Changes in Health Care Access, Behaviors, and Self-reported Health Among Low-income US Adults Through the Fourth Year of the Affordable Care Act

* Department of Health Law, Policy and Management Boston University School of Public Health, Boston, MA.

Jacob H. Bor

† Department of Global Health and Epidemiology, Boston University School of Public Health, Boston, MA.

* Department of Health Law, Policy and Management Boston University School of Public Health, Boston, MA.

† Department of Global Health and Epidemiology, Boston University School of Public Health, Boston, MA.

Correspondence to: Kevin N. Griffith, MPA, Department of Health Law, Policy and Management, Boston University School of Public Health, 715 Albany Street, T3-West, Boston, MA 02118-2526. ude.ub@tiffirgk.

Associated Data

Abstract

Background:

Health care access increased for low-income Americans under the Affordable Care Act (ACA). It is unknown whether these changes in access were associated with improved self-reported health.

Objective:

Determine changes in health care access, health behaviors, and self-reported health among low-income Americans over the first 4 years of the ACA, stratified by state Medicaid expansion status.

Design:

Interrupted time series and difference-in-differences analysis.

Subjects:

Measures:

Self-reported insurance coverage, access to a primary care physician, avoided care due to cost; self-reported general health, days of poor physical health, days of poor mental health days, and days when poor health limited usual activities; self-reported health behaviors, use of preventive services, and diagnoses.

Results:

Despite increases in access, the ACA was not associated with improved physical or general health among low-income adults during the first 4 years of implementation. However, Medicaid expansion was associated with fewer days spent in poor mental health (−1.1 d/mo, 95% confidence interval: −2.1 to −0.5). There were significant changes in specific health behaviors, preventive service use, and diagnosis patterns during the same time period which may mediate the relationship between the ACA rollout and self-reported health.

Conclusion:

In nationally-representative survey data, we observed 1 improvements in mental but not physical self-reported health among 3 low-income Americans after 4 years of full ACA implementation.

Keywords: health reform, insurance, Medicaid, access to care, self-reported health, health behaviors, health services research

The Affordable Care Act (ACA) increased health care access and utilization following its passage in 2010. These effects were concentrated among low-income households, who disproportionately benefited from the ACA’s flexibility for states to expand Medicaid income eligibility and the creation of federal and state insurance exchanges, with premiums capped on a sliding scale according to income. 1–4 Evidence on the ACA’s effects on health status has been mixed with reports showing large, 3,5 small, 6 or no improvements. 4,7 However, previous efforts have frequently relied on surveys with very low response rates, analyzed small subsets of states, did not account for pre-existing trends in outcomes, or had limited outcome periods which may have been insufficient for health effects to materialize.

Our objectives were to quantify changes in health care access and self-reported health status for low-income adults during the first 4 years of full ACA implementation, to assess the moderating role of state Medicaid expansion status, and to investigate changes in health behaviors and utilization patterns that may mediate these impacts. Our study addresses these questions using a large, high-quality nationally-representative survey, and contributes to the ongoing policy debate regarding the effects of the ACA, and health insurance in general, on the health status of low-income Americans.

METHODS

Data

Data were extracted for all nonelderly adults (18–64 y) surveyed in the 2011–2017 Behavioral Risk Factor Surveillance System (BRFSS). The surveys are designed to be representative of the United States noninstitutional population, and ~400,000 adults are surveyed each year. Data collection methodology, demographics, and limitations of this sample have been described elsewhere. 8,9

Although the ACA was enacted in 2010, many of the Act’s arguably most important provisions did not go into effect until 2014 such as the individual insurance mandate, Medicaid expansion, and creation of individual health insurance exchanges. 10,11 Our study period spans the 3 years preceding the full ACA rollout (2011–2013) and 4 years after (2014–2017). We excluded respondents aged 65 or older as well as respondents from US territories, states which had pre-existing Medicaid eligibility at or above 100% of the federal poverty level (FPL), and states which implemented the Medicaid expansion after 2014. 1,12 The latter restriction was included because the effects of health insurance are expected to be cumulative over time 13,14 ; inclusion of late expansion states may bias our results downwards. We included all states in sensitivity analyses. We used respondent household size and income to calculate an imputed percentage of the FPL following a procedure employed by Sommers et al. 2 Our analyses were restricted to respondents residing in households earning

Outcome Measures

We assessed trends in several measures of self-reported health care access, health status, health behaviors, utilization, and diagnoses. Our access outcomes included whether an individual had health insurance coverage, a personal doctor, or avoided care due to cost in the previous 12 months. We assessed 4 outcomes related to health status: general health (5-point Likert-scale; 1 = excellent, 5 = poor), and number of days in the previous month when physical health was poor, mental health was poor, or poor health limited usual activities. General health status was dichotomized into 1 = fair/poor and 0 = excellent/very good/good.

We also examined changes in behavioral risk factors, preventive service utilization, and diagnoses which could have implications for self-reported health. Behavioral risk factors included obesity (body mass index > 30), alcoholic drinks consumed per week, participation in exercise during the previous month, and daily smoking. Our preventive service measures included receipt of any of the following within the previous year: flu shot, regular checkup, mammogram, or pap smear. Our analyses were restricted to female respondents for the mammogram and pap smear outcomes. Finally, we quantified changes in whether respondents had ever been given a diagnosis of diabetes, high blood pressure, depressive disorders, or arthritis. The exact texts of all outcome measures are listed in Appendix Table A1 (Supplemental Digital Content 1, http://links.lww.com/MLR/B999). Hot-deck imputation was used to replace missing answers to specific survey questions and reduce potential nonresponse bias. 15

Analytic Approach

We first computed annual means for each outcome, stratifying our analysis by whether a state expanded Medicaid. Next, we estimated interrupted time series regression models to assess changes over time with the ACA implementation. We modeled calendar time using state-specific linear trends and flexibly modeled deviations from this trend in the post-ACA period with indicators for each year. Models additionally included demographic covariates for race/ethnicity, sex, home ownership, education, age, veteran status, household size, and presence of children in the household. We then estimated difference-in-differences (DID) models to assess changes in outcomes attributable to Medicaid expansion. All models were estimated as linear probability models using BRFSS sampling weights. 16 For more details on our methodology, please see the Appendix (Supplemental Digital Content 1, http://links.lww.com/MLR/B999).

RESULTS

Our final sample included 505,824 adult respondents in low-income households under 138% FPL. For characteristics of the study sample, please see Appendix Table A2 (Supplemental Digital Content 1, http://links.lww.com/MLR/B999). In presenting regression results, we focus on the adjusted comparisons between 2017 and 2013, as shorter-term outcomes have been reported previously. Unadjusted trends for all outcomes by state Medicaid expansion status are contained in Appendix Table A3 (Supplemental Digital Content 1, http://links.lww.com/MLR/B999).

Changes in Health Care Access

Households in expansion states reported greater pre-existing health care access in 2013 and experienced larger gains under the ACA compared with nonexpansion states ( Fig. 1 ). In adjusted DID models, Medicaid expansion was associated with a 10.3 percentage-point increase in coverage for respondents [95% confidence interval (CI): 4.5–16.2] from 2013 to 2017, a 4.5 percentage-point increase in having a personal doctor (95% CI: 0.9–8.1) and a 7.6 percentage point (95% CI: −13.0 to −2.2) reduction in avoiding care due to cost ( Table 1 ).

Changes in health care access for low-income households 2011–2017, by state Medicaid expansion status. The figure displays the percentage of noninstitutionalized US adults age 18–64 with household incomes of Source: Authors’ analysis of data from the Behavioral Risk Factor Surveillance System (BRFSS).

TABLE 1.

Adjusted Changes From 2013 to 2017 in Health Care Access for Low-income Adults Under the ACA, by Medicaid Expansion Status

‡ Difference between expansion and non-expansion states in changes over time, adjusted for covariates.

ACA indicates Affordable Care Act; CI, confidence interval.

Source: Authors’ analysis of data from the Behavioral Risk Factor Surveillance System (BRFSS).

Changes in Self-reported Health

Average self-reported health showed little variation across time of Medicaid expansion status ( Fig. 2 ). After adjustment for covariates and state-specific time trends, adults in nonexpansion states experienced 0.4 fewer days in poor physical health (95% CI: −0.4 to −0.0) and 1.2 additional days in poor mental health (95% CI: 0.8–1.5) in 2017 compared with 2013. In adjusted DID models, Medicaid expansion was associated with 1.1 fewer days spent in poor mental health (95% CI: −2.1 to −0.5) and 1.4 fewer days spent in poor physical health (95% CI: −5.7 to 2.9).

Trends in self-reported health status for low-income households 2011−2017, by state Medicaid expansion status. The exhibit displays the average responses for noninstitutionalized US adults age 18–64 with household incomes Source: Authors’ analysis of data from the Behavioral Risk Factor Surveillance System (BRFSS).

Changes in Health Behaviors, Preventive Care, and Diagnoses

Tables 2 and ​ and3 3 present adjusted changes in health behaviors, preventive care, and diagnoses. In adjusted DID models, changes in health behaviors were not significantly associated with residence in a Medicaid expansion state. Although the nationwide rates of diabetes, depression, and arthritis diagnoses did not significantly change during the study period, Medicaid expansion was associated with a 3.8 percentage point increase in the likelihood of respondents having received an arthritis diagnosis (95% CI: 0.7–6.8). Expansion was also associated with a 6.9 percentage point increase in the likelihood that female respondents received a pap test in the previous 12 months (95% CI 1.2-12.4).

TABLE 2.

Adjusted Changes From 2013 to 2017 in Self-reported Health for Low-income Adults Under the ACA, by Medicaid Expansion Status

‡ Difference between expansion and nonexpansion states in changes over time, adjusted for covariates.

§ Respondents were asked to rate their health on a 5-point Likert scale. Responses were dichotomized into “fair/poor” and “good/very good/excellent.”

‖ Respondents were asked to count how many days within the past 30 was their physical or mental health not good, or when poor health prevented their usual activities.

ACA indicates Affordable Care Act; CI, confidence interval.

Source: Authors’ analysis of data from the Behavioral Risk Factor Surveillance System (BRFSS).

TABLE 3.

Adjusted Changes From 2013 to 2017 in Health Behaviors, Preventive Care, and Diagnoses for Poor Households Under the ACA

The exhibit displays regression-adjusted changes in selected health behaviors, preventive service utilization, and diagnoses associated with the ACA rollout. All columns show regression estimates adjusted for covariates described in the text.

‡ Difference between expansion and nonexpansion states in changes over time, adjusted for covariates.

ACA indicates Affordable Care Act; BRFSS, Behavioral Risk Factor Surveillance System; CI, confidence interval.

Source: Authors’ analysis of data from the BRFSS.

In sensitivity analyses, we repeated our analyses including all states regardless of timing of Medicaid expansion or pre-existing Medicaid eligibility. The results were generally consistent, although expansion was no longer associated with pap tests or arthritis diagnoses (Appendix Tables A4–A6, Supplemental Digital Content 1, http://links.lww.com/MLR/B999). We also replaced the state-specific linear trends in our models with dummy variables for each state-year combination from 2011to 2013. The results were again generally consistent, although the DID effects of Medicaid expansion on insurance coverage and days spent in poor mental health were reduced (Appendix Tables A7–A9, Supplemental Digital Content 1, http://links.lww.com/MLR/B999).

DISCUSSION

We assessed changes in self-reported health care access, preventive service use, health behaviors, and health status through 4 years of full ACA implementation. Despite substantial improvements in health care access 2013 to 2017, we found no evidence of improved self-reported physical or general health among adults in low-income households, and no impact of Medicaid expansion on these outcomes. However, the Medicaid expansion was associated with 1.1 fewer days per month spent in poor mental health. The failure to find significant improvements in physical or general health differs from Courtemanche and colleagues previous assessment of the ACA using BRFSS data through 2016, which documented a slight increase in the proportion of respondents reporting general/excellent health. Notably, our models control for state-level time trends while Courtemanche et al 17 controlled for state-level insurance rates during 2013; this methodological difference may explain differences in our estimates of effect. Our results also comport with Baicker et al’s 18 study on the effects of the Oregon Medicaid experiment on clinical outcomes, which found improvements in mental but not physical health.

Why have access improvements under the ACA not led to greater improvements in self-reported health? First, health improvements should be anticipated following insurance expansions only if several intervening outcomes are achieved. 14,19 For instance, consumers must first enroll in a plan which covers necessary services they require from providers in their area. Although insurance coverage increased substantially, a large portion of households eligible for Medicaid or Marketplace subsidies remain uninsured. We observed smaller increases in the proportion of respondents with a personal doctor. Then, consumers must have the health literacy to make informed care decisions and access services as needed. Medicaid expansion was associated with increased likelihood of receiving a pap test, a regular check-up, or having a sigmoidoscopy, although estimates for these last 2 outcomes were insignificant. Our findings here are consistent with prior work showing increased utilization of certain services. 3,4,6 If these previous hurdles are met, the care delivered must be efficacious and not mitigated by adverse behavioral responses or other harmful care. Under this framework, insurance is a necessary but insufficient condition to achieve better health.

Second, it may take longer for many of the health benefits of preventive care to be realized. For example, there are few observable benefits of reduced blood pressure in the short run, but the major health benefits from reduced damage to your arteries, heart, brain, and other organs are experienced years or even decades down the line. 20

Third, individuals who gain health insurance under the ACA may have less incentive to guard against health risks due to the reduced financial burden of illness (known as ex ante moral hazard). Previous research has generally found little evidence that insured individuals take on more risky behaviors, 6,21 and our results suggest changes in health behaviors were not associated with Medicaid expansion.

Fourth, increases in access and utilization may result in new diagnoses of chronic conditions which lead to greater awareness of poor health; this could result in short-run reductions in perceived health, but long-run improvements in actual (and perceived) health. In adjusted DID models, Medicaid expansion was associated with an increased likelihood of a diagnosis for arthritis but no other conditions.

Limitations

This study has many of the standard limitations of observational study designs. First, the BRFSS has high response rates for telephone surveys (40%–50%) and includes sampling weights to minimize nonresponse bias. However, nonresponse bias may still occur if survey respondents differ from nonrespondents due to unobserved time-varying factors which are associated with both our outcomes and state Medicaid expansion status. Second, we are unable to follow individual respondents over time and observe changes in outcomes after gaining insurance. Third, we are unable to observe insurance source (eg, Medicaid, the ACA Marketplaces, or an employer). Fourth, our outcome variables are self-reported although these questions have been found to have both high reliability and validity. 22 Lastly, our DID estimates only have a causal interpretation if expansion states would have experienced identical changes in outcomes as nonexpansion states if they had not expanded Medicaid, after controlling for observed covariates and pre-existing time trends. Our interrupted time series results should only be interpreted as associations.

Summary

Prior evidence suggests the ACA provided substantial financial protection to US adults living in low-income households, particularly where Medicaid expanded. 1,23 However, the link from health insurance to health status has been more difficult to show and the literature is inconclusive. In large, nationally representative survey data, we find that ACA’s Medicaid expansion was associated with improvements in self-reported mental health for low income adults. In contrast, we did not find improvements in self-reported physical or general health through 4 years of full ACA implementation, despite large gains in insurance coverage.