the ability to understand the data and apply it is an essential skill. The leader’s role should be seen as a partner in the application of the analyzed data. Health care leaders should also have the knowledge to determine the validity or usability of the analysis. Data analysis creates the future of health care. In the analysis of the data, trends are identified, and those trends allow the assumption of predictability. The predictability assumptions create stronger models of care, which increases quality

outcomes and patient safety. Statistical analysis of the data is used to answer questions that improve population health.

Levels of Measurement In order to organize and analyze the data collected, four levels of measurement are used, which include nominal, ordinal, interval, and ratio. Measurement is a way to give meaning to numbers and give value to data collected. Each of the four levels of measurement are different and are not interchangeable. Any variable identified by a research study can be measured using one of the four levels of measurements. Nominal The nominal level of measurement is used to name or categorize things. Nominal data has to do with naming, categorizing, and labeling, such as gender, race, or marital status. Social security numbers are also considered nominal data because the number means nothing; it cannot be manipulated mathematically because it is just a label. Table 1.3 is an example of how nominal data can be organized. Nothing can be done with the number assigned to categorize the National Patient Safety Goal, it is just an alpha-numerical label assigned to each goal by The Joint Commission. Table 1.3

Example of Nominal Level of Measurement

⟵ Rank (Nominal Level) ⟶ National Patient Safety Goal Indicator Label

Use at least two ways to identify patients. For example, use the patient’s name and date of birth. This is done to make sure that each patient gets the correct medicine and treatment.

NPSG.01.01.01

Make sure that the correct patient gets the correct blood when they get a blood transfusion.

NPSG.01.03.01

Get important test results to the right staff person on time. NPSG.02.03.01

Ordinal The ordinal level of measurement defines the relationship between entities and assigns an order or ranking to each. The difference between ranks can be equal or unequal. The order or rank has no specific meaning. For example, the Centers for

Medicare and Medicaid Services (CMS) assigns a rank to each hospital that participates in meeting the National Patient Safety Goals (see Table 1.4). Table 1.4

Ordinal Level of Measurement

⟵ Name Ordinal Level ⟶ Hospital Quality Indicator

January 2018 National Rank

Hospital A NPSG.01.01.01 1

Hospital B NPSG.01.01.01 2

Hospital C NPSG.01.01.01 3

Interval The interval level of measurement assigns specified distances between the entities being measured. The length or amount of distance between intervals is defined and is equal. Each interval is the same and can be hours, months, inches, feet, or whatever is predetermined within the conceptual framework, which is the way the research is organized to reach the project’s purpose. For example, the number of occurrences for the National Patient Safety Goals are reported on a quarterly basis to evaluate hospital performance (see Table 1.5). Interval measurements can be manipulated mathematically, so analysis is possible using the interval level of measurement. Table 1.5

Interval Level of Measurement

⟵ Quarterly (Interval Level of Measurment) ⟶

1st Quarter 2nd Quarter 3rd Quarter 4th Quarter

Hospital A # of occurrences # of occurrences # of occurrences # of occurrences

• NPSG.01.01.01 3 5 4 7

• NPSG.01.03.01 2 3 7 6

• NPSG.02.03.01 4 6 6 7

Hospital B

• NPSG.01.01.01 2 3 4 5

• NPSG.01.03.01 0 1 2 2

• NPSG.02.03.01 0 1 3 5

Hospital C

• NPSG.01.01.01 4 5 3 4

• NPSG.01.03.01 2 2 2 2

• NPSG.02.03.01 3 6 3 3

Ratio The ratio level of measurement is very similar to the interval level of measurement, but the ratio level can define zero. The presence of zero makes it possible to compare ratios of measurements when analyzing data. The ratio level of measurement has meaning in results interpretation. The ratio is a level of measurement between two similar units. When using the ratio level of measurement “there is always an absolute zero that is meaningful. This means you can build a meaningful fraction (or ratio) with a ratio variable” (Trochim, 2006, para. 7). When applying this level of measurement, researchers can compare two values and their relationship to each other. For example,

when using the National Patient Safety Goals as an example, a reported value of zero is statistically important, as it indicates that the hospital has not met the National Patient Safety Goals. The comparison is the number of events to the number of times the hospital was compliant with that specific event. If a zero is reported in relation to National Patient Safety Goals, it means there has been no compliance with the quality indicators. If the hospital was compliant for each event, then the value reported would be 100%. The ratio unit of measurement can also be used when comparing units that are similar. There may be an instance when something occurred two or three times more than expected. If this occurs the ratio would be 2:1 or 3:1, which would give values of 200% and 300% respectively (see Table 1.6). As is demonstrated in Table 1.6, Hospital A was compliant with the first indicator three times out of three opportunities. To obtain a percentage, divide 3 by 3, and then multiply the answer by 100. A percent is a special ratio that compares any number to 100, with 100 representing one whole. Representing the unit whole by 100 makes it easier to make a comparison. If there are three opportunities to do something right, and it is done right three times, as a percentage, 100% compliance has been achieved. Table 1.6

Ratio Level of Measurement

⟵ Ratio Level of Measurment ⟶ # of Times Compliant With

Indicator / # of Opportunities Percentage Compliance

Hospital A

NPSG.01.01.01 3/3(3/3×100=100%)

100%

NPSG.01.03.01 2/2(2/2×100=100%)

100%

NPSG.02.03.01 4/5(4/5×100=100%)

80%

Hospital B

NPSG.01.01.01 2/2 100%

NPSG.01.03.01 0/1 0%

NPSG.02.03.01 0/1 0%

Hospital C

NPSG.01.01.01 4/5 80%

NPSG.01.03.01 2/2 100%

NPSG.02.03.01 3/3 100%

Using statistical levels of measurements, the data obtained from research studies can be organized and analyzed. Data that are nonnumerical variables can be organized using the nominal and ordinal levels of measurement. When analyzing numerical variables, interval and ratio levels of measurement are used. When discussing research and data, variables must be included in the discussion.

Variables Prior to analyzing data, it is important to define some key terms and identify types of data. Quantitative research collects numerical data, and qualitative research collects words as data. There are times when numerical variables can be changed and manipulated. For the researcher to begin a research study, he or she must define the measurement scale to be used and the type of statistical analysis required. The researcher must also determine the dependent and independent variables associated with the study to be performed. In research or data analysis, a variable is defined as a “characteristic of an item or individual” (Levin, Krehbiel, Berenson, & Viswanathan, 2014, p. 6). A variable is something that varies in size, amount, or degree. For a patient, this could be age, weight, and/or blood pressure. In research, the role of variables is used to determine how or why things vary and whether differences in one variable are somehow related to the differences in another variable (Polit & Beck, 2017). The independent variable is the variable that can be manipulated to determine the dependent variables’ value. An independent variable can stand alone. The dependent variable is dependent on the independent variable. The dependent variable changes only in relation to the independent variable. The extraneous variable is unknown when beginning a research study, but it shows itself in time. Many times, an extraneous variable is an unwanted variable and can skew the results of a research study. Extraneous variables can be categorized as situational, personal, or researcher-based.

• Situational: the physical situation changes for certain groups within the study. Examples include losing a job, death in the family, divorce, and moving.

• Personal: personality traits that members of the group may have that others do not. Examples include intelligence levels, critical-thinking skills, and depression.

• Researcher-Based: something the researcher does differently from one group in the research study or experiment to the next. Examples include telling one group something but not the other and spending more time with one group than the other.

Extraneous variables are why a researcher cannot state there is 100% certainty that there is causation within the research. One example of an extraneous variable is researcher bias or manipulation. Bias occurs when the researcher influences the results of the study. This can occur while planning the research, collecting data, or analyzing the results. Examples of extraneous variables include environmental influences that could affect subjects, subjects receiving unintentional clues or expectations from the researcher, participants having prior knowledge that could affect the outcome, and situational issues, such as noise, light, or other environmental distractions. Political or financial influences outside the study are other examples of bias that can impact the analysis and its outcome (Molyneux, Tsofa, Barasa, Niykuri, Waweru, Goodman, & Gilson, 2016). For example, a study will display bias if the government or the organization funding the research direct the outcome. Using inclusion and exclusion criteria are ways researchers attempt to control extraneous variables. Sometimes, it is the independent or extraneous variables that must be changed or adjusted for the research study to be successful. Health care leaders need to be cautious when reviewing research that claims to show direct causation. Extraneous variables are one component of research that can affect the data analysis and skew the outcomes. If a researcher is studying how to prevent patient falls, he or she must decide what fall precautions will be used to prevent falls. The fall precautions are the independent variables that will directly affect whether or not the patient falls (see Table 1.7). When researchers want to find new and innovative ways to prevent falls, they manipulate the independent variable to reduce falls. It is the independent variables’ effect on the dependent variable that is measured in research. Table 1.7

Types of Variables

Dependent Variable

Independent Variable Extraneous Variable

Falls • Call light accessible for patient

• Floor is wet • Pillow fell on floor

• Bed in low, locked position • Side rails up (maximum of 3) • Yellow gown • Yellow socks • Yellow leaf on patient's door

• Patient got up without alerting nurse

Medication Errors • Right patient • Right drug • Right dose • Right route • Right time • Bedside medication

verification

• Pharmacy stocked wrong medication • Nurse did not check proper patient

identification

Continuous Versus Categorical Variables

• Variables can be defined as continuous or categorical variables. Continuous variables are most often associated with quantitative research in which the data collected are numerical. The values of the data can change because they are numerical. Any type of data that can be counted are quantitative data. The levels of measurement used are usually the interval and ratio levels of measurement (see Table 1.8). Table 1.8

Continuous Variables

Patient Falls by Month

Hospital A Hospital B Hospital C

January 3 1 0

February 2 2 1

March 1 3 0

April 1 2 0

May 2 1 1

June 5 2 1

July 3 1 1

August 2 4 2

September 1 1 0

October 0 2 1

November 1 2 1

December 0 2 1

Categorical variables are most often associated with qualitative data and differences between groups or categories. The levels of measurement used are usually nominal and ordinal (see Figure 1.2). Figure 1.2

Example of Fall Data Results

Patient Falls 2017-2018 Patient Falls 2017-2018

2017 2018

Hospital A 21 15

Hospital B 23 18

Hospital A Hospital B Hospital C

Hospital C 9 9

2017 2018

Descriptive Versus Inferential Statistics The purpose of engaging in a research project is to identify an intervention or action to be imposed in specific situations to improve a clinical outcome. The results of the research method provide information that may describe the situation or allow the investigator to infer a relationship. When describing the population or variables associated with the research project, descriptive statistics such as mean, median, minimum, maximum, and standard deviation are explained. Descriptive statistics provides an estimation of specific variables being examined in the population of interest. Descriptive statistics also provides a summary of the basic data collected to represent the population or sample in the research process. Additionally, a variability of the descriptive is presented, usually the standard deviation. An educator may wish to identify the age of students returning to college to earn a degree in a health care-related field as their second career. In contrast to descriptive statistics, inferential statistics form opinions or assumptions of the population of interest. The goal of statistical inference is to establish a conclusion about not only the participants in the sample but also the population at large.

Key Terms

Bias: External and internal influences within a study that can affect the validity and reliability of the outcomes. Categorical Variables: Qualitative data associated with nominal or ordinal measures. Central Tendency: The central location of the data. It identifies the single value that represents the entire distribution. It is found by measuring the mean, mode, or median depending on the situation. Continuous Variables: A variable that has an infinite number of possible values. Data: Raw unorganized information from which conclusions can be made. Demographics: Statistical information regarding groups of persons. Dependent Variable: The outcome variable. It takes on different values in response to the independent variable. Descriptive Analysis: Statistics used to describe the basic features of the data in a study. This analysis allows data to be organized and categorized into frequencies. It also looks at frequency distributions and the measures of central tendencies. The analysis provides simple summaries about the sample and the measures. This form of analysis is the basis of virtually every quantitative analysis of data. Descriptive Statistics: Statistics that describe the basic features of the data in a study. They provide simple summaries about the sample and the measures. Epidemiology: The study of disease appearance, course, spread, and eradication. Extraneous Variable: A variable that can influence the relationship between the independent and dependent variables; can be controlled either through research design or statistical procedures; were not foreseen or known at the beginning of the study. Independent Variable: The experimental or predictor variable. It is manipulated in the research to observe the effect on the dependent variable. Inferential Statistics: Statistics designed to allow the researcher to infer characteristics regarding a population from sample population. Interval Level of Measurement: The variable has rank order and equal distances on the points of the scale. Mean: The average of the data collected; a component of descriptive analysis. To determine the mean, add up all the numbers, and divide the sum by the total number of numbers. Median: The middle of all the data. In order to determine the median, list all the data in numerical order, smallest to largest. The median is the middle value. Median is a component of descriptive analysis.

Nominal Level of Measurement: Used to name or categorize things; the first level of measurement. Ordinal Level of Measurement: Defines the relationship between things and assigns an order or ranking to each thing; the second level of measurement. Percent: A special ratio that compares any number to 100. Ratio Level of Measurement: A measurement level with equal distances between the points and a zero-starting point. Ratio: A comparison of any two numbers by division. Sample: The subset of the population to be studied. Samples can be chosen randomly, blindly, or double-blindly. Standard Deviation: The statistical determination of how far each data point deviates from the mean. To determine the standard deviation, determine the mean, subtract the mean, and square the result for each data point. The variance is found by calculating the average of the squared number. Statistics: A component of mathematics that looks at gathered data. Statistics covers how the data is collected as well as the analysis and the interpretation, presentation, and organization of the data. It has its only language, definitions, and theorems. Validity: The extent to which a concept, conclusion or measurement is well-founded and corresponds accurately to the real world. Variable: A data item such as characteristics, numbers, properties, or quantities that can be measured or counted. The value of the data item can vary or be manipulated from one entity to another. There are three different types of variables—dependent, independent, and extraneous.

References American Statistical Association. (n.d.). History of the ASA. Retrieved from

http://www.amstat.org/ASA/About/History.aspx Campbell, R. C. (2012). Statistics for biologists (3rd ed.). Cambridge, United

Kingdom: Cambridge University Press. Levin, D. M., Krehbiel, T. C., Berenson, M. L., & Viswanathan, P. K.

(2014) Business statistics: A first course (5th ed.). Chennai, India: Pearson. Fienberg, S. E. (1992). A brief history of statistics in thre and one half chapters: A

review essay. Statistical Science, 7(2), 208-225 Griffith, A. (2010). SPSS for dummies (2nd ed.). Hoboken, NJ: Wiley Publishing. Kopf, E. (1916). Florence Nightingale as statistician. Publications of the American

Statistical Association, 15(116), 388-404.

Molyneux, S., Tsofa, B., Barasa, E., Niykuri, M. M., Waweru, E. W., Goodman, C., & Gilson, L. (2016). Research involving health providers and managers: Ethical issues faced by researchers conducting diverse health policy and systems research in Kenya. Developing World Bioethics, 16(3), 168-177. doi: 10.1111/dewb.12130

Nightingale, F. (1969). Notes on nursing: What is is and what it is not. New York: Dover Publications.

Polit, D., & Beck, C. (2017). Nursing research: Generating and assessing evidence for nursing practice (10th ed.). Philadelphia, PA: Wolters Kluwer.

Trochim, W. M. K. (2006). Levels of measurement. Retrieved from https://socialresearchmethods.net/kb/measlevl.php