Despite the overall growth in antidepressant treatment, studies have reported persistent racial and ethnic disparities in use of antidepressants: among persons with a similar diagnosis of depression, the odds of antidepressant use were lower among blacks and Hispanics than whites by 20%−70% in the 1990s and early 2000s (
3,
4). Lack of access to health insurance and relatively low income in minority groups are among the main explanations for these disparities (
5). Some researchers thus suggest that increasing insurance coverage rates among racial-ethnic minority groups can narrow gaps in health care utilization (
6). Although the existence of racial disparities in the use of services in public insurance programs implies that other factors also contribute to the disparities (
7–
9), the literature generally indicates that increased insurance coverage reduces racial disparities in health care.
This study examined racial-ethnic disparities in the use of antidepressants among people with private coverage between 2006 and 2010. Private insurance is one of the main channels through which people obtain coverage in the United States. Further, the privately covered population will grow under the ACA, particularly among racial-ethnic minority groups. However, no research, to our knowledge, has analyzed the existence or degree of racial and ethnic disparities in private coverage. Our study thus aimed to provide information that could be used in exploring ways to address disparity issues in antidepressant use under the ACA.
Methods
We used multiyear data (2006–2010) of the Medical Expenditure Panel Survey Household Component (MEPS-HC), which includes five rounds of interviews over two-and-a-half years with a nationally representative sample of noninstitutionalized U.S. populations. The data include information on health care utilization and sociodemographic characteristics of respondents and their families.
We identified individuals with depression using MEPS-HC Medical Conditions files. The survey asked respondents to describe their medical conditions, and professional coders used the ICD-9-CM to record those self-reported conditions. We defined respondents as having depression if they had an ICD-9-CM code of 311 in any round of their interviews during a calendar year. The unit of analysis was person-year.
From MEPS-HC Prescribed Medicines files, we identified the following antidepressants: tricyclic antidepressants amitriptyline, amoxapine, clomipramine, desipramine, doxepin, imipramine, maprotiline, mirtazapine, nortriptyline, and protriptyline; selective serotonin reuptake inhibitors citalopram, escitalopram, fluoxetine, fluvoxamine, paroxetine, and sertraline; and others (bupropion, duloxetine, nefazodone, selegiline, trazodone, and venlafaxine). We constructed a binary indicator of whether the respondent had at least one of these medications during the same year when he or she reported a diagnosis of depression.
For a respondent’s race and ethnicity, we constructed three categories: non-Hispanic whites (for simplicity, we refer to this group as whites), non-Hispanic blacks (referred to as blacks), and Hispanics. We excluded other racial-ethnic groups, which accounted for 5.1% of the sample.
For insurance coverage, we categorized the sample into four groups: private coverage, Medicare, Medicaid, and the uninsured. The private coverage group excluded people who were eligible for either Medicare or Medicaid. Those who were eligible for both Medicare and Medicaid were classified into Medicare—the payer for prescription drugs during the study period.
Covariates included age, gender, education, family income, employment status, and health risk. Year dummies were also included to capture time-specific effects.
We used logistic regression to analyze variation across racial and ethnic groups in the use of antidepressants, controlling for health risk and sociodemographic factors. Using whites as the reference group, we obtained the adjusted odds ratios (AORs) of antidepressant use for blacks and Hispanics. We conducted the analysis separately for each insurance group.
We performed two additional analyses on private coverage. First, we limited the sample to privately insured people whose family income was between 125% and 399% FPL. This subgroup is likely to have similar characteristics to people who would purchase private coverage through exchanges under the ACA. This analysis thus provides information that is particularly applicable to understanding racial-ethnic disparities in private coverage under the ACA.
Second, we conducted an analysis by adding a cost-sharing index, which is measured by the ratio of a respondent’s out-of-pocket drug spending to his or her total drug spending during the given year. The purpose of including this index was to account for the possibility that minority groups, compared with whites, may enroll in plans with less generous drug benefits for antidepressants. Our index, based on actual out-of-pocket spending on drugs, may imperfectly capture generosity of plan benefits because actual spending is affected by benefit designs. However, this ratio would be an inappropriate measure in our analysis of antidepressant use only if cost-sharing substantially differed between antidepressants and other drugs. Although this differential cost sharing might be the case in some plans, it would lead to biased results only if race and ethnicity were systematically associated with choosing coverage that has such within-plan differential cost sharing across drug classes, which is very unlikely.
Standard errors were clustered by individual in all analyses to account for repeated interviews of the same person. Because the study used publicly available data, the institutional review board waived the need for approval.
Discussion
Main findings
We found that racial-ethnic disparities existed in antidepressant use among people with depression in private coverage between 2006 and 2010: the odds of using antidepressants were lower among blacks and Hispanics compared with whites by 50% and 30%, respectively. These are large gaps and somewhat alarming, particularly considering that no significant racial-ethnic disparities were found in other insured groups. Considering that racial-ethnic disparities have been a long-standing issue in health care, it is surprising that there has been no report or research on this issue in private coverage. It may have been assumed that racial-ethnic minority groups would not use health care on significantly different levels compared with whites when minority groups have access to private coverage. However, our analysis shows that is not the case.
Private plans may not have sought strategies to monitor and reduce racial and ethnic gaps in the use of antidepressants (or other health services) given that racial-ethnic minority groups account for a relatively small proportion of their members. This may partly explain our finding. In Medicaid, where people from racial-ethnic minority groups make up a relatively large share compared with other groups, racial-ethnic disparities in health service use have long been recognized. Interventions have been initiated to reduce the disparities that occur with Medicaid; such interventions include one-on-one outreach, education, and translation services (
12,
13). Some prior work reported few racial-ethnic disparities in the use of antidepressants among people with Medicaid coverage (
8), and our analysis also found no significant racial-ethnic disparities in antidepressant use.
Variation in unobserved generosity of benefits across private plans may be an alternative explanation for our finding. Unlike Medicare and Medicaid, which specify standard benefits, individuals in the private sector choose their own plans and benefits. Compared with whites, racial-ethnic minority populations may enroll in private plans with less generous benefits. In our analysis, inclusion of a cost-sharing index did not change the results. However, lacking plan-specific benefit information, we constructed that index based on actual out-of-pocket spending on all drugs. The index may thus have imperfectly captured benefit generosity for antidepressants. Further, some aspects of benefits may remain unobserved, such as utilization management (limiting medication quantity, for example) and breadth of drug formularies. Unmeasured benefit generosity may have influenced antidepressant use disproportionally in racial and ethnic minority groups. A recent study reported that the introduction of Medicare Part D particularly benefited racial-ethnic minority groups, partially supporting this possibility (
14). Our analysis focusing on a post–Part D period also found no significant disparities in Medicare.
Implications and future research
Our study has implications for the ACA provision. First, the higher level of antidepressant use among people with private insurance compared with those without insurance suggests that antidepressant use is likely to increase under the ACA, although it is not clear how much it might increase. Insurance coverage increases service utilization by lowering the price of services, which improves access to necessary care and creates financial incentives to use inefficient services. However, this potential increase in service use by coverage expansion will be reduced by the extent to which selection exists in private coverage: privately covered people may have greater need and preference for health services than currently uninsured populations. Thus the ACA may not increase antidepressant use to the level of current enrollees in private coverage.
Second, the wide racial-ethnic gap in antidepressant use among privately covered people (including persons with income levels similar to those who would enroll in private plans through exchanges) indicates that disparities are unlikely to be resolved solely by the expansion of coverage. These findings suggest that efforts to identify and reduce racial-ethnic disparities should be continued under the ACA. They also call for particular attention to private plans, which would experience an increase in racial and ethnic minority enrollees under the ACA.
The ACA stipulates that private plans must offer essential health benefits (similar to typical employer-based plan benefits) and that cost sharing will be subsidized for low-income families to limit their out-of-pocket expenditures; however, no provision specifies how benefits for each service are to be designed. Plans are likely to develop their own benefit schemes as long as their schemes meet the general requirement. Certain specifics of benefit designs could disproportionally affect racial and ethnic minority groups with mental health conditions. Evaluation of how benefit schemes in private coverage evolve and how minority populations’ access to health services improves would be necessary to reduce racial-ethnic disparities under the ACA.
Besides designing financial incentives for consumers, private plans’ initiatives to identify and remove any access barriers that may exist among racial-ethnic minority groups could also help mitigate disparities. The ACA provision requires collecting and reporting data on race-ethnicity, which would be used to analyze trends in disparities. This effort is encouraging, but it is not yet clear what interventions would be developed based on findings from those data. Data collection efforts should be accompanied by development and implementation of interventions to reduce racial-ethnic disparities.
Our study did not evaluate any specific approaches to reduce racial-ethnic disparities in antidepressant use. However, assessing the patterns of racial-ethnic disparities is a first step toward exploring strategies to close the gaps. An important next step would be to investigate reasons for less usage of antidepressants among racial-ethnic minority groups compared with whites (for example, collecting qualitative data from targeted interviews), which would provide useful information to develop specific interventions. Continuous monitoring of the pattern of racial-ethnic disparities and implementation and evaluation of any interventions would be essential to improve minority populations’ access to needed mental health care.
Limitations
Several cautions should be noted in interpreting the results. First, insurance expansion is likely to influence the diagnosis of depression itself, but our analysis did not incorporate that. However, descriptive data from MEPS showed that the prevalence of depression in each racial-ethnic group was similar between people without insurance and those with private coverage, indicating the uninsured are as likely as those with coverage to be aware of their condition. The uninsured may have obtained the diagnosis in formal health care settings (before losing coverage or through uncompensated care), or they may have self-diagnosed, given that depression is a symptomatic condition. In any case, the data suggest that not examining a possible increase in the diagnosis of depression under the ACA is unlikely to affect our results.
Second, we did not analyze whether antidepressants were used for mild symptoms that may not require medications (
15,
16). Information on the severity of depression was not available in the data. Although potential overuse of antidepressants is a health care quality concern, underuse in depression treatment is a commonly documented issue, and its consequences on health outcomes are greater than those of overuse (
17–
19).
Third, the low level of antidepressant use among people from racial-ethnic minority groups may partially reflect their preferences for other treatments (such as counseling) instead of medications (
20–
24). If this is the case, the differences in antidepressant use across racial-ethnic groups would be of less concern, and it would be important to ensure racial-ethnic minority groups’ access to alternative therapies. Exploring reasons for low rates of antidepressant use is an important topic to pursue in future research.
Fourth, we identified patients with depression and antidepressant use on the basis of self-reported information. Thus our data were subject to reporting errors. However, there is no evidence that those errors systematically differed across racial-ethnic groups, and their impact on our findings concerning racial-ethnic differences is unlikely to be significant.
Finally, characteristics of people from racial-ethnic minority groups who would purchase coverage under the ACA may not be the same as those currently enrolled in private coverage. This implies that our finding may not be fully applied to “newly” insured groups under the ACA. However, our subgroup analysis suggests that a similar pattern is likely to be observed among those people.