Chapter 3. Patterns of Occurrence

• The distribution of a disease within a population can be characterized by three basic questions: Who develops the disease? Where does the disease occur? When does the disease occur?
• A rapid and dramatic increase in the occurrence of a disease is referred to as an epidemic.
• Studies of migrant populations can be used to help distinguish whether a disease is more environmentally or genetically determined.

The patient resided with her husband and two young children in an apartment building. Tuberculin skin tests were administered to each of the family members at the time of the patient’s initial diagnosis, and results were positive for the patient’s husband and 3-year-old daughter. Although no evidence of clinically active tuberculosis was found in either the spouse or daughter, preventive therapy was administered to all three family members. Skin testing of 54 other residents of the apartment building revealed one other infected adult, who lacked evidence of active disease and received preventive antibiotic therapy. None of the tuberculin skin tests administered to the patient’s contacts at work were positive.

Tuberculosis is caused by mycobacteria transmitted on small airborne particles that are created when an individual with pulmonary tuberculosis coughs or sneezes. Air currents circulate these particles throughout an entire room or building. When a susceptible person inhales these particles, tubercle bacilli may become established in the lungs and spread throughout the body. Usually the host’s immune system contains this initial infection within a short period of time. A small proportion (10%) of patients will develop active clinical illness months or years later, when the mycobacteria begin to replicate and cause symptoms.

As shown in Table 3–1, environmental as well as personal factors affect the likelihood of tuberculosis transmission. Each of the environmental features listed tends to increase the concentration of mycobacteria in the air. Transmission also is promoted by (1) characteristics of the infected individual that contribute to greater release of mycobacteria and (2) characteristics of the susceptible person that diminish the immune response.

Public health officials in the United States have established the elimination of tuberculosis in this country as a national priority. An effective plan for the control of tuberculosis requires that persons who are infectious with active tuberculosis be identified early, isolated from susceptible persons, and treated with adequate antibiotic therapy. Control strategies also include screening for the presence of asymptomatic infection within groups at high risk of tuberculosis, followed by antibiotic therapy to prevent the development of active disease. On the basis of epidemiologic data, a number of groups with an elevated risk of developing tuberculosis have been identified (Table 3–2).

The Patient Profile illustrates many important points about tuberculosis. Because the patient had recently emigrated from an area with an elevated prevalence of tuberculosis, she was a member of a high-risk group. The presence of symptoms, in conjunction with cavitary lung lesions and tubercle bacilli in the sputum, indicated that the patient was highly infectious. Although the infectious state may end after several weeks of appropriate antibiotic treatment, relapse may occur unless therapy is sustained for at least 6 months.

The emergence of antibiotic-resistant tuberculosis has complicated the clinical management of this disease. Drug-resistant strains of tubercle bacilli arise from spontaneous chromosomal mutations; different mutations affect susceptibility to different drugs. Under conditions in which therapy is inadequate, because too few drugs are prescribed, the dosages are inadequate, or patient adherence to the prescribed regimen is poor, resistance to multiple antibiotic agents can arise. Recent surveillance data in the United States indicate that about 1 in 13 patients with tuberculosis has a strain resistant to one of the leading antibiotics, isoniazid. Foreign-born persons with tuberculosis are about twice as likely as those born in the United States to have antibiotic-resistant strains of the organism. Even more vexing is the challenge of multidrug-resistant tuberculosis, in which the organism is resistant to two or more antibiotics. Multidrug-resistant tuberculosis is particularly difficult to treat, often requiring an extended duration of therapy and the use of alternative antibiotics that are expensive and have higher risks of adverse effects.

Because of the rise in incidence of multidrug-resistant forms of tuberculosis, treatment recommendations for the disease have been modified as follows:

• In vitro testing of isolated tubercle bacilli for drug resistance and reporting of these results to the health department
• The use of a four-antibiotic regimen for initial treatment of tuberculosis infections
• Direct observation of initial therapy by a health care provider.

Directly observed therapy, involving ingestion of antibiotics by the patient in the presence of the health care provider or another designated person, is intended to ensure adherence to the prescribed antibiotic regimen. Surveillance data in the United States indicate that over half of tuberculosis therapy is now administered totally through directly observed therapy, with another one fourth of patients receiving a combination of directly observed and self-administered treatment.

Once a patient is diagnosed with clinically active tuberculosis, the patient’s close personal contacts should be tested for tuberculosis. The husband and two children of the woman in the Patient Profile were considered close contacts. Because asymptomatic infection was demonstrated in the husband and one child, preventive therapy was administered. The other child showed no signs of infection. Guidelines for preventive therapy dictate that children who are close contacts and have a negative skin test should receive antibiotics until a skin test repeated 12 weeks later is also negative. Residents of the patient’s apartment building and her contacts at work also were considered to be at high risk and were investigated for infection. The single infected resident was treated with preventive antibiotics, in accordance with established guidelines.

Broadly speaking, epidemiologic work can be divided into two main categories:

• 1.Descriptive epidemiology, which includes activities related to characterizing the distribution of diseases within a population; and
• 2.Analytic epidemiology, which concerns activities related to identifying possible causes for the occurrence of diseases.

Both types of epidemiology are fundamental to the prevention and control of diseases and to the advancement of medical knowledge. Descriptive patterns of disease occurrence often lead to hypotheses about disease causation that are tested in analytic investigations. Analytic studies may yield findings that help to explain descriptive patterns and to improve surveillance efforts.

• 1. Who develops tuberculosis?
• 2.Where does tuberculosis occur?
• 3.When does tuberculosis occur?

Collectively, these three questions serve as the basis for a descriptive investigation of tuberculosis. Answers to these questions characterize the distribution of tuberculosis by person, place, and time. As shown schematically in Figure 3–1, these features are the standard dimensions used to track the occurrence of a disease.

Person

A basic tenet of epidemiology is that diseases do not occur at random. In other words, not all persons within a population are equally likely to develop a particular condition. Variation of occurrence in relation to personal characteristics may reflect differences in level of exposure to causal factors, susceptibility to the effects of causal factors, or both exposure and susceptibility.

Typically, the minimal set of personal characteristics examined with respect to disease occurrence includes age, race, and gender. Because such information is routinely collected on the affected persons (cases), as well as the unaffected population from which the cases develop, epidemiologists rely on these characteristics to a great extent. The use of other attributes of interest, such as level of education and income, marital status, and occupation, is contingent on the availability of data.

The number of reported cases of tuberculosis by age in the United States during 2002 is shown in Figure 3–2. It should be emphasized that these data are derived from information reported by physicians, laboratories, and other health care providers. Tuberculosis is one of over 60 infectious diseases that currently are designated as notifiable at the national level within the United States. The Centers for Disease Control and Prevention (CDC) define a notifiable disease as follows:

A disease for which regular, frequent, and timely information on individual cases is considered necessary for the prevention and control of the disease.

The compulsory collection of information on selected infectious diseases was authorized at the national level in the United States and a number of other countries in the late 1800s. The list of nationally notifiable diseases is revised periodically in response to the emergence of pathogens that pose a threat to the health of the public. For example, the list of notifiable diseases in the United States was expanded to include West Nile encephalitis in 2002 and severe acute respiratory syndrome (SARS) in 2003.

Reporting of notifiable diseases in the United States typically begins with a clinician forwarding basic information on a newly diagnosed patient to the designated local or state health department. On a weekly basis, state and territorial officials transmit information about individual or aggregated cases of nationally notifiable diseases to the CDC. These reports follow a standard format, including information on the age, sex, and race of the patient, and date of occurrence of reported cases.

Although reporting of notifiable diseases is mandatory, and sanctions can be enforced for noncompliance, these sanctions are rarely applied. As a consequence, reporting is often incomplete, with wide-ranging estimates of completeness for various notifiable diseases. A number of factors probably affect the likelihood that a notifiable disease will be reported:

• 1. The clinical severity of the condition
• 2. Whether the affected individual consults a physician
• 3. The type of physician consulted (eg, private vs public provider, generalist vs specialist)
• 4. Any social stigma associated with the condition
• 5. Level of interest in the condition among clinicians
• 6. The physician’s knowledge of reporting requirements
• 7. Existence of an adequate definition of the condition for surveillance purposes
• 8. Availability and utilization of appropriate diagnostic laboratories
• 9. Availability of effective disease control measures
• 10. Interests and priorities of local and state health officials

For a disease such as tuberculosis, in which clinical manifestations can be serious, clear diagnostic criteria and effective treatments are available, and the risk of interpersonal transmission is high, complete reporting of diagnosed cases obviously is crucial. On the other hand, interpretation of reported numbers of cases must include the possibility of incomplete notification.

Returning to the data presented in Figure 3–2, and recognizing the possibility of incomplete reporting, it appears that the age group with the highest risk of tuberculosis is 25–44 years. This conclusion is incorrect, however, because it fails to take into account the varying sizes of the source populations across the different age groups. There were over 85 million persons in the 25- to 44-year-old age group, as compared with under 67 million persons in the 45- to 64-year-old category, and under 36 million persons in the group aged 65 years or older. By calculating incidence rates, it is possible to compensate for disparities in sizes of the source populations. As illustrated in Figure 3–3, the incidence of tuberculosis rises with age, reaching a level of 8.8 cases per 100,000 persons among those 65 years of age or older. The incidence among persons in the oldest age group is over 40% higher than that for the 25- to 44-year-old age group.

A number of factors contribute to the nonrandom relationship between incidence of tuberculosis and age.

• 1. The long latent period between infection and development of clinical symptoms means that the ages at detection of illness are expected to be skewed toward later life.
• 2. Because elderly individuals lived through time periods when the disease was more common, they are more likely to have been infected than younger persons (birth cohort effect).
• 3. Older persons are more likely to have other illnesses (eg, cancer, diabetes mellitus) that may make them more susceptible to tuberculosis.
• 4. The decline in immune function associated with the normal aging process may increase susceptibility.
• 5. Elderly persons are more likely to live in closed communal settings that are conducive to the spread of tuberculosis.

An equally striking nonrandom pattern of occurrence is seen when incidence is examined as a function of race or ethnicity (Figure 3–4). The highest incidence rate of tuberculosis in the United States is found among Asians and Pacific Islanders; it is more than 18 times greater than the incidence rate for white non-Hispanics. Foreign-born persons account for about 95% of tuberculosis cases among Asians and Pacific Islanders in the United States. As exemplified by the subject of the Patient Profile, many of these individuals acquire the infection in the high-risk country of origin, but do not develop symptomatic disease until they arrive in the United States. About three out of four tuberculosis cases among Hispanics in the United States also occur in foreign-born persons. Overall, foreign-born persons account for slightly more than half of all tuberculosis cases in the United States.

The high incidence rates of tuberculosis within certain minority groups in the United States reflect the influences of other risk factors. Tuberculosis is a disease that is associated with socioeconomic disadvantage. The combination of crowded housing, poor nutrition, inadequate access to preventive and therapeutic medical services, alcoholism, and injecting drug use, as well as any predisposing medical conditions, contributes to the high risk of tuberculosis among the poor. Since black and Native American/Alaskan Native populations in the United States have comparatively large percentages of disadvantaged persons, the incidence of tuberculosis in these communities is elevated.

The distribution of tuberculosis by gender is shown in Figure 3–5. The incidence of tuberculosis is almost 60% higher among males than among females. The higher occurrence of tuberculosis among males probably is related to differences in certain high-risk behaviors (eg, heavy alcohol consumption) as well as predisposing diseases (eg, AIDS, silicosis) between males and females.

Place

Variation in the place of occurrence of a disease can be evaluated at the national level (eg, across countries), at the regional level (eg, across states), or at the local level (eg, across communities). Certain countries, particularly those in the nonindustrialized parts of the world, have comparatively high rates of tuberculosis. The estimated incidence rates of this disease across various parts of the world during 2001 are shown in Figure 3–6.

Worldwide, it is estimated that over 8 million people develop tuberculosis each year. This number is an estimate because the actual reporting of numbers of new cases of tuberculosis as well as the sizes of the source populations are incomplete in many nonindustrialized countries. Accordingly, this information must be interpreted with caution. Nevertheless, 95% of all new tuberculosis infections are thought to occur in the developing world.

The region of the world where the incidence rate of tuberculosis is rising most rapidly is sub-Saharan Africa, which surpassed the previous leader, Southeast Asia, in 1995. The high rates of tuberculosis in nonindustrialized nations are attributable to poverty, malnutrition, crowded living conditions, inadequate preventive and therapeutic programs, and, particularly in sub-Saharan Africa, the high prevalence of infection with HIV. A growing proportion of tuberculosis occurrence worldwide is associated with HIV.

Even within an industrialized country such as the United States, variation in the incidence of tuberculosis is observed (Figure 3–7). The highest rates occur in urban areas, for example, San Francisco and New York (13 cases per 100,000 person-years). Comparatively low rates are found in rural states of the Midwestern and Western regions of the United States, such as North Dakota and Wyoming (each with less than 1 case per 100,000 person-years). These geographic patterns reflect differences in the underlying demographic characteristics of the various populations, including factors such as racial/ethnic composition and representation of immigrants from developing countries. In addition, the geographic distribution of other risk factors—poverty, malnutrition, crowded living conditions, substance abuse, and infection with HIV—probably influences the pattern of tuberculosis occurrence.

Time

The overall annual incidence of tuberculosis between 1980 and 2002 in the United States is depicted in Figure 3–8. During the first part of the 1980s, a consistent downward trend was observed, continuing a pattern that began many decades earlier. Between 1984 and 1989, the incidence of this disease remained fairly constant, with a rising incidence between 1989 and 1992. After 1992, the incidence rate began to fall again with a decline similar to that observed in the early 1980s. Another way to visualize this pattern is to compare the percentage change in incidence between the first and last years of successive 3-year time intervals (Figure 3–9). Between 1982 and 1984, the incidence of tuberculosis fell by 15%, with no change between 1985 and 1987. Between 1988 and 1990, the incidence increased by 13%, with a reversal to a 6% decrease between 1991 and 1993. Even greater decreases of 15% and 14% were observed in the next two time periods, respectively. A further reduction, albeit at a lower level (10%), occurred between 2000 and 2002.

The rise of tuberculosis occurrence in the United States during the late 1980s, followed by the subsequent fall in occurrence is a remarkable public health story. A number of factors probably contributed to the rise of this disease, including among others:

• 1. The emergence of a highly susceptible population of persons infected with HIV
• 2. Lapses in infection control practices in institutional settings
• 3. Increases in the number of immigrants arriving from countries with high rates of tuberculosis occurrence
• 4. The emergence of strains of tuberculosis resistant to multiple antibiotics
• 5. Failure of tuberculosis control programs to ensure that persons with active disease complete an adequate course of antibiotic therapy.

In response to the rise in the occurrence of tuberculosis in the United States, a massive public health effort was mounted to reverse this trend. The federal government appropriated substantial increases in funds to fight this disease. With the additional resources, public health agencies were able to expand programs for tuberculosis control. The essential components of these control efforts include:

• 1. Prompt identification of persons with active disease
• 2. Initiation of appropriate treatment for persons with active disease
• 3. Assurance that therapy is completed for persons with active disease
• 4. Tracing of the contacts of persons with active disease
• 5. Testing for infection among contacts
• 6. Completion of preventive therapy in eligible contacts.

The effectiveness of public health interventions was suggested by the progressive decline in incidence rates for tuberculosis after the interventions were begun. The argument that measures to control tuberculosis were responsible for the observed decline in incidence rates is strengthened by additional evidence. The rates of reduction were most substantial in areas that had the most successful control programs (as indicated by percentages of patients with elimination of organisms from sputum, completion of therapy for active cases, and tracing of contacts of active cases). Also, the proportion of organisms resistant to multiple antibiotics declined, with the greatest drop in areas with the highest rates of therapy completion. The decline in incidence rates of tuberculosis between 1992 and 2002 occurred across all age groups (Figure 3–10).

The time lag or latent period between exposure to a risk factor and diagnosis of a disease can be as short as a few hours (eg, staphylococcal food poisoning) or as long as decades (eg, infection to clinically active tuberculosis). Obviously, the greater the time between the occurrence of an initiating event and the recognition of disease, the more difficult it may be to establish the linkage between risk factor and disease occurrence. This task is made even more challenging if the risk factor is a weak determinant of the disease or if multiple different risk factors are involved.

As noted earlier, the incidence rate of tuberculosis in the United States ended a long-term continuous decline in 1984 (Figure 3–12). The solid line in Figure 3–12 depicts the number of tuberculosis cases that would have been expected if the pre-1984 rate of decline had continued. The squares in Figure 3–12 depict the actual numbers of patients with tuberculosis observed each year through 2002. It can be seen that the disparity between the actual and predicted numbers of cases is greatest in the early 1990s. After 1994, the slope of decline in the annual number of patients with tuberculosis appears similar to that prior to 1984. The rise in incidence of tuberculosis between 1984 and 1992 does not represent an epidemic in the conventional sense of a rapid rise in incidence. However, it does indicate a departure from the prior downward trend, and therefore represented a threat to the health of the public.

To develop hypotheses about possible causes of disease occurrence, the presence of a suspected risk factor can be measured in different populations and compared with the incidence of a particular disease. This type of comparison is referred to as an ecologic study because the analysis is at the level of an entire population rather than at the level of individual persons. Another name for this type of investigation is correlation study, because it seeks to determine the extent to which two characteristics (risk factor and disease occurrence) are related.

An example of ecologic data is shown in Figure 3–13. In this graph, the incidence rates of acquired immune deficiency syndrome (AIDS) in 15 states of the United States during 1997 are compared with corresponding incidence rates of tuberculosis for that same year. The states included in this analysis (Arkansas, California, Georgia, Idaho, Michigan, Mississippi, Nebraska, New Hampshire, North Carolina, Ohio, Oregon, Vermont, West Virginia, Wisconsin, and Wyoming) were selected because they represent diverse geographic areas with varying demographic characteristics.

In general, the states that had a high incidence of AIDS also had a high incidence of tuberculosis (eg, California and Georgia). At the other extreme, states with a low incidence of AIDS also tended to have a low incidence of tuberculosis (eg, Idaho, New Hampshire, and Wyoming). As a general rule, the incidence of AIDS was about two times greater than the corresponding incidence of tuberculosis. A few exceptions occurred, such as Arkansas, in which the incidence of AIDS (9.6 cases per 100,000 person-years) was only about 35% greater than the incidence of tuberculosis (7.1 cases per 100,000 person-years). An exception in the opposite direction was Wyoming, in which the incidence of AIDS (3.3 cases per 100,000 person-years) was eight times greater than the incidence of tuberculosis (0.4 cases per 100,000 person-years).

To assess the strength of the relationship between AIDS and tuberculosis incidence, a correlation analysis was performed (see Dawson and Trapp, 2004, Basic and Clinical Biostatistics). The correlation coefficient was 0.91, which indicates that the incidence rate (IR) of AIDS and incidence rate of tuberculosis in the study are strongly and positively related. (If AIDS and tuberculosis incidence rates were perfectly correlated, the correlation coefficient would be 1. If there was no correlation between these two incidence rates, the correlation coefficient would be zero.) The coefficient of determination, the square of the correlation coefficient, was 0.83. This means that over four fifths of the variability in the incidence of tuberculosis could be accounted for by knowing the incidence of AIDS in these 15 states.

A linear regression analysis of these data (see Dawson and Trapp, 2004) yielded the following equation:

The graph of this regression line is depicted in Figure 3–14. In this analysis, the effect of AIDS incidence in determining the incidence of tuberculosis is highly statistically significant. In other words, it is very unlikely that the observed relationship between the incidence rates of AIDS and tuberculosis occurred by chance alone. From the regression equation it can also be seen that in the absence of AIDS (AIDS incidence = 0 cases per 100,000 person-years), the expected incidence rate of tuberculosis is –0.8 cases per 100,000 person-years. Because it is impossible to have an incidence rate less than zero, we conclude that in these states, the tuberculosis incidence rate is negligible in the absence of AIDS. For every increase of one case per 100,000 person-years in the incidence of AIDS, the incidence of tuberculosis is expected to increase by 0.57 cases per 100,000 person-years.

The relationship depicted in Figure 3–14 is very striking and suggests that the occurrence of AIDS may influence the development of tuberculosis. This type of correlation analysis, however, is best viewed as a hypothesis-generating study, which means that it can be used to help formulate a hypothesis about the link between these two diseases, but it cannot establish a causal relationship between them. A correlation between incidence of AIDS and tuberculosis could occur for reasons other than a cause-and-effect relationship. For example, risk factors for both diseases (eg, injecting drug use) might be the true reason for the apparent association between AIDS and tuberculosis.

Studies that are designed to test the likelihood of a cause-and-effect relationship between a risk factor and a disease are termed hypothesis-testing investigations. The two approaches most commonly employed to test associations between risk factors and diseases are cohort and case–control studies. These research designs are described in Chapters 8: Cohort Studies and 9: Case–Control Studies, respectively. The effects of related variables, such as injecting drug use, can be considered in the design and analysis of these studies.

An important limitation on the epidemiologist’s ability to infer a causal explanation from a correlation study is the ecologic fallacy. This problem may occur when a suspected risk factor and disease occurrence are associated at the population level but not at the individual subject level. In other words, populations may have high incidence rates of AIDS (risk factor) and of tuberculosis (disease occurrence) without the same persons being affected by both conditions. This type of ecologic fallacy can be avoided only by making observations of risk factor and disease status on individual subjects. Methods of analytic epidemiology, such as cohort and case–control studies, involve observations on individuals, and thus are not subject to the hazards of ecologic reasoning. The link between the occurrence of AIDS and tuberculosis is now well established through hypothesis-testing studies, so it can be concluded that the correlation seen in Figure 3–14 is not attributable to an ecologic fallacy.

If environmental exposures early in life are critical, then the rate of occurrence may not be reduced among the migrants themselves (who were exposed prior to their departure), but should be diminished among their offspring born in the new location. A dramatic change in disease incidence within a single generation could not be explained on the basis of genetic changes. Approaches to the study of diseases that are thought to have genetic predispositions are presented in Chapter 11: Epidemiologic Studies of Genetics.

A number of studies have indicated that a progressive decline over time in the incidence of tuberculosis occurs among persons who migrate from high-risk areas (eg, Asia) to low-risk areas (eg, the United States and Western Europe). The basic pattern of change in incidence is shown in Figure 3–16. The incidence of tuberculosis is highest at the time of migration and falls rapidly in the next few years. The decline continues for many years, with smaller increments of change over time. The incidence among migrants does not fall to the level of the general population, however, presumably because of latent infections acquired prior to migration. Other factors that may contribute to the persistence of an elevated incidence of tuberculosis among migrants include the following:

• 1. Residence in migrant communities, thus maintaining a comparatively high rate of disease transmission
• 2. Crowded housing conditions
• 3. Poor nutritional status
• 4. Inadequate access to preventive or therapeutic medical services
• 5. Noncompliance with therapy
• 6. Presence of tuberculosis that is resistant to conventional antibiotics
• 7. Reinfection on return visits to the country of origin.

The overall decline in incidence of tuberculosis among migrants from high- to low-risk countries, however, provides strong circumstantial evidence about the importance of environmental determinants of this disease.

In this chapter, the basic approaches to descriptive epidemiology are presented, with a focus on the patterns of occurrence of tuberculosis. Description in epidemiology begins with the assumption that diseases do not occur at random. Typically, three standard questions are posed to characterize the nonrandom distribution of a disease:

• 1.Who gets the disease?
• 2.Where does the disease occur?
• 3.When does the disease occur?

These questions concern the elements of person, place, and time, respectively.

At a minimum, the personal attributes examined in relation to disease occurrence are the distributions by age, race, and sex. The incidence of tuberculosis in the United States increases with advancing age. The racial and ethnic groups with the highest occurrence of this disease are Asians and Pacific Islanders, blacks, Hispanics, Native Americans, and Alaskan Natives. In addition, males have higher incidence rates for tuberculosis than females.

The place of occurrence of a disease may be studied at the international, regional, or local level. Tuberculosis occurs with great excess in nonindustrialized countries. Even within an industrialized country such as the United States, substantial regional and local variation in incidence is reported.

Temporal patterns can be examined across years, months, or days, depending on the time course of the disease in question. For tuberculosis, a progressive decline in incidence over time was observed in the United States until 1984, with a leveling off through 1989, followed by a rise in incidence through 1992, and then a return to falling incidence rates. The resurgence of tuberculosis in the late 1980s in the United States was attributed to a variety of factors. The contributing factors included lapses in control efforts, increased immigration of persons from countries with high incidence rates of disease, the emergence of strains resistant to multiple antibiotics, and the rising occurrence of a predisposing condition—AIDS.

The concept of an epidemic as a rapid and dramatic increase in the incidence of a disease was introduced, using as an example the greater-than-expected occurrence of tuberculosis in the late 1980s. The use of ecologic (or correlation) studies was illustrated by a comparison of incidence rates of AIDS and tuberculosis in selected states. In an ecologic study, the overall amount of a risk factor (eg, AIDS) is related to the occurrence of a disease (eg, tuberculosis) across different populations. This type of correlation can be useful for generating hypotheses, but not for testing causal relationships. The influence of a background variable that is related to both the presumed risk factor and the outcome of interest can limit the utility of a correlation analysis. Furthermore, the ecologic fallacy can suggest a misleading conclusion when the risk factor and disease are related at the population level but not within particular individuals.

Finally, to distinguish genetic from environmental causes of disease, we discussed the use of studies of disease occurrence in relation to migration patterns. As applied to tuberculosis, persons who migrate from high-risk areas (eg, Asia) to low-risk areas (eg, the United States and Western Europe) experience a progressive decline in incidence over time. This pattern strongly indicates that the primary influences on the occurrence of tuberculosis are environmental.