Early Coverage, Access, Utilization, and Health Effects Associated With the Affordable Care Act Medicaid Expansions

Context The effect on health care and patient outcomes of the 2014 state Medicaid expansions for low-income adults is of interest to politicians, policymakers, insurers, and the public. Contribution Data from the National Health Interview Survey for 2010 to 2014 show that states implementing the expansions saw improvements in several outcomes, including insurance coverage and quality, health care utilization, and rates of diagnosis of key conditions. There was no improvement in self-reported health status among enrollees. Implication State Medicaid expansions seem to be achieving the broad goals of the Patient Protection and Affordable Care Act. Assessing the effects of such expansions over time, including perceptions of health status among participants, will be important. A key component of the Patient Protection and Affordable Care Act (ACA) was to expand Medicaid eligibility to adults earning up to 138% of the federal poverty level (FPL). Although this provision was originally intended to be enacted in all states, a U.S. Supreme Court decision gave states the option not to adopt it. Twenty-four states decided to forgo the expansion in 2014, affecting 6.7 million uninsured low-income adults who would otherwise have gained eligibility (1). Although a few states have since chosen to expand Medicaid, 19 have not done so at this time (2). Little is known about the effect of the ACA Medicaid expansions on access to care, utilization, and health. This represents a critical gap in knowledge as policymakers continue to debate whether to implement this policy. Early studies have indicated an increase in insurance coverage among low-income adults in states that expanded Medicaid compared with those that did not (36). Other investigators have analyzed previous state Medicaid expansions for low-income adults and found that these expansions were associated with lower mortality, greater coverage and access to care, higher utilization rates, and better self-reported health (710). A recent study found that low-income adults in states that expanded Medicaid under the ACA were less likely than those in nonexpansion states to report having no personal physician or no easy access to medicine (6). However, that study did not include information on several important outcomes, such as the use of health services and the diagnosis of chronic conditions. To date, there has been no direct analysis of the effect of the ACA Medicaid expansions on health care utilization and only limited analysis of the effect on access to care and health among persons gaining coverage. Methods Study Design We used a quasi-experimental difference-in-differences design that compared changes in outcomes for residents of expansion and nonexpansion states before and after the 2014 ACA Medicaid expansions. Our study period included the 4 years before (2010 to 2013) and the first year after (2014) the expansions. The difference-in-differences method adjusted for time-invariant differences in characteristics across the expansion and nonexpansion states as well as secular changes in outcomes over time. We defined expansion states as those implementing the ACA expansion by the end of 2014, with all other states serving as controls. The Medicaid expansions were effective on 1 January 2014 in all states except Michigan and New Hampshire. We excluded 5 states that already provided Medicaid or similar coverage to low-income adults during 2010 to 2013 (see Section 1 of the Supplement). Supplement. Additional Information Data This study used data for 2010 to 2014 from the National Health Interview Survey (NHIS), a nationally representative annual survey conducted by the National Center for Health Statistics (NCHS). In this cross-sectional survey, respondents are interviewed throughout the survey year. One advantage of this survey is its high response rate (>70%) (11). The study sample included nonelderly U.S. citizens (aged 19 to 64 years) in families with incomes less than 138% of the FPL, with information on race, ethnicity, age, sex, marital status, and educational attainment. Approximately 1% of respondents were excluded from the sample because of missing information on marital status or educational attainment; for those who were missing information on family income (9.5%), imputed values from multiple imputation files provided by NCHS were used (12). We excluded noncitizens from the analysis because not all of these persons are eligible for Medicaid (13). Our study used restricted-access state identifiers in the NHIS and was performed in a Census Research Data Center. The study was deemed exempt from review by the investigators' designated institutional review boards. We defined the postexpansion period to include respondent interviews in the second half of 2014 rather than all of 2014 because several of the outcome measures asked about respondent experiences during the previous 12 months. Figure 1 illustrates the timing of this 12-month look-back period for respondents interviewed in the second half of 2014 relative to the timing of the state ACA Medicaid expansions (green shading). For almost all respondents, the majority of the 12-month look-back period occurred during the period of expanded Medicaid eligibility. On average, respondents in expansion states in the second half of 2014 had received 8.7 months of exposure to the ACA Medicaid expansions at the time of the interview. Figure 1. 12-mo look-back for participant response relative to timing of 2014 ACA state Medicaid expansions, by National Health Interview Survey interview month. ACA = Patient Protection and Affordable Care Act. Outcome Measures All outcomes used in the analysis were based on self-reported information in the NHIS. The first set of outcomes was related to insurance coverage and health care utilization. Three binary coverage variables indicated no health insurance coverage (defined as coverage through Medicare, Medicaid, a private insurer, the military, or other government programs [excluding the Indian Health Service]), Medicaid coverage, and private insurance coverage at the time of the interview. In addition, we examined whether respondents reported that their health insurance or health care coverage was better than it had been 1 year before; this measure was not available for the 2010 survey year. The utilization measures were whether the respondent saw or spoke to a physician in general practice, family medicine, or internal medicine; saw or spoke to a medical specialist (excluding obstetrician-gynecologists, psychiatrists, and ophthalmologists); was hospitalized overnight (excluding in the emergency department [ED]); or visited a hospital ED during the previous 12 months. We next considered outcomes related to access to care, diagnoses of health conditions, and self-reported health. To measure access, we used 2 binary variables that indicated whether the respondent did not obtain necessary medical care because of cost or delayed care because of worry about cost within the previous 12 months. We also considered whether the respondent had a usual place of care for when they were sick or needed advice about their health and whether they reported not having a usual place of care because of the expense or a lack of insurance; the latter measure was not available for the 2010 survey year. Following prior work (10), we investigated whether the respondent reported ever having been diagnosed with diabetes, hypertension, or high cholesterol by a physician or health professional. The diagnosis of high cholesterol was available only for the 2012 and 2014 survey years. The health outcomes were whether the respondent reported their health to be very good or excellent, whether their health was better than it had been 1 year before, and whether they mentioned depression as a health problem. Some outcomes were available for all members of surveyed households, whereas others were available only for one sampled member of the household (Section 2 and Table 1 of the Supplement). Statistical Analysis Baseline sample characteristics and unadjusted means for each outcome for 2010 to 2013 (preexpansion period) and the second half of 2014 (postexpansion period) were estimated for expansion and nonexpansion states, with survey design variables provided by NCHS (11) used to account for the complex, multistage sampling design of the NHIS. In addition, we used an F test to evaluate differences in baseline characteristics for the 2 groups of states and unadjusted difference-in-differences for each outcome measure. We estimated a multivariate regression model to compare changes in outcomes for expansion and nonexpansion states. Our independent variable of interest was the interaction between a variable indicating that the state adopted the ACA Medicaid expansion and a variable indicating that the respondent was interviewed in the second half of 2014. The estimated coefficient on this term provided the average difference in outcomes in expansion and nonexpansion states in the second half of 2014 versus before the implementation of the ACA Medicaid expansions. Observations from the first half of 2014 remained in the sample and were indicated with a separate binary variable that was also interacted with the indicator for the Medicaid expansions. Our estimates were adjusted for race and ethnicity, marital status, number of children and adults in the family, educational attainment, and age, as well as state, half-year, and interview quarter fixed effects. The regression models used NHIS sampling weights, and we estimated HuberWhite robust SEs clustered at the state level to account for within-state correlation of the error terms and the state-level nature of the policy change (14). Additional details on the regression model are provided in Section 3 of the Supplement. We used linear probability models rather than nonlinear models to conduct this analysis. The primary drawback of using a linear model with binary outcome