Boland and Keller (Chapter 2) discuss the course and outcome of this disorder, describing the results of several large-scale longitudinal investigations that have monitored the course of depression over many years. Nezu, Nezu, McClure, and Zwick (Chapter 3) describe the most widely used interview-based and self-report measures of depression, and discuss important issues involved in the assessment of this disorder. Extending this discussion, Ingram and Siegle (Chapter 4) describe a number of methodological issues in the study of depressive disorders, and make several noteworthy recommendations concerning how research in this area might proceed most fruitfully. Klein, Durbin, Shankman, and Santiago (Chapter 5) then describe the nature of the relation between depression and various aspects of personality functioning. Finally, Johnson and Kizer (Chapter 6) discuss similarities and differences in the clinical phenomenology and psychosocial predictors of unipolar and bipolar depression.
Epidemiology of Depression
Ronald C. Kessler
The first modern North American general population epidemiological surveys that included information about depression were carried out in the late 1950s in the Midtown Manhattan Study (Srole, Langner, Michael, Opler, & Rennie, 1962) and the Stirling County Study (Leighton, Harding & Macklin, 1963). These early surveys used dimensional screening scales of nonspecific psychological distress to pinpoint respondents with likely mental disorders and then administered clinical interviews to these respondents. The outcome of primary interest was a global measure of mental disorder rather than individual diagnoses. No prevalence estimates of depression were reported. However, the screening scales in these studies included a number of items that assessed depressed mood and other symptoms that have subsequently come to be seen as part of the depressive syndrome. It is possible to make rough estimates about the prevalence and correlates of depressive disorders from these data (Murphy, Laird, Monson, Sobol, & Leighton, 2000).
In later surveys, variants on the screening scales used in the Midtown Manhattan and Stirling County studies were generally used without clinical follow-up. (See Link & Dohrenwend, 1980, for a review.) Scale scores were sometimes dichotomized in order to define “cases” of mental disorder based on some external standard of a clinically relevant cutpoint, although there was ongoing controversy about the appropriate decision rules for defining cases (Seiler, 1973). In order to resolve this controversy, structured diagnostic interviews appropriate for use in community surveys were developed in the late 1970s. The Diagnostic Interview Schedule (DIS; Robins, Helzer, Croughan, Williams, & Spitzer, 1981) was the first of these instruments. Dimensional screening scales continued to be widely used to screen for mental illness in primary care (Goldberg, 1972) and to assess symptom severity and treatment effectiveness among patients in treatment for mental disorders (Derogatis, 1977) even after the introduction of the DIS. However, psychiatric epidemiologists, influenced by the widely published results of the Epidemiologic Catchment Area Study (Robins & Regier, 1991), which was based on the DIS, abandoned the study of dimensional distress measures in favor of dichotomous caseness classifications in general population surveys. We now have had 2 decades of experience with community epidemiological surveys using fully structured diagnostic interviews like the DIS and the more recently developed CIDI (Robins et al., 1988), PRIME-MD (Spitzer et al., 1994), and MINI (Sheehan, Lecrubier,
Sheehan, Amorim, & Janavs, 1998). It is clear from this experience that fully structured diagnostic interviews, while very useful, are inadequate by themselves to provide the information needed by health policy planners on the magnitude of the problem of untreated seriousdepression. The reason for this is that the DSM and ICD criteria are so broad that close to half of the people in the general population receive one or more diagnoses on a lifetime basis (Kessler et al., 1994) and close to one-fifth at any one point in time (Kessler & Frank, 1997). With prevalences as high as these, the dichotomous caseness data provided in diagnostic interviews need to be supplemented with dimensional information on severity to be useful to health policy planners (Regier et al., 1998).
Unfortunately, the most recently available adult general population epidemiological data on the prevalence of major depression do not include dimensional severity measures.
This is an especially important omission in light of the suggestion by some commentators that the majority of community cases who meet criteria for major depression have fairly mild disorders (Regier, Narrow, Rupp, Rae, & Kaelber, 2000). The World Health Organization (WHO) is currently carrying out a massive worldwide epidemiological survey of mental disorders, known as the World Mental Health 2000 (WMH2000) Initiative, that aims to correct this problem by evaluating a wide range of mental disorders both with dichotomous diagnostic measures and with dimensional clinical severity measures (Kessler & Ustun, 2000). However, WMH2000 results will not be available for another 2 years.
The first section of this chapter presents a broad overview of the main findings in the literature regarding the descriptive epidemiology of major depression. The overview is brief because much of this literature has recently been reviewed elsewhere (Lewinsohn, Rohde, Seeley, Klein, & Gotlib, 2000; Merikangas, 2000; Blazer, 2000; Horwath & Weissman, 1995; Bland, 1997). The second section of the chapter addresses the issue of severity by reviewing available data on the consequences of depression as assessed in community surveys. The third section, finally, reviews epidemiological data on patterns of help seeking for depression.
DESCRIPTIVE EPIDEMIOLOGY
Point Prevalence
Community surveys that assess depression with symptom screening scales find that up to 20% of adults and up to 50% of children and adolescents report depressive symptoms during recall periods between 1 week and 6 months (Kessler, Avenevoli, & Merikangas, 2001). There is a U-shaped distribution of mean scores in these surveys in relation to age, with the highest scores found among the youngest and the oldest respondents and the lowest scores found among people in midlife (Kessler, Foster, Webster, & House, 1992). Point prevalence estimates for DSM major depression in surveys that use structured diagnostic interviews are considerably lower. Rates of current major depression are typically less than 1% in samples of children (reviewed by Merikangas & Angst, 1995), as high as 6% in samples of adolescents (reviewed by Kessler, Avenevoli, & Merikangas, 2001), and in the range 2–4% in samples of adults (WHO International Consortium in Psychiatric Epidemiology, 2000).
The discrepancy between the high prevalence of symptoms in screening scales and the comparatively low prevalence of depressive disorders means that many people have subsyndromal depressive symptoms. Recent epidemiological studies have started to investigate these subsyndromal symptoms using the diagnostic criteria for minor depression and recurrent brief depression (RBD) stipulated in DSM-IV-TR (American Psychiatric Association, 2000). Major depression (MD) requires 2 weeks of clinically significant dysphoria or anhe
Epidemiology of Depression 25
donia (or irritability among children) along with a total of five symptoms. Minor depression (mD), in comparison, requires two to four symptoms with the same severity and duration requirements as MD, while RBD requires the repeated occurrence of the same number and severity of symptoms as MD for several days each month over the course of a full year.
These recent studies have documented rates of subsyndromal depression among both adolescents (Gotlib, Lewinsohn, & Seeley, 1995; Kessler & Walters, 1998) and adults (Judd, Akiskal, & Paulus, 1997; Kessler, Zhao, Blazer, & Swartz, 1997) that are as high as, if not higher than, the rates of MD. In addition, a longitudinal study of adolescents followed into adulthood found that subsyndromal depression is a powerful predictor of the subsequent onset of MD (Angst, Sellaro, & Merikangas, 2000).
Subtypes
A number of proposals have been made to subtype the diagnosis of MD based on symptom profiles (reviewed by Kendell, 1976). The only stable subtyping distinction that has emerged consistently in empirical epidemiological studies, however, is between depression with vegetative symptoms (e.g., weight loss, insomnia, appetite loss) and reverse vegetative symptoms (e.g., weight gain, hypersomnia, appetite increase) (Davidson, Woodbury, Pelton, & Krishnan, 1988; Eaton, Dryman, Sorenson, & McCutcheon, 1989). Between onefourth and one-third of all people with MD have a reverse vegetative symptom profile, with some evidence that this atypical depression is more common among women than men and more strongly associated than vegetative depression with a family history of depression. There is little evidence, in comparison, that atypical depression is more persistent or severe than typical depression. Indeed, in one recent analysis of depression subtyping, typicality and severity emerged as separate and largely independent subtyping dimensions (Sullivan, Kessler, & Kendler, 1998).
Another important subtyping distinction concerns the existence of cyclical depression.
Two cycling depressive subtypes have been identified: seasonal affective disorder (SAD; Rosenthal et al., 1984) and premenstrual mood disorder (PMD; Halbreich, 1997). Community surveys find that 10% or more of people in the general population report seasonal variations in depressed mood and related symptoms (e.g., Booker & Hellekson, 1992; Rosen et al., 1990). Seasonal depression is typically most common in the winter months and more prevalent in northern than southern latitudes. However, the prevalence of narrowly defined DSM seasonal affective disorder, which requires a lifetime diagnosis of recurrent MD or mD and at least two-thirds of all episodes following a seasonal pattern, is much less common. Blazer, Kessler, and Swartz (1998) found that only 1% of the population meet narrowly defined criteria for SAD, representing only about 5% of all people with mD or MD. Among people with clinical depression, Blazer et al. found that SAD was somewhat more common among men than women and older than younger respondents.
Community surveys show that the majority of women report experiencing some symptom changes associated with their menstrual cycles (Pearlstein & Stone, 1998; Olive, 1991).
Only between 4 and 6%, however, report what appears to be a PMD (Sveindottir & Backstrom, 2000). A diagnosis of PMD requires a clear and recurring pattern of onset and offset of five or more mood and related symptoms at specific points in the majority of menstrual cycles over the course of a full year. Assessments with daily mood diaries over two or more menstrual cycles (Freeman, DeRubeis, & Rickels, 1996) typically show that only about half of the women who report cyclical mood problems actually suffer from PMD. The others have more chronic syndromal or subsyndromal mood disorders that are sometimes exacerbated by menstrual symptoms. There is currently a great deal of interest in PMD among depression researchers based on evidence of family aggregation with major depression (Yonkers, 1997) and responsiveness to selective serotonin reuptake inhibitors but not tricyclic antidepressants (Freeman, Rickels, Sondheimer, & Polansky, 1999). However, there is also controversy regarding appropriate diagnostic and assessment criteria (Severino, 1996). Community epidemiological data are scant due to the logistic complications created by the fact that a definitive diagnosis requires the collection of daily mood diaries across two or more menstrual cycles. Such diaries are typically collected only in clinical samples, although there are a few small community surveys that have collected diary data as well (e.g., Sveindottir & Backstrom, 2000). Given the existence of so many uncertainties in this area of investigation, a large representative epidemiological survey of PMS using dairy methods would be very valuable.
Lifetime Prevalence
Epidemiological surveys that administer diagnostic interviews generally assess lifetime prevalences of MD and estimate age of onset distributions from retrospective reports (e.g., Christie et al., 1988). Lifetime prevalence estimates of MD in U.S. surveys have ranged widely, from as low as 6% (Weissman, Bruce, Leaf, Florio, & Holzer, 1991) to as high as 25% (Lewinsohn, Rohde, Seeley, & Fischer, 1991). The only nationally representative general population data in the United States based on a structured diagnostic interview come from the National Comorbidity Survey (NCS; Kessler et al., 1994), where 15.8% of respondents met criteria for a lifetime MD episode and an additional 10.0% of respondents met criteria for lifetime mD (Kessler, Zhao, et al., 1997).
The wide variation in prevalence estimates across surveys is probably due to a combination of at least three factors. First, as discussed in more depth below, the prevalence of depression has probably increased in recent cohorts. This means that earlier surveys would be expected to have lower prevalence estimates than more recent surveys. Second, reluctance to admit depression has decreased in recent cohorts, which will also increase prevalences in more recent surveys. The third factor involves an important methodological difference between the diagnostic interviews that were used in early surveys based on the Epidemiologic Catchment Area (ECA) program (Robins & Regier, 1991), which uniformly produced very low prevalence estimates, and the more refined diagnostic interviews used in recent surveys such as the NCS, the Mental Health Supplement to the Ontario Health Survey (Offord et al., 1994), and the Mexican-American Prevalence and Services Survey (Vega et al., 1998). Both types of interviews use a stem–branch structure to assess mental disorders. In this approach, respondents are first asked one or more initial questions about core symptoms of the disorder under investigation. For example, a stem question for MD might be “Did you ever have a time lasting 2 weeks or longer when you felt sad or depressed most of the day nearly every day”? The respondents who are affirmative are then administered a more detailed set of follow-up questions that assess all criteria of the disorder. This same stem–branch approach is used to assess each of the dozen or more diagnoses evaluated in the surveys.
While both the type of interview used in the ECA and the type of interview used in the NCS were based on this stem–branch structure, only the NCS interview was designed to minimize the underreporting problems that methodological studies have shown to occur in interviews of this sort. A detailed discussion of the instrument design issues is presented elsewhere (Kessler, Wittchen, Abelson, & Zhao, 2000). In brief, methodological studies show that stem–branch questions are prone to two types of underreporting bias (Bradburn, Sudman, & Associates, 1979; Turner & Martin, 1984). One is that some respondents underreport stem questions once they recognize that positive responses lead to more detailed questions. The other is that most respondents fail to appreciate the cognitive complexity of he memory search involved in answering stem questions that require lifetime recall. These problems were addressed in the NCS by developing a Life Review Section near the beginning of the interview that included the stem questions for all the disorders assessed in the survey. The respondent instructions in this section were designed to facilitate and motivate active memory search. This entire section was administered before probing any positive stems, thus avoiding conscious nondisclosure once respondents recognized that positive stem responses led to further questioning. A field experiment carried out after the completion of the NCS randomly assigned respondents to either asked stem-and-branch questions in sequence throughout the interview or a version that included the Life Review Section and then validated diagnoses with clinician-administered reinterviews. Two important results emerged from this experiment. First, the Life Review Section was found to increase the prevalence estimates of depression and other disorders enough to explain the large observed differences in prevalence estimates between the ECA and the NCS. Second, the clinical reinterviews showed that the additional cases discovered with the Life Review Section were genuine cases of depression rather than false positives (Kessler, Wittchen, et al., 1998).
Based on these results, it seems safe to conclude that at least one out of every six adults in the U.S. population has met criteria for an MD episode at some time in their life and one in four has met criteria for either MD, mD, or recurrent brief depression. It is important to recognize, though, that these are estimates of prevalence-to-date risk rather than lifetime risk. Kaplan–Meier (KM) age-of-onset curves can be used to generate lifetime risk estimates. As shown in Figure 1.1, which presents KM curves separately for MD and mD based on the NCS data, the lifetime risk projections based on these curves are considerably higher than the lifetime-to-date prevalence estimates.
In evaluating the KM curves in Figure 1.1, it is important to recognize that the lifetime risk projections they generate are based on the assumption that conditional risk of first onset at given ages is constant across cohorts. This assumption is incorrect. As shown in FIGURE 1.1. Age of onset of major and minor depression in the NCS.
Figure 1.2, the KM curves for MD and mD combined in the NCS differ substantially by cohort. The same general pattern holds when we examine MD and mD separately (Kessler, Zhao, et al., 1997). This pattern of intercohort variation could be due to the risk of depression increasing in successively more recent cohorts. Other possible causes are that willingness to admit depression in a survey might have increased in recent cohorts (Kessler, 2000a) and that forgetting a past history of depression might increase with age (Giuffra & Risch, 1994). There is no way to adjudicate among these contending interpretations definitively with available data, although indirect evidence strongly suggests that at least part of the apparent cohort effect is due to a true increase in risk of depression in recent cohorts (Weissman & Klerman, 1992).
Course
Little longitudinal research has been done to study the course of depression in general population samples (but for important exceptions, see Angst & Merikangas, 1997; Lewinsohn et al., 2000). However, cross-sectional surveys consistently find that the prevalence ratio of 12-month MD versus lifetime MD is in the range between .5 and .6 (Kessler, McGonagle, Swartz, Blazer, & Nelson, 1993; Weissman et al., 1991). This means that between half and two-thirds of people who have ever been clinically depressed will be in an episode in any given year over the remainder of their lives. At least three separate processes contribute to the size of this ratio: the probability of a first episode becoming chronic; the probability of episode recurrence among people with a history who are not chronically depressed; and speed of episode recovery among people with recurrent episodes.
Epidemiological studies show that the first of these three processes is quite small, with only a small fraction of 1% of people in the population reporting a single lifetime depressive episode that persists for many years (Kessler et al., 1993). The prevalences of dysthymia FIGURE 1.2. Age of onset of major and minor depression by cohort in the NCS.
and chronic mD are somewhat higher, but still only in the range 3–4% combined in the total population (Kessler et al., 1994). Episode recurrence, in comparison, is very common, with more than 80% of people with a history of MD having recurrent episodes. In the NCS, the median number of episodes was seven among respondents with an age of first onset more than a decade prior to the interview. Moreover, over 90% of all episodes in the year prior to the interview were recurrences rather than first onsets. Speed of episode recovery, finally, appears to be highly variable, although the epidemiological evidence is slim. Only two large community surveys have studied speed-of-episode recovery. One found that 40% of cases of MD recovered by 5 weeks and over 90% by 1 year (McLeod, Kessler, & Landis, 1992). The other found that the median time to recovery was 6 weeks, with over 90% recovered within a year (Kendler, Walters, & Kessler, 1997). Very few of the people with short episodes ever come to clinical attention, which means that time to recovery is considerably longer in clinical samples (e.g., Brugha et al., 1990).
Distinguishing First Onset from Recurrence
It is important to distinguish between first onset and recurrence in studying the predictors of depression episode onset because the two types of episodes have different predictors (Lewinsohn, Allen, Seeley, & Gotlib, 1999). For example, women are nearly twice as likely as men to become depressed for the first time, while most epidemiological studies find no sex difference in recurrence risk (Kessler et al., 1993). History of depression is not only a powerful predictor of episode onset, but it is also strongly related to stressful events such as divorce and job loss and to presumed stress buffers such as social support and neuroticism (Kessler & Magee, 1993). Because of these relationships with all three of the main variables in models of the relationship between stress and depression (i.e., stress, stress buffers, and depression), failure to control for history of depression can lead to substantial bias in stress models. Yet most epidemiological studies that attempt to discover risk factors for episode onset of depression fail to include a control for history of prior episodes (Kessler, 1997a). A complicating factor is that risk of spontaneous recurrence of depressive episodes increases with number of prior episodes (Ghaziuddin, Ghaziuddin, & Stein, 1990). This means that it is important not only to control for history of depression in studies of the predictors of episode onset, but also to control for number of prior episodes and to estimate interactions between number of prior episodes and other predictors (Hammen & Gitlin, 1997).
Comorbidity
Studies of diagnostic patterns in community samples show that there is substantial lifetime and episode comorbidity between depression and other mental and substance use disorders. Indeed, comorbidity is the norm among people with depression. The ECA study found that 75% of respondents with lifetime MD also met criteria for at least one of the other DSM-III disorders assessed in that survey (Robins, Locke, & Regier, 1991), while the comparable proportion of DSM-III-R (American Psychiatric Association, 1987) comorbidity in the NCS was 74% (Kessler, 1995). There is controversy concerning the extent to which these high rates of comorbidity are artifacts of changes in the diagnostic systems used in almost all recent studies of comorbidity (Frances et al., 1992). In the United States, these systems, beginning with DSM-III (American Psychiatric Association, 1980) and continuing through DSM-IV-TR (American Psychiatric Association, 2000), dramatically increased the number of diagnostic categories and reduced the number of exclusion criteria so that many people who would have received only a single diagnosis in previous systems now receive multiple diagnoses.
differentiating information that could be useful in refining understanding of etiology, course, and likely treatment response (First, Spitzer, & Williams, 1990). However, it could also be argued that it had the unintended negative consequence of artificially inflating the estimated prevalence of comorbidity.
This uncertainty will presumably be resolved in the future by using established criteria to determine the validity of diagnostic distinctions (Cloninger, 1989). Until that time, though, we are left with a situation in which it appears that depression is highly comorbid with a number of other disorders. The strength of these comorbidities is remarkably consistent between the ECA and NCS surveys, the two largest general population surveys in the United States that have estimated comorbidities among DSM disorders (Kessler, 1995). The strongest lifetime comorbidities (odds ratios) of depression in both these surveys are with anxiety disorders, especially generalized anxiety disorder (6.0), panic disorder (4.0), and posttraumatic stress disorder (4.0), although less powerful but still significant comorbidities are found with a wide range of other mental disorders (Kessler, 1997b). Episode comorbidities are generally somewhat stronger, indicating that comorbidity is associated with recurrence risk (Kessler, 1995).
The majority of comorbid depression is temporally secondary in the sense that the first onset of depression occurs subsequent to the first onset of at least one other comorbid disorder, although this is more true among men than among women. Survival analysis of the cross-sectional NCS data using retrospective age-of-onset reports to determine temporal priority shows that a wide range of temporally primary anxiety, substance abuse, and other disorders predict the subsequent first onset of depression (Kessler et al., 1996). Time-lagged effects are strongest for generalized anxiety disorder (7.6) and simple phobia (4.2). There is little evidence of change in these odds ratios as a function of time since onset of the primary disorder. This absence of a time gradient is inconsistent with the hypothesis that secondary depression is a general exhaustion response to unremitting anxiety (Akiskal, 1990). At the same time, most of these odds ratios are confined to effects of active primary disorders as opposed to remitted primary disorders. This means that people who currently have these other disorders are at risk of depression. The fact that history of remitted anxiety is generally not associated with risk of depression suggests indirectly that anxiety is a risk factor rather than a risk marker. Two important exceptions, though, are early-onset simple phobia and panic, both of which appear to be markers rather than risk factors. The key evidence here is that people with a history of these disorders have elevated risk of subsequent first onset of depression even when the primary disorders are no longer active (Kessler et al., 1996).
THE CONSEQUENCES OF DEPRESSION
Psychiatric epidemiologists have traditionally been much more interested in estimating prevalences and discovering modifiable risk factors (e.g., Eaton & Weil, 1955) than in studying the consequences of mental illness (e.g., Faris & Dunham, 1939). This situation has changed in the past decade, though, as the managed care revolution and the rise ofevidence-based medicine have made it necessary to document the societal costs of illness (Gold, Siegel, Russell, & Weinstein, 1996). Depression has emerged as an important disorder in this new work. Indeed, the World Health Organization Global Burden of Disease (GBD) Study ranked depression as the single most burdensome disease in the world in terms of total disability-adjusted life years among people in the middle years of life (Murray & Lopez, 1996).
Role Impairment
It was noted in the introduction that the estimated prevalence of depression and other mental disorders in recent epidemiological surveys has been so high that some commentators have speculated that most must be mild cases (e.g., Regier et al., 2000). This speculation is superficially inconsistent with the GBD conclusion that depression is associated with more societal burden than any other condition. However, the GBD relied on expert opinion rather than epidemiological data to rank-order the impairments of chronic conditions. The expert raters were most familiar with clinical cases. It is possible that the cases found in community surveys are less seriously impaired.
The Medical Outcomes Study (Wells et al., 1989) collected data on this issue by screening samples of primary care patients for a small number of sentinel conditions that included MD and following these patients over time to evaluate their medical costs and role functioning. The role impairments caused by depression were comparable to those caused by seriously impairing chronic physical disorders. Similar results were found in the nationally representative general population sample assessed in the MacArthur Foundation’s Midlife Development in the United States (MIDUS) survey. The MIDUS results suggest that the role impairments caused by depression are comparable to those caused by such chronic physical disorders as arthritis, asthma, diabetes, and hypertension (Kessler, Mickelson, Barber, & Wang, 2001).
A substantial part of the role impairment caused by depression involves reduced work performance. A recent economic analysis of the costs of depression in the workplace estimated that the annual salary-equivalent costs of depression-related lost productivity in the United States exceeds $33 billion (Greenberg, Kessler, Nells, Finkelstein, & Berndt, 1996).
This is an underestimate of the overall workplace costs of depression because it excludes such potentially important components as the effects of depression on the performance of coworkers, industrial accidents, and turnover. It is important to note that these effects of depression on work performance disappear among remitted cases (Kessler & Frank, 1997), suggesting that effective depression treatment would reduce workplace costs. Simulations suggest that employers could recover between 45% and 90% of the direct treatment costs of depression in improved salary-equivalent work performance over the course of a single year (Kessler, Barber, et al., 1999). It is plausible to imagine that a complete cost accounting that considered the effects of depression on a broader set of workplace outcomes would show that the direct costs of depression treatment are fully offset by decreased indirect workplace costs. A definitive effectiveness trial to evaluate this hypothesis has not yet been carried out, although depression treatment trials have consistently documented significant effects of treatment on work outcomes (Mintz, Mintz, Arruda, & Hwang, 1992; Wells et al., 2000).
Role Transitions
It was noted in the discussion of the GBD study that depression has a unique constellation of characteristics leading to its rating by the World Health Organization as the single most burdensome chronic condition in the world among people in the middle years of life.
Perhaps the most important of these is early age of onset. The median age of onset of MD (see Figure 1.1) is in the mid-20s. This is at least 2 decades earlier than the median ages got-1.qxd 4/4/02 11:34 AM Page 31DESCRIPTIVE ASPECTS OF DEPRESSION 32 of onset of the chronic physical disorders that have prevalences and impairments comparable to those of depression. One important implication of this early age of onset is that depression, unlike most chronic physical disorders, occurs at a time in the life course when it can have a profound effect on critical life course role transitions. The latter include educational attainment, entry into the labor force, parenting, and marital timing and stability.
A series of analyses based on the NCS used retrospectively dated age of onset reports to estimate the effects of depression and other mental disorders on early life role transitions.
An investigation of the effects of early-onset depression on educational attainment found that depression prior to completing high school significantly predicted (odds ratio) high school dropout (1.5) and, among high school graduates, predicted failure to enter college (1.6) (Kessler, Foster, Saunders, & Stang, 1995). Depression as of the age of high school completion powerfully predicted college dropout among respondents who went to college (2.9). A separate investigation of the effects of early-onset depression on teen childbearing found that depression is associated with a 2.2 relative odds of teenage pregnancy among both girls and boys as well as with elevated rates of failure to contracept (Kessler, Berglund, et al., 1997). An investigation of the effects of early-onset depression on marital timing and stability, finally, found that prior depression predicts both teenage marriage (2.3) and subsequent divorce (1.7) (Kessler, Walters, & Forthofer, 1998).
It is important to appreciate that this constellation of truncated education, early childbearing, and marital instability are central components of welfare dependency. It is little wonder, then, that the welfare-to-work experiments that have been carried out in conjunction with recent state welfare reform programs have documented high rates of depression among welfare mothers and significant adverse effects of maternal depression on making a successful transition from welfare to work (Danziger et al., 2000; Olson & Pavetti, 1996).
This is another example of a case in which the societal costs of not treating depression may be greater than the costs of treatment. We are unaware, though, of any trial to evaluate the cost-effectiveness of providing mental health treatment as a component of the services provided to welfare mothers to facilitate the transition from welfare to work.
Other Adverse Consequences of Depression
It was noted in the last subsection that the financial savings to the employer due of increased work productivity with the remission of depression might approach or exceed the direct costs of treating depressed workers. The critical experiment needed to test this hypothesis has not yet been carried out. However, another type of experiment has been carried out that documents a cost saving of depression treatment for managed care.
Specifically, services research shows that people with untreated depression are often heavy users of primary care medical services for vaguely defined physical complaints. This observation has led some clinical researchers to speculate that systematic screening, detection, and treatment of primary care patients with depression might lead to an overall reduction in primary care costs. A series of experiments have shown that a partial offset effect of this sort exists (Katon et al., 1996; Katzelnick et al., 2000). The vast majority of depressed patients detected in primary care screening accept treatment for their depression. The average total cost of these patients to the managed care system exclusive of the cost of their depression treatment decreases significantly after their depression is treated.
This reduction partially offsets the cost of depression treatment over a follow-up period of 1 year. It is conceivable that the total costs of depression treatment are recovered over a longer time period, but long-term follow-up studies have not yet been carried out to determine whether this is the case.
EPIDEMIOLOGICAL STUDIES OF HELP SEEKING
Speed of Initial Treatment Contact
The findings reviewed above concerning the adverse effects of early-onset depression on role transitions raise an obvious question: Would timely treatment prevent these effects? We do not know the answer because the critical experiment has never been carried out. We do know, though, that timely treatment is the exception rather than the rule and that this is especially true for early-onset cases. This evidence comes from parallel studies of speed of initial treatment contact based on analysis of the NCS (Kessler, Olfson, & Berglund, 1998) and the Mental Health Supplement to the Ontario Health Survey (Olfson, Kessler, Berglund, & Lin, 1998). Both of these surveys asked respondents with a history of depression if they had ever sought treatment and, if so, their age of first obtaining treatment.
Comparisons of reported ages of onset with ages of first obtaining treatment were used to study patterns and correlates of delay in seeking treatment. The results were consistent in the two surveys in showing that delays in initial help seeking are pervasive. Only about onethird of the people who ever sought treatment did so in the same years as the first onset of their MD, while the median delay among those who did not seek immediate treatment was more than 5 years. Even more striking was the consistent finding that speed of contact is strongly related to age of onset. The vast majority of respondents who reported first onsets of depression in middle age or later sought treatment soon after the onset. Respondents with first onsets in early adulthood, in comparison, were much slower to seek treatment.
Respondents with child or adolescent onsets, finally, were by far the slowest of all, with median delays of more than a decade. It is not clear why this is the case, but one plausible hypothesis is that youngsters must rely on adults to initiate a treatment referral. Whatever the case may be, this is an especially disturbing pattern for two reasons. First, early-onset depression is often more severe than later-onset depression. Second, as noted above, earlyonset depression has powerful effects on critical developmental transitions that affect wellbeing throughout life. These results strongly suggest that special efforts are needed to reach out to children and adolescents with depression
Current Service Use
Turning from speed of initial lifetime help-seeking to treatment at a point in time, data from two nationally representative epidemiological surveys in the United States show that between one-third and one-half of the people who meet criteria for MD in a given year obtain some type of treatment for their depression during that year (Kessler, Zhao, et al., 1999; Wang, Berglund, & Kessler, 2000). A substantial proportion of this treatment occurs in the general medical sector. Unfortunately, analysis of the content of this treatment in comparison to published treatment guidelines (Agency for Health Care Policy and Research, 1993; American Psychiatric Association, 1993) shows that no more than 30% of these patients receive even minimally acceptable treatment (Katz, Kessler, Lin, & Wells, 1998; Wang et al., 2000). There is clear evidence that depression treatment that fails to conform with treat ment guidelines is associated with incomplete recovery and increased risk of recurrence (Melfi et al., 1998). These results show that advances in the development and implementation of treatment quality improvement programs are clearly needed.
Another development of great importance in depression treatment involves the rise of complementary and alternative (CAM) therapies. Three recently completed national surveys have documented that a substantial proportion of people with depression use a variety of CAM therapies, such as St. John’s wort and relaxation therapy, to treat their depression got-1.qxd 4/4/02 11:34 AM Page 33DESCRIPTIVE ASPECTS OF DEPRESSION 34 (Eisenberg et al., 1993; Eisenberg et al., 1998; Unutzer et al., 2000). In the most detailed of these surveys, which was carried out in 1997–1998, 54% of the respondents with selfdefined “serious depression” in the year prior to the interview reported that they used some form of CAM for their depression (Kessler et al., 2001). An alternative medicine professional, such as an energy healer or herbalist, was seen during that same year for the treatment of depression by 19% of the respondents with self-defined serious depression. This compares to 36% who reported seeing any conventional physician or mental health professional for their depression during that same time period. The patients who used CAM were more likely to see a conventional provider than those who did not use CAM, with 66% of the patients who saw a conventional professional for their depression also using CAM.
Importantly, only a small minority of CAM users who are also in treatment with a conventional provider tell the latter about their CAM use (Eisenberg et al., 1998). It is important to recognize that this type of unsupervised joint use of CAM and conventional therapy can be dangerous, as case studies show that some types of CAM can create potentially dangerous interactions with pharmacotherapies (Yager, Siegfried, & DiMatteo, 1999; Almeida & Grimsley, 1996). For example, recent case reports suggest that the mixture of St. John’s wort with selective serotonin reuptake inhibiters can induce a mild serotonin syndrome (Ernst, 1999). In vitro studies also suggest that hypericum extracts, which are commonly used herbal treatments for depression, are potent inducers of hepatic enzymes that can reduce plasma concentrations of a variety of concomitant prescription medications (Fugh Berman, 2000). Opening up lines of communication between conventional mental health professionals and patients with regard to CAM use is consequently of great importance.
FUTURE DIRECTIONS
Developmental Epidemiology
There is an increased interest in developmental studies of the onset and course of depression as part of a larger interest in developmental epidemiology (Angold & Costello, 1995). The realization that first onset of depression often occurs early in life and that gender differences in depression begin to emerge in midadolescence are fueling this interest. It is likely that future developmental epidemiological studies will collect blood or saliva samples that can be used to measure sex hormones in an effort to tease out the biological and social effects of pubertal status and timing. Two epidemiological studies of adolescents have already collect ed data of this type and has shown that increases in sex hormones appear to explain much of the emerging sex difference in depression in midpuberty (Angold, Costello, & Worth man, 1998; Patton et al., 1996). It is also important, though, that these future studies give equal attention to social changes that occur at about the same time. The importance of this equal treatment is illustrated nicely in a recent report from the National Longitudinal Study of Adolescent Health (Bearman, Jones, & Udry, 1998), which showed that the greater in crease in exposure to stresses associated with dating among girls than boys can explain much of the increasing sex difference in depression in midpuberty without reference to hor monal changes (Joyner & Udry, 2000).
Genetic Epidemiology
Psychiatric epidemiologists have been greatly interested in behavioral genetic studies of depression and other major mental disorders, with most of the focus being on twin and twin–family designs. Such studies use structural equation models to partition variances and got-1.qxd 4/4/02 11:34 AM Page 34Epidemiology of Depression 35 covariances into genetic and environmental components (Neale & Cardon, 1992). Although convincing data have been presented in these studies that depression is clearly heritable (Kendler et al., 1996), behavioral genetic studies have been disappointing in not ad vancing far beyond this basic fact. Some commentators on the future of psychiatric epidemiology have suggested that our greatest hope for a breakthrough in understanding the etiology of depression is likely to come from genetic epidemiology (Robins, 1992).
However, there is no indication that this promise has begun to be fulfilled in the nearly two decades since epidemiological studies based on genetically informative designs (i.e., twin–family and adoption designs) have been actively pursued. Linkage studies have been unable to identify a single specific gene or gene marker for any mood disorder. If such markers can be identified, integration of psychiatric epidemiology with population genetics would be valuable in a number of ways (Risch & Merikangas, 1996). It is not clear, though, when and if such markers will be identified.
Experimental Epidemiology
Epidemiology has played a major part in the development of many public health interventions. Important epidemiological contributions along these same lines are beginning to emerge in psychiatric epidemiology as well. Included here are studies that have documented effects of obstetrical complications on childhood-onset schizophrenia (Nicholson et al., 1999), of childhood nutritional deficits on conduct disorder (Neugebauer, Hoek, & Susser, 1999), and of childhood lead exposure on early-onset Alzheimer’s disease (Prince, 1998).
However, the enormous complexity of environmental etiological processes in bringing about mental disorders has led most psychiatric epidemiologists to focus their efforts on broad nonspecific risk factors such as stress, social support, social class, and gender that do not have clear intervention implications. As a result, psychiatric epidemiologists have been less actively involved in targeting interventions than epidemiologists working in other areas of research. (For an important exception, see Harris, Brown, & Robinson, 1999a, 1999b.)
As described in more detail elsewhere (Kessler, 2000b), the way in which analytic epidemiological research is carried out differs in important ways depending on whether the researcher sees the work as important for hypothesis testing or for guiding intervention development and targeting. If future psychiatric epidemiologists are to become closely involved in intervention work, changes will be needed in the types of questions asked, the kinds of analyses carried out, and the standards of proof required for epidemiological studies.
1 comments:
Post a Comment