Effects of Medicaid expansion on insurance coverage and health services use among adults with disabilities newly eligible for Medicaid

Abstract

Objective

To estimate the effects of Affordable Care Act (ACA) Medicaid expansion on insurance and health services use for adults with disabilities who were newly eligible for Medicaid.

Data Sources

2008–2018 Medical Expenditure Panel Survey data.

Study Design

We used the Agency for Healthcare Research and Quality (AHRQ) PUBSIM model to identify adults aged 26–64 years with disabilities who were newly Medicaid‐eligible in expansion states or would have been eligible in non‐expansion states had those states opted to expand. Outcomes included insurance coverage; access to care; receipt of primary care, outpatient specialty physician services, and preventive services; and out‐of‐pocket health care spending. To estimate the effects of Medicaid expansion, we used two‐way fixed effects models and a triple differences framework to compare pre‐post changes in each outcome in expansion and non‐expansion states for adults with and without disabilities.

Extraction Methods

We simulated Medicaid eligibility with the AHRQ PUBSIM model, which uses state‐specific Medicaid rules and MEPS data on family relationships, state of residence, and income.

Principal Findings

Among adults with disabilities who were newly eligible for Medicaid, Medicaid expansion was associated with significant increases in full‐year Medicaid coverage (35.9 percentage points [pp], p < 0.001), receipt of primary care (15.5 pp, p < 0.01), and receipt of flu shots (19.2 pp, p < 0.01), and a significant decrease in out‐of‐pocket spending (−$457, p < 0.01). There were larger improvements for adults with disabilities compared to those without disabilities in full‐year Medicaid coverage (11.0 pp, p < 0.01) and receipt of flu shots (18.0 pp, p < 0.05).

Conclusions

Medicaid expansion was associated with improvements in full‐year insurance coverage, receipt of primary and preventive care, and out‐of‐pocket spending for adults with disabilities who were newly eligible for Medicaid. For insurance coverage, preventive care, and some primary care measures, there were differentially larger improvements for adults with disabilities than for those without disabilities.

What is known on this topic

• Adults with disabilities often have additional health care needs but also face additional barriers to accessing care.

• Prior studies of low‐income adults with disabilities found Medicaid expansion was associated with more insurance coverage and fewer cost‐related barriers to care but no effects on access and service use.

What this study adds

• Among adults with disabilities who were newly eligible for Medicaid, Medicaid expansion was associated with increased receipt of primary and preventive care and reduced out‐of‐pocket health spending.

• Among newly eligible adults, Medicaid expansion differentially increased the probability of having full‐year Medicaid coverage and or receiving a flu shot for adults with disabilities relative to those without disabilities.

1. INTRODUCTION

In 2019, 24% of the adult population aged 18–64 years had a disability. ¹ This population faces significant barriers to health care. For example, in 2019 adults aged 18–64 years with disabilities were about two and a half times more likely than adults without disabilities to report that they could not see a doctor due to cost in the last twelve months (29.1% vs. 11.5%). ¹ Health insurance provides financial protection and access to care, and it is especially important for adults with disabilities, who are at increased risk of secondary health conditions ² , ³ and have higher health services use on average compared to adults without disabilities. ⁴

Public programs, including Medicaid and Medicare, are important sources of insurance coverage for adults with disabilities. Many adults with disabilities are eligible to enroll in Medicare or Medicaid due to their disabilities. Medicaid eligibility pathways specifically for people with disabilities are, however, limited to people meeting strict medical standards and having few assets. Gathering the documentation for the evaluation can be onerous and time‐consuming, and applicants often need to appeal initial, adverse decisions. The main Medicaid pathway is enrollment in Supplemental Security Income (SSI). In addition, working adults with disabilities are eligible to pay premiums for Medicaid coverage in most states, but the program parameters vary. ⁵

The 2010 Affordable Care Act (ACA) provided substantial federal matching funding for states to expand Medicaid eligibility to adults with family incomes ≤138% of the federal poverty line. For adults with disabilities, the advantages of this eligibility pathway are no disability medical evaluation requirements, no asset test requirements, and no premiums in all but a few states. These changes allow adults with disabilities who do not meet strict medical disability criteria to enroll. As of 2019, more than 60% of adults with disabilities aged 19–64 years had qualified for Medicaid without SSI, and some of those qualified through the Medicaid eligibility expansion under the ACA. ⁶

Even though Medicaid expansions may increase insurance coverage, improved access to care may not follow. People with disabilities face a health care system that may not be culturally competent with respect to disability and presents obstacles including inaccessible medical facilities or lack of individual accommodations. ⁷ People with mobility impairments may face physical barriers to obtaining care. ⁸ People with cognitive or sensory impairments may receive inaccessible communications about care. In addition, many physicians are not fully confident in their ability to provide the same quality of care to people with disabilities, and some would not welcome people with disabilities in their practice. ⁹ Many physicians also report little to no knowledge about their responsibilities for providing reasonable accommodations under the Americans with Disabilities Act (ADA) and cite lack of formal training in treating patients with disabilities as an important barrier. ¹⁰

In this study, we investigated the effects of ACA‐based Medicaid expansions on insurance coverage and health services use for adults with disabilities who were newly eligible for Medicaid (“newly eligible adults”). Our findings refine and extend the existing literature on the impact of Medicaid expansions for adults with disabilities. In a study applying difference‐in‐differences methods to 2007–2017 Medical Expenditure Panel Survey data, Dong and colleagues (2022) found that states' Medicaid expansions were associated with increased likelihood of having Medicaid coverage among nonelderly adults with disabilities, but they detected no effects on three measures of access and two measures of service use. ¹¹ Effects on overall insurance coverage and private insurance were sensitive to the inclusion of state‐specific trends in the regression. In a study using a triple‐differences framework and the 2010–2019 American Community Survey data, Stimpson and colleagues (2019) compared the effects of Medicaid expansions on adults with and without disabilities. Medicaid expansion increased the likelihood of having any Medicaid coverage and any coverage, in general, more for adults without than with disabilities. ¹² In a third study using 2006–2017 National Health Interview Survey data, Hill and colleagues (2021) found increased Medicaid enrollment and decreased cost‐related barriers to care associated with Medicaid expansion among nonelderly adults with work limitations who worked for pay. ¹³

The current study is innovative in three significant ways. First, we studied adults with disabilities who were not eligible for Medicaid prior to Medicaid expansions in their states. Prior studies included adults with disabilities who were already eligible for Medicaid, including those already enrolled in Medicaid. Hence these studies estimated the combined effect of Medicaid expansion on the newly eligible, those already enrolled, and those previously eligible (i.e., “welcome mat” effects). Prior studies, therefore, may have underestimated the effects of Medicaid expansions on the newly eligible. Second, to better isolate the effects of Medicaid expansion, we excluded adults above the federal poverty line, who became eligible for Marketplace subsidies in non‐expansion states. Third, we examined an array of important outcomes, many not included in prior studies: having full‐year Medicaid coverage above and beyond any point‐in‐time coverage, receipt of preventive care (flu shots), separate measures of primary care and specialty outpatient care, and absolute and relative measures of out‐of‐pocket spending. Fourth, we include two‐stage difference‐in‐differences as a sensitivity analysis to assess potential bias arising from Medicaid expansions rolling out in different states at different times over multiple years. ¹⁴

2. METHODS

2.1. Newly eligible sample

We drew our data from the 2008–2018 Medical Expenditure Panel Survey (MEPS), conducted annually by the Agency for Healthcare Research and Quality (AHRQ) since 1996. The MEPS is a nationally representative, longitudinal survey that collects detailed information on health care use and spending, health insurance, and disability and health status, as well as a wide variety of social, demographic, and economic characteristics for the U.S. civilian noninstitutionalized population. ¹⁵

We used the AHRQ PUBSIM model ¹⁶ , ¹⁷ , ¹⁸ to identify newly eligible adults aged 26–64 years who were either newly Medicaid eligible if they lived in expansion states or would have been eligible had their states opted to expand their Medicaid program. The PUBSIM model applies detailed state‐specific Medicaid eligibility to information from the MEPS about family relationships, state of residence, and amounts and sources of income to determine Medicaid eligibility for each MEPS sample member. To focus on the newly eligible in each expansion state, we excluded from the sample adults who would have been eligible for Medicaid prior to their state's expansion under the ACA. In each non‐expansion state in all data years, we excluded adults who were eligible for Medicaid under that state's 2014 rules, which include the ACA's 5% disregard, to sharpen the comparison across states. ¹⁸ , ¹⁹ Thus, in both types of states, we excluded those eligible for Medicaid through disability pathways, such as those related to SSI.

We excluded adults aged 19–25 years because they may have fallen under the ACA dependent coverage provision, which would have allowed them to remain covered by their parents' health insurance plan if that plan offers dependent coverage. ²⁰ We further excluded adults who were non‐US citizens, or were living in families with modified adjusted gross income (MAGI) above the ACA threshold of 100% of the federal poverty level (FPL). We excluded adults in families with MAGI between 100% and 138% of the FPL because these individuals would have been newly eligible for marketplace coverage in non‐expansion states and excluding them allows us to more accurately identify the effects of the Medicaid expansions.

2.2. Expansion status

We defined Medicaid expansion at the state‐year level according to tracking performed by the Kaiser Family Foundation. ²¹ For the 2008–2018 study period, there were a total of 32 expansion states and 19 non‐expansion states (see Appendix Table A1). Five states expanded Medicaid eligibility prior to the ACA, so there were few newly eligible beneficiaries in those states. Six of the expansion states adopted some of the expanded eligibility criteria prior to 2014 and were excluded in a sensitivity analysis. ²² Nineteen states expanded Medicaid on January 1, 2014, and seven states expanded later in 2014 or in 2015, or 2016. Of the 19 non‐expansion states in our control group, seven expanded Medicaid after our study period, and the remaining 12 states had not expanded Medicaid as of January 2022.

To allow for a phase‐in period in expansion states (see Figures 1 and 2), we excluded from our analyses both the first year in the post period. We excluded the last year in the pre‐expansion period for all states to account for early expansions. That is, we excluded 2013 and 2014 for all non‐expansion states and 90% of the observations in expansion states.

FIGURE 1.

Proportion of adults with disabilities newly eligible for Medicaid with 12 months of any continuous coverage. Sample limited to adults newly eligible for Medicaid in families with modified adjusted gross incomes below the poverty line. The y‐axis is measured in proportions. Time (x axis) is measured in years relative to expansion for expansion states and 2014 for non‐expansion states. In non‐expansion states, the sample is adults who would be newly eligible for Medicaid if the state were to expand coverage. Adults with disabilities had a limitation in major activities of work, housework, or school; a limitation in physical or cognitive functioning; received help or supervision with activities of daily living or instrumental activities of daily living; or had serious difficulty hearing or seeing. Dashed lines represent 95% confidence intervals. Source: Agency for Healthcare Research and Quality, Medical Expenditure Panel Survey 2008–2018 [Color figure can be viewed at wileyonlinelibrary.com]

FIGURE 2.

Proportion of adults with disabilities newly eligible for Medicaid with 12 months of continuous Medicaid coverage. Sample limited to adults newly eligible for Medicaid in families with modified adjusted gross incomes below the poverty line. The y‐axis is measured in proportions. Time (x axis) is measured in years relative to expansion for expansion states and 2014 for non‐expansion states. In non‐expansion states, the sample is adults who would be newly eligible for Medicaid if the state were to expand coverage. Adults with disabilities had a limitation in major activities of work, housework, or school; a limitation in physical or cognitive functioning; received help or supervision with activities of daily living or instrumental activities of daily living; or had serious difficulty hearing or seeing. Dashed lines represent 95% confidence intervals. Source: Agency for Healthcare Research and Quality, Medical Expenditure Panel Survey 2008–2018 [Color figure can be viewed at wileyonlinelibrary.com]

2.3. Disability status

Our main measure of disability was a binary indicator of whether a person had a limitation in major activities of work, housework, or school; a limitation in physical or cognitive functioning; received help or supervision with activities of daily living or instrumental activities of daily living; or had serious difficulty with hearing or seeing. We tested two alternative measures of disability based, respectively, on major activity limitation and daily activity or functioning and found nearly identical results (not shown) on our insurance outcome measures. On other outcome measures, the results (not shown) were qualitatively similar but often stronger for the group defined by major activity limitation.

2.4. Outcome measures

We constructed three binary measures of health insurance coverage from the monthly health insurance indicators available in the MEPS: full‐year (all months eligible for MEPS) coverage of any type, full‐year Medicaid coverage, and full‐year private health insurance coverage. We conducted sensitivity analyses using alternative criteria of 1 month or more of coverage and coverage of 6 months or more and obtained qualitatively similar results with each of these alternatives.

We constructed two binary measures of access to care: whether the person had a usual source of care (USC), and whether the person was either delayed in receiving or unable to obtain needed medical care or prescription drugs.

Our measures of ambulatory care use were binary (any primary care physician visits and any specialty physician visits) and continuous (number of visits to primary care physicians and specialty physicians visits). Primary care physicians are general practice, family practice, internal medicine, pediatrics, osteopathy, and OB\GYN. We also included a binary measure of emergency department (ED) use.

Measures of preventive care were binary indicators of whether a person received a flu shot in the last year, had their blood pressure checked in the past 2 years, and had their cholesterol checked in the past 5 years.

Finally, we assessed two measures of out‐of‐pocket spending: a continuous measure of amounts spent for all health care in dollars and a binary indicator of whether the person was part of family that collectively spent more than 5% of their family income on medical care, a measure of out‐of‐pocket burden for families with low incomes consistent with limits on out‐of‐pocket spending in Medicaid. ²³ In this calculation, the denominator is the maximum of total family income and $1000. ²⁴

2.5. Statistical methods

We adapted the standard two‐way fixed effects (TWFE) difference‐in‐differences model of the effects of the Medicaid expansion, which compares changes in expansion states after policy implementation to any changes observed in non‐expansion states. ¹⁹ , ²⁵ , ²⁶

$$Y_{i,t}=\alpha +{\beta }_1{\text{POST}}_T\times {\text{EXPANSION}}_S+{\beta }_2{\text{POST}}_T\times {\text{EXPANSION}}_S\times {\text{DISABILITY}}_{i,t}+{\gamma }_1{\text{DISABILITY}}_{i,t}+{\gamma }_2{\mathit{\text{POST}}}_T\times {\text{DISABILITY}}_{i,t}+{\gamma }_3{\text{EXPANSION}}_S\times {\text{DISABILITY}}_{i,t}+{\gamma }_4{\mathrm{X}}_{i,t}+{\theta }_S+{\tau }_t+{ϵ}_{i,t},$$

where $Y_{i,t}$ is person i's outcome in year t; ${\text{POST}}_T$ is a measure of whether the person was observed after the state‐specific expansion year in states that expanded, or 2014 in non‐expansion states; ${\text{EXPANSION}}_S$ indicates the type of state; ${\text{DISABILITY}}_{i,t}$ is the person's disability status; ${\mathrm{X}}_{i,t}$ is a vector of personal, family, and geographic characteristics; ${\theta }_S$ and ${\tau }_t$ are state and year fixed effects; $\alpha$, $\beta$, and $\gamma$ are coefficients; and ${ϵ}_{i,t}$ is the error term. ${\text{POST}}_T$ and ${\text{EXPANSION}}_S$ by themselves are excluded because state and year fixed effects are already in the model. The coefficient ${\beta }_1$ measures the standard difference‐in‐differences effect of the Medicaid expansion on the outcome for newly eligible adults without disabilities. The coefficient ${\beta }_2$ on the three‐way interaction term between post‐expansion, expansion‐state, and disability measures any differential effect of the expansion on adults with disabilities.

We obtained most estimates using ordinary least squares regression, which allows for direct interpretation of estimated coefficients as percentage‐point changes in the mean probability of the dependent variable event occurring (e.g., receiving a flu shot) for binary outcomes. For number of visits and out‐of‐pocket expenditures, we used two‐part models, with logit models for any visits and any spending, and generalized linear models (GLM) for the second part. ²⁷ For visits, the second parts are Poisson models with log links. For out‐of‐pocket spending, the second part is a GLM gamma model with a log link. Average marginal effects were derived from the estimated coefficients.

All adjusted difference‐in‐differences models included controls for gender, age, residence in a metropolitan statistical area (MSA), race/ethnicity, highest level of education, parenthood, household size, marital status, interview language, MAGI as a percent of the federal poverty level, number of chronic conditions reported, poor/fair perceived physical health, whether the sample member self‐reported their information, and state and year fixed effects. We performed all analyses in Stata/MP 17, applying the MEPS public use file weights and using the SVY commands to account for the complex, stratified design of the survey.

We also conducted event study analyses to determine whether any trends preceded Medicaid expansion, numerous sensitivity analyses, and a falsification test, the results of which are described below, and tables are in the appendix.

3. RESULTS

The total unweighted study sample size was 11,261, representing an average annual weighted estimate of about 10 million adults aged 26–64 years with low family incomes during the study period. Table 1 reports sample characteristics, comparing adults with disabilities who were newly eligible for Medicaid in expansion states (n = 1399) and would have been newly eligible in non‐expansion states (n = 1106). The comparison group of adults aged 26–64 without disabilities who were newly eligible for Medicaid regardless of disability status is also shown in Table 1 (n = 3905 in expansion states, n = 4806 in non‐expansion states). Overall, about 51% of all adults resided in expansion states at the time they were surveyed and about 25% of the sample met study criteria for having one or more disabilities.

TABLE 1.

Descriptive statistics among adults newly eligible for Medicaid aged 26–64, 2008–2018

		Adults with disabilities		Adults without disabilities
	All adults	Expansion	Non‐Expansion	Expansion	Non‐Expansion
	(n = 11,261)	(n = 1399)	(n = 1106)	(n = 3950)	(n = 4806)
Pre‐period outcome mean (omitting last year of pre‐period)
Any insurance > = 1 month	0.49	0.60	0.48**	0.49	0.45**
Any insurance full year	0.30	0.38	0.22***	0.31	0.27**
Medicaid > = 1 month	0.19	0.40	0.20***	0.18	0.13**
Medicaid full year	0.09	0.22	0.03***	0.09	0.05***
Private > = 1 month	0.32	0.24	0.30*	0.33	0.34
Private full year	0.21	0.16	0.19	0.23	0.22
Have usual source of care	0.57	0.76	0.72	0.54	0.49**
Barrier	0.19	0.34	0.38	0.14	0.13
Flu shot	0.25	0.34	0.34	0.23	0.21
Blood pressure check <2 yrs	0.83	0.92	0.94	0.80	0.81
Cholesterol check <5 yrs	0.78	0.85	0.89*	0.75	0.75
Any primary care doctor visit	0.44	0.60	0.65	0.39	0.38
# of pimary care doctor visits	1.33	2.49	2.30	0.99	1.02
Any specialist doctor visit	0.31	0.57	0.58	0.24	0.23
# of specialist doctor visits	1.47	4.10	2.90**	0.80	0.82
Any emergency department visit	0.19	0.31	0.31	0.15	0.17**
Mean out‐of‐pocket spending	726	1417	1380	480	558
Out‐of‐oocket spending >5% Family income	0.40	0.49	0.58**	0.32	0.32
Post‐period outcome mean (omitting first year of post‐period)
Any insurance > = 1 month	0.72	0.91	0.57***	0.83	0.57***
Any insurance full year	0.49	0.73	0.31***	0.57	0.36***
Medicaid > = 1 month	0.43	0.75	0.26***	0.60	0.19***
Medicaid full year	0.26	0.59	0.04***	0.39	0.09***
Private > = 1 month	0.34	0.22	0.35**	0.31	0.41***
Private full year	0.21	0.14	0.25**	0.17	0.25***
Have usual source of care	0.60	0.77	0.70*	0.58	0.51**
Barrier	0.14	0.25	0.32	0.09	0.08
Flu shot	0.30	0.43	0.26**	0.30	0.25
Blood pressure check <2 yrs	0.86	0.97	0.91**	0.82	0.82
Cholesterol check <5 yrs	0.83	0.87	0.90	0.80	0.81
Any primary care doctor visit	0.43	0.64	0.57	0.39	0.33**
# of primary care doctor visits	1.43	2.77	1.89**	1.22	0.91**
Any specialist doctor visit	0.32	0.58	0.54	0.25	0.21*
# of specialist doctor visits	1.57	3.74	3.05	1.17	0.53***
Any emergency department visit	0.21	0.34	0.30	0.15	0.17
Mean out‐of‐pocket spending	521	473	1115***	346	517**
Out‐of‐pocket spending >5% family income	0.27	0.33	0.46***	0.20	0.24*
Sample characteristics all years (omitting last year of pre‐period and first year of post‐period)
Expansion	0.51	1.00	0.00	1.00	0.00
POST	0.42	0.46	0.47	0.39	0.41
Disability	0.25	1.00	1.00	0.00	0.00
Work, school, or housework limitation	0.18	0.72	0.68*	0.00	0.00
Other limitation	0.20	0.84	0.84	0.00	0.00
Female	0.51	0.51	0.57**	0.48	0.53***
Age 26–34	0.31	0.15	0.14	0.37	0.35
Age 35–44	0.22	0.18	0.18	0.20	0.26***
Age 45–54	0.24	0.33	0.33	0.21	0.20
Age 55–64	0.24	0.34	0.35	0.22	0.19**
White (omitted)	0.57	0.67	0.60**	0.60	0.50***
Hispanic	0.13	0.08	0.09	0.14	0.16
Black	0.22	0.16	0.24**	0.18	0.29***
Asian	0.03	0.02	0.01**	0.05	0.03**
Other	0.04	0.06	0.06	0.03	0.03
Married	0.31	0.22	0.27*	0.29	0.36***
Parent of minor child	0.26	0.09	0.14**	0.20	0.42***
Family size	2.69	2.07	2.19	2.70	3.06***
Non‐English interview	0.06	0.02	0.03	0.06	0.08
Proxy reported	0.43	0.33	0.31	0.49	0.44**
No HS diploma (omitted)	0.18	0.19	0.21	0.16	0.17
HS diploma	0.42	0.42	0.46	0.41	0.42
Some college	0.26	0.29	0.25	0.23	0.28***
Bachelors	0.10	0.07	0.06	0.14	0.10***
Masters or higher	0.03	0.03	0.02	0.04	0.02**
MAGI as %FPL	43.81	31.73	32.19	45.49	50.07*
# chronic conditions	1.31	2.41	2.46	0.94	0.94
Poor/fair health	0.26	0.59	0.62	0.14	0.15
MSA	0.80	0.79	0.80	0.83	0.78
Physicians per 1000	2.27	2.47	1.92***	2.55	2.01***
2008	0.10	0.10	0.11	0.10	0.09
2009	0.12	0.12	0.11	0.12	0.12
2010	0.12	0.10	0.09	0.13	0.12
2011	0.12	0.11	0.12	0.11	0.13*
2012	0.12	0.10	0.09	0.12	0.13
2013	0.01	0.02	0.00***	0.02	0.00***
2014	0.00	0.00	0.00**	0.01	0.00**
2015	0.11	0.11	0.14	0.10	0.11
2016	0.12	0.13	0.13	0.12	0.11
2017	0.11	0.12	0.11	0.10	0.10
2018	0.09	0.10	0.10	0.08	0.09

Note: Sample limited to adults newly eligible for Medicaid in families with modified adjusted gross incomes below the poverty line. Both the first year in the post period and the last year in the pre‐expansion period is excluded for all states (2013 and 2014 for all non‐expansion states). In non‐expansion states, the sample is adults who would be newly eligible for Medicaid if the state were to expand coverage. Adults with disabilities had a limitation in major activities of work, housework, or school; a limitation in physical or cognitive functioning; received help or supervision with activities of daily living or instrumental activities of daily living; or had serious difficulty hearing or seeing. ***p < 0.01, **p < 0.05, *p < 0.10 expansion versus non‐expansion.

Abbreviations: FPL, federal poverty level; MAGI, modified adjusted gross income; MSA, metropolitan statistical area.

Source: Agency for Healthcare Research and Quality, Medical Expenditure Panel Survey 2008–2018.

Figures 1 and 2 show trends in unadjusted insurance coverage in event time for adults with and without disabilities in expansion and non‐expansion states. In line with findings in the literature, insurance coverage for both samples increased in expansion states with the implementation of Medicaid expansion in 2014, while coverage for those in non‐expansion states remains relatively constant. These figures also provide evidence for the parallel trends assumption necessary for our primary analyses. Prior to the expansion, insurance coverage in both expansion and non‐expansion states trended similarly, implying they would have continued their paths but for Medicaid expansion. Evidence for unadjusted parallel trends for our other outcomes can be found in similar graphs in Appendix Figure A1 and a more formal test of adjusted trends with an event study is detailed below.

3.1. Difference‐in‐differences regression estimates

Table 2 provides TWFE difference‐in‐differences estimates for the insurance outcomes (Tables A2 and A3 in the appendix show full regression model results). The results for the full newly eligible population are consistent with the unadjusted difference‐in‐differences estimates calculated from the sample means (not shown). Medicaid expansion was associated with a significant increase in the likelihood of having any source of full‐year insurance coverage of 29.6 percentage points (p < 0.001) among adults with disabilities and of 16.7 percentage points (p < 0.001) among those without disabilities. Thus, there was a significant 12.9 percentage point (p < 0.05) larger increase in any full‐year coverage associated with Medicaid expansion for adults with disabilities compared to adults without disabilities.

TABLE 2.

Adjusted difference‐in‐differences estimates of insurance coverage, adults newly eligible for Medicaid aged 19–64, 2008–2018

	Full‐year of coverage			1 month or more of coverage
	Difference	Std. Err	p‐value	Difference	Std. Err	p‐value
Any type of coverage
No disability	0.167	(0.031)	0.000	0.216	(0.026)	0.000
Disability	0.296	(0.047)	0.000	0.257	(0.047)	0.000
Difference	0.129	(0.057)	0.024	0.041	(0.051)	0.433
Medicaid
No disability	0.250	(0.022)	0.000	0.357	(0.025)	0.000
Disability	0.359	(0.038)	0.000	0.285	(0.045)	0.000
Difference	0.110	(0.040)	0.007	−0.072	(0.048)	0.134
Private coverage
No disability	−0.081	(0.028)	0.004	−0.094	(0.029)	0.001
Disability	−0.051	(0.044)	0.242	−0.031	(0.055)	0.575
Difference	0.029	(0.055)	0.595	0.064	(0.065)	0.329

Note: Sample limited to adults newly eligible for Medicaid in families with modified adjusted gross incomes below the poverty line (N = 11,261). Both the first year in the post period and the last year in the pre‐expansion period is excluded for all states (2013 and 2014 for all non‐expansion states). In non‐expansion states, the sample is adults who would be newly eligible for Medicaid if the state were to expand coverage. Adults with disabilities had a limitation in major activities of work, housework, or school; a limitation in physical or cognitive functioning; received help or supervision with activities of daily living or instrumental activities of daily living; or had serious difficulty hearing or seeing.

Abbreviation: Std. Err: standard error.

Source: Agency for Healthcare Research and Quality, Medical Expenditure Panel Survey 2008–2018.

Medicaid expansion was also associated with similar increases in full‐year Medicaid coverage. Those with disabilities had a significant increase in full‐year Medicaid coverage of 35.9 percentage points (p < 0.001) while those without disabilities had a significant increase of 25.0 percentage points (p < 0.001). Therefore, those with disabilities had a significant 11.0 percentage point (p < 0.01) larger increase in full‐year Medicaid coverage than those without disability.

Changes in full‐year private coverage associated with Medicaid expansion did not appear to follow the same pattern as Medicaid coverage and any coverage. While there was a significant decrease in full‐year private coverage among adults without disability associated with expansion (−8.1 percentage points, p < 0.01), we did not detect group‐specific changes for adults with disabilities or differential effects for those with disabilities compared to those without disabilities.

Table 3 presents the results of our analysis of health care use and out‐of‐pocket spending. The adjusted triple‐differences results are presented here with the full regression results appearing in the Appendix, Tables A4 and A5. While the point estimates are similar to the unadjusted difference‐in‐differences estimates (not shown), the adjusted effects of Medicaid expansion on having a usual source of care and barriers to care for newly eligible adults with or without disabilities were not statistically significant. However, when examining preventive services use as measured by receipt of a flu shot, we found evidence of a differential effect of Medicaid expansion for those with disabilities. Relative to adults without disabilities, those with disabilities had a significant, 18.0 percentage point (p < 0.05) greater increase in the probability of getting a flu shot (disabilities: 19.2 pp, p < 0.01; no disabilities: 1.2 pp, p = 0.742).

TABLE 3.

Adjusted difference‐in‐differences estimates of access, use and out‐of‐pocket spending, adults newly eligible for Medicaid Aged 19–64, 2008–2018

	Difference	Std. Err	p‐value
Have usual source of care
No disability	0.017	(0.033)	0.609
Disability	0.062	(0.049)	0.207
Difference	0.045	(0.057)	0.432
Report unable to get or delay in needed care (n = 10,334)
No disability	−0.024	(0.021)	0.259
Disability	−0.016	(0.045)	0.723
Difference	0.008	(0.049)	0.872
Flu shot (n = 8688)
No disability	0.012	(0.038)	0.742
Disability	0.192	(0.058)	0.001
Difference	0.180	(0.070)	0.011
Blood pressure check (n = 8704)
No disability	−0.005	(0.030)	0.876
Disability	0.083	(0.031)	0.008
Difference	0.088	(0.043)	0.042
Cholesterol check (n = 8357)
No disability	−0.019	(0.032)	0.567
Disability	0.029	(0.044)	0.509
Difference	0.048	(0.053)	0.365
Any primary care doctor visit
No disability	0.045	(0.030)	0.130
Disability	0.155	(0.055)	0.005
Difference	0.110	(0.059)	0.061
# of primary care doctor visits
No disability	0.391	(0.153)	0.011
Disability	0.921	(0.251)	0.000
Difference	0.529	(0.284)	0.063
Any specialist doctor visit
No disability	0.023	(0.025)	0.350
Disability	0.070	(0.054)	0.200
Difference	0.047	(0.057)	0.417
# of specialist doctor visits
No disability	0.998	(0.303)	0.001
Disability	0.278	(0.588)	0.637
Difference	−0.720	(0.585)	0.219
Any emergency department visit
No disability	0.002	(0.020)	0.928
Disability	0.055	(0.043)	0.208
Difference	0.053	(0.047)	0.265
Mean out‐of‐pocket spending ($)
No disability	−191	(72)	0.008
Disability	−457	(168)	0.007
Difference	−266	(196)	0.176
Proportion with out‐of‐pocket spending >5% family income
No disability	−0.043	(0.027)	0.111
Disability	−0.023	(0.056)	0.684
Difference	0.021	(0.061)	0.738

Note: Sample limited to adults newly eligible for Medicaid in families with modified adjusted gross incomes below the poverty line (N = 11,261). Both the first year in the post period and the last year in the pre‐expansion period is excluded for all states (2013 and 2014 for all non‐expansion states). In non‐expansion states, the sample is adults who would be newly eligible for Medicaid if the state were to expand coverage. Adults with disabilities had a limitation in major activities of work, housework, or school; a limitation in physical or cognitive functioning; received help or supervision with activities of daily living or instrumental activities of daily living; or had serious difficulty hearing or seeing.

Abbreviation: Std. Err: standard error.

Source: Agency for Healthcare Research and Quality, Medical Expenditure Panel Survey 2008–2018.

Among adults with disabilities, Medicaid expansion was also associated with significant gains in the likelihood of visiting a primary care doctor (15.5 pp, p < 0.01) as well as in the mean number of times visiting a primary care doctor (0.9 visits, p < 0.001), though we did not find these gains to be significantly different than those for adults without disabilities. We did not observe statistically significant differences in likelihood of specialist physician or ED visits or in the mean number of times a specialist was visited for adults with disabilities (but adults without disabilities had a significant, increase in specialty doctor visits).

Finally, Medicaid expansion was also associated with a significant reduction in out‐of‐pocket spending for adults with disabilities (−$457, p < 0.01). There was not a significant net difference in the reduction in out‐of‐pocket spending between adults with and without disabilities.

3.2. Sensitivity analyses

We conducted a falsification test and several sensitivity analyses to assess the validity and robustness of our estimation approach. First, we created a pseudo implementation date of January 2012 for all expansion states and conducted our main analysis treating 2008–2011 as the pre‐period and 2012 as the post‐period. This falsification analysis tests the comparability of expansion and non‐expansion state trends in the outcome in the pre‐implementation period. If trends are diverging in the pre‐implementation period, then the results of the falsification analysis would be statistically significant, and our identification strategy would be called into question. In general, most of these results are not statistically significant and did not show any consistent patterns in terms of signs or magnitudes (Appendix Table A6). Taken together with the insignificant results from the interaction of the expansion indicator with year indicators reported below, these results affirm our model's assumptions.

Second, we assessed whether staggered policy adoption biases our estimates using two approaches: a two‐stage TWFE difference‐in‐differences estimator, which provides consistent estimates accounting for staggered expansion rollout ¹⁴ , ²⁸ and excluding states that expanded after 2014. The estimates from both tests are similar to the main results (Appendix Tables A7 and A9).

Additional analyses included: limiting the sample to childless adults, excluding early expansion states, removing health status measures (except for ${\text{DISABILITY}}_{i,t}$) from the control variables, and standard difference‐in‐differences models for adults with disabilities. Excluding the early expansion states attenuated the differential impacts of expansions on adults with disabilities' Medicaid enrollment and flu shots. Otherwise, the results of the sensitivity analyses were similar to the main results.

Finally, when we re‐estimated our models using a standard event study approach (Appendix Table A8), interacting our expansion indicator with event‐time indicators. This allows us to further evaluate the parallel trend assumption, with significant differences in the pre‐expansion indicators suggesting potentially divergent trends in the outcomes. With respect to full‐year insurance coverage, two pieces of evidence supporting our primary analytic approach emerged. First, in comparisons of annual changes in the pre‐expansion period, there were no significant differences by disability status, providing additional support for the pre‐period parallel trends assumption. Second, consistent, significant overall and disability group‐specific increases in insurance coverage did not emerge until after the first post‐period year (2014 in most cases), which supported our decision to exclude these observations as a transitional period before the expansion was fully implemented. Similarly, the results for the use and spending outcomes are consistent with the primary DID analysis and do not call into question the parallel trends assumption.

4. DISCUSSION

Low‐income adults with disabilities who do not qualify for Medicaid through disability‐specific pathways are an understudied population. In the current study, we examined the effects of ACA‐driven Medicaid expansion on insurance coverage and health services use for this population. Medicaid expansion was designed as a broad‐based health reform without specific focus on addressing the needs of adults with disabilities. Nonetheless, our analyses show that low‐income adults with disabilities who were newly eligible for Medicaid benefited differentially from Medicaid expansion in several ways: increased insurance coverage, and increased use of preventive and primary care. Low‐income adults with disabilities also benefited from decreased out‐of‐pocket spending.

While similar effects have also been identified in prior studies, ¹¹ , ¹² , ¹³ the findings from the current study advance our understanding of Medicaid expansion's impact on adults with disabilities and low family incomes in multiple ways. First, our study may have found more differential impacts because we refined the analysis to better identify the population affected by Medicaid expansion. We excluded adults who were already eligible for Medicaid without the expansions, and we excluded adults above the poverty line, who were eligible for Marketplace subsidies in non‐expansion states; inclusion of these groups would muddy the comparison group.

Second, our study focuses on whether adults had insurance coverage for a full calendar‐year. For adults with disabilities, who often have disproportionate needs for primary and specialty health services (including services that require ongoing care for sustained periods), consistent coverage is critical. Our results show that adults with disabilities and low family incomes made large, significant gains in full‐year coverage in association with Medicaid expansion—from any source and through Medicaid in particular—and that those gains were significantly larger than the gains experienced by adults without disabilities. Full‐year Medicaid coverage results in an elimination of churning not just on and off any insurance coverage but also between sources of insurance that often have different benefit profiles and degrees of cost‐sharing. Avoiding those changes and gaps in coverage means greater financial protection (which we observed signs of in our out‐of‐pocket spending analysis), less hassle and confusion when navigating the health care system, and potential future benefits. ²⁹ We also observed that Medicaid expansion was associated with a significant reduction in private coverage for newly eligible adults without disabilities but not those with disabilities (both full‐year and 1 month or more of coverage). These decreases were more than offset by increases in Medicaid coverage and resulted in significant increases in coverage from any source. Private coverage estimates for newly eligible adults with disabilities were also negative in sign but were smaller in magnitude and not statistically significant.

Third, our study adds to the literature new findings on differential improvements in preventive care use for adults with disabilities and low family incomes. Preventive care is as important for people with disabilities as for people without disabilities and may be more important for this population. For example, some people with disabilities have chronic conditions that increase the risk of more severe outcomes from influenza, so attaining a higher vaccination rate among people with disabilities is especially important. ³⁰ We found that Medicaid expansion was associated with a large, significant increase in the probability of receiving a flu shot among adults with disabilities and that these gains were substantially larger than any changes observed for adults without disabilities. No significant changes in probability of receiving a flu shot were found for adults without disabilities. In fact, the large, differential improvement for adults with disabilities appeared to come at a time when overall flu shot rates were declining.

Fourth, we studied specialty care, which is important for many people with disabilities. Despite the increase in insurance coverage, use of specialty care did not significantly increase. The increased use of primary care we identified may have reduced the need for some specialty care by addressing health concerns at an earlier stage. Additional research is needed to determine the extent of unmet needs for specialty care, uncover the barriers adults with disabilities face in obtaining specialty care, and develop strategies to overcome those barriers.

The findings from this study have several policy implications. First, our findings contribute novel evidence to a large and growing body of literature showing that large‐scale policy interventions targeting insurance can effectively increase coverage for a wide range of vulnerable populations. Examining a nationally representative, community‐based sample of adults with disabilities, we found that there remained a population of considerable size still newly eligible for Medicaid coverage following enactment of the ACA. Among these individuals, we determined that not only did expansion lead to greater likelihood of having any coverage, but it was also associated with increased odds of having consistent coverage for a full year or more. Therefore, interventions that expand coverage and simplify eligibility rules can be expected to improve financial protection over the longer term, which is particularly important for adults with disabilities who often have greater health care needs. Still, about 73% of newly adults with disabilities in expansion states had full‐year coverage in the post‐expansion period, leaving room for future interventions to further expand access to continuous coverage for this population. Second, our results suggest that Medicaid expansion may be necessary but not sufficient for driving improvement in key measures health services use for this population. Despite the coverage gains we observed, we did not detect increases in having a usual source of care or decreases in having unmet needs (being unable or having delays). Additional, complementary interventions targeting other factors (e.g., network adequacy, accommodations for patients with disabilities, and social determinants of care) are likely needed on top of consistent coverage to ensure reliable access to quality care. Third, despite increased coverage, greater odds of primary care use, and lower average out‐of‐pocket spending, we observed that about a third of newly eligible adults with disabilities in expansion states had emergency department visits in the post‐expansion period, and Medicaid expansion was not associated with a significant reduction in ED visits for this population. If a goal of health reform is to reduce adverse events and episodes of acute care when possible, additional interventions appear necessary.

Despite its strong quasi‐experimental design, our study has limitations. Our TWFE estimator cannot account for any time‐varying factors that change contemporaneously with the treatment. In the case of Medicaid expansion, many states expanded in January 2014, and at the same time other ACA provisions, such as the insurance marketplaces, went into effect. While we have limited the sample to those newly eligible for Medicaid and not eligible for marketplace subsidies, the results remain potentially biased. Other ACA provisions might result in attenuation bias in the estimates of Medicaid expansion. In particular, the ACA required states' Medicaid programs to provide preventive care with a strong evidence base (including influenza vaccines) to newly eligible adults without any cost sharing, so our study identifies the combined effects of eligibility expansion and more generous coverage for these services. Some states also eliminated cost sharing for these services for adults eligible for Medicaid through traditional pathways, and our focus on a sample of newly eligible adults minimizes the potential for that change to affect our estimates. Finally, because we limited the sample to newly eligible adults with incomes <100% FPL and applied other exclusion criteria to isolate the effects of Medicaid expansion, our sample may not have been large enough to detect small to moderate effect sizes; similarly, our sample was not large enough to study potentially heterogeneous impacts across adults with specific types of disabilities.

5. CONCLUSION

Our study suggests that, among newly eligible adults, ACA Medicaid expansions differentially benefited adults with disabilities relative to adults without disabilities by increasing full‐year insurance coverage, increasing the use of preventive and primary care, and by reducing out‐of‐pocket spending.

Supporting information

Appendix S1. Supporting information.

Creedon TB, Zuvekas SH, Hill SC, Ali MM, McClellan C, Dey JG. Effects of Medicaid expansion on insurance coverage and health services use among adults with disabilities newly eligible for Medicaid. Health Serv Res. 2022;57(Suppl. 2):183‐194. doi: 10.1111/1475-6773.14034