| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
RESEARCH ARTICLE |
Rush 1 Institute for Healthy Aging
2 Alzheimer's Disease Center
Departments of 3 Internal Medicine
4 Preventive Medicine
5 Neurological Sciences
6 Behavorial Sciences; Rush University Medical Center, Chicago, Illinois.
Address correspondence to Kimberly A. Skarupski, PhD, MPH, Rush Institute for Healthy Aging, 1645 W. Jackson, Suite 675, Chicago, IL, 60612-3227. E-mail: Kimberly_Skarupski{at}rush.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
= 0.467 on the log scale, p<.001). The differences were larger for men than for women. After controlling for age, sex, time, education, income, and related interaction terms, the baseline race effect was reduced by almost half ( = 0.225, p<.001) but remained robust. The racial differences increased slightly over time. Our findings support heightened awareness of depression in older Black populations.
The U.S. Department of Health and Human Services (U.S. DHHS; 2003
), in its publication Healthy People 2010, states two overarching goals for the first decade of the 21st century: (a) increase quality and years of healthy life and (b) eliminate health disparities among different segments of the population. Health disparities between non-Hispanic Whites and members of ethnic minority groups across all ages have been widely documented. There is clear evidence for racial or ethnic health disparities in older populations, and most research has focused on differences in physical health (Levine et al., 2001; U.S. DHHS, 2000). However, whether there are racial or ethnic disparities in mental health outcomes is less clear (George & Lynch, 2003). For the purposes of this article, we examine racial disparities in mental health among older adults, with a specific focus on Black–White differences in depressive symptoms.
Depression is the most prevalent mental disorder in the U.S. population (Robins & Regier, 1991). According to recent data from the National Comorbidity Survey Replication (NCS-R) study, 16.2% of the adult U.S. population reported a lifetime history of major depressive disorder (MDD), and 6.6% reported an episode of MDD during the past 12 months (Kessler et al., 2003). In contrast to MDD, it is unclear whether older adults experience higher or lower symptom levels of depression relative to younger adults. On the basis of cross-sectional research, studies have either reported a negative age slope for symptom levels (Henderson et al., 1998; Newmann, 1989), or a curvilinear age slope, with highest symptom levels among younger and older adults (Kessler, Foster, Webster, & House, 1992). In contrast, a recent longitudinal study found evidence for an increase in symptom levels with older age, both cross-sectionally and over time (Fiske, Gatz, & Pedersen, 2003). Furthermore, in that study, age-related increases in symptoms were neither limited to somatic symptoms of depression nor solely attributable to illness events. However, the generalizability of those findings may be limited given that they were based on twin registry data.
Research on racial differences in mental health outcomes in general, and depression in particular, has produced largely conflicting findings. Studies focusing on symptom levels suggest that, in the general adult population, Blacks tend to report higher levels of depressive symptoms than do Whites (Frerichs, Aneshensel, & Clark, 1981; Jackson-Triche et al., 2000; Sayeta & Johnson, 1980; Ulbrich, Warheit, & Zimmerman, 1989). However, regarding older adults, there is no clear understanding of the existence or magnitude of disparities in depression. Although some researchers have found that, on average, older Whites report higher symptom levels than do Blacks (Callahan & Wolinsky, 1994; Gallo, Cooper-Patrick, & Lesikar, 1998), there is some evidence that this race effect varies by geographic region (Gallo et al.). Interestingly, in their study, Gallo and colleagues found that although older Blacks were less likely than Whites to report sadness or depressive symptoms, they were more likely to think about death. Still, others have found little evidence for differences in depressive symptom levels between older Blacks and Whites (Berkman et al., 1986; Blazer, Landerman, Hays, Simonsick, & Saunders, 1998). Clearly, the issue of Black–White differences in depressive symptoms among older adults is complex and as yet ambiguous.
Adding to the uncertainty is the fact that most research in this area has been based on data from the 1980s, such as the Epidemiologic Catchment Area (ECA) Program (Gallo et al., 1998) or the Established Populations of the Epidemiologic Studies of the Elderly (EPESE) studies (Berkman et al., 1986; Blazer et al., 1998). However, recent evidence suggests that, over the past two decades, racial disparities have increased for life expectancy and relative mortality (Levine et al., 2001). For the ambitious goals set forth in Healthy People 2010 (U.S. DHHS, 2003) to be achieved, racial disparities in mental health outcomes should be examined with the use of more recent data.
Therefore, our primary purpose in this study is to examine and describe racial differences in depressive symptoms among older adults, using data from a recent population-based, longitudinal survey of older adults. In this analysis, we focus on average depressive symptom levels, rather than diagnostic categories of depression, and test specifically whether older Blacks report higher levels of depressive symptoms compared with older Whites. In general, racial disparities in health are thought to be in large part a function of the substantial differences in socioeconomic status (SES) between Blacks and Whites (Williams, 1998). There is clear evidence that lower SES is associated with elevated levels of depression (Berkman et al., 1986; Blazer et al., 1998; Kessler et al., 1994). Thus, as a secondary objective, we test the degree to which racial differences in depressive symptoms persist after we have accounted for differences in SES.
|
METHODS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
; Evans et al., 2003). Assessments were conducted at approximately 3-year intervals, including a baseline interview from 1993 to 1997 (N = 6,158), and two follow-up interviews conducted between 1997 and 2000 (N = 4,320), and 2000 and 2003 (N = 2,943), respectively. All data were collected by trained interviewers in the participants' homes. The interviews included structured questions about sociodemographic characteristics, health, and lifestyle, as well as performance-based tests of physical and cognitive function. The Institutional Review Board of Rush University Medical Center approved the study and all participants provided written, informed consent.
Measures
We based an assessment of depressive symptoms on the 10-item version of the Center for Epidemiologic Studies Depression scale (CES-D; Kohout, Berkman, Evans, & Cornoni-Huntley, 1993). This abbreviated CES-D is derived from the original 20-item version (Radloff, 1977), and it has been found to have acceptable reliability and a similar factor structure compared with the original version (Kohout et al.). Item responses are coded in a yes–no format, yielding a summary measure (CES-D) with a range from 0 to 10 after one sums across the individual items. Of the 10 items, 6 had to be nonmissing for the summary score to be nonmissing.
In a secondary analysis, we computed a dichotomous CES-D variable to represent participants with an elevated level of depressive symptoms. We split the CES-D variable into two levels: scores of 3 or lower and scores of 4 and higher. A score of 4 or higher on this version of the CES-D has shown reasonable specificity and sensitivity in identifying older adults with major depression (Irwin, Artin, & Oxman, 1999).
Sociodemographic variables that we considered in the analyses included age (in years); sex; education (in years of schooling completed); income; and race. We assessed income by having respondents select 1 of 10 categories representing a range of personal income during the past month or year. We recoded responses to a rank-order variable ranging from 1 to 10, with the lowest category representing an income of less than $5,000/year and the highest category representing an income of over $75,000/year. For the purpose of our analysis, we centered the age, education, and income variables at their approximate median values.
Analysis
We used t tests for continuous variables and chi-square tests for categorical variables to compare age, sex, education, and income differences between Blacks and Whites at baseline. We used generalized estimating equation (GEE) models to model the vector of CES-D scores from each person as a function of age, sex, time since baseline, race, and other predictor variables. GEE is particularly suitable for the analysis of these data, as it offers a choice of link functions to model the outcome variable and accounts for the within-person correlation across repeated measurements (Diggle, Heagerty, Liang, & Zeger, 2002). Because of the highly skewed distribution of the CES-D scores, we considered these scores as a count variable of the number of symptoms and specified a log link function and Poisson error structure (Allison, 1999; SAS Institute, 2000). We assumed the "working" within-person correlation to be identical for each pair of times of observation (exchangeable error structure); the estimates from GEE models are robust to the choice of working correlation matrix.
We tested the primary hypothesis of Black–White differences in depressive symptoms in three sequential models. In the first model, after we adjusted for baseline age, male sex, and time since baseline, we entered terms for race and the Race x Time since baseline interaction. Preliminary analysis suggested that there was a slightly curvilinear pattern of change in depressive symptoms over time that varied by sex. To account for this pattern, we also included a squared term for Time since baseline (time2) as well as Time (since baseline) x Male sex and Time2 x Male sex interaction terms. We also considered baseline marital status (married vs. nonmarried) for the initial model, but this variable was neither significantly related to depressive symptoms nor affected Black–White differences in the outcome, and therefore we did not include it. In this first model, the main race effect represents the average difference in symptoms between Blacks and Whites at baseline. The Race x Time interaction represents the deviation in symptom levels as a function of time since baseline (i.e., during follow-up) for Blacks from the change in symptom levels among Whites. This term is therefore a test of whether Black–White differences in symptoms increase or decrease over time. In the second model, we added an interaction term for Race x Sex to see if Black–White differences in depressive symptoms at any point in time were different between men and women. In the third model, we added education and income to test the degree to which Black–White differences in depressive symptoms were modified by SES.
In a secondary series of analyses, we examined the association between race and elevated level of depressive symptoms by replacing the continuous CES-D score dependent variable with the dichotomous CES-D variable. We did this in order to facilitate comparison of our results with other published research that has examined depression by using a dichotomous CES-D variable. We modeled the dichotomous CES-D variable by using GEE with a logit link function and binomial error structure, with the category of three symptoms or fewer as the referent category (Irwin et al., 1999). In addition, there are two CES-D items that may confound depressive symptoms with perceptions of discrimination; namely the "I felt that people disliked me" and the "people were unfriendly" items (Barnes, Mendes de Leon, Wilson, et al., 2004). In an attempt to reduce the likelihood that a Black–depressive symptoms association would be due to disproportionate endorsement by Blacks of the "dislike" and "unfriendly" CES-D items, we removed the items from the composite score and reran the multivariate models. There were no substantive differences between the models, so we retained the items in the composite CES-D score.
We examined model assumptions both graphically and analytically and found them to be adequately met. We performed all longitudinal analyses by using the GENMOD procedure of SAS Version 8 (SAS Institute, 2000).
|
RESULTS |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
2 (1, N = 5937) = 9.1340, p =.003, than the nonparticipants. There were no differences in the proportion of Blacks and Whites between the participants and nonparticipants, 2 (1, N = 5937) = 0.0796, p =.778. Table 1 presents the demographic profile of the final sample (n = 4,275) with corresponding p values from the t tests (two-tailed) and chi-squares. On average, Blacks were younger, t(3096) = 13.14, p <.001, had fewer years of education (10.9 years vs. 13.8 years), t(3603) = 27.27, p <.001, and had lower average household income, 2 (9, N = 4152) = 564.3444, p <.001, than Whites. For both Blacks and Whites, female individuals composed the majority of the population—61% of Blacks and 63% of the Whites, 2 (1, N = 4275) = 1.2999, p =.254.
|
|
= 0.467, p <.001), after adjustment for age, male sex, and the follow-up time. As shown by the Black race x Time interaction term, Black–White differences in CES-D scores tended to increase during follow-up ( = 0.025, p =.004), although the magnitude of the longitudinal race effect was small relative to overall (baseline) differences in CES-D scores between Blacks and Whites. The other terms in the model indicate that CES-D scores were higher with greater baseline age ( = 0.025, p <.001), and were lower among men at baseline compared with women ( = –0.285, p <.001). The significant Time and Time2 terms reflect the curvilinear pattern of increase followed by a decrease in CES-D scores over time, consistent with the mean unadjusted scores shown in Table 2. The Time x Male sex and Time2 x Male sex interaction terms indicate that the curvilinear pattern of CES-D scores over time was significantly more pronounced among men than women.
|
= 0.143, p =.043), such that Black–White differences in CES-D scores were greater among men than women. We used the results of this model to illustrate the magnitude of Black–White differences in depressive symptom levels as a function of sex and time since baseline by plotting predicted CES-D scores for a 75-year-old person (Figure 1). The figure shows that Black men and women had approximately 60% higher predicted CES-D scores than White men and women, respectively. The figure also shows the clear gender effect, with significantly higher symptoms levels among women than men by race. Finally, the figure shows the curvilinear trend of increase and decrease in depressive symptom levels during follow-up, which is more pronounced among men.
|
= 0.423, p <.001 in Model 2 to = 0.225, p <.001 in Model 3. The Black race x Male sex interaction term was also reduced and became nonsignificant, from = 0.143, p =.043 in Model 2 to = 0.065, p =.359 in Model 3, resulting in a decrease in estimated Black–White differences in CES-D scores at baseline among men from an estimated 0.565, p <.001 to an estimated 0.291, p <.001 (which we calculated by summing the estimates for Black race and Black race x Male sex). As we expected, education ( = –0.039, p <.001) and income ( = –0.054, p <.001) were both inversely associated with symptom levels; that is, both fewer years of formal education and lower income were associated with more depressive symptoms. However, education and income did not have an impact on the longitudinal effect associated with race ( = 0.025, p =.005), indicating that these SES variables did not modify the slight increase in Black–White differences over time. The longitudinal, curvilinear, and male sex trends in depressive symptoms remained relatively unchanged across the three models. We performed secondary analyses to determine if the inference from the models was similar by using the dichotomous-level CES-D variable versus the continuous-level CES-D variable. Table 4 shows the unadjusted prevalences of elevated depressive symptoms, which are similar to the mean CES-D scores observed in Table 2. Blacks reported a higher prevalence of elevated depressive symptoms than Whites. In the total population, the prevalence of elevated depressive symptoms ranged from 17.3% to 23.6% for Blacks and 8.8% to 12.2% for Whites. Women reported higher prevalence than men, with the prevalence of elevated depressive symptoms ranging from 10.7% to 25.2% for all women compared with 5.5% to 20.9% for all men. Overall, the prevalence of elevated depressive symptoms is higher at Cycle 2 than at baseline or Cycle 3. In the secondary analyses, we also ran the same three regression models as already described and depicted in Table 3, but we used the dichotomous CES-D as the outcome variable (data not presented here). In the first regression model, we observed the same race effect for depressive symptoms such that, at baseline, after we adjusted for age, sex, and time since baseline, Blacks had greater odds (odds ratio, or OR = 2.24; 95% confidence interval, or CI = 1.87, 2.70) of having four or more depressive symptoms than whites. Racial differences increased slightly but not significantly over time (OR = 1.03; 95% CI = 0.99, 1.07). Similar to the continuous CES-D scores, racial differences were greater among men than women, although the modifying effect of gender was not significant (OR = 1.23; 95% CI = 0.90, 1.68) for the dichotomous CES-D outcome. As in the primary analysis, racial differences in elevated CES-D levels were reduced by almost half after we controlled for education and income.
|
|
DISCUSSION |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
; Dunlop, Song, Lyons, Manheim, & Change, 2003), in our study adjusting for demographic confounders including age, sex, education, and income decreased the overall difference by race but did not eliminate it altogether. In fact, after adjustment, the association between race and depression dropped approximately 50% but was still robust. The literature regarding the course of late-life depressive symptoms is mixed and our data showed equally mixed patterns over time. Although our baseline association of age with depressive symptoms was positive, suggesting an increase in symptoms with increasing age, the longitudinal pattern was more ambiguous. Among Whites, there did not appear to be consistent changes in depressive symptoms over time; however, among Blacks, there was an increase in symptoms over time, especially during early stages of follow-up. This resulted in a widening of racial differences in depressive symptoms over time. This longitudinal race effect remained consistent, regardless of other covariates entered into the models, particularly education and income. Overall, our data indicated a curvilinear pattern in the change of depressive symptoms over time, which varied by sex. It is unclear whether this pattern represents meaningful deviations from either stable symptom levels or a linear increase over time. It is possible that the initial rise in reported symptoms at the first follow-up is due to increasing morbidity as the cohort ages. We can also speculate that there was a greater disclosure of negative emotions during the first follow-up, as a result of a greater sense of familiarity with the interviewers. The subsequent decline in symptom levels at the second follow-up is perhaps attributable to increasingly selective attrition of more depressed participants as a result of mortality. More frequent observations at smaller intervals may be required to investigate the exact nature of the longitudinal course of depressive symptoms in older populations. Finally, we found that Black–White differences in depressive symptoms were greater among men than women.
The results from our study are unique and contrary to the existing research. Although other studies have found that older Whites have either more symptoms than older Blacks (Gallo et al., 1998) or that there are no differences (Blazer et al., 1998), our data show the alternative, that older Blacks had higher depressive symptoms than older Whites. There are two possible explanations for our results. First, our data are more recent. CHAP baseline data collection started in 1993 with follow-up data collection through 2003. Our data may reflect recent changes in mental health status that have gone unnoticed to this point. Most of the existing research in this area was conducted with data that were collected in the 1980s; thus, the differences we observed in this study may be a reflection of a recent widening of the mental health gap between Blacks and Whites. Second, our results may be symptomatic of the increase in overall physical health disparities. In fact, recent evidence suggests that racial disparities for major health outcomes increased during the 1980s and 1990s (Levine et al., 2001).
There is ample documentation of racial or ethnic health disparities (e.g., see U.S. DHHS, 2000, 2003, 2004). Arguably, the most common explanation for health disparities has been the profound inequalities in socioeconomic resources (Williams, 1998). In our data, the association between race and depressive symptoms was decreased by approximately half following an adjustment for education and income; however, the association remained robust and significant. Other studies have also found that SES fails to fully account for health disparities (Geronimus, Bound, Waidmann, Hillemeier, & Burns, 1996), most likely because the most commonly used indicators of SES (e.g., education and income) do not fully capture the differences in economic status between races (Williams & Collins, 1995). For example, with equivalent levels of education, Blacks have lower earnings than Whites; furthermore, with equivalent levels of income, Blacks may have less wealth and purchasing power than Whites (Williams & Collins, 1995). Other explanations for the health disparities have included access to health care and treatment issues (Cohen, 2003; U.S. DHHS, 2001) and psychosocial conditions such as social resources and networks, early life experiences including discrimination, and neighborhood or environment issues (Barnes, Mendes de Leon, Wilson, et al., 2004; Barnes, Mendes de Leon, Bienias, & Evans, 2004).
Disparities in mental health outcomes have not been as widely documented as disparities in physical health; nonetheless, depression is a vital component of overall quality of life in older adults and has tremendous health and economic implications. In addition to its status as a mental disorder in its own right, depression is considered a key risk factor for important physical health outcomes, including mortality (Barefoot & Schroll, 1996; Simonsick, Wallace, Blazer, & Berkman, 1995; Whooley & Browner, 1998), common chronic conditions (Barefoot & Schroll; Everson-Rose et al., 2005; Ford et al., 1998; Jonas & Mussolino, 2000), and overall poor health (Bruce, Seeman, Merrill, & Blazer, 1994). There are also serious financial ramifications for depression. The financial costs related to depression have been estimated at $44 billion in the United States per year (Greenberg, Stiglin, Finkelstein, & Berndt, 1993), and research has shown that elderly, depressed patients have approximately 50% higher total health care costs than their nondepressed peers, regardless of chronic morbidity (Katon, Lin, Russo, & Unutzer, 2003).
Healthy People 2010's objective to "increase the proportion of adults with mental disorders who receive treatment" (U.S. DHHS, 2004) presumes identification of depression, although depression in older adults is often underdiagnosed, misdiagnosed, or dismissed as a normal part of aging. Our data indicate a need for aggressive identification and treatment of depression in older adults, particularly older Blacks.
Our study has several strengths. First, our data come from a population-based sample of community-dwelling older adults that was drawn from a census of a community population. Second, our biracial sample has good representation of the Black population (62%) with adequate power to detect differences. Third, we used three waves of data spread approximately 3 years apart whereas other studies in this area have typically relied on fewer follow-ups, which may be inadequate to capture meaningful change over time.
There are limitations to our study. First, CHAP is a study of Chicago's urban-dwelling, biracial, older adult population. Therefore, our sample may not be representative of older adults living in smaller cities, rural areas, or other geographic areas of the country, or of older adults from other ethnic groups. Second, only three waves of data at 3-year intervals were available for analysis, which may still be too sparse for us to examine changes in symptom levels over time. Third, we used the 10-item CES-D, which may not fully capture the range of depressive symptoms in this population; the mean symptoms reported over time ranged from only 1.1 to 2.1. Furthermore, there is a growing body of evidence of unique measurement properties of the CES-D across diverse populations (Blazer et al., 1998; Callahan & Wolinsky, 1994; Long-Foley, Reed, Mutran, & DeVellis, 2002). For example, it is well known that Blacks have more disadvantaged social conditions than Whites, including racial discrimination (Kessler, Mickelson, & Williams, 1999; Ren, Amick, & Williams, 1999). However, after removing 2 CES-D items that may confound depressive symptoms with perceptions of discrimination (i.e., "I felt that people disliked me" and "people were unfriendly") and rerunning the models, we found no substantive differences, suggesting that those items did not explain the observed racial differences in health.
In summary, we found a significant race difference in depressive symptoms in which Blacks have higher levels of depressive symptoms—approximately 60%—than Whites among older adults. Our data showed that a substantial portion of the race effect—roughly 50%—was due to measures of SES; however, unlike in other studies, the race effect did not disappear after we adjusted for SES measures. Furthermore, we observed a small but stable longitudinal effect of worsening symptoms over time as well as more pronounced symptoms for men than women. The association of race with depressive symptoms in older adults is complex, and the exact origin of these differences will have to be explored in more detail. Depression in older adults is an important issue for clinicians, service organizations and providers, health care plans, and policy makers, and those groups should be especially mindful of the increased likelihood of depressive symptoms among their older Black populations.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
Received for publication July 13, 2004. Accepted for publication December 22, 2004.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  R. S. Wilson, S. E. Arnold, T. L. Beck, J. L. Bienias, and D. A. Bennett Change in Depressive Symptoms During the Prodromal Phase of Alzheimer Disease Arch Gen Psychiatry, April 1, 2008; 65(4): 439 - 445. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
