Content area
Epidemiology speaks to the risk of disease, injury, or death. The purpose of epidemiology is to predict disease (or injury or death) based on risk factors or exposures. Results are communicated with caution, to avoid unjustly labeling a risk factor as a "cause" of disease. Although an individual may be at increased risk for a given disease, there is often no way to predict if or when that individual will actually contract the disease. Rather, epidemiology is a tool to assess populations and improve the health of the general public.
(ProQuest: ... denotes formula omitted.)
Given the worldwide focus on a possible pandemic, epidemiology is a common topic in the news. In addition to reports about widespread outbreaks of communicable diseases, news media frequently address obesity, heart disease, mental illness, and other chronic health problems. Occasionally, more obscure topics, such as cancer clusters, danger from electromagnetic fields, and radiation from cell phones, receive media coverage.
Occupational health nurses are often asked their opinions about epidemiologic issues. They must sufficiently understand epidemiology to assist workers in interpreting media reports. Additionally, nurses in occupational settings must recognize situations in the workplace that warrant further investigation and intervention. This article provides the basics of epidemiology and its application to occupational and environmental health nursing practice.
EPIDEMIOLOGY DEFINED
Traditionally, epidemiology is the study of the distribution and determinants of disease and death in populations (Aschengrau & Seage, 2003; Friis & Sellers, 1999; Last, 2001; Szklo & Nieto, 2000). Over time, the definition of epidemiology has expanded to include the study of the distribution and determinants of injuries and disability. Any health phenomenon can be included in epidemiology studies. Epidemiology explores the distribution of health conditions in human and other populations, and the causes of specific health conditions. Epidemiological studies focus on specific populations (e.g., inmates in a state prison system, children attending daycare centers, patients in Veterans Administration hospitals, residents of a specific geographic region, or employees in a defined industry).
The aim of epidemiology is to describe the health status of populations, explain the etiology (i.e., causal factors and modes of transmission) of diseases, and predict the occurrence of diseases in the future, including where and under what circumstances the diseases (deaths or injuries) may occur and who will be affected. To achieve such goals, epidemiologists must gather substantial information and compare diseased or exposed individuals with healthy individuals.
The era of modern epidemiology began more than 100 years ago when John Snow, a British physician, plotted an outbreak of cholera on a map of London to identify the source of contamination. When Dr. Snow removed the handle from a pump that was providing contaminated water, the cholera outbreak ended (Hennekens & Buring, 1987; Friis & Sellers, 1999). Since then, the science and practice of epidemiology continue to be refined and expanded with the use of sophisticated study designs, technology, and statistical methods.
LANGUAGE OF EPIDEMIOLOGY
Epidemiology uses a unique vocabulary to describe the frequency and occurrence of disease. The vocabulary is so unique that it merits its own dictionary (Last, 2001). Basic awareness of some of the terminology used in epidemiology enables occupational health nurses to understand and effectively communicate health-related situations at the worksite or in the community.
Incidence, the number of new events in a specific population during a specific period (Last, 2001), is a term often used in epidemiology. It is expressed as a number, and is different from incidence rate. Incidence rate is mathematically calculated with the number of new cases divided by the total population at risk. For example, the incidence of stubbed toes among construction workers in a hypothetical construction company may be 42 per year (Table 1). Because 1,000 construction workers may be at risk for stubbing their toes during this year, the incidence rate is 42 per 1,000 (or 4.2%). Government incidence rates are often reported "per 10,000" or "per 100,000," depending on the size of the incidence.
Prevalence, the number of existing cases in a given population at a given time, is another common term. Prevalence is more commonly used with chronic diseases. For example, once asthma or heart disease is diagnosed, the individual always has the condition. The prevalence "rate" is actually a proportion where the numerator includes only those who are diagnosed and the denominator includes both those diagnosed and those at risk for the disease (Friis & Sellers, 1999; Last, 2001). Prevalence continues to rise over time if a condition is not abated and if the population remains constant (i.e., individuals with the condition do not die). Rising prevalence could indicate longer survival after diagnosis, or an increase in the number of individuals affected by the condition.
The purpose of using a rate or proportion, rather than just the incidence or prevalence, is to standardize the numbers for comparison. As Table 1 indicates, if the population of lettuce pickers is larger or smaller than the population of construction workers, it is difficult to tell which group has a greater problem with toe stubbing. Incidence rates allow for a reliable comparison.
Ratios compare two groups that are not necessarily related to one another. For example, a ratio of males to females, Hispanics to non-Hispanics, or smokers to nonsmokers may give information about how different groups are affected by injury or disease in the same time period. If 13 men and 2 women experienced injuries in a given week, the ratio of men to women for the number of injuries is 13 to 2.
...
Crude rates are based on the actual number of events in a population during a given time period (Friis & Sellers, 1999). Crude rates do not often give enough information about a population, so specific rates and adjusted rates are also used. For instance, it may be beneficial to describe the death rate for a specific cause, or the injury rate during a specific activity. Calculating adjusted rates for specific work groups, age groups, or job categories may provide additional information.
Adjusted rates are a better indicator of risk than crude rates because they focus on subsets of populations. Direct adjustment is used when the rate for a population is known. For example, when an occupational health nurse wants to know how one division is performing compared to the whole company, direct adjustment would use the whole company's rate. Indirect adjustment uses a known, standard rate. When the population's rate is unknown, the occupational health nurse may want to use a national or state rate for comparison. For instance, if 6% of the participants in a company's health screening indicate they are smokers, can the nurse make any generalizations about the smoking rate of the company's population if a smoking assessment has not been done? Does 6% represent the true percentage of smokers? Or does the company have the same (higher) smoking rate as the rest of the state (e.g., approximately 23%), and are many of the smokers avoiding the health screening? If such an assessment has not been done, the state's smoking rate may be reasonably applied in this situation. It is not important for most occupational health nurses to know the differences between direct and indirect adjustment. Rather, it is valuable to understand that different methods can be used, depending on the circumstances.
One division in a hypothetical prefabricated wood manufacturing company had 15 lost-time injuries last year, with 150 days lost. The safety committee wondered whether these losses were acceptable. One committee member mentioned an employer who never experienced more than 10 lost- time injuries with 100 days lost in a year. Another committee member blamed the high numbers on a merger with another company. To solve the dilemma, the committee first calculated the rates for injuries and lost time. Once rates were established, the committee compared these rates with those of other divisions, the whole company, previous years, and national statistics.
With the use of the information presented in Table 2, it is possible to calculate the expected number of injuries this year, based on last year's rates. The number of employees this year is multiplied by last year's injury rate and divided by 100. On the basis of last year's injury rate, the committee could have expected 20 lost-time injuries, rather than the 15 that actually occurred.
The committee still wondered whether the division was experiencing injuries comparable to other manufacturing companies. The committee used the Bureau of Labor Statistics (BLS) to find the injury rates for manufacturing companies. The BLS website showed an incidence rate of 4.0 per 100 workers for prefabricated wood manufacturing in 2004 (BLS, 2006). If the hypothetical division had experienced an injury rate comparable to the one reported at the national level, the safety committee would have documentation of 40 injuries.
The standardized mortality ratio can be used for disease and injury, as well as death (Friis & Sellers, 1999). This ratio compares the number of observed events with the number of expected events. The hypothetical manufacturing company from Table 1 saw 15 lost-day injuries. If it had the same injury rate as the BLS, it would have had 40 lost-day injuries. Thus, this company's lost-time injury rate was only 37.5% of the national rate. From this example, it appears that the manufacturing company's safety program is effective this year.
DESCRIPTIVE EPIDEMIOLOGY
To understand diseases, injuries, and death, it is important to examine patterns of person, place, and time. What are the characteristics of individuals affected by the disease? Where do injuries occur? When do injuries occur, and under what circumstances? Establishing patterns of disease and injury provides a basis for planning, implementing, and evaluating health services. Accurate data are necessary for securing resources to solve health problems.
Gathering information about person, place, and time is essential for establishing patterns of injury and illness in the workplace. Assembling and reporting this information usually requires an electronic database. Some of the information may be available from the human resources database. Most commercially available occupational health software systems have the capability of collecting pertinent information. Occupational health nurses must first determine how data will be used. This will dictate what data are collected. For instance, many injuries are the result of "inattention." But what causes inattention? How many hours was the individual working before the incident? How many hours of sleep did the individual have the night before? Is the individual experiencing stress at home or work? This type of additional information can be used to plan interventions that actually address the underlying causes of accidents and injuries.
Epidemiology can provide a picture of the risk for the disease or injury. Epidemiologists are cautious about indicating what caused a disease. True causes must possess three attributes: association, time order, and direction (Aschengrau & Seage, 2003). Epidemiologists ask whether a risk factor is necessary and sufficient to cause a disease. For communicable diseases, the criteria of necessary and sufficient association are fairly easy to understand. One does not get malaria unless one is bitten by a mosquito carrying the malaria parasite. A child must be exposed to someone with chickenpox to contract the disease.
However, the criteria of necessary and sufficient are more difficult to establish with chronic disease. For instance, individuals develop lung cancer even if they do not smoke cigarettes, and some smokers may never be diagnosed with lung cancer. Scientists recently identified human papilloma virus as the necessary condition for cervical cancer. All women who develop cervical cancer have had a human papilloma virus infection, but not everyone who contracts human papilloma virus will develop cervical cancer (Rydstrom & Tornberg, 2006).
For the time order attribute to be present, the causal agent must precede the disease (e.g., tainted potato salad must be ingested before food poisoning occurs).
The third attribute, direction, indicates an asymmetrical relationship between exposure and outcome (Aschengrau & Seage, 2003). That is, a causal agent may lead to disease, but not vice versa (e.g., smokers have heart disease more often than nonsmokers but heart disease does not cause smoking).
In 1965, Austin Bradford Hill developed nine guidelines to assist scientists in determining whether associations are causal (Aschengrau & Seage, 2003; Friis & Sellers, 1999; Hill, 1965):
* Strength of association: Repeated epidemiological studies are necessary to provide supporting evidence for causality. Strength of association results from quality studies showing a strong relationship between the factor being considered and the disease. When the association between the risk factor and the disease is marginal, it is more difficult to assert a causal relationship.
* Consistency: Consistent findings from one study to another, a variety of populations, and different locations or circumstances increase the likelihood of a causal relationship between factor and disease.
* Specificity: Cause leads to a single effect and a specific effect only has one cause. This concept works well with infectious diseases, but has limited use with chronic diseases, where one harmful agent can cause several diseases, and the same disease can result from several causes.
* Temporality: Exposure to the agent happens before the disease occurs.
* Biological grathent: A phenomenon in which the consequences are greater with increased exposure. For instance, smoking two to three packs of cigarettes per day results in disease more frequently than smoking one pack per day.
* Plausibility: Answers the question, "Does this make sense?" When science supports epidemiology, causality is more likely. For example, a study found an association between male pattern baldness and risk of prostate cancer. However, without additional information about the relationship among hormone levels, baldness, and prostate cancer, the study was nonfunctional.
* Coherence of explanation: This guideline is similar to biological plausibility in that findings should not conflict with known facts.
* Experiments: Variables are well controlled. With experiments, fewer "by chance" factors influence the outcome.
* Analogies: Comparable situations exist and outcomes are similar. For example, exposure to a harmful agent during pregnancy (e.g., rubella) can result in birth defects. Scientists were able to use this knowledge analogically to show that thalidomide causes birth defects when pregnant women ingest the drug.
STUDY DESIGNS
Although most occupational health nurses are not likely to do epidemiological research at work, it is important to understand some basics of study design to read and understand research studies. Each study design has pros and cons, and not all studies are equal.
The quickest and easiest epidemiological study is correlation, or ecological, studying relationships between variables. This method is used when research funds are limited, or little is known about a situation. For instance, a community interested in gathering evidence to support curbing the number of liquor licenses will look at the number of crimes committed within 1 mile of liquor stores. Another study may investigate how many vehicle crashes occur within a half-mile of schools, to gain support for decreased school speed limits. It is difficult to draw conclusions from correlation studies because the researcher typically uses existing data. Origins of the data may be unknown. A host of other factors may not be considered (e.g., weather conditions when vehicle crashes occur). Correlation studies are often a starting point for further, more comprehensive research.
Case studies, descriptions of a single case, are used in epidemiology because they do not require many resources. The first AIDS cases were reported in case studies. Researchers provide as much descriptive information as possible and make recommendations for additional studies.
Cross-sectional studies attempt to gather information about a population at a specific time (i.e., a "snapshot"). With cross-sectional studies it is impossible to establish cause-and-effect relationships. However, this method of data collection can be used to describe the magnitude of a problem. This approach is not appropriate for studying rare diseases.
Case-control studies are relatively quick, easy, and cost-effective, and are therefore appropriate for studies of rare diseases. In the case-control method, researchers identify cases, then find appropriate "controls" (without the disease). The groups are compared to determine essential differences. The challenge for this method is selecting appropriate controls. Typically, cases and controls are matched by age, gender, socioeconomic status, race or ethnic group, and other known factors. The researcher may be familiar with theories about risk factors associated with the disease (from literature reviews or previous studies), and attempts to measure those risk factors in the subjects. Case-control studies provide an odds ratio, but no direct estimate of the risk of contracting the disease, given the exposure.
Cohort studies are large studies that randomly select subjects from a whole population. Cohort studies provide stronger evidence of an association between an exposure (risk factor) and a disease (or death). The findings are generalizable to the target population. However, cohort studies are expensive and take a long time to conduct. An increased likelihood of losing subjects exists because they move away or lose interest in participating. Cohort studies do provide an estimation of risk for contracting the disease (relative risk) based on exposure to the risk factor.
Intervention studies are carefully controlled experiments that do something to one group of subjects and nothing to another group of subjects. Intervention studies are considered the gold standard because the research is well controlled. Typically, subjects are randomly assigned to an intervention or control group. Drug studies fall into this category. Intervention studies are costly and time intensive. As with cohort studies, participants may drop out.
STUDY ANALYSES
Epidemiology studies use a 2 × 2 table for analysis (Table 3). A 2 × 2 table is used to calculate odds ratio and relative risk. The odds ratio describes the probability of exposure among cases compared with the probability of exposure among controls (Last, 2001). An odds ratio of 1.0 means that the probability of exposure is the same for cases and controls (Friis & Sellers, 1999). The relative risk is the ratio of the risk of getting the disease among the exposed to the risk of getting the disease among the unexposed (Last, 2001). A relative risk of 1.0 means that the chance of getting the disease is equal among the exposed and the unexposed (Friis & Sellers, 1999).
To illustrate how the odds ratio and relative risk are used and interpreted, an example of meter readers with stubbed toes is used (Table 4). In this study, meter readers who stubbed their toes were nearly 144 times more likely to be wearing flip-flops at the time of injury compared with meter readers who did not stub their toes (odds ratio interpretation). Meter readers who were wearing flipflops were more than 65 times more likely to stub their toes than those not wearing flip-flops (relative risk interpretation). Use of odds ratio or relative risk is dictated by the study design. Odds ratio is used for case-control studies, and relative risk is used for cohort studies.
BIAS IN EPIDEMIOLOGY STUDIES
The quality of a study determines its usefulness and applicability to real-life settings. Every study has limitations, but researchers attempt to minimize bias through careful study design. To objectively read and apply research findings, occupational health nurses should be aware of some common sources of bias.
Selection bias occurs when every individual in a population under study does not have an equal opportunity to participate in the study. If participants are contacted by phone, individuals without phones do not have an opportunity to participate. If work surveys are handed out during first shift, those working third shift will not be represented. Selection bias is more common in case-control studies than in cohort studies.
Information bias comes into play when individuals are asked to recall information (e.g., diet or health histories), when an interviewer misinterprets observations or answers, or when subjects are lost to follow-up (i.e., information is lost about when they contracted the disease).
Misclassification occurs when a clear, defined standard for placing a subject into one category or another is missing. For example, a disease is misdiagnosed or exposure or non-exposure is not clearly documented. Having a dermatologist examine the head can minimize misclassification of baldness, rather than asking the participant if he is bald. How would flexibility or strength be measured and classified? Does a "fever" mean an oral temperature over 99.60F, or does the temperature have to be above 1010F to be classified as a fever? Does fever mean the same thing to a lay person and a health professional? What is a "near miss" traffic accident? Failure to clarify such terms can lead to misclassification.
Confounding is a distortion of risk factor effect on the outcome, caused by another factor that is associated with both the risk factor and the disease (Last, 2001). For example, the risk of heart disease increases with age, so if the researcher is studying cholesterol as a risk factor for heart disease, age must be controlled, because age can also affect cholesterol levels.
IMPLICATIONS FOR OCCUPATIONAL HEALTH NURSES
In occupational and environmental health, epidemiology tends to focus on injuries and exposures. For occupational health nurses who want to examine situations in their workplaces or communities, but choose not to engage in controlled research, it is essential to standardize incidence and prevalence by using rates. Injury statistics can be compared in several ways. The BLS uses 2,000 worker-hours to equal 1 employee (i.e., full-time equivalent). By converting payroll hours to full-time equivalents, individual companies can compare their results with BLS standards. When companies use only the number of employees in the human resources database, overtime hours and part-time employees are not taken into account. In some companies, employees work enough overtime to equal more full-time equivalents than the official number of employees.
It is important to determine what is in the denominator when trying to standardize rates. For instance, statistics about the dangers of flying and driving abound. How is risk calculated? Is the focus on the number of deaths per miles traveled, or per trips taken? Should trips taken in a specific make or model of vehicle be considered? Does it matter if the trips are during winter or summer, or in the northeast rather than the southwest? The media may report only what they understand; the details may be lost. Occupational health nurses may find value in reading original studies to understand design and bias.
Occasionally, the media will report a cancer cluster in a community. When a state's epidemiologist concludes that residents in the area are at no significant increased risk for the disease, residents may be outraged at a perceived "whitewash" of the problem. However, environmental exposures are extremely difficult to validate. Correlation does not prove causation, and every study design has limitations. As such, it is challenging to demonstrate how an exposure to one hazard in the community plays a more substantial role in causing a disease than any other exposures.
SUMMARY
Epidemiology speaks to the risk of disease, injury, or death. The purpose of epidemiology is to predict disease (or injury or death) based on risk factors or exposures. Results are communicated with caution, to avoid unjustly labeling a risk factor as a "cause" of disease. Although an individual may be at increased risk for a given disease, there is often no way to predict if or when that individual will actually contract the disease. Rather, epidemiology is a tool to assess populations and improve the health of the general public.
Occupational and environmental health nursing, perhaps more than any other nursing specialty, uses the principles of epidemiology on a regular basis. Occupational health nurses translate numbers into meaningful information for management and employees. Nurses working in occupational settings should consider epidemiology part of the foundation of their practice. Whether the occupational health concern is the prevalence of depression, the incidence of specific "struck-by" injuries, or the rate of noise-induced hearing loss among workers, occupational health nurses will make complex phenomena more comprehensible by understanding the principles of epidemiology and putting them to work.
REFERENCES
Aschengrau, A., & Seage, G. R. III. (2003). Essentials of epidemiology in public health. Sudbury, MA: Jones and Bartlett.
Bureau of Labor Statistics. (2006). Table SNR03. Retrieved June 21, 2006, from www.bls.gov/iif/oshwc/osh/os/ostbl477.txt
Friis, R. H., & Sellers, T. A. (1999). Epidemiology for public health practice (2nd ed.). Gaithersburg, MD: Aspen.
Hennekens, C. H., & Buring, J. E. (1987). In S. L. Mayrent (Ed.), Epidemiology in medicine. Philadelphia: Lippincott Williams & Wilkins.
Hill, A. B. (1965, May). The environment and disease: Association or causation? Proceedings of the Royal Society of Medicine, 58, 295-300.
Last, J. M. (2001). A dictionary of epidemiology (4th ed.). Oxford: Oxford University Press.
Rydstrom, C, & Tornberg, S. (2006). Cervical cancer incidence and mortality in the best and worst of worlds. Scandinavian Journal of Public Health, 34(3), 295-303.
Szldo, M., & Nieto, F. J. (2000). Epidemiology: Beyond the basics. Gaithersburg, MD: Aspen.
by Eileen Lukes, MS, RN, COHN-S, CCM, FAAOHN
ABOUT THE AUTHOR
Ms. Lukes is a PhD (epidemiology) candidate at the University of Arizona, Tucson, AZ.
Copyright SLACK INCORPORATED Jan 2007