Abstract
Objectives. We examined prereform patterns in insurance coverage, access to care, and preventive services use by race/ethnicity in adults targeted by the coverage expansions of the Patient Protection and Affordable Care Act (ACA).
Methods. We used pre-ACA household data from the Medical Expenditure Panel Survey to identify groups targeted by the coverage provisions of the Act (Medicaid expansions and subsidized Marketplace coverage). We examined racial/ethnic differences in coverage, access to care, and preventive service use, across and within ACA relevant subgroups from 2005 to 2010. The study took place at the Agency for Healthcare Research and Quality in Rockville, Maryland.
Results. Minorities were disproportionately represented among those targeted by the coverage provisions of the ACA. Targeted groups had lower rates of coverage, access to care, and preventive services use, and racial/ethnic disparities were, in some cases, widest within these targeted groups.
Conclusions. Our findings highlighted the opportunity of the ACA to not only to improve coverage, access, and use for all racial/ethnic groups, but also to narrow racial/ethnic disparities in these outcomes. Our results might have particular importance for states that are deciding whether to implement the ACA Medicaid expansions.
The existence of persistent racial/ethnic disparities in insurance coverage, access to medical care, and adherence to preventive services recommendations has been well documented.1–7 Reducing disparities such as these is a core objective of the Patient Protection and Affordable Care Act (ACA; Pub L No. 111–148). In this study, we used the Medical Expenditure Panel Survey (MEPS) to take a detailed look at disparities in coverage, access to care, and preventive services use among adults aged 19 to 64 years, and we examined these disparities against the backdrop of the coverage provisions of the ACA.
The ACA contains numerous provisions that directly or indirectly seek to reduce disparities in health care. As a result of ACA coverage provisions, more than half of all states and the District of Columbia (DC) now provide Medicaid up to 138% of poverty,8 and subsidized Marketplace coverage is available in all states and DC, between 100% and 400% of poverty, to those who are ineligible for public coverage and who are not offered affordable employer-sponsored insurance (ESI). The ACA includes preventive services mandates (for non-grandfathered, non–self-insured private plans) and federal incentives for states to cover preventive screening under Medicaid and to improve care coordination. The ACA also contains numerous other provisions aimed at reducing disparities, including enhanced federal monitoring, incentives to increase provider cultural competency, and more.9,10
The effect of these provisions on disparities in health care will depend in large measure on what Eisenberg and Power11 termed the “voltage drops” associated with incomplete take-up of coverage by eligible individuals, the care-seeking behavior of newly insured individuals, and the ability of the health care system to accommodate increased demand.12 The full extent of these voltage drops may be unknown for several more years, because the full effects of the ACA are not expected to be observed until 2016 or beyond, especially for services that are not recommended on an annual basis. Nevertheless, it is possible to gain insights into the potential impact of the ACA by examining the pre-reform disparities in coverage, access, and preventive care use across the groups the ACA targeted for coverage expansions.
We examined households from 2005 to 2010, a period before implementation of the main coverage provisions of the ACA. We examined the extent to which minorities were disproportionately represented among the groups targeted for coverage expansions, and we identified pre-reform disparities, by ACA coverage groups, in coverage, access to care, and preventive services use. Our analysis examined the opportunity of the ACA not only to improve coverage, access, and use for all racial and ethnic groups, but also to narrow the racial/ethnic disparities in these outcomes. The study took place at the Agency for Healthcare Research and Quality (AHRQ) in Rockville, Maryland.
MEPS is a nationally representative survey of the civilian noninstitutionalized population of the United States that is sponsored by the AHRQ.13 We restricted our sample to adult citizens and legal immigrants aged 19 to 64 years. To increase precision, we pooled data from 6 years of MEPS (2005–2010). MEPS collects detailed data on income, assets, family structure, and state of residence, which enabled us to simulate eligibility for public coverage and other ACA relevant subgroups. MEPS also collects information on individuals’ usual source of care, receipt of preventive services, and demographic characteristic information (e.g., age and race/ethnicity).
We used the PUBSIM model to simulate public coverage eligibility. The PUBSIM model, which was developed by AHRQ researchers, uses detailed state-and-year specific Medicaid eligibility rules and information from the MEPS on the family income, assets, state of residence, family structure, and so on of adults to simulate eligibility for Medicaid, the Children’s Health Insurance Program (CHIP), and state-funded public health insurance programs.14 Pre-ACA eligibility for public coverage is simulated using the 2005 to 2010 eligibility rules. Eligibility for Medicaid under the ACA is simulated using the modified adjusted gross income (MAGI) of the potential eligibility units for adults and using information on states’ decisions, as of early 2014, of whether to expand Medicaid. Eligibility for Marketplace subsidies is simulated based on MAGI and eligibility for ESI. Further details of the PUBSIM model, including the descriptions of different types of public coverage, the methods of forming potential eligibility units and calculating MAGI, and the data sources, have been previously published.14
Our analysis examined access to care and use of preventive services by race/ethnicity and by ACA relevant coverage groups. We defined the groups hierarchically, and the groups were mutually exclusive. The first group was the “pre-ACA public eligibles,” which included individuals enrolled in or simulated to be eligible for free or highly subsidized public coverage, including Medicaid, Medicare for those younger than 65 years, and CHIP. Some adults are eligible for public coverage through CHIP parent coverage.15 The next group was the “pre-ACA ESI eligibles,” which consisted of those generally much higher on the socioeconomic spectrum, who were eligible for ESI (themselves or through their spouses), enrolled in Tricare or ChampVA coverage, or enrolled in private coverage held by policyholders outside their reporting unit. In the nonexpansion states, we excluded from this group those who had access to private coverage but had MAGI income less than 100% of poverty and were uninsured. It is possible that the offers these individuals had were not affordable.
After defining these 2 groups of pre-ACA eligibles, we next identified 5 subgroups targeted by ACA coverage expansions. The first group was individuals in Medicaid expansion states with MAGI up to 138% of poverty, who would be newly eligible for Medicaid. The second group was those in nonexpansion states with MAGI up to 100% of poverty, who would have been eligible had their states expanded Medicaid. The last 3 groups were a “high-subsidy” Marketplace group (with MAGI at or less than 200% of poverty), a “low-subsidy” group (with MAGI more than 200% and up to 400% of poverty), and a “no-subsidy” group (with MAGI more than 400% of poverty).
Our coverage measure was a dichotomous indicator for having public or private insurance at any point during the first several months of the year (i.e., the first round of data collection in the calendar year). Our measure of access to care was an indicator of whether the individual had a usual source of care (other than an emergency room). Data on preventive services use came from a series of questions asked about receipt and timing of specific preventive health services. We applied United States Preventive Services Task Force (USPSTF) recommendations with regard to frequency of the preventive services.16,17 We examined 6 measures of USPSTF-recommended services: flu shot (men and women aged 19–64 years, within the previous year), blood cholesterol screening (men aged 35–64 years and women aged 45–64 years, within the previous 5 years), blood pressure screening (men and women aged 19–64 years, within the previous 2 years), colorectal cancer screening (men and women aged 50–64 years, a blood stool test within the previous year, or a sigmoidoscopy or colonoscopy within the previous 5 years), Papanicolaou test (women aged 19–64 years with no hysterectomy, within the previous 3 years), and mammogram (women aged 40–64 years, within previous 2 years). In addition, although the USPSTF does not recommend any general medical examination, we nevertheless examined whether the adult had a “general checkup” during the year. We did this because the majority of patients and physicians think it is important to receive annual general medical checkups, and because checkups might proxy some types of recommended care that were not measured explicitly by MEPS.18,19
We presented estimates by race and ethnicity for the following single-race categories: non-Hispanic Whites (“Whites”), non-Hispanic Blacks (“Blacks”), and Hispanics. Sample size limitations precluded presenting separate estimates for non-Hispanic Asians and others (including multirace) by ACA relevant coverage groups.
All estimates (percentages) are population-weighted and age-standardized, whereby the age distribution in each “cell,” defined by race/ethnicity and by ACA relevant subgroup, matched the age distribution of the full sample. SEs and the Wald tests of significance of the differences across subgroups accounted for the complex survey design of the MEPS. All differences discussed in the Results section were significant at the 5% level unless otherwise stated.
Figure 1 presents the distribution by ACA relevant subgroups by race/ethnicity (see data available as a supplement to the online version of this article at http://www.ajph.org for point estimates and SEs). Compared with Whites, larger proportions of Blacks and Hispanics would be newly eligible for Medicaid or would be eligible for Marketplace coverage with high subsidies (i.e., MAGI less than 200% of poverty). Summing across expansion and nonexpansion states, individuals targeted by Medicaid expansions represented 6.4% of Whites versus 10.3% of Blacks and 13.7% of Hispanics. However, approximately 3 in 5 Blacks targeted by ACA Medicaid expansions lived in states that have not yet expanded Medicaid (derived from estimates in Figure 1). For Whites and Hispanics, the ratio was flipped, with approximately 3 in 5 living in a state that had expanded Medicaid.
FIGURE 1— Simulated eligibility for the Affordable Care Act Medicaid expansions and marketplace coverage by race/ethnicity: United States, 2005–2010.
Note. ESI = employer-sponsored insurance; MAGI = modified adjusted gross income.
Figure 2 presents insurance coverage rates by ACA coverage group and by race/ethnicity (data available as a supplement to the online version of this article at http://www.ajph.org). We found 4 key findings.
FIGURE 2— Age-adjusted racial/ethnic differences in the percentages of nonelderly adults with insurance coverage, by Affordable Care Act relevant subgroups: United States, 2005–2010.
Note. ESI = employer-sponsored insurance; MAGI = modified adjusted gross income.
First, coverage rates varied substantially across ACA coverage categories for all racial/ethnic groups, including non-Hispanic Whites. The lowest rates occurred among the ACA-targeted groups, which were groups in which Blacks and Hispanics were disproportionately represented. Second, differences in coverage rates existed between Whites and the other 2 racial/ethnic groups within nearly all ACA groups (the exception was the Black–White difference among prereform public eligibles). Third, percentage point differences between Whites and Hispanics were even larger than those between Whites and Blacks, but this difference was not significant among the no-subsidy Marketplace eligibles. Fourth, the percentage point coverage differences across race/ethnicity were largest among those groups targeted by the ACA. For instance, whereas the difference in coverage rates between Whites and Hispanics with access to ESI was 10.3 percentage points (93.9% vs 83.6%), the corresponding gap among those eligible for low Marketplace subsidies was 26.1 percentage points (45.0% vs 18.9%). If coverage rates in the Medicaid expansion groups were to rise to the levels observed among prereform eligibles, and if those eligible for Marketplace subsidies were to reach the coverage rates observed among ESI eligibles, then the overall disparities in coverage would be greatly reduced.
Figure 3 shows the shares of individuals with a usual source of care (USC) by ACA coverage group and by race/ethnicity (data available as a supplement to the online version of this article at http://www.ajph.org). As with coverage, USC rates varied substantially across ACA groups, with the lowest rates occurring among the ACA-targeted groups (in which Blacks and Hispanics were disproportionately represented). Moreover, racial/ethnic differences in USC rates existed within all of the ACA groups (although the Black–White difference in USC rates among pre-ACA public eligibles was small and only significant at the 10% level), with the largest differences between Whites and Hispanics and among those groups targeted by the ACA. For instance, whereas the difference in USC rates between Whites and Hispanics with access to ESI was 9.8 percentage points (79.9% vs 70.1%), the corresponding gap among those eligible for high Marketplace subsidies was 25.9 percentage points (62.0% vs 36.1%). If, despite the inevitable voltage drops in insurance take-up, care-seeking, and system capacity, USC rates were to rise to the levels we observed in the pre-reform public eligibles (for the Medicaid expansion groups) and pre-reform ESI eligibles (for the Marketplace subsidy groups), then overall racial/ethnic disparities would greatly diminish.
FIGURE 3— Age-adjusted racial/ethnic differences in the percentages of nonelderly adults with a usual source of care, by Affordable Care Act relevant subgroups: United States, 2005–2010.
Note. ESI = employer-sponsored insurance; MAGI = modified adjusted gross income.
The general checkup estimates presented in Figure 4 revealed a somewhat different disparity pattern than what we observed for coverage and USC. As in Figures 2 and 3, checkup rates were substantially lower among the groups targeted by the ACA, and substantial disparities existed within these groups. In addition, we found that the White versus Hispanic differences were larger than those between Whites and Blacks. However, in Figure 4, the percentage point differences by race and ethnicity within these groups were more uniform across ACA groups, and many of the Black–White differences were not statistically significant (data available as a supplement to the online version of this article at http://www.ajph.org). These results highlight the potential of the ACA coverage provisions to increase checkup rates in the targeted groups. However, even if these groups reach the checkup rates we observed among pre-ACA public or ESI eligibles, this shift would not appreciably narrow the prereform percentage point disparities we observed within the targeted groups.
FIGURE 4— Age-adjusted racial/ethnic differences in the percentages of nonelderly adults with a general checkup within the past year, by Affordable Care Act relevant subgroups: United States, 2005–2010.
Note. ESI = employer-sponsored insurance; MAGI = modified adjusted gross income.
We observed yet another pattern regarding the pre-ACA distribution of blood cholesterol screening (data available as a supplement to the online version of this article at http://www.ajph.org). We observed some large differences across ACA groups, but relatively small racial/ethnic differences within these groups. Thus, apart from the fact that Blacks and Hispanics were disproportionately represented among the groups targeted by the coverage expansions of the ACA, our estimates suggested that there was relatively little pre-ACA disparity to be reduced.
The remaining preventive care measures were blood pressure screening, colorectal cancer screening, Papanicolaou test, mammography, and flu shots. With the exception of flu shots, our results for these other measures resembled aspects of those presented previously (data available as a supplement to the online version of this article at http://www.ajph.org). For blood pressure screening, White versus Black differences were small in all ACA groups, whereas the White versus Hispanic disparity followed the pattern for USC in Figure 3 (with the lowest rates and the largest disparities observed among the groups targets by the ACA coverage expansions). The highest rates we observed were among Whites and Blacks with pre-ACA eligibility for public or ESI coverage (with rates more than 90%). The lowest rate we observed was among Hispanics in the high-subsidy Marketplace group, at 63.8%.
Some of the largest differentials we observed were with respect to colorectal cancer screening. Whereas Whites and Blacks with pre-ACA ESI eligibility had 52.3% and 53.1% screening rates, respectively, Hispanics in that same ACA coverage group had only a 37.5% screening rate. Moreover, these rates all dropped significantly among the ACA target groups. The White and Hispanic screening rates among high-subsidy Marketplace eligibles were 36.6% and 19.7%, respectively.
Papanicolaou test and mammograms both followed the pattern we observed for cholesterol screening. Rates of adherence were generally high, with only modestly lower rates among the groups targeted by the ACA coverage expansions. Moreover, we saw no evidence of pre-ACA disparities; Whites in many categories had lower adherence rates than Blacks or (in a few cases) Hispanics.
Finally, adherence to flu shot recommendations was fairly low for all coverage groups and for all races/ethnicities (data available as a supplement to the online version of this article at http://www.ajph.org). Flu shot rates among pre-ACA public eligibles were 32.0%, 29.1%, and 29.9% for Whites, Blacks, and Hispanics, respectively. Among pre-ACA ESI eligibles, the flu shot rate for Whites was quite similar at 32.2%; however, rates were lower for Blacks (23.9%) and for Hispanics (27.7%). Thus, we observed a substantially larger Black–White disparity among pre-ACA ESI eligibles than we did among pre-ACA public eligibles. There were also large differences in flu shot rates between those with pre-ACA eligibility (public or ESI) and those targeted by the ACA coverage expansions, with significant disparities across race/ethnicity within each coverage group.
We investigated patterns of racial and ethnic disparities in coverage, access to care, and preventive services use among ACA relevant subgroups of the nonelderly adult population. Our results reinforced previous evidence that minorities were disproportionately represented among the groups targeted by the ACA coverage expansions.20,21 In addition, for all 3 racial/ethnic groups we examined, coverage, access, and use rates were significantly lower among the subgroups targeted by the ACA compared with the groups who were prereform eligible for either public or ESI coverage. We observed disparities, especially between non-Hispanic Whites and Hispanics, within ACA coverage groups for most of the measures we examined. Percentage point disparities for many of the measures we studied were largest within the groups targeted by the ACA affordable coverage provisions.
Although we did not simulate postreform outcomes, our results clarified the opportunities for the ACA to reduce disparities with regard to coverage, access to care, and preventive services use. Early evidence from 2014 suggested that the ACA helped to narrow racial/ethnic disparities in coverage.22–25 Whether the potential of the ACA for reduced disparities in access and preventive services use will also be achieved depends on the voltage drops faced by the targeted groups.
It is plausible that groups that obtained access to Medicaid (in states that expanded the program) would achieve coverage, access, and use levels similar to those of pre-ACA public eligibles, especially because ACA enhanced the federal match rate for Medicaid coverage and enhanced federal incentives for states to provide preventive services to adults.26 However, supply side constraints, such as availability of physicians who would accept new Medicaid patients, might limit the possible gains in access and preventive services use, at least in the short run.
The extent to which coverage, access, and use rates among Marketplace subsidy eligibles would mirror those among pre-reform ESI eligibles is also difficult to predict. Recent studies suggested that low- or moderate-income individuals who purchased health plans through the Marketplaces faced similar premium costs and deductibles as adults in the same income ranges with employer-provided coverage.27,28 Moreover, public cost-sharing reductions are available to those under the 200% poverty rate. At the same time, families in the subsidy-eligible groups will generally have lower incomes than those with prereform ESI eligibility, and Marketplace subsidy eligibles with incomes more than 200% of poverty receive lower premium subsidies and are ineligible for public cost-sharing reductions. Supply side constraints may represent additional voltage drops that affect the coverage, access, and use of Marketplace subsidy eligibles.
Several limitations to our study should be noted. First, eligibility indicators for public coverage and for Marketplace subsidies were subject to error, because they were simulated rather than observed. Second, data from the postreform period were not yet available, and therefore, the eligibility estimates did not reflect possible subsequent changes in the distributions of income, insurance coverage, and other demographic characteristic variables. Third, we did not account for potential endogeneity of the population categories defined by income and insurance coverage. Fourth, we did not examine other measures of access to care, such as linguistic, cultural, and health literacy barriers, and timeliness of care.
Despite these caveats, the magnitude of the coverage, access, and use differences we observed—both across ACA groups and by race/ethnicity within these groups—highlighted the opportunity for the ACA coverage expansions to reduce racial and ethnic disparities. Our results might have particular importance for states deciding whether to implement the ACA Medicaid expansions.
Acknowledgments
We thank our anonymous reviewers for their detailed reviews, helpful comments, and suggestions.
Human Participant Protection
This study was covered under Chesapeake institutional review board Agency for Healthcare Research and Quality protocol, Secondary Analysis of Confidential Data From the Medical Expenditure Panel Survey (CRRI 0504015).
