The Impact Of State Policies On ACA Applications And Enrollment Among Low-Income Adults In Arkansas, Kentucky, And Texas

: States are taking variable approaches to the Affordable Care Act (ACA) Medicaid expansion, Marketplace design, enrollment outreach, and navigator programs. We surveyed nearly 3000 low-income adults in late 2014 to compare experiences in three states with markedly different ACA policies: Kentucky, which expanded Medicaid, created a successful state Marketplace, and supported outreach efforts; Arkansas, which enacted the private option and a federal-state partnership Marketplace, but with legislative limitations on outreach; and Texas, which did not expand Medicaid and passed onerous restrictions on navigators. We found that application rates, successful enrollment, and positive experiences with the ACA were highest in Kentucky, followed by Arkansas, with Texas performing worst on most outcomes. Awareness of the ACA was low – less than half of adults had heard some or a lot about the law. Navigator assistance was the strongest predictor of successful enrollment, while Latinos were much less likely to complete the process. Twice as many respondents felt the ACA had helped them as hurt them, though advertising was strongly associated with perceptions of the law’s impact. While the ACA is a national law, state policy choices have had major impacts on enrollment experiences among low-income adults and their overall perceptions of the ACA.


INTRODUCTION
2][3] Today, Medicaid covers over 68 million individuals, most of whom live under or near the poverty level. 4In 2014, the Affordable Care Act (ACA) expanded Medicaid eligibility to 138% of the federal poverty level (FPL) in participating states, offering the possibility of health insurance to millions of low-income uninsured adults who were previously excluded from the program.However, the Supreme Court ruling in 2012 made this expansion optional.More than half of states have moved forward with expansion, some have decided not to expand, and others are still debating the issue. 5Even among states expanding coverage, there are distinct differences in approach.Several states have opted to expand Medicaid by using federal dollars to purchase private insurance for low-income adults, rather than traditional Medicaid coverage.This so-called "private option," first adopted by Arkansas, has become a potential model for several other states. 6yond the decision of whether to expand Medicaid, the ACA provides states with substantial discretion in other areas regarding insurance coverage.Sixteen states and the District of Columbia have launched state-based Marketplaces, while the remaining states chose to partner with the federal government to run their Marketplaces or defaulted to the federal Marketplace. 7ates also have flexibility to determine how they want to regulate or employ insurance navigators.Navigators are individuals tasked with providing consumers with unbiased information about health insurance plans to best meet their needs and assisting individuals with the application process. 8Their efforts can take place through in-person meetings, online communications, or phone conversations, and they are trained, certified, and funded by the Marketplaces in which they operate.In addition, navigators are responsible for increasing awareness about their state's Marketplace by conducting outreach campaigns and information sessions to highlight offerings.However, a number of states including Texas have imposed restrictions that substantially hinder the work of navigators by requiring additional training beyond the federal requirements, as well as more extensive background checks including fingerprinting -which critics have called burdensome and unnecessary. 9 is unclear the extent to which these different approaches to Medicaid expansion, Marketplace design, and navigator programs will impact overall insurance enrollment and experiences with the ACA among low-income Americans.Medicaid historically has struggled to achieve high enrollment rates among eligible individuals, and the ACA sought in part to address this challenge with enrollment assistance and a streamlined application process. 10,11[14][15][16][17] Arkansas used the private option and a federal-state partnership Marketplace, along with an inperson assistance program, 15 but passed legislative prohibitions against state-sponsored outreach. 18Texas neither expanded Medicaid nor established a state Marketplace, did not create an in-person assistance program, 15 and enacted restrictive regulations on navigators. 9Our objective was to assess the relative experiences with the ACA among low-income adults in these states, and to explore predictors of application for Medicaid or Marketplace coverage, completion of enrollment, and perceived harms or benefits of the ACA.

Survey Design
We conducted a random-digit telephone survey of low-income adults in Arkansas, Kentucky, and Texas.Adults eligible for the study had to be U.S. citizens between the ages of 19 and 64, with a family income less than 138% of FPL.The sampling frame included both cell phones and landlines, and interviews were available in both English and Spanish.We purposely oversampled black respondents in Kentucky (where their prevalence is lowest) to ensure that we had at least 100 observations for this group in each of the three states.
The survey explored several issues related to health care and the ACA, as well as basic demographic information.In the survey, the ACA was described as "the national health reform law, sometimes referred to as Obamacare or the Affordable Care Act."We included both of these terms since prior research has shown that public opinion can be biased by survey questions that only use one term or the other. 19Respondents were asked a series of questions about their experiences with and perceptions of the ACA, as described in more detail below.When possible, questions were drawn from previous established surveys, [20][21][22] and all questions were pilot-tested with a subsample of respondents before the survey instrument was finalized.(See Appendix for survey questions and approach to missing values).
Surveys were completed in November and December 2014, nearly a full year after the expansion of Medicaid eligibility (in expanding states) and the beginning of Marketplace subsidies available under the ACA.The overall response rate was 23%, using the American Association for Public Opinion Research's RR3 definition.Data were weighted to population estimates in each state for the low-income 19-64 year-old citizen population, based on the 2012 American Community Survey and National Health Interview Survey, for the following characteristics: age, gender, education, race/ethnicity, marital status, geographic region within each state, population density, and cell phone versus land-line status.Further details on the survey methodology have been published previously. 23

Analysis
We conducted two sets of analyses.First, we compared a range of measures related to low-income adults' experiences with the ACA across the three study states.These outcomes were awareness of the ACA coverage expansions, application for Medicaid or Marketplace coverage in 2014, receipt of application assistance from a navigator or social worker, successful completion of the application process, overall quality of the application experience, exposure to advertisements in favor or opposed to the ACA, and whether the ACA had "helped you, hurt you, or had no direct impact."Unadjusted comparisons of these outcomes across the states were conducted using survey-weighted chi-square tests.
Next, we used multivariate logistic regression to examine demographic and policy predictors for three of these key outcomes: whether a person applied for Medicaid/Marketplace coverage; whether -among those who applied -a person successfully completed the application process to obtain coverage; and whether the ACA had helped or hurt them.Potential predictor variables included in these regressions were the state of residence, awareness of the ACA expansions, exposure to positive or negative advertisements, age, sex, marital status, race/ethnicity, education, income, rural vs. urban status, cell phone vs. landline interview, political affiliation, and two measures of health -"fair or poor" self-reported health and the presence of one or more chronic conditions assessed in the survey (see Appendix).
For the analysis of whether a person who had applied for coverage successfully completed the process, we included two additional questions asked only of this subsample (n=1070).First, we asked whether they had received assistance from a navigator or social worker (a term we used to encompass roles such as non-navigator assisters and certified application counselors, 8 terms likely unfamiliar to most respondents).Second, we asked at which location(s) they had applied for coverage -Marketplace website, Medicaid/public assistance office, doctor's office/hospital, or somewhere else, and respondents were allowed to select one or more categories.
Results were reported as adjusted odds ratios, which we then converted into predicted probabilities for ease of interpretation.All analyses were conducted using Stata 12.1.

Limitations
One significant limitation of this study is a lower response rate than those obtained by inperson government interview surveys.However, our response rate of 23% is significantly higher than that of many political polls and health care surveys, including the Gallup tracking poll, 20,24 and recent research suggests that the bias from low-response rates in polls can be largely mitigated through the use of appropriate demographic weighting, as we have done here. 25,26other limitation is that our analysis uses self-reported outcomes, subject to bias depending on the respondents' perspectives.This may have led those who already favored the ACA to describe it in more positive terms, and vice versa among those opposed to the law.In part, our multivariate analysis was designed to tease apart some of these factors by considering both behavior (whether someone applied) and attitudes (whether they thought the ACA helped or hurt them), and by adjusting for political ideology.However, unmeasured influences may have produced additional biases in survey responses.Overall, our data can only show associations and we cannot be sure that our findings are causally related to the different policies in these states.
Finally, our study sample was drawn from just three states, which may limit the generalizability of our findings.However, the three states we studied here in many ways represent three distinct points along the ACA implementation spectrum, providing a valuable set of case studies for evaluating the experiences of low-income adults under the health reform law.

RESULTS:
Our sample consisted of nearly one thousand low-income adults in each of the three study states (Exhibit 1), for a total of 2801 adults.The sample across all three states had similar distributions of age, gender, and marital status.Texas had a more urban population than the other states, and it also had a much higher share of Latino respondents than did Arkansas or Kentucky.
Exhibit 2 summarizes the ACA-enrollment related outcomes for the three states.
Awareness about ACA coverage options was highest in Kentucky, but even there, only 51% of respondents had heard "some" or "a lot" about the ACA coverage options.
Application rates for Medicaid or Marketplace coverage were higher in Arkansas (43%) and Kentucky (44%) than Texas (33%).Among those who applied, application assistance from a navigator or social worker was most common in Kentucky (46%) and least common in Texas (32%).More than half of applicants (50-65%) said their overall application experience was "good" or "excellent."In all three states, most applicants (87-96%) reported they had been able to successfully complete the application process.
Of those who said they completed the application during the prior 12 months, 53% reported having Medicaid coverage, 19% reported having Marketplace coverage, 18% said both Medicaid and Marketplace coverage, and 7% were uninsured at the time of the survey.The majority (68%) of those who failed to complete the application process were uninsured at the time of the survey, while some had acquired other insurance in the interim (most commonly employer coverage, for 14%).
On average for all three states, roughly 20% of respondents had heard more advertising against the law than in support of it, while 10% recalled more positive ads, and approximately two-thirds said that they had heard equal numbers of both.Negative ads were most frequently reported in Arkansas.
Exhibit 3 identifies several characteristics that were associated with applying for Medicaid or Marketplace coverage: living in an expansion state (Kentucky or Arkansas), having greater awareness of the ACA, female gender, and having an income below 50% of the FPL.
Results were similar if we excluded individuals who said they had heard or read nothing about the 2014 coverage expansions.
Exhibit 4 shows factors related to successfully completing the application for coverage, among those who applied.Navigator assistance, applying through the Medicaid office, higher levels of education, and female gender were all associated with higher application completion rates.Conversely, Latino applicants had much lower completion rates.Individuals in Kentucky had significantly higher completion rates than those in Texas and Arkansas, even after multivariate adjustment.In a sensitivity analysis limited to adults with incomes above 100% of FPL (who are eligible for subsidized coverage in all 3 states), Kentucky residents still had significantly higher completion rates than individuals in the other two states.
In terms of how respondents felt that the ACA had impacted their lives, a large plurality in all three states (48-66%) said that the law had not had any impact on them personally.Of those directly impacted, roughly twice as many said they had been helped by the law than had been hurt by it.Exhibit 5 shows that older adults and people who reported hearing mostly negative advertisements were more likely to say that the ACA had hurt them, while individuals in the two expansion states, Democrats, blacks, and those reporting hearing mostly positive ads were more likely to say that the law had helped them.
Among those who did not apply for insurance despite having no other coverage, the most common reasons for not applying were that they thought coverage would cost too much (21%) or they did not know enough about the coverage options (19%).Only 1% said they did not think their state's Medicaid / Marketplace coverage would be good insurance.Explanations for not applying did not differ significantly by state (Appendix Table 1).

DISCUSSION
The ACA was enacted as a national health reform law, but states have had substantial discretion in its implementation.Among low-income adults in three Southern states, ACArelated experiences varied widely and demonstrated strong associations with key state policy decisions.We found that application rates, successful enrollment, and positive experiences with the ACA's 2014 expansions were highest in Kentucky, followed by Arkansas, with Texas performing most poorly on these outcomes.This corresponds to the general pattern of state-level engagement and support for the ACA coverage expansions in these three states: Kentucky established a well-functioning state Marketplace and conducted aggressive outreach, [12][13][14]16 while Arkansas adopted the innovative private option and a partnership-style Marketplace, but also enacted a legislative prohibition against state-funded outreach to support enrollment efforts. 18Both expansion states also enabled their Marketplace websites to link consumers directly to information on navigator programs, 15 and recent survey data suggests these two states experienced the largest declines in their uninsured rates in 2014 of any states in the country.27 At the other extreme, Texas did not expand Medicaid, did not participate at all in Marketplace planning in the state, and enacted restrictive regulations for navigators.9,15 The fruits of these different policy efforts are evident in our findings.
Beyond these critical state-level differences, our findings also shed light on some of the key individual-level factors in who applied for ACA coverage, who completed the application process, and how low-income adults perceived the law's overall impact.

Who Applied?
Overall, fewer than half of individuals in our sample applied for Medicaid/Marketplace coverage.Lack of information was a key barrier: our multivariate results demonstrated a strong relationship between level of awareness and whether a person had applied for coverage.
Awareness of the ACA expansions remained somewhat low even at the end of 2014, with just 40-50% saying they had heard "a lot" or "some" about the new ACA coverage options, adding to previous evidence that information barriers about the law remain a challenge especially for lowincome adults. 23other barrier to applying was the perception that coverage would cost too much, consistent with a recent report from the Kaiser Family Foundation. 28However, for individuals with incomes under 138% of FPL in the expansion states, coverage is available without a premium.This suggests that many individuals did not understand the coverage they would likely be eligible for under the ACA, again pointing to challenges related to lack of knowledge about the law.

Completing the Application Process
Among those adults who tried to apply for coverage, the results were generally encouraging, with completion rates over 85% in all three states -resulting in more than half of applicants obtaining Medicaid and roughly one-third reporting Marketplace coverage (though some reported having both Medicaid and Marketplace insurance, suggesting some degree of confusion about coverage types).Kentucky had the highest completion rates, followed by Arkansas, with Texas lowest.While it is to be expected that expansion states should have more success at getting applicants into coverage than the non-expanding state, we found that application success rates for Kentuckians remained significantly higher even among applicants with incomes between 100-138% of FPL, who were eligible for subsidized coverage in all three states including Texas.
The strongest overall predictor of completing the application process was receiving assistance from a navigator or social worker, which increased the probability of successful enrollment by almost ten percentage points.To our knowledge, our study is the first to quantify the potential impact of navigator assistance on application completion rates.Navigator assistance was most common in Kentucky and least common in Texas, as expected, given the states' different approaches to the navigator program.
We also noted that applicants who applied for coverage at the state Medicaid office were more likely to complete the process than those who did so via the Marketplace or a medical facility.Further research exploring how the experience of applying for coverage differs across those locations would be valuable for understanding why some people successfully completed the process and others did not.Difficulties understanding or navigating the Marketplace webpage may be one contributing factor, particularly among lower-income and less educated populations.Another consideration is that some Marketplaces (including healthcare.gov) had trouble transferring information from website applications to Medicaid offices, which may have delayed or even prevented some applicants from obtaining coverage. 29key demographic predictor of failure to complete the application process was Latino ethnicity.Federal enrollment statistics have suggested that Latino participation has been slow, 30 and our study suggests that the application process (as opposed to never applying in the first place) may be a key obstacle.One possibility is that this could be due to language barriers.However, in a sensitivity analysis, we found that having completed our survey in Spanish was not associated with significantly lower completion rates, and Latino ethnicity remained a strong predictor even after adjustment for language.
Another possibility is that concerns about immigration enforcement kept some individuals from finishing their applications.Even though all members of our sample were selfreported U.S. citizens, prior research has documented a "chilling effect" for Medicaid enrollment among households with mixed citizenship status. 31Furthermore, there have reportedly been challenges with verification of immigration status, which may be causing difficulties for legal immigrants attempting to enroll in coverage. 32The fact that most of our Latino sample resided in Texas makes it difficult to generalize to other states, but racial and ethnic differences in enrollment need to be monitored closely by policymakers to prevent widening health care disparities.

Advertisements and the Perceived Impact of the ACA
ACA-related advertisements appeared to have little effect on whether individuals applied or completed the application process.However, ads were an important predictor of individuals' perceptions of whether the law had helped or hurt them.We found that negative advertisements were most common in Arkansas, consistent with a recent media analysis that found high numbers of anti-ACA ads during Arkansas' contested 2014 Senate election. 33The fact that advertisements were significant predictors of attitudes towards the ACA may reflect that ads were effective in influencing perceptions of the law, or alternatively, that people already inclined to support or oppose the law were more likely to recall ads consistent with those views.
In terms of overall assessment of the law's impact, race, income, and political affiliation were important predictors.Blacks, lower-income individuals, and Democrats were significantly more likely to say that the ACA had helped them, while Republicans and older adults were more likely to say that the law had hurt them.However, even after adjusting for these demographic features, as well as policy factors such as ACA awareness and advertising, one of the strongest predictors of positive views of health reform was the state of residence.As one would expect, low-income adults in the two expansion states -Kentucky and Arkansas -were much more likely than those in Texas to say the law had helped them.Nonetheless, pluralities in all three states felt the law had not directly impacted them in 2014, suggesting the ACA has not yet reached many of those who might benefit from it.

Conclusion
The first year of the Affordable Care Act's major coverage expansions is now complete.
Although the legislation was national in scope, our study demonstrates that state policy decisions are having a critical impact not only on eligibility but also on who chooses to apply for coverage and whether they are able to complete the process.Navigator programs and other forms of direct application assistance appear to be a valuable approach to improving the effectiveness of the coverage expansion, and states that enacted restrictions on navigators are likely having significant detrimental effects on their low-income residents' ability to obtain coverage.Meanwhile, our findings suggest that effective implementation -as in Kentucky and Arkansasand positive outreach efforts -most evident in Kentucky -can have major effects on the experiences and perceptions of low-income adults under the ACA.

Experiences of Low-Income Adults in the 2014 Open Enrollment Period
NOTES: P-values represent chi-square test for significant differences in each demographic variable across the three states.SOURCE: Authors' analysis of three-state survey of low-income adults, conducted in November-December 2014.Exhibit 2:

Predictors of Completing Application Process for Medicaid/Marketplace Coverage in 2014 (n=1070)
Odds Ratios are from multivariate logistic regression controlling for all the variables listed in the table.Predicted Probabilities were obtained using the "margins" command in Stata.† Options for this variable were not mutually exclusive; respondents could indicate one or more.SOURCE: Authors' analysis of three-state survey of low-income adults, conducted in November-December 2014.

Predictors of Perceived Harm or Benefit from the ACA (n=2,771) Variable "ACA Helped Me" "ACA Hurt Me" Odds Ratio Predicted Probability Odds Ratio Predicted Probability
Odds Ratios are from multivariate logistic regression controlling for all the variables listed in the table.Predicted Probabilities were obtained using the "margins" command in Stata.**p<0.01,*p<0.05SOURCE: Authors' analysis of three-state survey of low-income adults, conducted in November-December 2014.In the past year, would you say you have seen (more advertisements and news coverage in support of the national health reform law), (more advertisements and news coverage opposed to the national health reform law), or about equal numbers of both?Since October 1, 2013, have you tried to get health insurance for yourself through either [State Marketplace Name] or [State Medicaid Plan Name]?This could be by mail, in person, by phone, or on the internet.1 Yes, have tried to get insurance 2 No, have not tried to get insurance What is the main reason you did not try to get insurance through [State Marketplace Name] or [State Medicaid Plan Name] since October 2013?Is it because…?(READ LIST.ENTER ONE ONLY) 1 You already have health insurance 2 You didn't think you would be eligible for health insurance 3 You thought insurance would cost too much 4 You don't think you need health insurance 5 It is too hard to sign up or you didn't know how to apply 6 You don't think [State Marketplace Name] or [State Medicaid Plan Name] is good insurance 7 You don't want government help to get health insurance 8 You didn't know enough about these options IF 4 = "YES", ASK QUESTIONS E THROUGH I: 6) You mentioned that you signed up or tried to get insurance through either [State Marketplace Name] or [State Medicaid Plan Name] in the past year.Thinking back to when you applied, did you receive any assistance applying for coverage from a social worker or insurance navigator specifically trained to help sign people up for coverage under the national health reform law? 3)