Work Requirements and Medicaid Disenrollment in Arkansas, Kentucky, Louisiana, and Texas, 2018

Abstract

Objectives. To identify risk factors for Medicaid disenrollment after the implementation of Arkansas’s work requirements.

Methods. Using a 2018 telephone survey of 1208 low-income adults aged 30 to 49 years in Arkansas (expansion state with work requirements implemented in June 2018), Kentucky (expansion state with proposed work requirements blocked by courts), Louisiana (expansion state without work requirements), and Texas (nonexpansion state), we assessed Medicaid disenrollment rates among the age group targeted by Arkansas’s policy.

Results. The Medicaid disenrollment rate was highest in Texas (12.8%), followed by Arkansas (10.5%), Kentucky (5.8%), and Louisiana (2.8%). Over half of those who disenrolled in Texas and Arkansas became uninsured, compared with less than a quarter in Kentucky and Louisiana. In multivariate models, Arkansas had significantly higher disenrollment compared with the 3 comparison states; men and non-Hispanic Whites experienced higher disenrollment than women and racial minorities. In Arkansas, having a chronic condition was associated with higher disenrollment.

Conclusions. As states debate work requirements and Medicaid reforms, our findings provide insights for policymakers about which populations may be most vulnerable to losing Medicaid coverage.

Medicaid churn (i.e., disruptions in coverage) is common, and low-income nonelderly adults who disenroll from Medicaid often become uninsured and experience reduced access to care.¹ Prior work has shown that Medicaid expansion is associated with reduced churning.^2–5 Better understanding the characteristics of those who disenroll is especially relevant now that 20 states have proposed work requirements, which may exacerbate disenrollment. In Arkansas, the first state that implemented work requirements, over 18 000 individuals lost coverage before a federal judge halted the program, and recent research found that the work requirement was associated with an increased uninsured rate.⁶ Starting in June 2018, Arkansans were disenrolled if they did not meet monthly online reporting requirements for 3 months, demonstrating at least 80 hours per month of work or another qualifying community engagement activity (e.g., job training or community service) or an exemption (e.g., disability or pregnancy).

As policymakers consider alternative approaches in Medicaid, it is important to understand risk factors for disenrollment. Although some individuals disenroll after losing eligibility, prior literature shows that increased administrative burdens also lead to higher churn among those still eligible.⁷ Demographic and health characteristics may be risk factors, as individuals in rural areas may have fewer job opportunities or those with less education may face challenges reporting.⁸ The potential impact of these policies on racial disparities is also an important public health issue. We examined Medicaid disenrollment rates in Arkansas and 3 comparison states after implementation of work requirements and assessed risk factors for disenrollment.

METHODS

We conducted a random digit-dialing telephone survey among US citizens aged 30 to 49 years in Arkansas (n = 610), Kentucky (n = 212), Louisiana (n = 144), and Texas (n = 242) with self-reported family income below 138% of the Department of Health and Human Services’ federal poverty level. We focused on this age group because Arkansas implemented work requirements for those aged 30 to 49 years, and we oversampled for Arkansas because it was the only state with work requirements in effect at the time of the survey.

In Kentucky, work requirements were approved in early 2018 but were blocked in June 2018, before our survey was conducted, and were never implemented. Louisiana is an expansion state without work requirements. As a nonexpansion state, Texas is another useful comparator, because in some states, expansion with work requirements has been proposed as a compromise between traditional expansion and nonexpansion.

Conducted in late 2018 on landlines and cellphones in English and Spanish, the survey had a response rate of 14%, similar to those of other widely cited telephone surveys.⁹ We weighted survey estimates to account for potential nonresponse bias, and prior survey waves have been validated against federal government survey data.¹⁰

Outcomes

Our primary outcomes were Medicaid coverage and uninsured rates at the time of the survey, and the share who reported losing Medicaid in the previous 12 months. We then calculated state-specific disenrollment rates by dividing the share that lost Medicaid by the share that had ever been on Medicaid in the past year.

Statistical Analysis

Multivariate linear regression models identified predictors of disenrollment among those who had been on Medicaid (n = 766). We included state, race/ethnicity, gender, income, education, urban versus rural residence, presence of at least 1 of 9 common chronic conditions (hypertension, coronary artery disease, asthma, chronic obstructive pulmonary disease, diabetes, depression, anxiety, substance use disorder, and cancer), and self-reported fair or poor health. We estimated robust standard errors.

RESULTS

In our sample of 1208 low-income citizens aged 30 to 49 years, current Medicaid coverage was reported by 59.6% of Arkansans, 63.4% of Kentuckians, 63.4% of Louisianans, and 43.7% of Texans. By state, 7.0% of Arkansans, 3.9% of Kentuckians, 1.8% of Louisianans, and 6.4% of Texans reported dropping out of Medicaid in the past 12 months. For overall uninsurance rates, 14.6% of Arkansans, 6.9% of Kentuckians, 9.7% of Louisianans, and 29.3% of Texans reported being uninsured.

The Medicaid disenrollment rate was highest in Texas (12.8%), followed by Arkansas (10.5%), Kentucky (5.8%), and Louisiana (2.8%; Table 1). Over half of those who disenrolled in Texas and Arkansas reported being uninsured at the time of the survey, compared with 20.4% in Kentucky. In Louisiana, there was a single disenrolled individual, who obtained other coverage.

TABLE 1—

Characteristics of Low-Income Adults Aged 30 to 49 Years Who Were Ever on Medicaid in the Past Year: 4 US States, November–December 2018

	Arkansas (n = 408), %	Kentucky (n = 149), %	Louisiana (n = 91), %	Texas (n = 118), %
Disenrolled from Medicaid in the past year	10.5	5.8	2.8	12.8
Currently on Medicaid	89.5	94.2	97.2	87.2
Female	63.5	65.2	71.4	68.5
Race/ethnicity
Hispanic	3.4	1.2	4.8	43.7
Non-Hispanic White	62.7	87.1	37.2	24.5
Non-Hispanic Black	28.4	8.8	54.9	29.6
Other non-Hispanica	5.6	2.9	3.2	2.2
< high school diploma	16.7	16.0	35.6	29.0
Income < 50% FPL	38.8	27.6	41.2	29.9
Rural	48.5	56.2	29.5	15.1
Any chronic health conditionb	80.0	83.6	77.1	74.3
Fair or poor health status	49.6	43.7	51.1	50.0

Note. FPL = federal poverty level (per Department of Health and Human Services). Respondents who did not report current Medicaid coverage were asked if they had disenrolled from Medicaid in the past 12 months. Using a χ² test comparing respondents in Arkansas with those in the 3 control states, we found that disenrollment rates did not significantly differ (P = .22).

Source. Authors’ analysis of data from a November–December 2018 telephone survey of 766 US citizens aged 30 to 49 years with self-reported family incomes less than 138% of the FPL who reported that they were ever on Medicaid in the previous year (either currently on Medicaid or disenrolled from Medicaid in the past 12 months).

^a“Other non-Hispanic” included Asian, Native Hawaiian, Pacific Islander, American Indian, Alaska Native, or self-reported other race not previously listed.

^bChronic health conditions included hypertension, coronary artery disease, asthma, chronic obstructive pulmonary disease, diabetes, depression, anxiety, cancer, or substance use disorder.

In the multivariate model, living in Arkansas was associated with a significantly higher disenrollment rate, compared with living in Kentucky, Louisiana, or Texas (P < .05; Table A, available as a supplement to the online version of this article at http://www.ajph.org). In all 4 states, men (compared with women) and non-Hispanic Whites (compared with non-Hispanic Blacks) were more likely to disenroll. In Arkansas, those with any chronic condition were more likely to disenroll (compared with those with none). Results were similar when we used multivariate logistic regression (Table B, available as a supplement to the online version of this article at http://www.ajph.org).

DISCUSSION

Our results suggest that state policies may play an important role in shaping Medicaid disenrollment rates. The disenrollment rate was significantly higher in Arkansas, after adjustment for covariates, than in the control states. This is consistent with other published evidence that many in Arkansas lost coverage in association with the new work requirements.^6,11,12 We found higher disenrollment rates among non-Hispanic Whites and men. These findings are similar to Medicaid disenrollment patterns before enactment of the Affordable Care Act,⁸ and suggest that these groups may be less attached to the health care system or public insurance. Qualitative research may help further unpack these differences in churning rates in Arkansas.

Those with any chronic condition were also more likely to disenroll in Arkansas, raising concerns about potential adverse health impacts. Although our sample sizes for the 3 control states were small and not powered to detect differences between them, the pattern of disenrollment rates was suggestive. The highest rate was in Texas, a nonexpansion state, consistent with studies showing that expansion itself reduces churn.^2–5 The lowest rate was in Louisiana, which expanded Medicaid and had no work requirement waiver.

Limitations

Our questions about disenrollment applied to a small sample of respondents. We also only asked individuals whether they disenrolled from Medicaid; because Arkansas expanded through use of the private option (i.e., using Medicaid funds for subsidized Marketplace coverage), we may not have fully captured those who lost Marketplace coverage, which could underestimate the disenrollment rate. We were unable to evaluate how much churn was directly due to work requirements, since prior work showed that many beneficiaries were unaware of the policy, suggesting likely confusion about any coverage losses.⁶ Finally, our survey relied on self-reported data, which is not as accurate as enrollment data for studying churn. However, we were able to identify uninsured individuals.

Implications and Conclusions

Access to care is an important public health challenge, and improving Medicaid retention is a key strategy to increase access. Our study found that Medicaid work requirements in Arkansas—which substantially increased the administrative burden of remaining enrolled—were associated with worsened Medicaid retention, compared with states without work requirements. Of note, those with chronic conditions were more likely to lose Medicaid in Arkansas, which may raise concerns about the distributional impacts and health effects of work requirements. As Medicaid reforms such as work requirements continue to be considered, it is important to understand how they may affect enrollment.