Depression and health-related quality of life in ethnic minorities seeking care in general medical settings

Journal of Affective Disorders 58 (2000) 89–97 www.elsevier.com/ locate / jad Research report Depression and health-related quality of life in ethnic minorities seeking care in general medical settings a , b c d*Maga E. Jackson-Triche , J. Greer Sullivan , Kenneth B. Wells , William Rogers , c cPatti Camp , Rebecca Mazel aDepartment of Psychiatry and Biobehavioral Sciences, University of California at Los Angeles, Sepulveda Veterans Affairs Medical Center, 16111 Plummer Street OOP-B, Sepulveda, CA 91343, USA bDepartment of Psychiatry, University of Arkansas Medical Center, Arkansas, AR, USA cDepartment of Psychiatry and Biobehavioral Sciences, University of California at Los Angeles, RAND Corporation, Los Angeles, CA, USA dDepartment of Biostatistics, Tufts University Medical School, USA Received 8 December 1998; accepted 1 March 1999 Abstract Background: To examine ethnic groups differences in (a) prevalence of depressive disorders and (b) health related quality of life in fee-for-service and managed care patients (n 5 21 504) seeking care in general medical settings. Methods: Data are from the Medical Outcomes Study, a multi-site observational study of outpatient practices. The study screened patients of clinicians (family practice, internal medicine, cardiology, diabetology and endocrinology) for four chronic medical conditions; depression, coronary heart disease, hypertension and diabetes. A brief eight-item depression screener followed by the Diagnostic Interview Schedule-Depression Section (DIS) for screener positives identified depressed patients (n 5 2195). The Short Form Health Survey (SF-36) assessed health-related quality of life. Patient self-report determined ethnicity. Results: Before adjusting for demographic factors, African-Americans and Hispanics had highest rates of depressive symptoms. Asian-Americans had the lowest. After adjusting for demographics (particularly gender and income), we found ]] few statistically significant differences in prevalence or severity of depression. However, among the depressed, Whites were the most, and African-Americans the least likely to report suicidal ideation ( p , 0.01), and Hispanics and Whites were more likely to have melancholia ( p , 0.01). African-Americans reported the poorest quality of life. Limitations: DSM III criteria (though few changes in DSM IV), and relatively small sample size of Asian-Americans compared to other groups. Conclusions: Gender and socioeconomic status are more significant factors than ethnicity in determining risk for depressive disorder. However, ethnic differences in symptom presentation, and health-related quality of life could have clinical and social consequences, and merit further study. Ó 2000 Elsevier Science B.V. All rights reserved. Keywords: Depression; Ethnicity; Outpatient; Health-related quality of life; General medicine *Corresponding author. E-mail address: [email protected] (M.E. Jackson-Triche) 0165-0327/00/$ – see front matter Ó 2000 Elsevier Science B.V. All rights reserved. PI I : S0165-0327( 99 )00069-5 90 M.E. Jackson-Triche et al. / Journal of Affective Disorders 58 (2000) 89 –97 1. Introduction providers towards an appropriate level of suspicion for depression in ethnic minority group patients, and In many health care systems primary care pro- to add information about the possible differential viders function as gatekeepers for all medical care, effects of depression on quality of life of ethnic including psychiatric care (Eisenberg, 1985; Weiner minority group patients. et al., 1994). Because depression is widely recog- nized as a common, debilitating illness that often presents in primary care settings, the Agency for 2. Methods Health Care Policy and Research has developed depression guidelines for primary care (Agency for 2.1. Study design Health Care Policy and Research, 1993; Katon and Schulberg, 1992; Kessler et al., 1985; Williams et al., The MOS is an observational study of adult 1995). Yet, little is known about whether health care outpatients receiving care in one of three health care systems, and specifically primary care providers, systems: (1) large multispecialty group with mixed should alter their approaches to detection and treat- fee-for-service and prepaid coverage, (2) Health ment, to make them more appropriate for different Maintenance Organization, or (3) solo-practice with ethnic groups. Although, community epidemiological mixed prepaid and fee-for-service and small single studies, have found little variation by ethnicity in specialty group in Boston, Chicago and Los Angeles. prevalence of depression, few studies have addressed A representative sample of physicians in each system the issue of ethnic group differences in prevalence of care was asked to participate in the study. for patients seeking care in general medical settings Physician specialties included in this analysis are (Brown et al., 1995; Kessler et al., 1994; Somervell family practice, internal medicine, cardiology, dia- et al., 1989; Vega et al., 1998; Weissman and Myers, betology and endocrinology. 1971). Also, despite some evidence of ethnic vari- The patients of those providers who agreed to ation in clinical presentation, little is known about participate in the study were screened for one of four how severity of depression varies by ethnic group chronic medical conditions: depression, coronary (Cooper-Patrick et al., 1994; Gary et al., 1989; heart disease, hypertension and diabetes. Detailed Golding et al., 1990; Jones and Gray, 1986; Munoz descriptions of the sampling strategy have been et al., 1990; Raskin et al., 1975; Roberts, 1992; Ying, reported in previous articles (Rogers et al., 1993; 1988). Ethnic differences in severity and health- Tarlov et al., 1989; Wells et al., 1995; Wells et al., related quality of life could imply the need for 1992; Wells et al., 1989). Ethnicity was determined providers to modify their expectations and treatment by patient self-report. approaches. A two-stage screening process accomplished case To examine the prevalence of various types of finding for depression. Patients first completed an depressive disorders, their severity, and their effect eight-item screener that measured intensity of depre- on health-related quality of life across ethnic groups, ssive symptoms over the previous two weeks, and we analyzed Medical Outcomes Study (MOS) data. episodes of depressed mood over the previous twelve The MOS was an observational study of adults months (Burnam et al., 1988). Patients who ex- receiving outpatient care in Boston, Chicago and Los ceeded the brief screener cut-off score, and who Angeles in the mid-1980s (Tarlov et al., 1989). All were eligible for enrollment in the longitudinal phase 21 504 subjects who participated in the MOS were of the study, were contacted for a follow-up tele- insured, either through fee-for-service or prepaid phone interview. Those who agreed were given the plans. This paper is among the first to examine the depression section of the Diagnostic Interview prevalence of depression across ethnic groups in an Schedule (DIS) by telephone (Wells et al., 1988). Of insured general medical population, and we feel that the 21 504 patients screened in the study, 3536 the importance of this issue was sufficient to merit exceeded the eight-item screener cut off score. Of reanalysis of this data set, despite it being several these, 2195 completed the follow-up DIS interview. years old. Our results are intended to help guide A subset of the DIS-positive patients were given the M.E. Jackson-Triche et al. / Journal of Affective Disorders 58 (2000) 89 –97 91 Structured Interview Version of the Hamilton Depres- the cut-point for depression on the screening instru- sion Rating Scale (SI-HRS) (Hamilton, 1967; Potts ment. This DIS used DSM-III criteria to detect et al., 1990). affective disorder. Previous studies found it equiva- Of the 489 who were administered the SI-HRS, lent to face-to-face administration for lifetime disor- 366 (75%) were found to be currently depressed, and der (Robins et al., 1981; Wells et al., 1988). 123 (25%) met criteria for lifetime depression only. The overall response rate was 70%. Among those 2.2.3. Severity completing the DIS, a sub-sample was enrolled in a The study derived measures of severity from the longitudinal study and completed a self-administered depression section of the DIS, and chose those questionnaire (n 5 1216). reflecting factors important in determining both We created five groupings. The first three prognosis and treatment (Wells et al., 1995). Items included;(1) subjects who had ‘‘any symptom’’ of included; the number of lifetime and one-year symp- depression on the screener, (2) subjects who had toms of major depression or dysthymia, both major ‘‘sub-threshold’’ depressive symptoms (i.e., were depression and dysthymia in the past year, more than screener-positive but did not meet criteria for a one episode of depression, number of suicidal symp- depressive disorder), and (3) subjects who met toms in the previous year, and melancholia in the criteria for any current depressive disorder. Subjects previous year. defined as having any current depressive disorder met the following criteria: (a) lifetime diagnosis of 2.2.4. Health-related quality of life major depression or dysthymia as listed in DSM-III; The Short Form Health Survey (SF-36) is a 36- (b) and episode of depression or dysthymia during item instrument designed to measure health status in the past twelve months; (c) no remission from the clinical and general populations (Stewart et al., 1993; most recent depressive episode. Remission was Ware and Sherbourne, 1992). It measures eight key defined as eight weeks or more with two or less health concepts: physical functioning; role limita- symptoms of depression. We separately estimated tions due to physical problems; social functioning; prevalence of (4) major depressive and (5) bodily pain; general mental health; role limitations dysthymic disorder in the full sample, using the due to emotional problems; vitality, and general positive predictive value of the screener for each health perception. The form can be administered by disorder separately. Persons with a lifetime diagnosis self, telephone or by an in-person interviewer. Inter- of mania, or depression due only to grief were nal reliability and item-discriminant validity are excluded. excellent (McHorney et al., 1994). This measure is only available on subjects enrolled in the longi- 2.2. Measures tudinal study (n 5 1216). 2.2.1. Eight-item screener 2.3. Statistical methods The MOS eight-item screener is a brief self-report instrument based on the DIS (two-items) and the We used a multi-stage procedure to estimate true Center for Epidemiologic Studies Depression Scale (unadjusted) and adjusted prevalence of Depressive (CES-D; six items), designed to screen for depres- disorder by ethnic group status. First, among all sive disorders. This instrument has been found to outpatients, we estimated the probability of exceed- have high sensitivity (86%), specificity (95%) and ing the cut-point on the first stage depression positive predictive value (37%), and a low false screener for each ethnic group using logistic regres- negative rate in primary care patients (Burnam et al., sion. We averaged individual predicted probabilities 1988). using the parameters of this regression to obtain prevalence of depressive symptoms (i.e., exceeding 2.2.2. The telephone DIS the first-stage screener cut-point). Among those Trained personnel administered the depression exceeding the cut-point and eligible for and accept- section of the DIS by telephone to patients exceeding ing the second-stage DIS, we estimated the con- 92 M.E. Jackson-Triche et al. / Journal of Affective Disorders 58 (2000) 89 –97 ditional probability of having current depressive a chronic, persistent course. In order to decide what disorder by ethnic status using logistic regression. factors were critical to be included as covariates, we We used the parameters of that regression analysis to conducted preliminary sensitivity analyses with dif- generate a predicted conditional probability for each ferent combinations of demographic variables and individual in the full sample. To do this, we assigned determined that most of the variance between ad- a value of ‘‘0’’ to those who were negative on the justed and unadjusted findings was due to gender and first stage screener, and we assumed that the positive income, (measured as total family income). Overall predictive value for the first stage screener was the conclusions were not sensitive to the combination of same within an ethnic group for persons who were demographic factors used in adjustment, so final screener-positive and did or did not complete the models were based on contrasting unadjusted and DIS. Then, for each person in the sample, we then fully adjusted results. computed the product of the estimated probabilities Data were weighted for probability of response by for the first and second stages and averaged these clinician and patient at enrollment and at administra- predictions within ethnic group, correcting those tion of each instrument, so the reference population averages for the known sensitivity of the first-stage is all outpatients eligible for the MOS. Inferential screener based on data from independent samples statistics were corrected for the cluster sampling and (Burnam et al., 1988). We applied the same estimate weighting using the method of Huber, also know as of sensitivity across ethnic groups. The resulting the method of Liang and Zeger (Huber, 1967; Liang estimate is the unadjusted prevalence of current and Zeger, 1986). Significance tests between specific depressive disorder for an ethnic group. Ethnic status pairs of ethnic groups were carried out using t-tests is correlated with demographic socioeconomic status, based on regression parameters. While we did not so we conducted a parallel analysis using socio- formally correct for multiple statistical comparisons, demographic characteristics (age, sex, income and we interpret results in light of this issue. Where education) as covariates in each logistic regression, applicable, a test for an overall difference among all and used the actual characteristics of each individual ethnic groups was conducted using an F-test. For when calculating the individual probabilities at each analyses among depressed patients, sample sizes stage. This yielded an adjusted (for these demo- varied by item depending on whether the item was graphic factors) prevalence of current depressive answered on the questionnaire. disorder. To estimate differences in average severity of depression, or level of health-related quality of life 3. Results (HRQOL) among depressed persons by ethnicity, we estimated logistic (for dichotomous variables), or Of the 21 504 patients screened for depression, least-squares (for continuous variables) regression, 16% (n 5 3536) exceeded the screener cut-point using each severity /HRQOL measure as the depen- which indicated a high probability (positive predic- dent variable and ethnic status as the independent tive value 37%) of having current depressive disor- variable (and for adjusted analyses, including age, der. The majority of these had depressive symptoms gender, income and education as covariates). The but no current disorder, and 5% of the full sample model parameters were used to generate predicted (n 5 1092) had current depressive disorder. These unadjusted (no covariates) and adjusted (covariates) results are consistent with other estimates suggesting level of severity /HRQOL for each individual, and that the prevalence of major depression in general predictions were then averaged within ethnic group. medical settings lies between 5% and 9% (Katon and The subsample for these analyses is all patients with Schulberg, 1992; Regier et al., 1993). current depressive disorder who completed the sec- Demographic characteristics of patients with de- ond-stage DIS. pressive symptoms are presented in Table 1. African- Depression may affect functioning, level of in- Americans and Hispanics were significantly younger, come and, to some extent, educational attainment, had fewer years of education, and lower incomes especially when depression occurs early in life or has than Whites. African-Americans were more likely to M.E. Jackson-Triche et al. / Journal of Affective Disorders 58 (2000) 89 –97 93 Table 1 aDemographic characteristics of depressed (CESD 1 ) (unadjusted) F score Ethnicity White African Hispanic Asian Other All (2713) American (219) American (86) (3536) (454) (64) Mean age ***9.27 43.6 ab 40.1 a 39.0 b 40.0 43.5 42.7 % Male *2.42 31.5 a 25.0 a 31.1 31.4 27.0 30.4 Median adjusted income ($1000) ***40.95 22.9 abc 14.3 ad 15.2 be 21.4 def 16.5 cf 21.0 Mean years of education ***27.31 13.4 abcd 12.7 aef 11.7 beg 14.3 cfgh 12.1 dh 13.2 Ù Ù% Employed *3.05 65.0 a 69.5 74.1 a 73.2 64.3 66.4 % Retired ***5.43 12.2 ab 6.7 a 5.7 b 6.8 10.7 10.9 % Married ***10.13 48.0 ab 37.4 acde 60.2 bc 58.8 d 55.0 e 47.7 Ù ÙMean family size ***16.70 2.3 abc 2.7 ad 2.7 b 2.5 3.1 cd 2.4 Log of family size ***18.50 0.64 abc 0.81 a 0.85 b 0.73 0.92 c 0.67 Mean days in bed **4.33 1.3 a 2.1 a 1.5 1.0 1.8 1.4 Ù ÙCurrent health perception ***15.18 61.4 abc 53.1 ad 53.7 be 64.7 def 52.8 cf 59.7 a Notes: % or mean with the same letter indicates a significant difference between the groups sharing this letter ( p , 0.05). A letter Ùfollowed by a carrot ( ) indicates marginal significance ( p , 0.10). * indicates significant F score at p , 0.05. ** indicates significant F score at p , 0.01. *** indicates significant F score at p , 0.001. be female and single than White patients. Asian- cans. Asian-Americans had a lower estimated preval- Americans had significantly more years of education ence of disorder than all other ethnic groups. While than Whites. Some of these demographic differences African-Americans and Whites did not differ sig- (chiefly gender and income) accounted for differ- nificantly from each other in estimated prevalence of ences in estimated prevalence of depression. current depressive disorder, African-Americans had a Differences in depression prevalence by ethnic higher prevalence of subthreshold symptoms. His- group status are presented in Table 2. In unadjusted panic patients had a particularly high prevalence of analyses, African-Americans and Hispanics were dysthymic disorder (6.8%) relative to Whites (3.1%). more likely to have depressive symptoms than Asian-Americans had a low (0.9%) prevalence of Whites or Asian-Americans. Hispanics had a higher either type of depressive disorder. Except for Asian- estimated unadjusted prevalence of current depres- Americans, who had a lower prevalence of each sive disorder than either Whites or African-Ameri- category of depressive disorder, ethnic group differ- Table 2 aPrevalence of depression by ethnicity Ethnicity Screener % Any depressive % Sub-threshold % Any depressive % Major % Dysthymia sample symptoms symptoms disorder depression n Unadj Adj Unadj Adj Unadj Adj Unadj Adj Unadj Adj White 16 989 14.5 ab 15.3 9.5 a 9.9 5.0 ac 5.4 a 3.4 a 3.7 a 3.1 abe 3.4 a African-American 2533 18.5 ac 15.7 12.6 a 10.7 6.0 bd 5.0 b 3.7 b 3.1 b 3.9 c 3.2 b Hispanic 1009 20.6 bd 16.3 11.2 9.1 9.4 abf 6.9 c 4.8 c 3.5 c 6.8 ad 5.0 c Asian-American 527 12.9 cd 11.4 11.3 10.4 1.6 cdef 1.4 abcd 0.9 abcd 0.8 abc 0.9 bcdf 0.8 abcd Other 446 18.1 15.7 10.7 9.4 7.5 f 6.2 d 3.6 d 3.1 7.0 ef 5.7 d All combined 21 504 15.3 15.3 10 10 5.3 5.3 3.5 3.5 3.4 3.4 Ethnicity Effect (F ) 13.87*** 1.89 a Note: * p , 0.05, ** p , 0.01, *** p , 0.001. Unadj 5 unadjusted, Adj 5 adjusted. Adjusted means adjusted for age, gender, site and income. Cells that share a letter in a column are significantly different ( p , 0.05). 94 M.E. Jackson-Triche et al. / Journal of Affective Disorders 58 (2000) 89 –97 Table 3 aAdjusted severity of depression by ethnic status Ethnicity % Melancholia % . 2 Depression % Prior attempted suicide % With suicidal ideation Adj Adj Adj Adj White 12.3 65.2 17.0 40.1 a African-American 6.8 59.8 12.0 19.2 a Hispanic 11.0 69.7 13.4 22.4 Asian-American 0.8 48.6 10.8 28.6 Other 2.9 59.4 7.4 37.9 All Combined 11.1 64.5 15.8 35.9 Sub-sample (n) 2142 1092 2142 2143 Ethnicity Effect (F ) 3.83** 1.16 2.22 4.43** a Note: * indicates p , 0.05, ** indicates p , 0.01, *** indicates p , 0.001. Adj 5 adjusted. Adjusted means adjusted for age, gender, site and income. Cells that share a letter in a column are significantly different ( p , 0.05). ences were not significant after demographic adjust- While these levels of significance are not high ment. enough to survive a formal multiple comparisons Table 3 presents adjusted severity of depression by test, such a consistent pattern suggests that such a ethnic status. In both the adjusted and unadjusted formal test may be too conservative. The general analyses, there were no significant ethnic differences pattern across domains is for Whites and African- in percent with a prior suicide attempt, or in the Americans to have the poorest HRQOL. Asian- percent with recurrent depression. Whites reported Americans had the poorest role functioning due to more suicidal ideation ( p , 0.01) than other groups. physical health. Unadjusted analyses showed similar Table 4 presents our analyses of HRQOL. For trends. Whites appeared healthier in unadjusted than these analyses, precision was relatively poor for most in adjusted analyses (i.e., their HRQOL is worse than comparisons between specific ethnic groups (com- would be expected given their sociodemographic parisons of White and African-Americans have the characteristics). best precision), so we focus on the tests of overall Table 5 shows ethnic group differences in adverse differences (Ethnicity Effect F-test) by ethnic status. life events. African-Americans had significantly Among the eight domains of HRQOL, six adjusted more ( p , 0.01) adverse life events, especially ad- analyses show a main effect ( p , 0.05) of ethnicity. verse financial events. They also had more bereave- Table 4 aHealth related quality of life measurement of depressed Ethnicity SF-36 Scales Physical Lack of Role limits due Role limits due Emotional Social Energy/ General health functioning bodily pain to physical health to emotional health well-being functioning fatigue perceptions Unadj Adj Unadj Adj Unadj Adj Unadj Adj Unadj Adj Unadj Adj Unadj Adj Unadj Adj White 74.3 73.7 64.4 63.8 50.1 49.2 48.8 48.2 56.6 56.4 69.7 69.0 45.8 45.6 54.4 53.9 African-American 69.2 a 72.1 58.6 61.1 50.7 53.9 54.0 56.0 59.5 60.1 64.9 67.1 45.9 46.6 53.2 55.1 Hispanic 80.6 a 79.2 65.3 66.2 62.8 62.8 50.8 53.5 59.2 60.5 69.0 71.8 50.6 51.2 62.4 62.5 Asian-American 83.8 83.3 75.2 74.1 40.6 42.2 61.0 60.7 64.2 63.7 82.4 82.1 52.3 51.4 64.0 65.4 Other 81.2 86.6 65.5 68.4 56.9 61.4 59.7 63.3 62.6 63.9 66.4 70.7 52.8 55.2 58.8 62.4 All combined 74.1 74.1 63.7 63.7 50.8 50.8 50.0 50.0 57.3 57.3 69.0 69.0 54.8 46.3 54.8 54.8 Sub-sample (n) 1177 1177 1212 1212 1184 1184 1173 1173 1210 1210 1216 1216 1190 1194 1190 1190 Ethnicity Effect (F ) 3.01* 3.53** 2.47* 1.36 1.56 2.78* 1.17 2.40* 1.58 2.49* 1.84 1.05 2.26 2.48* 2.26 3.51** a Note: * indicates p , 0.05, ** indicates p , 0.01, *** indicates p , 0.001. Unadj 5 unadjusted, Adj 5 adjusted. Adjusted means adjusted for age, gender, site and income. Cells that share a letter in a column are significantly different ( p , 0.05). Each scale is standardized on a 0–100 scale, with 100 meaning perfect health. M.E. Jackson-Triche et al. / Journal of Affective Disorders 58 (2000) 89 –97 95 Table 5 aAdverse life events in depressed patients by ethnic status Ethnicity Adverse life events Mean No. of Mean No. of Mean No. of Mean No. of % With bereavement Amount of change Amount of time events overall financial events relationship events bereavement events event in past year overwhelmed by change Unadj Adj Unadj Adj Unadj Adj Unadj Adj Unadj Adj Unadj Adj Unadj Adj White 2.3 a 2.3 0.49 a 0.48 0.44 a 0.46 0.72 0.73 18.8 19.3 2.2 2.2 3.9 3.9 African-American 2.9 a 2.7 0.55 0.59 0.73 a 0.63 0.79 0.74 29.5 26.5 2.1 2.1 3.7 3.8 Hispanic 2.5 2.2 0.61 0.58 0.48 0.36 0.59 0.50 19.6 18.2 2.2 2.2 3.9 3.9 Asian-American 0.9 1.1 0.34 a 0.35 0.28 0.39 0.10 0.12 18.5 23.4 3.1 3.0 3.7 3.6 Other 3.2 3.0 1.03 1.00 0.63 0.55 0.63 0.58 35.4 34.8 2.1 2.1 3.9 3.9 All combined 2.4 2.4 0.52 0.52 0.48 0.48 0.71 0.71 20.6 20.6 2.2 2.2 3.9 3.9 Sub-sample (n) 1073 1073 1071 1071 1045 1045 1073 1073 1067 1067 1051 1051 1049 1049 Ethnicity Effect (F ) 4.13** 2.17* 2.80* 2.86* 4.87*** 2.17* 1.74 2.37 3.01* 1.98 2.96* 2.70* 0.59 0.28 a Note: *indicates p , 0.05, ** indicates p , 0.01, *** indicates p , 0.001 Unadj 5 Unadjusted, Adj 5 adjusted. Adjusted means adjusted for age, gender, site and income. Cells that share a letter in a column are significantly different ( p , 0.05). ment events, but these results were marginally the screener would be much lower for Asian-Ameri- significant. cans, if the DIS is valid for this group. This clearly is an area where further cross-cultural research is needed. 4. Discussion Is the clinical presentation, in terms of indicators of severity and HRQOL of the depressed, similar This study revealed large and consistent demo- across ethnic groups? Here we are more cautious graphic differences by ethnic group, and demon- with conclusions, because of limited precision for strates the need to carefully control for these factors specific inter-ethnic group comparisons. On two in any analysis comparing ethnic groups. Otherwise important indicators of depression severity (recurrent there is a substantial risk of attributing observed depression, number of depressive symptoms) there differences to ethnicity alone (Williams, 1986). The were no marked ethnic differences. Recurrent de- results of this study raise new questions about pression was common across ethnic groups. How- specific ethnic differences in functioning, markers of ever, we found ethnic differences in prevalence of severity, and patterns of depression among the suicidal ideation and melancholia (Cooper-Patrick et depressed. Asian-American patients had a distinctly al., 1994). Whites had a very high risk for suicidal different pattern of depression than other ethnic thoughts and melancholia, while Hispanics had a groups. Despite having a comparable level of depres- high risk of melancholia relative to African-Ameri- sive symptoms (after sociodemographic adjustment), cans. Ethnic differences in suicidal ideation, though, Asian Americans had a lower prevalence of both were not accompanied by significant differences in major depression and dysthymic disorder. The pre- history of suicide attempts. valence finding is consistent with some community This study suggests that, with the exception of studies of Asian samples in the USA and other poorer role functioning due to physical health for countries (Nakane et al., 1991; Takeuchi et al., Asian-Americans, depressed Whites and African- 1998). However, the similar prevalence of symp- Americans have the lowest levels of functioning and toms, but lower prevalence of disorder raises ques- well being. Further research is needed to confirm this tions about both appropriateness of diagnostic finding. Ethnic differences in HRQOL could have criteria, and screening instrument sensitivity for significant clinical care and social cost implications Asian-Americans. The characteristics of our first (e.g., target symptoms, work performance). stage screener may differ markedly for the Asian- We examined ethnic group differences in adverse Americans. Indeed, the positive predictive value of life events, a major risk factor for depression and 96 M.E. Jackson-Triche et al. / Journal of Affective Disorders 58 (2000) 89 –97 poor HRQOL, and found that African-Americans a suggestion of higher prevalence of some specific had significantly more adverse life events, especially severity markers (suicidal ideation and melancholia) financial losses, than Whites. It is possible that the in Whites and Hispanic patients relative to African- pattern of worse HRQOL among depressed African- Americans. Whites and African-Americans appear to Americans relates to a greater number of stressful have worse HRQOL across multiple domains. life events or possibly other community risk factors The implications for clinical practice is that clini- not controlled for in our analyses (Otten et al., 1990; cians should have at least as high an index of Sherbourne et al., 1992). This, also, is an area for suspicion for depression in African-Americans and future research. Hispanics as Whites, and should keep in mind that This study has limitations. The data were collected income and gender are commonly confounded with in mid-1980s and, although the basic criteria for ethnic status. However, there appear to be significant diagnosing depression have changed little in the past ethnic differences in the functioning and well being few years, this study used DSM III. Patterns of of the depressed. This is an important area for future utilization may have changed, altering the prevalence research. and/or severity of depression in a consecutive sample of outpatients seeking care in a general medical practice. Nevertheless, such data on preval- References ence and symptom presentation by ethnicity are rare, and we hope this study to be an important baseline Agency for Health Care Policy and Research, 1993. Clinicalfor further investigation. We studied particular urban Practice Guideline: Depression in Primary Care, Detection and areas and managed care organizations, matched by Diagnosis, Vol. 1, AHCPR. Agency for Health Care Policy and Research, 1993a. Clinicalgeographic area with fee-for-service practice loca- Practice Guideline: Depression in Primary Care, Treatment oftions. The results may not generalize to other regions Major Depression, Vol. 2, AHCPR.or types of practices. We had limited precision for Brown, D.R., Ahmed, F., Garey, L.E., Milburn, N.G., 1995. Major some ethnic comparisons and were unable, because depression in a community sample of African-Americans. Am. of sample sizes, to examine potentially significant J. Psychiatry 152 (3), 373–378. differences within groups. We relied on a multi-stage Burnam, M.A., Wells, K.B., Leake, B., Landsverk, J., 1988. Development of a brief screening instrument for detectingestimation process to determine prevalence of de- depressive disorders. Med. Care. 26 (8), 775–789.pressive disorders. This required assumptions about Cooper-Patrick, L., Crum, R.R., Ford, D.E., 1994. Identifying stability of characteristics of the first-stage screener Suicidal ideation in general medical patients. JAMA 272 (22), within and across ethnic groups, and about screener 1757–1762. positives who did, and did not, complete the second Cooper-Patrick, L., Crum, R.R., Ford, D.E., 1994. Characteristics of patients with depression who receive care in general medicalstage. Results for Asian-Americans, particularly in and specialty mental health settings. Med. Care 32 (1), 15–24.terms of prevalence of depressive disorders, must be Eisenberg, J.M., 1985. The internist as gatekeeper: Preparing the viewed with caution because of the relatively small general internist for a new role. Ann. Intern. Med. 102, 537– sample size and uncertainty about the characteristics 543. of the screener in this sample. The strengths of the Gary, L.E., Brown, D.R., Milburn, N.G., Ahmed, F., Booth, J., 1989. Depression in Black American Adults: Findings from thestudy are the range of general medical practices Norfolk Area Health Study, Institute Urban Affairs, Washing-represented, breadth of health status measures, and ton, DC. variability in patient ethnicity. Golding, J.M., Karno, M., Rutter, C.M., 1990. Symptoms of majorIn summary, the prevalence of depressive symp- depression among Mexican-Americans and non-Hispanic toms and disorder in patients seeking general medi- Whites. Am. J. Psychiatry 147, 861–866. Hamilton, M., 1967. Development of a rating scale for primarycal care in private practice appears to be similar depressive illness. Br. J. Social Clin. Psychol. 6 (4), 278–296.across ethnic groups, once underlying differences in Huber, P.J., 1967. The behavior of maximum likelihood estimatessociodemographic status (primarily gender and in- under non-standard conditions. Proceedings of the 5th Berkeley come) are taken into account. The results for Asian- Symposium in Mathematics and Statistics and Probability 1, American outpatients must be viewed with caution 221–233. because of the previously noted limitations. There is Jones, E.J., Gray, B.A., 1986. Problems diagnosing schizophrenia M.E. Jackson-Triche et al. / Journal of Affective Disorders 58 (2000) 89 –97 97 and affective disorders among blacks. Hosp and Comm Psychi- Somervell, P.D., Leaf, P.J., Weissman, M.M., Blazer, D.G., Bruce, atry. 37 (1), 61–65. M.L., 1989. The prevalence of major depression in black and Katon, W., Schulberg, H., 1992. Epidemiology of depression in white adults in five United States communities. Am. J. Epi- primary care. Gen. Hosp. Psychiatry 14, 237–247. demiol. 4, 725–735. Kessler, L.G., Cleary, P.D., Burke, J.D., 1985. Psychiatric dis- Stewart, A.L., Sherbourne, C.D., Wells, K.B., Burnham, M.A. et orders in primary care: Results of a follow-up study. Arch. al., 1993. Do depressed patients in different treatment settings Gen. Psychiatry 42, 583–587. have different levels of well-being and functioning? J. Consult. Kessler, R.C., McGonagle, K.A., Zhao, S., Nelson, C., Hughes, Clin. Psychol. 61, 849–857. M., Eshleman, S., Wittchen, H., Kendler, K.S., 1994. Lifetime Takeuchi, D.T., Chung, R.C., Lin, K., Shen, H., Kurasaki, K., and 12-month prevalence of DSM-III-R psychiatric disorders Chun, C., Sue, S., 1998. Lifetime and twelve-month prevalence in the United States. Arch. Gen. Psychiatry 51, 8–19. rates of major depressive episodes and dysthymia among Liang, K.Y., Zeger, S.L., 1986. Longitudinal data analysis using Chinese Americans in Los Angeles. Am. J. Psychiatry 155 generalized linear models. Biometrika 73, 13–22. (10), 1407–1414. McHorney, C.A., Ware, Jr. J.E., Lu, J.F., Sherbourne, C.D., 1994. Tarlov, A.R., Ware, Jr. J.E., Greenfield, S., Nelson, E.C., Perrin, The MOS 36-item Short-Form Health Survey (SF-36). III. E., Zubkoff, M., 1989. The Medical Outcomes Study: An Tests of data quality, scaling assumption, and reliability across application for monitoring the results of medical care. JAMA diverse patient groups. Med. Care. 32 (1), 40–66. 262 (7), 925–930. Munoz, R.A., Boddy, P.P., Prime, R., Munoz, L., 1990. Depres- Vega, W.A., Kolody, B., Aguilar-Gaxiola, S., Alderete, E., sion in the Hispanic community: Preliminary findings in Catalano, R., Caraveo-Anduaga, J., 1998. Lifetime prevalence Hispanic general medical patients at a community health of DSM-III-R psychiatric disorders among urban and rural center. Ann. Clin. Psychiatry 2, 115–120. Mexican Americans in California. Arch. Gen. Psychiatry 55, Nakane, Y., Radford, M., Yan, H., Wang, X., Lee, H.Y., Min, S.K., 771–778. Michitsuji, S., Ohtsuka, T., 1991. Comparative study of Ware, J.E., Sherbourne, C., 1992. The MOS 36-item Short Form affective disorders in three Asian countries. II: Differences in Health Survey (SF-36): Conceptual framework and item selec- prevalence rates and symptom presentation. Acta Psyhciatr. tion. Med. Care. 30 (6), 473–483. Scand. 84 (4), 313–319. Weiner, M.F., Edland, D.S., Luszczynska, H., 1994. Prevalence Otten, Jr. M.W., Teusch, S.M., Williamson, D.F., Marks, J.S., Jr. et and incidence of major depression in Altzheimer’s Disease. al., 1990. The effect of known risk factors on the excess Am. J. Psychiatry 151 (7), 1006–1009. mortality of black adults in the US. JAMA 263 (6), 845–850. Weissman, M.M., Myers, J.K., 1971. Rates and risks of depressive Potts, M.K., Daniels, M., Burnam, M.A., Wells, K.B., 1990. A symptoms in a United States urban community. Acta Psyhciatr. structured interview version of the Hamilton Depression rating Scand. 57, 219–231. scale. J. Psychiat. Res. 24, 335–350. Wells, K.B., Burnam, M.A., Camp, P., 1995. Severity of depres- Raskin, A., Crook, T.H., Herman, K.D., 1975. Psychiatric history sion in prepaid and fee-for-service general medical and mental and symptoms differences in black and white depressed health specialty practices. Med. Care. 33 (4), 350–364. patients. J. Consult. Clin. Psychol. 43, 73–80. Wells, K.B., Burnam, M.A., Leake, B., Robins, L.N., 1988. Regier, D.A., Narrow, W.E., Rae, D.S., Manderscheid, R.W., Agreement between face-to-face and telephone administered Locke, B.Z., Goodwin, F.K., 1993. The de facto US mental and versions of the depression section of the NIMH Diagnostic addictive disorders service system. Arch. Gen. Psychiatry 50, Interview Schedule. J. Psychiatr. Res. 22, 207–220. 85–89. Wells, K.B., Burnam, M.A., Rogers, W.H., 1992. Course of Roberts, R.E., 1992. Manifestation of depressive symptoms depression for adult outpatients: Results from the Medical among adolescents: A comparison of Mexican Americans with Outcomes Study. Arch. Gen. Psychiatry 49 (10), 788–794. other minority populations. J. Nerv. Ment. Dis. 180 (10), Wells, K.B., Hays, R.D., Burnam, M.A., Rogers, W.H., Greenfield, 627–633. S., Ware, Jr. J.E., 1989. Detection of patients receiving prepaid Robins, L.N., Helzer, J.E., Croughan, J., Ratcliff, K., 1981. or fee-for-service care. JAMA. 262, 3298–3302. National Institute of Mental Health Diagnostic Interview Williams, D.H., 1986. The epidemiology of mental illness in Schedule: Its history, characteristics and validity. Arch. Gen. Afro-Americans. Hosp. Comm. Psychiatry 37, 42–49. Psychiatry 38, 381–389. Williams, J.W., Kerber, C.A., Mulrow, C.D., Medina, A., Aguilar, Rogers, W.H., Wells, K.B., Meredith, L.S., Sturm, R., Burnam, C., 1995. Depressive disorders in primary care: prevalence, M.A., 1993. Outcomes for depressed outpatients under prepaid functional disability and identification. JGIM 10, 7–12. or fee-for-service financing. Arch. Gen. Psychiatry 50, 517– Ying, Y.W., 1988. Depressive symptomatology among Chinese- 525. Americans as measured by the CES-D. J. Clin. Psychol. 44, Sherbourne, C.D., Meredith, L., Rogers, W., Ware, J., 1992. Social 739–746. support and stressful life events: Age differences and their effects on health related quality of life among the chronically ill. Qual. Life Res. 1, 235–246.