| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
TOPIC 4. ECONOMIC STATUS AND HEALTH INEQUALITIES |
1 Department of Mental Health, Bloomberg School of Public Health, The Johns Hopkins University, Baltimore, Maryland.
2 Department of Anthropology and Sociology, Western Carolina University, Cullowhee, North Carolina.
Address correspondence to Richard Miech, The Johns Hopkins University, Room 853, Hampton House, 624 North Broadway, Baltimore, MD 21205. E-mail: rmiech{at}jhsph.edu
Abstract
Objectives. Higher levels of psychopathology among people with lower socioeconomic status and among women persist as cohorts age. In this analysis, we examine whether the persistence of these disparities as a cohort ages results from (a) a single set of people within a disadvantaged group who have chronic psychopathology or (b) continually changing sets of people within a disadvantaged group who have psychopathology of short duration.
Methods. Data for this analysis come from the Epidemiologic Catchment Area Study, which includes two psychological assessments (depressive syndrome and psychological distress) in a population sample of adults in Baltimore, Maryland, collected 13 years apart.
Results and Discussion. Results indicate that the persistence of disparities across education resulted primarily from one single set of respondents with chronic psychopathology over the 13 years of the survey, while the persistence of disparities across sex involved new sets of women as the cohort aged. We discuss implications of these results for theory and policy.
Mental health disparities across socioeconomic status and sex stem at least partly from social influences (Turner & Lloyd, 1999
; Turner, Wheaton, & Lloyd, 1995). The motivation to identify these influences spans multiple disciplines. In sociology, they represent an opportunity to examine social structure and identify socially based adversities and disadvantages—known as "stressors"—that are unequally distributed across social groups and shape individual well-being (Pearlin, 1989). In epidemiology, their analysis contributes to the effort to identify the risk factors for disease (Szklo & Nieto, 2000). In psychiatry and psychology, their analysis can contribute potential clues to the etiology of mental illness (Gilman, Kawachi, Fitzmaurice, & Buka, 2003). These various theoretical interests have generated hundreds of analyses that examine social disparities in mental health and attempt to identify the environmental sources of these disparities.
The process by which these disparities persist as cohorts age has important implications in the search for the responsible social factors, implications that we draw out in the Discussion section. On the one hand, the process may involve chronic psychopathology among a single set of people within a disadvantaged group. In this scenario, the social factors that foster mental health disparities have an effect on mental health early in the life course, and the effect lingers as a cohort ages. These effects continue independently of the environmental factors that fostered them.
Reasons why psychopathology may be chronic and long-lasting (Kennedy, Abbott, & Paykel, 2003) and continue independently of its precipitating events may potentially be biological, social, or a combination of both. One biological reason is that people who experience a severe episode of mental illness such as depression undergo biological changes that make recovery difficult and predispose individuals to future episodes (McEwen, 2003; Post, 1992). Another biological possibility is that some people may have a mental illness diathesis due to genetic and/or personality factors (Kendler, Prescott, Myers, & Neale, 2003; McLeod & Kessler, 1990). At the same time, the persistence may also stem from social influences; Caspi, Elder, and Bem (1987) articulate a process of "cumulative causation" in which individuals with mental disorder may evoke responses from their social environment that serve to reinforce their mental state.
In contrast to a process in which one set of people exhibit chronic psychopathology, an alternate process that can also explain the persistence of mental health disparities as cohorts age points to continually changing sets of people within a disadvantaged group. In this scenario, an elevated prevalence of mental illness in a disadvantaged group serves as a queue. On one end of the queue are new sets of people on whom social influences continually foster new onset of psychopathology over time. At the other end of the queue are individuals who recover from psychopathology after a limited amount of time. As a result of this queuing process, an elevated level of psychopathology in a disadvantaged group may be represented by new sets of members over long periods of time because all those who exhibited psychopathology at baseline may have recovered and those who exhibit psychopathology at a later follow-up developed it over the intervening period.
This study empirically examines whether persistence of mental health disparities as a cohort ages stems from either chronic psychopathology or, instead, a queuing process. A queuing process leads us to expect that disparities at the follow-up survey consist of people who did not show symptoms of psychopathology at the baseline interview many years previously. Therefore, in the analyses below, we test whether disparities at the follow-up survey are present among people who screened negative on a baseline psychopathology screener. We run parallel models using baseline screeners of different symptom severity to assess the influence of the baseline screener cutoff level on the study results and conclusions.
DATA AND METHODS
Data
Data for this study come from the Epidemiologic Catchment Area (ECA) Follow-Up, which is a multistage probability sample of 3,481 adults aged 18–96 years residing in East Baltimore, Maryland, in 1981 (Eaton et al., 1997). The study set out to reinterview all sample members in 1993–1996, at which point 848 study members had died, 415 were not located, and 298 declined to participate (for a detailed description, see Badawi, Eaton, Myllyluoma, Weimer, & Gallo, 1999). In all, 1,920 members were interviewed in the follow-up, which represents 73% of the survivors.
We restrict the analysis of this sample to respondents younger than 60 years at follow-up because of evidence that the constellation of symptoms that represent mental illnesses such as depression are significantly different among the elderly population as compared with younger populations (Gallo, Rabins, & Anthony, 1999). Specifically, among individuals who experience depression, elderly persons are significantly less likely to report the symptom of "sadness" than younger adults, a finding that appears to be an aging and not a cohort effect (Gallo & Rabins, 1999).
In the baseline ECA, 1,801 respondents were in the age range to be younger than 60 years at the follow-up. Of these respondents, 68 died during the interim, 316 were not located, and 204 declined to participate. In all, 1,213 members in the target age range were interviewed in the follow-up, which represents 70% of the survivors. The analyses that follow are based on the 1,171 study members who provided complete information for the measures used in this study.
Measures
Psychological distress at the follow-up survey is assessed using the self-reported General Health Questionnaire (GHQ; Goldberg, 1972; Goldberg & Hillier, 1979). The GHQ is a widely used self-reported measure of psychological distress in clinical and population settings and consists of symptoms of depression, anxiety, insomnia, somatic symptoms, and social dysfunction. The GHQ does not portray a specific psychiatric disorder but helps identify individuals who may have mental illness. As used in the ECA and the ECA follow-up, the GHQ consists of 20 items and is about 80% specific and 75% sensitive for depression (Samuels, Nestadt, Anthony, & Romanoski, 1994). Psychological distress scores in this study are continuous and are the sum of the 20 GHQ items (the total ranges from 20 to 78). In the analysis pool of this study, the 20-item GHQ has an internal consistency reliability of.89 (using Cronbach's alpha) at both baseline and follow-up.
Depressive syndrome at the follow-up survey is a dichotomous measure assessed with criteria from the Diagnostic and Statistical Manual of Mental Disorders (3rd ed; DSM; American Psychiatric Association, 1980). The concept and measurement of depressive syndrome is discussed at length by Eaton, Muntaner, Bovasso, and Smith (2001). In brief, it is measured as the presence of three or more DSM depressive symptoms in the last month, including sadness, problems with eating, sleeping, concentrating, moving, unusual fatigue, unreasonable guilt, and suicidal thoughts or behaviors. Depressive syndrome has a higher prevalence than clinical depression, which allows greater statistical power. It is also related to important risk factors in a manner suggesting that it captures the same etiological process as major depressive disorder (Chen, Eaton, Gallo, & Nestadt, 2000). Both depressive syndrome and major depressive disorder are related to important outcomes in a similar way (Judd, Paulua, Wells, & Rapaport, 1996). In the algorithm used in this analysis, all reported symptoms are included even if the respondent discounts them owing to presumed cause by medication, drugs, alcohol, physical illness, or injury because respondents can be too quick to discount symptoms in this manner, and agreement on diagnoses with psychiatrists is better if all symptom reports are included in the algorithm (Eaton, Neufeld, Chen, & Cai, 2000).
The analysis includes variables to indicate respondents who had mental illness symptoms at the baseline assessment. The study results consist of parallel analyses that compare results using a dichotomous baseline depression screener that is measured two ways: (a) the presence of three or more DSM criteria for depression (the same criteria that define depressive syndrome at follow-up); and (b) the presence of two or more DSM criteria for depression. The results also includes parallel analyses that compare models using a baseline distress screener that is measured in two ways: (a) respondents in the top 5% of the psychological distress distribution (a score of 49 or higher on the GHQ); and (b) respondents in the top 10% of the psychological distress distribution (a score of 42 or higher on the GHQ).
Female is coded 1 for female respondents and 0 for males. Less than 12 years of education is coded 1 for respondents who report 11 or fewer years of education and 0 otherwise. Age is assessed with the dichotomous variables age 40–49 years (37% of the sample) and age 50–59 years (21% of the sample); respondents aged 30–39 years, who are 39% of the sample, are the reference category in the models.
Statistical Analysis
The regression analyses used standard logistic regression and ordinary least squares (OLS) regression equations, and focused on the mental health of the study members at the last survey wave in 1993. Analyses that focused on the dichotomous DSM-based measure of depressive syndrome used logistic regression. Analyses that focused on the continuous GHQ measure of mental illness used OLS regression.
The main part of the analysis examined the extent to which mental health disparities across education and sex at the follow-up resulted from chronic psychopathology or a queuing process. The test of these competing processes centers on the group of respondents who had no depressive syndrome symptoms at the baseline survey. If a queuing process is at work, then new disparities will form in this group over the course of the survey. In contrast, if chronic psychopathology accounts for disparities as cohorts age, then these disparities formed only once early in the life course and no new disparities should form among respondents with little or no psychopathology at the baseline assessment. As described in more detail in Results, we empirically address this research question using multiplicative interaction terms.
RESULTS
Bivariate Associations
Table 1 presents measures of mental health at the baseline assessment and the follow-up, stratified by educational attainment and sex. All measures are distributed across education and sex as expected, and in all cases, the differences across education and sex are statistically significant. The prevalence of depressive syndrome for women was about twice as a high as it was for men, at both the baseline and the follow-up survey. Similarly, respondents with less than 12 years of education also had a prevalence of depressive syndrome that was about twice as high as it was among respondents with 12 or more years of education at both the baseline and the follow-up survey. Differences in psychological distress were as expected: Women had more psychological distress than men, respondents with low education had more psychological distress than those with higher education, and the levels of disparities across sex and education were about the same at the baseline survey and at follow-up. The baseline screeners for depression and psychological distress also have a higher prevalence rate among women and respondents with less education.
|
|
Model 2 of Table 2 presents results from the regression equation that included these additional three variables, using a baseline screener of respondents who reported three or more DSM symptoms. Neither of the interaction terms of the screener variable with the education and sex variables was statistically significant. This result would, at first, suggest that knowledge of the 5% who had high depressive symptoms at baseline was not informative for analysis of health disparities at the follow-up 13 years later, and thus not provide support for disparities as a result of chronic psychopathology. However, this analysis has low statistical power both because the dependent variable is dichotomous and also because the baseline screener indexes a small number of respondents.
Model 3 of Table 2 is parallel to Model 2 and differs only in the definition of the baseline screener, which uses the more lenient criteria of only two or more DSM symptoms to test positive. These results are consistent with a process of chronic mental illness for disparities across education. The coefficient for education was not statistically significant, indicating that among the 88% of respondents who tested negative on the depression screener at the baseline survey, educational status did not predict new cases of depressive syndrome at the follow-up assessment. The sum of the education coefficient and the interaction term that included education was significantly greater than 0 (p <.01 using a Wald test, not shown), indicating that among respondents who tested positive on the depression screener, respondents with low educational attainment were more likely to report depression syndrome at the follow-up survey. Taken together, these findings indicate that the overall association of education with depressive syndrome observed in Model 1 resulted from the small group of individuals with low education who tested positive on the depression screener at the baseline assessment 13 years earlier, and did not represent new cases of depressive syndrome that developed over the course of the survey.
The results in Model 2 of Table 2 indicate a substantively different process for disparities across sex. The coefficient for sex is significantly different from zero, indicating that among respondents who tested negative on the depression screener at baseline, women were more likely than men to report onset of depressive syndrome at the follow-up assessment. The sum of the sex coefficient and the interaction term that included sex was not significantly different from zero (p =.48, Wald test, not shown) and indicated that among respondents who tested positive on the depression screener, women were no more likely than men to report depression syndrome at the follow-up. Taken as a whole, this pattern of results indicates that the overall association of sex with depressive syndrome observed in Model 1 was in large part the result of new sex disparities that formed over the course of the survey.
For the results in Model 3 of Table 2, we used Wald tests to examine whether the estimated coefficients for education significantly differed from the estimated coefficients for sex. In Model 1, the estimated education coefficient did not significantly differ from the estimated sex coefficient (p =.91), indicating that these overall associations with depressive syndrome had similar magnitudes. In Model 3, the magnitude of the education coefficient differed significantly from the sex coefficient (p <.03), indicating that education was a significantly smaller predictor of depressive syndrome than sex among respondents who tested negative on the baseline screener for depressive symptoms. Also, in Model 3, the sum of the education coefficient and its associated interaction coefficient differed significantly from the sum of the sex term and its associated interaction coefficient (p <.001), indicating that education was a significantly stronger predictor of depressive syndrome than sex among respondents who tested positive on the baseline screener for depressive symptoms. Thus, whereas both sex and education have about the same overall association with depressive syndrome at the follow-up survey, these factors differ significantly from each other as predictors of depressive syndrome among the subgroups of respondents who did and did not have depressive symptoms at the baseline assessment.
Table 3 presents results from a parallel analysis that substitutes psychological distress as the measure of psychopathology for depressive syndrome. These results showed the same pattern of findings, regardless of whether the baseline screener indicated the respondents in the top 5% of the distress distribution or the top 10%. Low education did not significantly predict elevated levels of distress at the follow-up survey among respondents who tested negative on the distress screener at baseline, but was a strong predictor of higher levels of distress at the follow-up among respondents who tested positive on the screener. In contrast, sex showed the opposite pattern of results in comparison to education. Among respondents who tested negative on the baseline distress screener, women reported significantly higher levels of distress than men, whereas among respondents who tested positive on the baseline distress screener, women had about the same likelihood of reporting distress as men. As in the analysis of depressive syndrome, the magnitude of the coefficients of education and sex did not significantly differ from each other in their overall association in Model 1 (p =.17), but in Models 2 and 3, they did differ from each other as predictors of distress among respondents who tested negative on the baseline screeners (p <.01) and those who tested positive (p <.001).
|
This study set out to examine whether mental health disparities across education and sex represent either chronic psychopathology or a queuing process. This analysis is well suited to address this topic for two reasons. First, the analyses are based on a longitudinal study of mental health with assessments 13 years apart. Prospective information on mental illness from a population sample over such an extended period of time is uncommon and allows the analysis to consider chronic psychopathology without the biases introduced by retrospective reports. Second, the study includes parallel analyses of mental illness measured both using the criteria of the DSM (American Psychiatric Association, 1980), which is widely used by clinicians, and also items from the GHQ (Samuels et al., 1994), which is widely used by researchers. The inclusion of both measures widens the scope of this study and produces results that directly inform both the existing clinical and research literature.
The results indicate that the processes that underlie mental health disparities differ across education and sex. As a cohort ages, mental health disparities across education are in large part due to a process of chronic psychopathology among a single set of individuals. These disparities at the follow-up were specific to a small group of individuals—about 10% of the sample or less—that displayed psychopathology 13 years earlier at the baseline assessment. In contrast, mental health disparities across sex are in large part due to a queuing process. The excess of depressive syndrome and psychological distress of women in comparison to men that is present at the last wave of the survey is due in large part to onset of new depressive syndrome or increases in psychological distress that occurred during the 13 years since the baseline assessment.
Before discussing the implications of these findings in detail, it is important to note the limitation that the data do not contain yearly prospective information on mental health for the 13 intervening years between assessments. Consequently, the analyses in this article do not address many of the details of the incidence and chronicity of mental illness over the course of the survey, such as the hazard rate of mental illness for each year or the decomposition of chronic mental illness into components of remission and duration. Instead, we focus our research question on mental health disparities at the follow-up survey and examine the extent to which they stem from new incident cases of mental illness that were not present at the baseline assessment or chronic mental illness among respondents who displayed at least some symptoms of mental illness 13 years previously. This study does not examine the processes that occurred within the two time points of the survey.
Conclusion
The theoretical and policy implications of these findings differ for mental health disparities across education in comparison with mental health disparities across sex. A major theoretical framework that guides analysis of health disparities is the stress paradigm, which posits that social factors play a major role in their etiology. However, this framework provides little direction as to what stage or stages in the life course these social factors exert their impact on mental health. The results of this study indicate that the social factors responsible for disparities across education most likely have an impact at a stage in the life course before adulthood, when they foster chronic psychopathology. In contrast, the results of this study indicate that the social factors that foster disparities across sex constantly operate throughout adulthood and continually affect new groups of women as previously affected groups recover.
The policy implications of these findings suggest that treatment may be practical for the reduction of disparities across education but not across sex. Our results indicate that treatment or prevention programs addressed at the mental health of individuals before adulthood could, if they were effective, potentially have a lasting influence on mental health disparities across education over the life course. To the extent that mental health disparities across education represent the aftermath of a process that first occurred prior to or in early adulthood, interventions that effectively prevent this process from taking place could potentially cut off disparities before they start. In contrast, treatment programs aimed at women in early adulthood would not have a lasting effect on sex disparities as cohorts age. In the context of a queuing process, treatment-based interventions to reduce sex disparities would be a Sisyphean task because soon after one set of women was successfully treated, a higher onset of psychopathology for women as compared with men would ensure that another group would soon fill their place and the sex disparity would still persist. An effective approach to the reduction of mental health disparities across sex is a somewhat daunting task that will require policies and programs aimed at the social influences that produce them, keeping in mind that disparities may well represent the effect of different social influences at different stages of the life course.
Acknowledgments
This research was supported by grants MH47447 and MH61877 from the National Institute of Mental Health.
References
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
