Racial and Ethnic Disparities in Treatment and Treatment Type for Depression in a National Sample of Medicaid Recipients

Access to mental health care services is critically important to timely and accurate diagnosis and efficacious treatment of depression. This is particularly true for Medicaid beneficiaries, the majority of whom have low incomes and who are more likely than individuals from higher socioeconomic strata to experience economic barriers to mental health care (1). Furthermore, a disproportionate number of African Americans and persons from other racial-ethnic minority groups in the United States are Medicaid beneficiaries. African Americans and persons from other racial-ethnic minority groups have worse access to mental health care and receive lower-quality mental health care, compared with Caucasians. Among adult Medicaid beneficiaries, racial disparities in treatment rates for depression and treatment modalities for depression are understudied (1).

One six-state analysis of racial-ethnic differences in receipt of Medicaid-funded mental health services found lower utilization among African Americans and Hispanics (1). Also, Caucasians were more likely than African Americans and Hispanics to receive their treatment in community-based settings, and African Americans and Hispanics were more likely than Caucasians to receive their care in inpatient settings, emergency departments, and outpatient settings (1). Some studies have found that follow-up care after emergency mental health services is lower for African Americans than for other racial-ethnic groups, which further exacerbates mental health disparities among African Americans (2). Explanations for this disparity are unclear because of limitations of Medicaid claims research; however, socioeconomic factors, including justice system involvement, should be explored. These findings suggest that racial-ethnic disparities in mental health treatment found in the general population also exist among Medicaid beneficiaries. Medicaid plays a significant role in the lives of adults diagnosed as having depression—a diagnosis that is responsible for approximately half of state expenditures on mental health (3). Further work is needed to understand potential variations in depression treatment by race in order to effectively achieve racial-ethnic equity in mental health outcomes.

Medicaid beneficiaries with depression have more severe symptoms, are more likely to receive minimally adequate care, and are more expensive to treat because of related general medical comorbidities, compared with patients with other health care coverage (4, 5). A 2010 study examining the quality of counseling and drug therapy received by Medicaid beneficiaries diagnosed and treated in specialty mental health care settings found that African Americans received lower rates of minimally adequate depression care than their Caucasian counterparts (4). Thus, although the Medicaid population includes many low-socioeconomic-status, high-risk individuals, race may operate as an additional risk factor for inadequate depression treatment in this high-risk population.

The purpose of this secondary data analysis was to describe disparities in depression treatment and treatment modality among adult Medicaid beneficiaries with depression from a nationally representative sample of Medicaid claims data, which included 28 states and the District of Columbia. In addition, we explored secondary factors, including income, neighborhood-level variables, and general medical comorbidities. Further work is needed to understand the mechanisms for these disparities. There have been no previous studies of racial-ethnic disparities in depression treatment in the Medicaid population in a nationally representative sample, and the results of this study may help inform national trends. We hypothesized that disparities in treatment would exist by racial-ethnic subgroups and that persons from minority groups would be less likely than Caucasians to receive treatment for depression.

Methods

Study Population

We conducted a cross-sectional secondary analysis to examine racial disparities in depression treatment and treatment modality among adults with depression. The study population was drawn from 2008–2009 Medicaid Analytic eXtract (MAX) files obtained from the Centers for Medicare and Medicaid Services for 28 states (Alabama, Arizona, Arkansas, California, Colorado, Connecticut, Florida, Georgia, Illinois, Indiana, Louisiana, Maryland, Massachusetts, Michigan, Mississippi, Missouri, New Jersey, New Mexico, New York, North Carolina, Ohio, Oklahoma, Pennsylvania, South Carolina, Tennessee, Texas, Virginia, and Washington) and the District of Columbia. Medicaid enrollees were included in this study based on the following criteria: 24 months of continuous enrollment or until death from January 1, 2008, to December 31, 2009; ages 18 to 64; and at least one billed claim from the inpatient file or at least two billed claims from the outpatient file with a diagnosis of depression based on ICD-9-CM codes 296.20–296.25, 296.30–296.35, 298.0, and 311. The codes were chosen on the basis of those used in previous studies of depression that used administrative claims data (6). (The diagnoses corresponding to these codes are listed in an online supplement to this article.) MAX files provide the primary and secondary diagnosis codes for outpatient claims and all diagnosis codes for inpatient claims. The Morehouse School of Medicine Institutional Review Board approved this study, and all patient records and information were anonymized and deidentified.

Measures

The primary outcome of this study was type of treatment for depression, which consisted of four categories: medication only, therapy only, medication and therapy, and no treatment. The secondary outcome was treatment for depression (yes-no), which was based on the primary outcome. The National Drug Code variable from prescription claims was used to identify medication for depression (see online supplement). Therapy for depression was identified by any of the following CPT codes in inpatient or outpatient claims: 90785, 90832, 90834, 90837, 90839, 90840, 90847, 90853, 96101–96103, 96105, 96111, 96116, 96118–96120, and 96150–96155.

The main predictor was race-ethnicity, which was categorized into four groups: Caucasian, African American, Hispanic, and other (Asian, American Indian, Alaska Native, Pacific Islander, and multiple races or unknown).

Covariates were sex, age, comorbidity, and five zip code–level variables of residence (median annual personal income, proportion of African-American residents, proportion of Hispanic residents, proportion of residents with less than a high school diploma, and Gini coefficient). We categorized age into four groups (18–24, 25–39, 40–54, and 55–64). Comorbidity was calculated by using the Elixhauser Comorbidity Index, a validated approach that summarizes disease burden and predicts risks by using administrative claims data (7). The index was categorized into four groups (0, 1, 2, and ≥3 comorbid conditions). Zip code–level covariates were acquired from 2010 U.S. Census data. We categorized median annual personal income into four groups (<$20,000, $20,000–$29,999, $30,000–$39,999, ≥$40,000), proportion of African-American residents into three groups (<10%, 10%−20%, and >20%), proportion of Hispanic residents into three groups (<10%, 10%−20%, and >20%), proportion of residents with less than a high school diploma into three groups (<15%, 15%−30%, and >30%), and Gini coefficient into three groups (<0.4, 0.4–0.5, and >0.5). In this study, the Gini coefficient was used as a measure for inequality of income, and a value of 0 expressed perfect equality.

Statistical Analysis

Descriptive statistics analyses were performed for the overall population and for each of the race-ethnicity groups. Using type of treatment for depression as the primary outcome, we performed univariate and multivariate multinomial logistic regression with no treatment as the reference group. Using treatment for depression (yes-no) as the secondary outcome, we performed univariate and multivariate binomial logistic regression. Crude odds ratios (ORs) and adjusted odds ratios (AORs) were generated for univariate and multivariate models respectively; 95% confidence intervals (CI) of ORs and p values were also calculated. Multivariate logistic regression was conducted on the basis of three different models: race-ethnicity, sex, and age were adjusted in model 1; race-ethnicity, sex, age, and comorbidity were adjusted in model 2; and race-ethnicity, sex, age, comorbidity, and zip code–level covariates were adjusted in model 3. Covariates were selected on the basis of known associations with the outcome of interest—any treatment or type of treatment for depression. There are established differences in both receipt and type of treatment for depression by age (8), gender (9), and race-ethnicity (10–12). In addition, the presence of comorbid conditions (13) and neighborhood-level social deprivation (14, 15) have been found to be associated with differences in depression treatment (16). All p values were two-sided, and a p value <0.05 was considered statistically significant. SAS, version 9.4., was used to perform all analyses.

Results

In this secondary data analysis of a nationally representative sample of Medicaid claims data (28 states and the District of Columbia), which accounted for 80% of all Medicaid beneficiaries and 90% of Medicaid beneficiaries from racial-ethnic minority groups, a total of 599,421 beneficiaries had a diagnosis of depression and constituted the final sample. Characteristics of the sample population and of their zip codes of residence are shown in Table 1. Overall, the sample was 23.1% African American, 52.7% white, 9.0% Hispanic, and 15.1% from other racial-ethnic groups. Of the study population, 75.7% was female, and the sample was distributed across the age spectrum, with the bulk of participants between ages 25 and 54. Comorbidities were common, as shown by the Elixhauser Comorbidity Index. In the sample, 58.2% were from zip codes with a median income of $20,000 to $29,999, 53.4% were from zip codes that were <10% African American, and the largest percentage (47.5%) lived in zip codes where 15%−30% had less than a high school education. Most of the Medicaid beneficiaries (68.1%) lived in zip codes with a Gini Index between 0.4 and 0.5. A Gini Index of 0 indicates complete equality, and a Gini Index of 1 indicates complete inequality. There were notable differences between racial-ethnic groups. Compared with Caucasians, African Americans tended to have more comorbidities, live in poorer zip codes with a higher proportion of African-American and less educated residents and a higher level of income inequality. Compared with Caucasians, Hispanics were more likely to live in zip codes with poorer and less educated residents. Hispanics tended to live in zip codes with more poverty, less education, and more Hispanic residents, compared with Caucasians.

Overall 14.3% of depressed Medicaid beneficiaries had no record of treatment, including 19.9% of African Americans, 15.2% of Hispanics, and 11.9% of Caucasians (Table 1). Table 2 presents the ORs from logistic regression models for race differences (Caucasians as the reference group) for receipt of any treatment, adjusted for covariates. Compared with white Medicaid beneficiaries, African Americans, Hispanics, and those from other racial-ethnic groups were less likely to receive treatment for depression, and these associations changed very little after adjustment for individual-level and zip code–level covariates. After full adjustment, African Americans were about half as likely as Caucasians to receive treatment (AOR=0.52), Hispanics were about a third as likely (AOR=0.71), and those from other racial-ethnic groups were about a fifth as likely (AOR=0.84).

Table 3 shows the ORs for racial-ethnic differences (Caucasians as the reference group) in type of treatment from multinomial logistic regression models. After full adjustment, compared with Caucasians, African Americans were less likely to receive medication alone (AOR=0.52), more likely to receive therapy alone (AOR=1.57), and less likely to receive therapy together with medication (AOR=0.76). Compared with Caucasians, Hispanics were less likely to receive medication alone (AOR=0.71) and less likely to receive therapy plus medication (AOR=0.72). For Hispanics, the difference in receipt of therapy alone compared with Caucasians was not significant after full adjustment. After full adjustment, compared with Caucasians, beneficiaries of other racial-ethnic groups were less likely to receive medication alone (AOR=0.84), more likely to receive therapy alone (AOR=1.45), and less likely to receive medication and therapy (AOR=0.88). For receipt of therapy alone, a substantial change in the ORs toward the null (from 1.98 to 1.57) occurred for African Americans when we adjusted for zip code–level variables in model 3. For receipt of therapy plus medication by African Americans, the same adjustment in model 3 led to a reduction in the ORs from 0.95 to 0.76, compared with Caucasians.

Discussion

This cross-sectional secondary analysis of data from Medicaid beneficiaries identified significant racial and ethnic disparities in rates of treatment and treatment modality among adults with a diagnosis of depression, despite the fact that patients in the sample had the same health care coverage through Medicaid. Compared with Caucasians diagnosed as having depression, African Americans with this diagnosis were about half as likely to have received treatment and Hispanics with this diagnosis were about three-quarters as likely. Compared with Caucasians, African Americans who received treatment were about half as likely to receive medication alone, and more likely (AOR=1.57) to receive therapy alone. This study of a nationally representative sample, encompassing 80% of all Medicaid enrollees and 90% of all Medicaid beneficiaries from racial-ethnic minority groups, is the first to identify these disparities between racial-ethnic subgroups in rates and types of treatment. Results affirm the findings of smaller, state, regional, and local studies that have found similar disparities (1, 17, 18), and the findings can support the development of interventions to achieve equity in depression treatment across racial-ethnic subgroups.

A host of potential drivers of disparities in treatment utilization operate at multiple levels. These include cultural differences in care-seeking behaviors; stigma related to mental illness; provider implicit and explicit bias; and structural factors at the health system, community, and policy levels that perpetuate racial-ethnic differences in treatment for depression (19–21). However, it is important to note that the inequalities we found in uptake and modality of treatment for depression across racial-ethnic groups persisted when individual and neighborhood-level factors that might have affected access to care or treatment were included in the models—including medical comorbidities, neighborhood income inequality, neighborhood poverty, and neighborhood educational context. At the patient level, there are known cultural differences associated with care-seeking behavior for mental health treatment. Stigma and mistrust of mental health professionals have been shown to deter treatment seeking (19, 22) by persons from racial and ethnic minority groups. Differences in treatment modality may also be related to culture. Research suggests that persons from racial and ethnic minority groups are less likely than Caucasians to believe that antidepressant medication is acceptable (11). However, a different study found that African Americans and Caucasians had similar beliefs regarding the effectiveness of medication and therapy, whereas Hispanics were less likely than Caucasians to believe that medication and therapy were effective (23).

Some care models that have been successful at overcoming cultural stigma and improving access to mental health treatment include the integrated primary care–behavioral health model. In this model, behavioral health services are embedded into the primary care practice (24). This approach helps identify and treat the substantial number of patients with mental health conditions who receive care in primary care but who do not receive care from mental health providers (25, 26). One policy barrier to successful implementation of this model is the inability to bill Medicaid for primary care and mental health visits on the same day. Although some state Medicaid programs allow same-day billing of mental health and primary care in the same location, many state programs continue to ban this practice (27).

This study had several limitations. Although the Medicaid beneficiaries in this sample had comparable health care coverage, geographic differences in access to care may have existed with respect to the location of mental health services available to Medicaid beneficiaries in a community. For example, rural Medicaid beneficiaries have poor access to mental health services (28–30). In addition, the capacity of mental health providers to care for Medicaid patients may be limited, because Medicaid reimbursement for mental health services is lower than that of private insurance or self-pay patients, and as a result, there are narrow networks of mental health specialists serving Medicaid patients (31). This study used health care claims data, which provide information on each claim about the type of service and the diagnosis codes that support the episode of care. Prescription drug claims note the type of drug prescribed and prescriptions filled by the patient. Similarly, for a therapy claim to be generated, the patient would have had to attend the appointment. Because of this limited information, it was not possible to include measures of care quality, unless we had used health care utilization or prescribing patterns as a proxy for quality of care. It was also not possible to measure whether a treatment such as medication or therapy was recommended by a clinician but declined by a patient.

In addition, there was insufficient information to assess the timing of the diagnosis, which has an impact on treatment uptake and modality. It cannot be assumed that all diagnoses captured in this data set occurred at the same time, which is a limiting factor in interpreting findings. It is also possible that some patients in the no treatment category may have sought treatment that was not recorded in the claims data, such as use of therapists who did not accept Medicaid or self-medication through the use of supplements. It is unknown whether use of these services would be different by race-ethnicity in the population enrolled in Medicaid, which is of low socioeconomic status. There may have been differences in the providers available to different racial-ethnic groups, such as the availability of race-, sex-, or age-concordant providers, and thus differences in providers’ interpersonal bias could affect treatment planning. Unfortunately, we did not have access to physician demographic data to examine whether these potential differences in providers contributed to the observed disparities in treatment.

Finally, the data are from 2008–2009, which was the most current data set we could access at the time. This period predates the Affordable Care Act (ACA) and Medicaid expansion in many states and policy interventions that have been associated with improved mental health outcomes (32). However, many of the states that have not expanded their Medicaid programs are home to large populations of racial and ethnic minority groups. At a national level, disparities in access to and utilization of behavioral health treatment have persisted, even after implementation of the ACA (33), suggesting that the disparities identified in this study may still be present.

Addressing mental health access and treatment disparities at the population level requires targeted policy and practice strategies. States that expanded their Medicaid program under the ACA have seen improvements in the overall mental health of their residents and reductions in depression diagnoses among adults with chronic conditions (32). Medicaid expansion has led to improved access to mental health services and mental health outcomes among Medicaid beneficiaries (34). More research is needed to understand whether and to what extent these policies have affected racial and ethnic behavioral health disparities. Health care systems as a whole must adopt evidence-based practices, such as measurement-based care, to better understand the experiences of persons from racial and ethnic minority groups receiving mental health services and to inform practice improvements that may increase the uptake of service use and improve the quality of care.

Conclusions

This analysis of depression treatment rates among Medicaid beneficiaries diagnosed as having depression, found racial and ethnic disparities across 28 states and the District of Columbia. Compared with Caucasians, African Americans and Hispanics appeared to receive treatment for depression less often and were less likely to be prescribed medication for depression. These findings may provide the basis for exploring potential drivers of these differences that can serve as points of intervention to remove barriers to care for depression, improve quality, and address disparities.