Medicaid Expansion and Racial-Ethnic Health Care Coverage Disparities Among Low-Income Adults With Substance Use Disorders

Acquisition of health insurance has been connected with increased access to primary care (1, 2) and preventive health services (3) and with decreased mortality rates after onset of major general medical illnesses (4, 5). Yet despite established health benefits, ethnic and racial disparities in health care coverage persist, with Black and Hispanic individuals having lower rates of coverage than White individuals (6). Because racial and ethnic differences in coverage are associated with disparities in health care access (7), there is keen interest in evaluating whether health insurance reform can help narrow gaps in coverage and service access.

The Medicaid expansion provision of the Affordable Care Act (ACA) has increased health care coverage of low-income adults (8). Some (9, 10), but not all (11–13), studies of ACA Medicaid expansion have found that income-eligible Hispanic or Black adults had greater increases in health care coverage compared with White adults, thereby narrowing racial-ethnic coverage gaps. It is not known, however, whether Medicaid expansion has reduced racial-ethnic coverage gaps among adults with substance use disorders.

Compared with adults in the general population, adults with substance use disorders are more likely to have low income (14) and lack health insurance (15, 16). In keeping with broader trends, Black and Hispanic adults with substance use disorders are also less likely than their White counterparts to have health care coverage (17) or to receive specialty substance use treatment (17, 18) and therefore have potentially the most to gain from reforms that increase coverage. Because significant Black-White differences in substance use treatment have been observed among uninsured individuals, but not among those who are insured, it is possible that increasing health care coverage might lower treatment disparities (17).

Among low-income adults with substance use disorders, Medicaid expansion was associated with greater increases in Medicaid coverage and greater declines in uninsured rates in expansion states, compared with nonexpansion states (19, 20). Among patients with substance use disorders, the percentage of those with Medicaid coverage increased and the percentage of those without insurance decreased (21). It is not known whether gains in coverage occurred among Black, Hispanic, and White individuals to a similar extent and whether such gains helped to narrow disparities in coverage. To examine these issues, we assessed changes in health care coverage in nationally representative samples of Black, White, and Hispanic low-income adults with substance use disorders after the 2014 ACA Medicaid expansion.

Methods

Data Source

The National Survey on Drug Use and Health (NSDUH) is an annual cross-sectional national and state representative survey of the civilian noninstitutionalized U.S. population, conducted by the Substance Abuse and Mental Health Services Administration (22). The NSDUH yields national and state-level representative estimates of substance use disorders for the civilian noninstitutionalized population. Individuals without a household address, active duty military personnel, and institutional residents are excluded from the sample. The institutional review board at RTI International approved the protocol for NSDUH data collection. The annual mean weighted overall response rate for the 2008–2019 NSDUHs ranged from 64.9% to 75.6% (total N=749,033; Ns for the racial-ethnic subsamples were unavailable).

Participant Characteristics

The NSDUH yields estimates of past-year DSM-IV alcohol, cannabis, cocaine, and heroin dependence or abuse disorders. Respondents were classified by self-reported race (“Which of these groups describes you? White, Black, or African American, etc.”) and ethnicity (“Are you of Hispanic, Latino, or Spanish origin or descent?”). For simplicity, we refer to non-Hispanic White as White and non-Hispanic Black as Black in the following. The NSDUH also collects self-reported information on age, sex, family income, marital status, education level, employment status, general health status (fair or poor, good, very good, and excellent), and state of residence.

Medicaid Expansion State Residence

The 26 states and the District of Columbia that expanded Medicaid by the end of 2014 were considered expansion states, and the remaining 24 states were considered nonexpansion states (a list of these states is available in a table in the online supplement to this article). Sensitivity analyses examined the effects of Medicaid expansion when states that expanded before and after 2014 were removed.

Outcome Variables

Health insurance and substance use treatment were the outcome variables. After a description of each type of health care coverage, respondents were asked if they were currently covered by Medicare, Medicaid, private health insurance, military health care (TRICARE, Civilian Health and Medical Program of the Uniformed Services [known as CHAMPUS], Civilian Health and Medical Program of the Department of Veterans Affairs [known as CHAMPVA], or U.S. Department of Veterans Affairs), or any other type of coverage. Responses were hierarchically grouped into Medicaid, other public insurance (Medicare or military health care), private insurance, and no insurance.

Respondents were asked about past-year substance use treatment. Treatment included services within a hospital, rehabilitation facility, mental health center, emergency department, private physician’s office, other organized settings, and self-help attendance. Insurance-eligible treatment excluded self-help. Services provided in prisons or jails, which are not reimbursed by insurance, were not included.

Study Samples

The study sample was limited to low-income adults ages 18–64 years who met diagnostic criteria for past-year DSM-IV alcohol, cannabis, cocaine, or heroin use disorder and were of White, Black, or Hispanic race-ethnicity. Following the ACA Medicaid income eligibility threshold, we defined low income as self-reported household income of ≤138% of the federal poverty level.

Statistical Analysis

Background characteristics of the low-income White, Black, and Hispanic adults with substance use disorders were compared with chi-square tests, overall and within each racial-ethnic group, by residence in expansion and nonexpansion states. Overall differences in health coverage and substance use treatment were compared preexpansion (2008–2013) and postexpansion (2014–2019) across White, Black, and Hispanic adults by using multivariable logistic regression analyses. A difference-in-differences design (23) assessed differences in trends in health care coverage among racial-ethnic groups between the pre- to postexpansion periods. For each dichotomous outcome, the model included effects of expansion period (pre vs. post), racial-ethnic group, and the interaction of expansion period and racial-ethnic group while adjusting for age and sex. These differences were further investigated with a three-way interaction effect of expansion state by expansion period by racial-ethnic group. None of these differences were statistically significant (data not shown).

Adjusted proportions were estimated for White, Black, and Hispanic respondents during the pre- and postexpansion periods (24). Adjusted risk differences (ARDs) with 95% CIs were used to examine the statistical significance of differences between expansion and nonexpansion states. The difference was considered statistically significant if CIs did not include 1. Adjusted proportions of White, Black, and Hispanic adults with substance use disorders without health insurance or with Medicaid coverage were also estimated for every 2-year increment between expansion and nonexpansion states.

All analyses were performed with SAS software, version 9.4. Analyses using SAS-callable SUDAAN accounted for the complex survey design and sampling weights of NSDUH. All reported annualized percentages were weighted by survey weights to provide U.S. population estimates.

Results

Background Characteristics

Most low-income adults with substance use disorders in each racial-ethnic group were men who were ages <35 years and had never married (Table 1). As shown in Table 1, the racial-ethnic groups significantly differed from each other in sociodemographic composition, self-reported health status, likelihood of having received substance use treatment in the past year, and the proportion that resided in an expansion state.

Significant differences were noted in background characteristics within each racial-ethnic group by residence in expansion or nonexpansion states (see table in the online supplement). Among White and Hispanic individuals, those in nonexpansion states were significantly more likely to be married than those in expansion states. The educational level of White and Black persons in expansion states was significantly higher than that of their counterparts in nonexpansion states. For White and Hispanic individuals, past-year alcohol use disorder was significantly more common among nonexpansion state residents than among those in expansion states, whereas the reverse was true for past-year cannabis use disorder, which was significantly more common among White and Hispanic residents of expansion states than among their counterparts in nonexpansion states. For White and Black individuals, past-year heroin use disorder was significantly more common among expansion state residents than among nonexpansion state residents.

Overall Trends in Coverage and Treatment After Expansion

Compared with the preexpansion period, during the postexpansion period each racial-ethnic group had a significant decline in the percentage that was uninsured, and these declines did not significantly differ across the three groups (Table 2). Medicaid coverage during this period also increased in all three groups. However, the increase was significantly smaller for Black adults (9.1%) than for White adults (15.0%).

During this period, the percentage of Hispanic adults with private insurance significantly increased. We observed no significant differences among the three racial-ethnic groups in the percentage of adults who received past-year substance use treatment or insurance-eligible treatment.

Trends in lack of coverage by expansion state and race-ethnicity.

For White and Hispanic low-income adults with substance use disorders, but not for Black individuals, the overall decrease in being uninsured was significantly larger for those residing in expansion states, compared with those in nonexpansion states (Table 3). During the postexpansion period, the percentage of low-income adults with substance use disorders who were uninsured was more than twice as high in nonexpansion states, compared with expansion states, for Black people (33.3% vs. 13.0%) and for White people (31.8% vs. 13.3%) and approximately twice as high for Hispanic people (51.8% vs. 25.7%).

For low-income White adults with substance use disorders, the decline in being uninsured between 2012–2013 and 2014–2015 was significantly larger in expansion states than in nonexpansion states (Figure 1). Between 2014–2015 and 2016–2017, the percentage of White adults without coverage significantly declined in both expansion and nonexpansion states. Between 2012–13 and 2014–15, Black individuals had a significant decline in being uninsured in both expansion and nonexpansion states, as well as an earlier decline (2008–2009 to 2010–2011) in nonexpansion states. For Hispanic individuals in expansion states, uninsurance rates significantly declined from 2012–2013 to 2014–2015, continuing to 2016–2017. In nonexpansion states, uninsurance rates among Hispanic individuals significantly declined before Medicaid expansion (2008–2009 to 2010–2011).

In sensitivity analyses, removing states that expanded Medicaid before 2014 (see table in the online supplement) or between 2015 and 2019 (see table in the online supplement) had little effect on these trends. In the latter analysis, however, the decline in being uninsured for Black adults in nonexpansion states was no longer statistically significant. A sensitivity analysis that also controlled for state-period unemployment rate yielded results similar to the main findings, except that the decline in being uninsured among Hispanic adults in nonexpansion states was significant (ARD=−9.81, 95% CI=−18.76, −0.86) (see table in the online supplement).

Trends in Medicaid by expansion state and race-ethnicity.

In expansion states, Medicaid coverage significantly increased between the preexpansion (2008–2013) and postexpansion (2014–2019) periods for all three racial-ethnic groups of low-income adults with substance use disorders (Table 3). Among White but not among Black or Hispanic individuals in nonexpansion states, Medicaid coverage also significantly increased during this period. For all three racial-ethnic groups, however, Medicaid coverage increased significantly more in expansion states than in nonexpansion states.

For White individuals in Medicaid expansion states, but not in nonexpansion states, Medicaid coverage significantly increased following ACA implementation (2012–2013 to 2014–2015) (Figure 2), and the results of the difference-in-differences analyses were also statistically significant. We also noted a significant increase in Medicaid coverage among White persons in nonexpansion states from 2014–2015 to 2016–2017. For Black persons in expansion states, the increase in Medicaid coverage from 2014–2015 to 2016–2017 was significantly larger than the change in Medicaid coverage in nonexpansion states. For Hispanic individuals in expansion states, Medicaid coverage significantly increased from 2012–2013 to 2014–2015. The sensitivity analyses that excluded early and late expanding states yielded similar results, as did the analysis that also controlled for state-period unemployment, except that the increase in Medicaid coverage in nonexpansion states for Black persons was also significant (ARD=5.86, 95% CI=0.35–11.38) (see tables in the online supplement).

Other Trends During the Study Period

We observed no significant changes in private insurance or other health care coverage between expansion and nonexpansion states among all individuals in our sample before or after implementation of Medicaid expansion. In expansion states, the percentages of White individuals who received any treatment or insurance-eligible treatment for substance use disorders were significantly lower during the postexpansion period than during the preexpansion period (Table 3). A similar result was obtained in expansion states when the analysis excluded late-expanding states and when the analysis controlled for state-period unemployment but not when the analysis excluded early-expanding states, in which these decreases were not statistically significant (see table in the online supplement).

Discussion

During the first 6 years after ACA Medicaid expansion, low-income Black, White, and Hispanic individuals with substance use disorders became more likely to have Medicaid coverage and less likely to be uninsured. However, these gains were not uniform across expansion and nonexpansion states or across racial-ethnic groups and did not translate into increased substance use treatment.

Implementation of Medicaid expansion coincided with greater gains in Medicaid coverage for income-eligible Black, White, and Hispanic adults with substance use disorders who resided in expansion states, compared with these groups in nonexpansion states. These patterns, which are consistent with trends in the general population (9), underscore the adverse effects on health insurance coverage for states that opted out of Medicaid expansion.

Among low-income adults with substance use disorders in nonexpansion states, only White individuals had significant gains in Medicaid coverage after implementation of the ACA Medicaid provisions. It is possible that simplification of Medicaid enrollment, which was required in all states under the ACA without regard to expansion status, contributed to differential Medicaid coverage gains in nonexpansion states of White individuals over Black and Hispanic individuals. Streamlined Medicaid enrollment application processes introduced by the ACA included removing face-to-face interviews, discontinuing asset tests, and introducing online and telephone enrollment options (25). Racial-ethnic differences in online access (26) or how enrollment implementation was conducted in nonexpansion states may have differentially disadvantaged low-income non-White individuals with substance use disorders from gaining coverage.

For the minority of Hispanic adults who are undocumented, fears related to immigration status may have also contributed to their low level of Medicaid coverage and lack of significant gains in coverage in nonexpansion states. After expansion, similar coverage gains occurred in expansion states both for people in households with mixed immigrant status and for those in nonmixed households, but in nonexpansion states, coverage gains were significantly lower for mixed than nonmixed households (27). These findings, which are broadly consistent with our results for low-income Hispanic adults with substance use disorders, raise the possibility that legal concerns about exposing undocumented family members to immigration enforcement actions could have impeded Medicaid enrollment of Hispanic adults with substance use disorders, especially in nonexpansion states. In nonexpansion states, compared with expansion states, state governments are more likely to take legislative actions that restrict access or place greater administrative burdens on immigrants’ access to public benefits and services (28).

Within expansion states, we noted a variation in the timing of the increase in Medicaid coverage among the racial-ethnic groups. Specifically, Medicaid coverage significantly increased earlier for White and Hispanic adults (2012–2013 to 2014–2015) than for Black adults (2014–2015 to 2016–2017). The reasons for the delay in Medicaid uptake among Black individuals are not known. In other contexts, however, new innovations, which over time become widely distributed, may create temporary disparities as groups with more information, influence, resources, and social capital gain faster access (29). COVID-19 vaccination rates provide an example for attenuation of initially large early racial-ethnic disparities in uptake over time (30).

Implementation of Medicaid expansion also coincided with reductions in the percentage of uninsured low-income Black, White, and Hispanic adults with substance use disorders. In expansion states, the percentage of uninsured adults significantly declined across all three racial-ethnic groups; uninsurance rates declined also in nonexpansion states, but only for Black and White adults, not Hispanic adults. Moreover, a higher percentage of Hispanic individuals, compared with Black or White individuals, was uninsured throughout the study period. Besides the previously mentioned potential for immigration status concerns, state differences in Medicaid outreach and application assistance efforts also may play a role in the higher rate of uninsured Hispanic adults (31). In addition, most Spanish-language Medicaid enrollment applications are highly complex (32), which could impede enrollment of some Hispanic individuals with lower levels of health insurance literacy.

As in previous studies (19, 20), we found no evidence that increased health care coverage reform translated into greater use of substance use treatment. These findings may be partially explained by a limited availability of substance use treatment programs, including those that accept Medicaid (33); stigma associated with substance use treatment (34); and a belief that treatment will not be effective (35). Nevertheless, health care coverage confers a range of important health and economic benefits, including benefits for low-income adults with substance use disorders (36).

This study had several important limitations. First, other ACA policy provisions, such as subsidies for low-income individuals to purchase marketplace plans (19) or the dependent coverage provision that required private insurers to allow children to remain on their parents’ plans up to age 26 (37), may have contributed to the observed trends. Second, it was not possible to quantify the amount or quality of treatment, and because of a redesign of the NSDUH in 2015, we could not assess trends in treatment of prescription opioid use disorder or use of stimulants or sedatives. Third, NSDUH queried respondents about substance use treatment or counseling without specifically mentioning medication treatments, and medication treatments may therefore have been undercounted. Fourth, different results might have been obtained had the analyses focused on the entire populations of the three racial-ethnic groups rather than on the low-income individuals who were the target of the Medicaid expansion. Fifth, constraints on survey sample size limited statistical power to detect small but nonetheless potentially important policy effects. Sixth, insurance coverage might help avert development of substance use disorders via early intervention (38) or by reducing financial strain (39), effectively removing from this analysis individuals who might otherwise have had substance use disorders. Finally, the NSDUH sampling frame excluded institutionalized and homeless adults; both groups are at increased risk for substance use disorders. Inclusion of these adults might have altered the observed racial-ethnic trends and patterns.

Conclusions

After enactment of the main provisions of the ACA, health care coverage increased among low-income Black, White, and Hispanic adults with substance use disorders. Unlike the general population, however, for whom racial-ethnic disparities in health care coverage decreased (9), low-income Hispanic adults with substance use disorders remained significantly less likely to be insured, compared with their Black or White counterparts. Substance use treatment rates remained low for all three groups. Policy and service reforms are needed to further extend the reach of Medicaid coverage to these vulnerable populations. Greater investment is needed in culturally sensitive and linguistically appropriate outreach for Medicaid enrollment in Hispanic communities. Efforts are also needed to increase the number of Medicaid-certified substance use treatment programs and expand integration of Medicaid-reimbursed substance use treatment into primary care.