Racial and Ethnic Disparities in Depression Care in Community-Dwelling Elderly in the United States

Abstract

Objectives. We investigated racial/ethnic disparities in the diagnosis and treatment of depression among community-dwelling elderly.

Methods. We performed a secondary analysis of Medicare Current Beneficiary Survey data (n = 33 708) for 2001 through 2005. We estimated logistic regression models to assess the association of race/ethnicity with the probability of being diagnosed and treated for depression with either antidepressant medication or psychotherapy.

Results. Depression diagnosis rates were 6.4% for non-Hispanic Whites, 4.2% for African Americans, 7.2% for Hispanics, and 3.8% for others. After we adjusted for a range of covariates including a 2-item depression screener, we found that African Americans were significantly less likely to receive a depression diagnosis from a health care provider (adjusted odds ratio [AOR] = 0.53; 95% confidence interval [CI] = 0.41, 0.69) than were non-Hispanic Whites; those diagnosed were less likely to be treated for depression (AOR = 0.45; 95% CI = 0.30, 0.66).

Conclusions. Among elderly Medicare beneficiaries, significant racial/ethnic differences exist in the diagnosis and treatment of depression. Vigorous clinical and public health initiatives are needed to address this persisting disparity in care.

Depression is a significant public health concern for older Americans.¹ It has been estimated that 6.6% of older Americans experience an episode of major depression during 1 year.² If untreated or undertreated, depression can significantly diminish quality of life³ and increase mortality.⁴ Depression can complicate several comorbid general medical conditions that are common in older populations, including congestive heart failure,^5,6 diabetes,⁷ and arthritis.⁸ Antidepressant treatment and psychotherapy have been shown to be effective in increasing rates of remission for depression in older adults.⁹

Several studies during the 1990s identified racial/ethnic differences in the diagnosis and treatment of depression, both in the general adult population and among the elderly.^10–14 Although there was a general increase in rates of depression diagnosis and antidepressant use during this period, some studies suggest that these increases are not consistent across racial/ethnic subgroups¹⁵ and that disparities in the treatment of diagnosed depression are persistent.^10,16,17

More recent studies (often combining the nonelderly and elderly adult population rather than considering these groups separately) have provided mixed findings. Some evidence indicates that minority group members with depression continue to receive less mental health care than do non-Hispanic Whites, and some studies suggest that mental health treatment differences by race/ethnicity may have worsened in the early 2000s.^18–21 By contrast, 1 recent national study reported that although overall increases in treatment rates were modest in the 2000s, there were significant increases in treatment rates among African Americans, possibly narrowing the racial/ethnic gap among adults in general.²² However, this study did not examine disparities in treatment separately among older adults and was founded on household reported conditions that are only modestly related to provider diagnoses.²³ The pattern of diffusion of depression treatment may differ between elderly and nonelderly adults. Consequently, racial/ethnic differences in diagnosis and treatment among the elderly remain a potentially important public health concern.

We investigated whether there are racial/ethnic differences (1) in the rate of diagnosis of depression among the elderly, controlling for sociodemographic characteristics and depression symptoms (depressed mood and anhedonia) reported on a 2-item screener; and (2) in the treatment provided to those diagnosed with depression by a health care provider, adjusting for these covariates. In a large, nationally representative sample, we examined whether relationships between race/ethnicity and depression diagnosis or depression treatment are mediated by insurance coverage and perceived access to medical care, depression symptoms, and severity, or by other global measures of health such as self-reported health status and impairment in daily activities.

METHODS

The Medicare Current Beneficiary Survey (MCBS) is a 4-year rotating panel combining information from beneficiary interviews and Medicare claims for a nationally representative sample of Medicare beneficiaries, including those living in long-term care facilities. Beneficiary surveys are conducted at 4-month intervals. Respondents are sampled from the Medicare enrollment file. The sample is stratified by age (< 45, 45–64, 65–69, 70–74, 75–79, 80–84, and ≥ 85 years) and drawn within zip codes designated as primary sampling units. The oldest-old (≥ 85 years) and the disabled (≥ 64 years) were oversampled to allow detailed analysis of these subpopulations. For community dwellers (our study population), response rates for initial interviews ranged from 80% to 90%; once the first interview was completed, participation rates in subsequent rounds were > 95%. The survey includes questions on health care use and costs, health status, medical and prescription drug insurance coverage, access to care, and use of services by beneficiaries. Claims are only available for services financed through traditional indemnity plans and are not available for services financed by Medicare Managed Care plans; thus, our sample is restricted to the former.

Participants

Our study population included Medicare beneficiaries 65 years old and older. To standardize the observation period, we restricted the sample for each year to respondents living in the community for the entire year. We excluded individuals who died, became eligible for Medicare during the year, or were institutionalized at some point during the year. We excluded beneficiaries with a diagnosis of bipolar disorder because appropriate treatment strategies may differ for unipolar and bipolar mood disorders.²⁴ We also excluded enrollees with missing data on either race/ethnicity or depression symptom. The unit of analysis was the person-year, and because MCBS is a rotating panel, each person could contribute up to 3 observations to our analytic data set. This resulted in a study sample of 12 353 unique persons contributing 33 708 person-years of observations between 2001 and 2005. The sample was predominantly non-Hispanic White; 8.3% were African American, 1.9% were Hispanic, and 2.5% were non-Hispanic other.

Outcomes

An indicator variable for depression diagnosis describes whether the person had a medical care claim during the observation year with depression listed as a diagnosis. We conceptualized treatment, our second outcome measure, as receipt of either psychotherapy or antidepressant medications in the same observation year. Medication containers and explanations of benefits are reviewed, and detailed data on filled prescriptions are recorded during interviews. We identified antidepressant use from those survey responses and psychotherapy from the procedure codes indicated on the Medicare claims for professional services. The list of codes and drugs are available from the corresponding author upon request.

Independent Variables

The conceptual framework developed by Kilbourne et al. for “advancing health disparities research within the health care system” guided us in identifying and organizing our explanatory variables.²⁵ Because there is a growing consensus that groups other than those defined by race/ethnicity are at risk for being medically underserved, we controlled for gender, age (young-old vs old-old), and rural residence (areas other than metropolitan statistical areas). Race/ethnicity was self-identified; therefore, it reflects the underlying cultural perceptions and beliefs of the respondent.²⁶

We also focused on the effect of key factors that are potentially mutable (e.g., income, perceived access to care, organization and financing of health services such as availability of medical insurance supplementing Medicare, and quality of supplemental prescription drug coverage). For single respondents, if personal income was less than 150% of the federal poverty line for a 1-person family, we categorized the respondent as poor. We categorized married respondents as poor if the couple's income was less than 150% of the federal poverty line for a 2-person family. We categorized supplemental insurance as Medicaid; all other coverage (i.e., employer-sponsored or self-purchased private insurance); or none. We measured quality of supplemental prescription coverage (none, limited, or comprehensive) by the proportion of self-reported total prescription drug costs that were paid out-of-pocket, with less than 30% indicating comprehensive coverage. We constructed perceived access to care measures (financial cost barrier, service barrier, and dissatisfaction with care) using established survey response patterns.²⁷ We defined financial cost barriers as difficulties getting needed health care or seeing a doctor because of 1 or more listed cost-related reasons. Service barriers included trouble getting needed health care because of a lack of transportation to the doctor or hospital or difficulty getting an appointment; not seeing a doctor because the enrollee could not get an appointment soon enough, no doctor was available, or the enrollee had no transportation; or dissatisfaction with the waiting time, the location of the doctor, or the paperwork. We classified enrollees as dissatisfied with care if they were dissatisfied or very dissatisfied with information about their diagnosis, quality of medical care received, doctor's concern for overall health, follow-up care after initial treatment, time spent with the doctor, the doctor's thoroughness, the doctor's attitude, or the doctor's competence.

Control Variables

Two questions on depression symptoms were available in the survey. The first question assessed sadness and was worded as “In the past 12 months, how much of the time did you feel sad, blue, or depressed?” Possible responses included all the time (4), most of the time (3), some of the time (2), a little of the time (1), or none of the time (0). The second question assessed anhedonia with a binary response format and was worded as “In the past 12 months, did you have 2 weeks or more when you lost interest or pleasure in things that you usually cared about or enjoyed?” We assigned a score of 2 to the presence of anhedonia. Finally, we generated a depression symptom score ranging from 0 (no symptom) to 6 (sad all the time and lost interest in things). We categorized scores lower than 4 as low symptoms; 4 as medium; and 5 and 6 as high. These questions were similar to a 2-item depression screener (Patient Health Questionnaire or PHQ-2) with acceptable psychometric properties.²⁸ The PHQ-2 inquires about the frequency of depressed mood and anhedonia over the past 2 weeks, scoring each as 0 (not at all) to 3 (nearly every day).

We controlled for measures of health, including (1) self-reported health status (categorized as excellent, very good, or good vs fair or poor) and (2) impairment in activities of daily living (ADL; bathing or showering, dressing, eating, getting in and out of bed or a chair, walking, and using the toilet) or instrumental activities of daily living (IADL; using a telephone, light housework, heavy housework, preparing meals, shopping for personal items, and managing money).

Statistical Methods

We computed Rao-Scott χ² statistics to test for differences in the distribution of race/ethnicity, diagnosis rates, and treatment rates by explanatory variables (Table 1). In multivariate analyses, we estimated logistic regression models to assess the adjusted association of each covariate with the probability of being diagnosed and treated for depression (Table 2). The MCBS sampling design is a multistage probability sampling with 3 stages. The unit of analysis was the person-year, and each person could have contributed up to 3 observations to the data set. We used the SURVEYFREQ and SURVEYLOGISTIC procedures in SAS version 9.2 (SAS Institute, Cary, NC) to account for the complex sampling design and the within-person correlation across time. We weighted all calculations (except sample sizes) to reflect national estimates.

TABLE 1—

Population Characteristics (Demographic, Socioeconomic, Coverage or Access, and Clinical Indicators) and Outcomes, Stratified by Race/Ethnicity: Medicare Current Beneficiary Survey, United States, 2001–2005

Population Characteristics	%a (No.)	Race/Ethnicity					Outcomes
		Non-Hispanic White, %	African American, %	Hispanic, %	Non-Hispanic or Other, %	P b (Selected Two-Way Comparisonsc)	Depression Diagnosis Rates, %	Sample Size With Depression Diagnosis, No.	Treatment Rates Among Diagnosed, %
All	100.0 (33 708)	…	…	…	…	…	6.20	2122	71.5
Race/Ethnicityde
Non-Hispanic White	87.3 (29 402)	…	…	…	…	…	6.43	1910	73.0
African American	8.3 (2875)	…	…	…	…	…	4.23	131	60.3
Hispanic	1.9 (656)	…	…	…	…	…	7.17	49	63.4
Non-Hispanic other	2.5 (775)	…	…	…	…	…	3.78	32	39.8
Genderd						.009 (1≠2; 1 = 3; 2 = 3; 1 = 4)
Men	43.3 (14 675)	43.9	38.4	43.4	40.4		3.88	597	69.3
Women	56.7 (19 033)	56.2	61.6	56.6	59.6		7.96	1525	72.3
Age						.001 (1≠2; 1≠3; 2≠3; 1 = 4)
65–74	48.6 (14 589)	48.2	53.3	38.1	54.7		6.03	891	74.3
75–84	40.0 (13 904)	40.4	35.0	49.5	36.5		6.16	869	69.0
≥ 85	11.4 (5215)	11.4	11.8	12.4	8.9		7.03	362	68.9
Locatione						< .001 (1 = 2; 1≠3; 2 = 3; 1≠4)
Nonmetro	27.1 (10 541)	28.4	22.2	10.1	10.2		6.29	661	67.6
Metro	72.9 (23 164)	71.6	77.9	89.9	89.7		6.16	1461	73.0
Educatione						< .001 (1≠2; 1≠3; 2≠3; 1≠4)
No high school degree	29.1 (10 438)	25.3	54.7	72.5	43.1		6.61	683	67.0
High school graduate	36.5 (12 138)	38.4	24.1	16.9	25.7		6.28	781	72.0
> high school	34.1 (11 014)	36.0	20.4	10.7	30.5		5.74	647	75.7
Incomede						< .001 (1≠2; 1≠3; 2≠3; 1≠4)
< 150% below poverty line	32.3 (11 491)	27.7	60.4	78.4	63.2		7.14	821	67.2
≥ 150% below poverty line	67.7 (22 217)	72.3	39.7	21.7	36.8		5.75	1301	74.1
Supplemental health insuranced						< .001 (1≠2; 1≠3; 2≠3; 1≠4)
Medicaid	12.2 (4335)	8.0	34.2	54.4	52.4		9.10	388	72.5
Other supplemental insurance	78.9 (26 347)	84.0	47.0	33.3	38.6		6.02	1615	71.9
None (Medicare only)	8.9 (3026)	7.9	18.8	12.3	9.0		3.75	119	63.5
Prescription coveragede						< .001 (1≠2; 1≠3; 2 = 3; 1≠4)
Comprehensive coverage	42.6 (14 186)	41.0	49.0	56.7	63.8		7.53	1072	77.4
Limited coverage	34.5 (11 635)	35.8	27.4	22.6	20.2		6.05	723	67.0
No coverage	19.0 (6626)	19.4	18.6	17.1	9.2		4.54	311	64.2
Perceived access to care
Cost barrierd						< .001 (1≠2; 1 = 3; 2 = 3; 1 = 4)
No	94.7 (31 932)	94.9	92.2	93.2	94.4		6.06	1973	71.7
Yes	5.3 (1776)	5.1	7.8	6.8	5.6		8.52	149	69.2
Service availability barrierd						.299
No	89.8 (30 316)	89.8	89.7	87.6	91.2		6.04	1853	71.0
Yes	10.2 (3392)	10.2	10.3	12.4	8.8		7.55	269	75.0
Dissatisfaction with cared						< .001 (1≠2; 1≠3; 2 = 3; 1 = 4)
No	88.0 (29 664)	87.6	91.0	91.8	89.6		5.84	1759	71.2
Yes	12.0 (4044)	12.4	9.0	8.2	10.4		8.80	363	72.9
Clinical indicators
Self-reported health statusd						< .001 (1≠2; 1≠3; 2 = 3; 1≠4)
Fair or poor	20.6 (7222)	18.9	33.0	37.6	23.9		10.99	780	71.0
Excellent, very good, or good	79.2 (26 388)	80.8	66.8	62.4	75.8		4.94	1333	71.7
Average ADL impairment	0.70	0.52 (0.49, 0.55)	0.78 (0.70, 0.85)	0.95 (0.75, 1.14)	0.64 (0.51, 0.76)	…	…	…	…
Average IADL impairment	0.62	0.46 (0.43, 0.48)	0.68 (0.61, 0.75)	0.83 (0.68, 0.98)	0.60 (0.50, 0.70)	…	…	…	…
Depression symptomsd						< .001 (1 = 2; 1 = 3; 2≠3; 1 = 4)
Low	92.3 (31 012)	92.5	91.4	85.7	92.1		4.97	1574	71.0
Medium	5.0 (1742)	4.8	5.9	8.3	5.7		18.22	320	72.6
High	2.7 (954)	2.7	2.7	6.0	2.1		25.68	228	73.4
Diagnosis among subpopulation  diagnosed with depressione						.585
Major depressive episode	23.4 (488)	27.1	24.4	31.1	26.0		…	488	84.72
Other depression diagnosis	76.6 (1634)	72.9	75.6	68.9	74.0		…	1634	67.49

Note. ADL = activities of daily living; IADL = instrumental activities of daily living. We treated ADL and IADL as continuous variables, thus reported values are means with 95% confidence intervals.

^aPercentages are weighted and therefore reflect national estimates. Subsample sizes may not add up to 33 708 as a result of missing data. Percentages may not add up to 100 as a result of missing data or rounding.

^bIndicates prob > Rao-Scott χ² test statistic.

^cIndicates 0.0125 > prob(Rao-Scott χ² test statistic). We calculated α(0.0125) according to Bonferroni, such that familywise error is 0.05. We calculated test statistics for non-Hispanic Whites versus African Americans; non-Hispanic Whites versus Hispanics; African Americans versus Hispanics; non-Hispanic Whites versus others.

^dIndicates that depression rates are statistically significantly different between subgroups, P < .05.

^eIndicates that treatment rates among those diagnosed with depression are statistically significantly different between subgroups, P < .05.

TABLE 2—

Patterns of Depression Diagnosis: Medicare Current Beneficiary Survey, United States, 2001–2005

Population Characteristics	Model 1,a AOR (95% CI)	Model 2,b AOR (95% CI)	Model 3,c AOR (95% CI)	Model 4,c AOR (95% CI)	Model 5,c AOR (95% CI)
Race/Ethnicity
Non-Hispanic White (Ref)	1.00	1.00	1.00	1.00	1.00
African American	0.60 (0.46, 0.78)	0.52 (0.30, 0.90)	0.52 (0.40, 0.68)	0.50 (0.38, 0.65)	0.53 (0.41, 0.69)
Hispanic	1.02 (0.68, 1.53)	0.82 (0.43, 1.54)	0.80 (0.52, 1.21)	0.79 (0.52, 1.21)	0.79 (0.52, 1.19)
Non-Hispanic other	0.59 (0.40, 0.88)	0.66 (0.29, 1.50)	0.44 (0.29, 0.65)	0.46 (0.31, 0.68)	0.49 (0.33, 0.72)
Supplemental health insurance
Medicaid (Ref)	…	…	1.00	1.00	1.00
Other supplemental insurance	…	…	0.71 (0.59, 0.84)	0.88 (0.74, 1.04)	0.92 (0.77, 1.10)
None (Medicare only)	…	…	0.21 (0.17, 0.28)	0.26 (0.20, 0.34)	0.27 (0.21, 0.35)
Prescription coverage
Comprehensive coverage (Ref)	…	…	1.00	1.00	1.00
Limited coverage	…	…	0.73 (0.66, 0.81)	0.76 (0.69, 0.84)	0.76 (0.68, 0.84)
No coverage	…	…	0.76 (0.64, 0.90)	0.82 (0.69, 0.97)	0.82 (0.69, 0.97)
Perceived access to care
Cost barrier
No (Ref)	…	…	1.00	1.00	1.00
Yes	…	…	1.45 (1.20, 1.74)	1.23 (1.02, 1.49)	1.09 (0.89, 1.33)
Service availability barrier
No (Ref)	…	…	1.00	1.00	1.00
Yes	…	…	1.12 (0.97, 1.29)	1.00 (0.86, 1.15)	0.97 (0.84, 1.12)
Dissatisfaction with care
No (Ref)	…	…	1.00	1.00	1.00
Yes	…	…	1.43 (1.27, 1.61)	1.26 (1.12, 1.42)	1.16 (1.02, 1.31)
Clinical indicators
Self-reported health status
Fair or poor (Ref)	…	…	…	1.00	1.00
Excellent, very good, or good	…	…	…	0.53 (0.47, 0.59)	0.60 (0.53, 0.67)
ADL impairment	…	…	…	1.22 (1.15, 1.31)	1.12 (1.05, 1.20)
IADL impairment	…	…	…	1.19 (1.11, 1.27)	1.13 (1.06, 1.21)
Depression symptoms
Low (Ref)	…	…	…	…	1.00
Medium	…	…	…	…	3.05 (2.64, 3.52)
High	…	…	…	…	4.04 (3.39, 4.80)

Note. ADL = activities of daily living; AOR = adjusted odds ratio; CI = confidence interval; IADL = instrumental activities of daily living. AORs are from logistic regression models in which the dependent variable is the binary indicator of depression diagnosis during the year of observation.

^aControlling for demographic characteristics.

^bControlling for demographic characteristics and socioeconomic status.

^cControlling for demographic characteristics, socioeconomic status, coverage, and access variables.

RESULTS

Table 1 presents the population characteristics of the 4 racial/ethnic groups. The most pronounced difference was in their income: approximately one quarter of non-Hispanic Whites reported low incomes, whereas a majority of the African Americans and others were poor, and 78% of the Hispanics were poor. Hispanics had the lowest education levels, followed by African Americans. Many African Americans and the majority of Hispanics and others, but only 8% of non-Hispanic Whites, were dually enrolled in Medicaid. There were no substantial differences by gender or perceived access to care. Non-Hispanic Whites and others had fewer ADL and IADL impairments than did African Americans and Hispanics and were less likely to rate their health as fair or poor. Hispanics reported the highest levels of depression symptoms, but the difference was not substantial, although it was statistically significant.

The overall depression diagnosis rate was 6.2%. Rates varied significantly (P < .05) by race/ethnicity (6.4% for non-Hispanic Whites, 4.2% for African Americans, 7.2% for Hispanics, and 3.8% for others; Table 1). Depression diagnosis rates also varied by gender, income, health insurance, prescription drug coverage, perceived access to care, heath status, and symptom level (Table 1). Treatment rates were associated with education, income, prescription drug coverage, residence (urban vs rural), and depression diagnosis (major depressive disorder vs other). These variables also varied significantly by race/ethnicity and could have mediated the association between depression care indicators and race/ethnicity (Table 1). Thus, we estimated a series of nested logistic regression models to examine whether controlling for these factors (Tables 2 and 3) explain some of the association between race/ethnicity and the outcome variables. We successively added controls for membership in other traditionally underserved groups (model 1); socioeconomic characteristics (education and income, in model 2); insurance and perceived access to services (model 3); clinical indicators (model 4), and depression symptoms (model 5).

TABLE 3—

Patterns of Depression Treatment Among Enrollees Diagnosed With Depression: Medicare Current Beneficiary Survey, United States, 2001–2005

Population Characteristics	Model 1,a AOR (95% CI)	Model 2b AOR (95% CI)	Model 3c AOR (95% CI)	Model 4c AOR (95% CI)	Model 5c AOR (95% CI)
Race/Ethnicity
Non-Hispanic White (Ref)	1.00	1.00	1.00	1.00	1.00
African American	0.51 (0.33, 0.78)	0.50 (0.42, 0.58)	0.45 (0.38, 0.53)	0.42 (0.35, 0.49)	0.45 (0.30, 0.66)
Hispanic	0.59 (0.29, 1.21)	0.69 (0.52, 0.92)	0.58 (0.43, 0.79)	0.57 (0.42, 0.77)	0.61 (0.30, 1.25)
Non-Hispanic other	0.23 (0.10, 0.52)	0.43 (0.32, 0.58)	0.34 (0.25, 0.46)	0.35 (0.25, 0.47)	0.16 (0.07, 0.39)
Supplemental health insurance
Medicaid (Ref)	…	…	1.00	1.00	1.00
Other supplemental insurance	…	…	0.75 (0.66, 0.86)	0.96 (0.84, 1.1)	0.85 (0.60, 1.19)
None (Medicare only)	…	…	0.63 (0.56, 0.72)	0.8 (0.70, 0.92)	0.71 (0.43, 1.17)
Prescription coverage
Comprehensive coverage (Ref)	…	…	1.00	1.00	1.00
Limited coverage	…	…	0.69 (0.64, 0.74)	0.71 (0.66, 0.77)	0.57 (0.46, 0.72)
No coverage	…	…	0.48 (0.43, 0.55)	0.52 (0.46, 0.58)	0.55 (0.41, 0.73)
Perceived access to care
Cost barrier
No (Ref)	…	…	1.00	1.00	1.00
Yes	…	…	1.44 (1.26, 1.64)	1.19 (1.04, 1.35)	0.87 (0.60, 1.28)
Service availability barrier
No (Ref)	…	…	1.00	1.00	1.00
Yes	…	…	1.23 (1.13, 1.35)	1.07 (0.98, 1.18)	1.04 (0.82, 1.33)
Dissatisfaction with care
No (Ref)	…	…	1.00	1.00	1.00
Yes	…	…	1.33 (1.22, 1.45)	1.16 (1.06, 1.27)	0.94 (0.73, 1.20)
Clinical indicators
Self-reported health status
Fair or poor (Ref)	…	…	…	1.00	1.00
Excellent, very good, or good	…	…	…	0.57 (0.53, 0.62)	1.14 (0.92, 1.41)
ADL impairment	…	…	…	1.35 (1.29, 1.42)	1.09 (0.96, 1.24)
IADL impairment	…	…	…	1.28 (1.21, 1.34)	1.09 (0.95, 1.25)
Diagnosis
Major depressive episode	…	…	…	…	1.00
Other	…	…	…		0.39 (0.29, 0.52)
Depression symptoms
Low (Ref)	…	…	…	…	1.00
Medium	…	…	…	…	1.10 (0.83, 1.46)
High	…	…	…	…	1.16 (0.79, 1.69)

Note. ADL = activities of daily living; AOR = adjusted odds ratio; CI = confidence interval; IADL = instrumental activities of daily living. AORs are from logistic regression models in which the dependent variable is the binary indicator of depression treatment utilization (either antidepressant fill or psychotherapy utilization) during the year of observation. As a result of cell size considerations, we limited analyses to the subsample of non-Hispanic Whites and African Americans.

^aControlling for demographic characteristics.

^bControlling for demographic characteristics and socioeconomic status.

^cControlling for demographic characteristics, socioeconomic status, coverage, and access variables.

When we adjusted for inclusion in other medically underserved subgroups on the basis of gender, age, and geographic location, the racial/ethnic differences in diagnosis rates persisted. The odds of receiving a depression diagnosis were lower for African Americans than for non-Hispanic Whites (model 1; adjusted odds ratio [AOR] = 0.60; 95% confidence interval [CI] = 0.46, 0.78). This pattern remained after controlling for income and education (model 2), insurance coverage and perceived access to care (model 3), and general health measures (model 4; AOR = 0.50; 95% CI = 0.38, 0.65). The findings were also robust, controlling for depression symptoms (model 5; AOR = 0.53; 95% CI = 0.41, 0.69). In fact, including symptoms had little effect on the odds ratio for African Americans, suggesting that little of the bivariate difference could be explained by differences in self-reported depression symptoms. Comparing model 1 with model 5, it is also of interest that the racial/ethnic disparity was undiminished after adjusting for the full set of covariates. We observed a similar pattern for non-Hispanic others but not for Hispanics. In all 5 models, the odds of receiving a depression diagnosis were not statistically different for Hispanics than for non-Hispanic Whites.

We operationalized the second outcome, treatment of elderly with a depression diagnosis, as receipt of either psychotherapy or antidepressants. As we observed for diagnosis rates, treatment rates and modalities differed by race/ethnicity: 27.0% of non-Hispanic Whites versus 39.6% of African Americans did not receive any treatment (P < .05; Table 1). Treatment modalities were different across racial/ethnic subgroups as well (P < .001; data not shown). Among non-Hispanic Whites, 57.9% were treated with antidepressants alone, 4.3% with psychotherapy only, and 10.8% with both antidepressants and psychotherapy. Antidepressant use rates were lower among African Americans (52.5%) than among Whites (68.7%); whereas 58.0% of Hispanics used antidepressants. Rates of psychotherapy use among African Americans (18.0%) and Whites (15.0%) were not substantially different.

Racial/ethnic differences in depression treatment were robust, regardless of the control variables included in the model (Table 3). African Americans were approximately half as likely to receive treatment as were non-Hispanic Whites, controlling for sociodemographic characteristics, perceived access to care, and global health (model 4; AOR = 0.42; 95% CI = 0.35, 0.49). Differences in symptoms and disease severity did not explain the gap we found (model 5; AOR = 0.53; 95% CI = 0.41, 0.69). Treatment difference could not be explained by fewer symptoms among African Americans. We also observed similar patterns for Hispanics and non-Hispanic others.

DISCUSSION

During the years 2001 through 2005, minorities were less likely to receive a depression diagnosis and be treated for it than were non-Hispanic Whites. These differences remained after adjusting for depression symptoms and severity, suggesting that there may be disparities, as defined by the Institute of Medicine.²⁹ Our findings are consistent with the thesis that there is continuing underrecognition and undertreatment of depression among minority elders, net of differences in underlying symptoms, which persisted into the first decade of the 21st century despite overall increases in diagnosis and treatment rates.^30,31

Differences in depression diagnosis rates among racial/ethnic groups may be the result of both differences in underlying rates of pathology and underdiagnosing of depression in certain groups. Data on the underlying rates of pathology among adult populations are inconsistent. Some community-based epidemiological studies on adults report that African Americans have lower rates of depression than do non-Hispanic Whites^32–35; whereas the Epidemiologic Catchment Area Study found that the prevalence of depression is similar across racial/ethnic groups.^36,37 Although the MCBS does not include a full-scale depression measure (such as the Patient Health Questionnaire-9), adjustment for self-reports of depressed mood and anhedonia in the 2-item screener did not affect the racial/ethnic differences in diagnosis rates, suggesting that differences in symptoms (as identified by self-report) do not explain the gap and that there is a need to look at other factors, which may include racial/ethnic differences in depression help-seeking patterns,³⁸ differences in access to health care that are not captured by our explanatory variables, or differences in providers’ clinical detection of depression.³⁹

Evidence suggests that help-seeking patterns differ by race/ethnicity, contributing to the gap in depression diagnosis rates. Stigma, patient attitudes, and knowledge also may vary by race/ethnicity.^40,41 A recent vignette study found that African Americans were more likely than were their non-Hispanic White counterparts to believe that mental health problems would improve on their own.⁴² Low-income African Americans who were engaged in psychotherapy reported that stigma, dysfunctional coping behavior, shame, and denial could be reasons some African Americans do not seek professional help.⁴³ African Americans and Hispanics are more likely than are Whites to seek depression care from nonmedical providers, such as pastors or lay counselors.^44–46

Racial/ethnic disparities in depression diagnosis rates may also result from racial/ethnic differences in the patient–physician relationship during the clinical encounter. African Americans report greater distrust of physicians and poorer patient–physician communication than do White patients.^47,48 Communication difficulties may contribute to lower rates of clinical detection of depression among depressed African Americans because the diagnosis of depression depends to a considerable degree on communication of subjective distress. Such communication difficulties are more common among African American than among non-Hispanic White patients.^49,50 Race/ethnicity concordant visits, which are presumed to be less common for non-Whites, also have been characterized by better communication.⁵¹

Racial/ethnic differences in the clinical presentation of depression may further explain the lower rates of depression detection among African American patients. Studies of adult populations suggest that symptom presentation for mental health disorders varies by race/ethnicity.⁵² Symptom presentation by African Americans may differ from what most clinicians are trained to expect on the basis of clinical stereotypes, resulting in clinical misdiagnoses.^14,30 African Americans may be more likely to present with predominantly somatic and neurovegetative depression symptoms and less prominent mood or cognitive symptoms, which may complicate detection and diagnosis.¹⁴ In a randomized clinical trial of depression treatment in primary care, depressed African Americans were more likely to have symptoms of poor physical health, pain, and somatization than were their non-Hispanic White counterparts.⁵³

Finally, monetary factors may also play a role. Among Medicare beneficiaries, African Americans are substantially less likely than are non-Hispanic Whites to have private supplemental insurance that covers charges larger than standard Medicare-approved amounts.^54,55 Differences in provider reimbursement may favor increased clinical detection of depression in White patient groups if higher payment rates result in longer visits. Our data were collected before the implementation of the Medicare Part D drug benefit. As expected, both diagnosis and treatment rates were higher for those who had comprehensive prescription drug coverage than for enrollees with limited or no coverage. Yet prescription drug coverage was not mediating the racial/ethnic gap, suggesting that disparities may have persisted following Medicare Part D implementation.

Various factors may contribute to racial/ethnic differences in the treatment of those diagnosed with depression. Access barriers to their preferred mode of treatment may contribute to lower rates of treatment among African Americans. Some evidence suggests that African Americans and Latinos are less likely to accept antidepressant treatment than are non-Hispanic Whites.^56–59 Consequently, we operationalized treatment as the receipt of psychotherapy or antidepressants (the Institute of Medicine defines a disparity as differences in the medical treatment provided to members of different racial/ethnic groups that were not justified by the underlying health conditions or treatment preferences). We found that the use of psychotherapy was limited among both non-Hispanic Whites and African Americans. It was not possible to determine whether low rates of psychotherapy stemmed from patient preferences or from access barriers that could differentially affect non-Whites, such as high out-of-pocket costs⁶⁰ or a limited supply of mental health providers serving their community.⁶¹ Geographic-level differences in the supply of mental health services, particularly psychosocial services, may be a significant source of treatment differences: there may be inadequate access to mental health care in poor communities, where non-Whites are more likely to live.⁶²

Most interventions to reduce racial/ethnic differences in depression care attempt to do so by enhancing access to care, screening, or improving processes of care through process improvement strategies.⁶³ Studies have found that multicomponent chronic disease management interventions have improved depression outcomes for non-White populations, with case management as a critical component.⁶³ For example, in the IMPACT study, we observed improved outcomes and eliminated disparities among older adults with depression through an intervention that provided case management, patient education, and psychotherapy; non-Whites enrolled in this intervention had outcomes similar to those of non-Hispanic Whites.⁶⁴ Quality improvement programs for depressed primary care patients have also improved health outcomes and the unmet need for appropriate care among Latinos and African Americans relative to Whites.⁶⁵ Public policy options to combat these disparities include public financial incentives for primary care doctors and psychiatrists to practice in poor communities and incentives to increase the proportion of disadvantaged racial/ethnic groups in the health care workforce. Incorporating cross-cultural education into health professional training may also reduce these differences in diagnosis and treatment.⁶⁶

Limitations

First, some of the patients classified as untreated may have received counseling for their depression from non-Medicaid or non-Medicare providers (e.g., pastors or lay counselors). Second, the cell sizes for Hispanics and others were relatively small, which could make the estimates imprecise and underpowered. The survey items measuring depression symptoms and the response formats were similar to those in the validated PHQ-2 but were not exactly the same. Most importantly, the MCBS items refer to the past year, whereas the PHQ-2 refers to the past 2 weeks. Fourth, our findings are founded on fee-for-service enrollees. Managed care enrollees are slightly younger, reported better health status, have fewer limitations, and are more likely to live in urban areas.⁶⁷ Fifth, the MCBS does not capture stigma and other cultural factors that may mediate racial/ethnic differences in depression care. Finally, we used broad definitions of depression treatment, and for many patients, use of any psychotherapy or antidepressant treatment is not necessarily adequate depression treatment.⁶⁸

Conclusions

Our results document the substantial race/ethnicity-related differences that have persisted for depression care of community-dwelling elderly Medicare beneficiaries. Efforts are needed to reduce the burden of undetected and untreated depression and to identify the barriers that generate disparities in detection and treatment. Promising approaches include providing universal depression screening and ensuring access to care in low-income and minority neighborhoods. An increase in the reimbursement of case management services for the treatment of depression may also be effective. Continued surveillance and research documenting racial/ethnic differences in depression diagnosis and treatment among the elderly is also necessary to evaluate whether progress in eliminating any disparities continues.