Historical Perspective
First-Generation Studies
The earliest formal first part of the 20th century. They were generally of limited scale, relied on institutional records, and used small groups of informants for their data. These were “convenience” studies in which, instead of initiating the surveys themselves, epidemiologists assembled health data from those persons who had already received treatment for a medical problem or had committed suicide. A notable exception was Brugger’s study in Thuringia, in which he attempted to use census tracts to identify mental illness in a defined population. Ferris and Dunham’s relatively large pre–World War II study examined the geographic distribution of patients with mental disorders in mental hospitals in the Chicago area. They found that manic-depressive illness was distributed equally throughout the geographic area, whereas schizophrenia clustered in the lower socioeconomic areas.
Second-Generation Studies: The Stirling County & Midtown Manhattan Studies
In comparison to pre–World War II investigations, the studies that followed World War II took advantage of the considerable health information gathered on the military forces during the war. This was the beginning of the “community survey” era of epidemiology. Postwar studies--such as the Stirling County (Nova Scotia) Study and the Midtown Manhattan Study (which are discussed in this chapter) and the Baltimore Study of mental illness in an urban population--were second-generation studies that attempted to determine prevalence rates of mental illness by interviewing community residents using nonpsychiatrist clinical interviewers. The postwar studies examined general health as well as psychiatric disorders and tended to gather and interpret rates of symptom presentation in groups rather than assess the presence of discrete cases. The gathering of often-isolated symptoms, or the finding of psychopathology or emotional illness using the data collection system of the World War II era, was not helpful to health planners or policymakers.
Contributing to the ascension of morbidity associated with chronic disease (including that of mental disorders) was more important than was the mortality and morbidity associated with acute, generally infectious, disorders. Difficulty with case identification in the community continued to preclude the determination of prevalence rates for specific clinical disorders.
The initial paradigms for these newer studies were often quite different. The Stirling County Study attempted to determine rates for qualitatively different disorders as well as for overall impairment. The Midtown Manhattan Study assumed that mental disorders were on a continuum and--reflecting the thinking at that time (that mental illness differed in degree and not kind)--that all clinical manifestations of illness could be evaluated in terms of functional impairment. The overall prevalence of psychiatric impairment from both of these studies was approximately 20%. Dohrenwend and colleagues reviewed most of the major European studies and, despite some individual study differences, found a prevalence rate of psychiatric disorder of 16–25%, quite similar to the North American studies. Leighton and colleagues demonstrated in Stirling County that the mental illness of the individual could be influenced, for benefit or detriment, by the attributes of the community, thus ushering in an emphasis, during the late 1950s and early 1960s, on social psychiatry. During this period, Hollingshead and Redlich studied a population of individuals receiving psychiatric treatment in the New Haven, Connecticut, area and found that socioeconomic class was an important correlate of both the presence of a psychiatric disorder and the kind of treatment provided.
Third-Generation Studies
Third-generation epidemiologic studies were based on more advanced epidemiologic and statistical techniques and on a move toward scientific or evidence-based medicine. These studies began with the important development of operational criteria for mental disorders. Newer methodological techniques helped address the increasing need for more exact rates of specific disorders for specific persons in specific settings. Indeed, effective treatment has been shown to be related directly to accurate and thus specific assessment and diagnosis. Similarly, appropriate mental health policy planning for the unique health needs of persons with various psychiatric disorders depends greatly on an accurate and precise definition of boundaries between disorders. Further, research into the etiology and thus effective treatment and, it is hoped, eventual prevention of psychiatric disorders must derive from the specificity of operational criteria. Otherwise, the blurring that has occurred between symptom patterns can lead only to similar blurring in the assessment of treatment and prevention effectiveness.
The Present State Examination (PSE), designed as an instrument that demonstrated diagnostic reliability when used cross-nationally, was used by the World Health Organization in the early 1970s in its study of schizophrenia in a number of countries. About this same time, Feighner and colleagues published diagnostic criteria for 14 major psychiatric illnesses. These criteria both provided a framework for the comparison of data gathered by different research groups and facilitated communication among investigators. This work was followed by the development of the Research Diagnostic Criteria by Spitzer and his colleagues, which led to the American Psychiatric Association’s DSM-III. DSM-III was a clear departure from its predecessors in that the specificity and boundaries that demarcated a foundation of this evolving instrument led to specific case and non-case determinations. Many etiologic assumptions in DSM-II that were not demonstrated by empirical research were abandoned in DSM-III. Although interview instruments have been derived from each of these criteria sets, the Diagnostic Interview Schedule (DIS), associated with the development of DSM-III, was the first instrument designed for use by (trained) lay interviewers in community-based epidemiologic studies, a decision based on cost-benefit considerations. The DIS became the preferred instrument for use in most large epidemiologic studies during the 1980s, such as the Epidemiologic Catchment Area (ECA) study (see next section).
The NIMH Environmental Catchment Area Study
The National Institute of Mental Health’s ECA study was the most comprehensive and sophisticated epidemiologic study accomplished in its time in the United States. When it was undertaken between 1980 and 1984, its purpose was to provide the best estimates, for the United States, of the prevalence of alcohol and drug abuse and other mental disorders, based on a formalized criteria set (DSM-III) rather than global impairment. Unlike previous studies, this investigation included not only data from institutional and community samples but also longitudinal data and information on disease severity. ECA investigators explored the specific demographic, biological, psychosocial, and environmental factors that might influence the presence and the severity of a mental disorder (ie, the biopsychosocial model). The study not only allowed investigators to follow-up on possible clinical change but also assessed the service utilization of both mental health and general health services. The ECA study has assisted greatly in the planning for future health care service needs--including physical resources, financing, personnel, and educational requirements. The ECA study also confirmed the capability of DSM-III criteria to discriminate among mental disorders and generally helped sharpen the nosology of mental illness.
Although the methodology of the ECA study was a great improvement on previous work, the use of DSM-III as the basis for case identification tended to emphasize reliability rather than validity. DSM-III diagnostic criteria, in contrast to DSM-II criteria, were intended to enhance diagnostic reliability, which, although necessary, is insufficient to establish diagnostic validity. Further, because lay interviewers were utilized in the ECA study, in-depth, qualitative data could not be collected.
Compared to previous surveys, the ECA study found lower rates of virtually all disorders, except for severe cognitive impairment in younger age groups. Women exhibited higher rates of mental disorders than did men, although there were important differences in the rates for specific disorders. Men had higher rates of substance abuse and antisocial personality disorder, and women had significantly higher rates for anxiety-based, affective, and somatization disorders. Men and women exhibited similar rates for schizophrenia and manic episodes. The ECA study showed that individuals with comorbid conditions were more likely to receive treatment than were those with a single disorder; yet, less than one third of persons with mental illness, a substance abuse disorder, or both received any treatment. An important methodological finding was that, in comparison to international studies (after adjusting for differences in diagnostic categories and time frames), disease rates based on the DIS were found to be essentially compatible with previous epidemiologic studies based on the PSE.
The National Comorbidity Study
The National Comorbidity Study (NCS) was the first attempt in the United States to estimate the prevalence of specific psychiatric disorders, with and without comorbid substance abuse, in a national population sample. The NCS was designed to further the findings of the ECA study, but in contrast to the ECA study (which was drawn from local and institutional groups), the NCS had a national focus. The NCS sought risk factors as well as prevalence and incidence rates, in contrast to the ECA study, which focused only on the latter. Table 5-2 provides comparison data from the two studies. With its national focus, the NCS made possible regional comparisons, including rural and urban differences, and it was possible to establish, on a national basis, more precise investigations into unmet mental illness treatment needs. Furthermore, the NCS was referenced to the DSM-IIIR rather than DSM-III and also contained some questions that would allow comparison to the future DSM-IV and the International Classification of Diseases, 10th edition (ICD-10). The NCS found a higher prevalence of mental disorders in the U.S. population, and this prevalence was aggregated in approximately one sixth of the population (ie, in individuals who had three or more comorbid disorders).
In the NCS, risk profiles were constructed for depression alone and when found in association with other psychiatric disorders: 4.9% of persons studied were found to have current major depression (ie, within the last 30 days). The lifetime prevalence of depression was 17.1%. Multiple studies, often using quite different instruments, have found reasonably uniform results on depression: a point prevalence of about 3%, a 6-month to 1-year prevalence of about 6%, and a lifetime prevalence of approximately 16%.
Risk factors for both current and lifetime depression were as follows: being female; having a lower level of education; and being separated, widowed, or divorced. The rates for all time frames and the demographic distributions were higher than those found in the ECA study. The fact that a different method of case identification was used probably explains most of the difference in prevalence between the NCS and the ECA study. The sample from the NCS was younger, and younger persons are known to have a higher prevalence of major depression. The NCS also suggested that although “pure” depression may have a strong biogenetic contribution, comorbid depression may be more environmentally determined. Furthermore, as in the ECA study and other international investigations, more recent birth cohorts were found to be at increased risk for major depression.
Many explanations have been offered for the striking finding of high estimates of childhood depression and the unexpectedly low estimates of depression in the elderly: methodological limitations including the bias found in diagnostic instruments for the assessment of psychopathology in both children and the elderly, differential morbidity, faulty sampling, response-biased memory, institutionalization, and selective migration. Some or all of these explanations may play a role.
Continued investigation of the NCS data, building on the findings of the Medical Outcome Study (which showed that depressive symptoms themselves were a significant risk factor for other diseases--see next section), found that major and minor depression were not distinct entities but were actually on a continuum. Further, and also using the NCS data, the lifetime prevalence of major and minor depression associated with seasonal affective disorder (SAD) was found to be much lower (1%) than that found in previous studies. This was probably because the instrument used more accurately reflected DSM-IIIR criteria for SAD. Another study, using the NCS data, found a significant lifetime association between panic disorder and depression in patients who first present with panic disorder and a less powerful but statistically valid association for those who first present with depression. An investigation from Germany, using the revised version of the Composite International Diagnosis Interview on a community sample of adolescents and young adults, found that agoraphobia and panic disorder had “marked differences in symptomatology, course, and associated impairments” and were not necessarily linked, a finding at odds with some earlier studies. If confirmed, this study, which used a more sophisticated epidemiologic design than was available in much earlier studies, will demonstrate a more precise separation between several disorders previously considered to be closely related. This finding may lead to a more definitive basis for both prevention and treatment strategies.
More narrowly focused epidemiologic studies have contributed to increased understanding of psychiatric disorders associated with social conditions. Through a population survey, Breslau and colleagues demonstrated that posttraumatic stress disorder occurred in 9.2% of the population following exposure to trauma. Not only was this prevalence lower than that reported previously, but the most common trauma experienced was the unexpected death of a loved one, not the usually reported combat, rape, or other serious physical assault. Bassuk and colleagues investigated the prevalence of mental illness and substance abuse disorders among homeless and low-income-housed mothers, compared to the prevalence of these disorders among all women in the NCS, and found the prevalence of trauma-related disorders among poor women to be significantly higher than that found among women in the general population.
