THE EFFECT OF EARLY MEDICAID EXPANSION ON MENTAL HEALTH IN CONNECTICUT AND MINNESOTA

Transcription

1 THE EFFECT OF EARLY MEDICAID EXPANSION ON MENTAL HEALTH IN CONNECTICUT AND MINNESOTA A Thesis submitted to the Faculty of the Graduate School of Arts and Sciences of Georgetown University in partial fulfillment of the requirements for the degree of Master of Public Policy In Public Policy By Jordan A. Messner, B.A. Washington, DC March 22, 2016

2 THE EFFECT OF EARLY MEDICAID EXPANSION ON MENTAL HEALTH IN CONNECTICUT AND MINNESOTA Jordan A. Messner, B.A. Thesis Advisor: Yuriy Pylypchuk, Ph.D. ABSTRACT Americans living in poverty face higher rates of mental illness compared to their higherincome counterparts, and Medicaid is the largest payer for mental health services in the United States. The purpose of this study was to assess how early Medicaid expansion under the Affordable Care Act in Connecticut and Minnesota impacted mental health outcomes. Through difference-in-difference estimation, I assessed the impact of early Medicaid expansion in Connecticut and Minnesota on mental and general health and access to care using data from the Behavioral Risk Factor Surveillance System in 2007 and Medicaid expansion only had a significant effect on the probability of not being able to see a doctor due to costs in Minnesota, reducing the risk by 7.39 percentage points. It had a marginally significant positive effect on self-reported general health in Minnesota as well. In this study, early Medicaid expansion did not have a significant impact on mental health. ii

3 TABLE OF CONTENTS Introduction and Motivation... 1 Literature Review and Institutional Background... 3 Methodological Approach and Underlying Model Empirical Model and Estimation Strategy Description of Data Results Policy Implications, Limitations and Caveats Conclusions References Appendices iii

4 LIST OF TABLES Table 1: Descriptive Statistics for Controls Table 2: Descriptive Statistics for Primary Outcomes Table 3: Descriptive Statistics for Secondary Outcomes Table 4: Primary Outcomes, Connecticut Expansion Table 5: Primary Outcomes: Time Since Last Checkup, Connecticut Expansion Table 6: Secondary Outcomes, Connecticut Expansion Table 7: Unadjusted Difference-in-Difference, Number of Days in Past 30 Days Mental Health Not Good, Connecticut Expansion Table 8: Primary Outcomes, Minnesota Expansion Table 9: Primary Outcomes: Time Since Last Checkup, Minnesota Expansion Table 10: Secondary Outcomes, Minnesota Expansion Table 11: Unadjusted Difference-in-Difference, Number of Days in Past 30 Days Mental Health Not Good, Minnesota Expansion iv

5 INTRODUCTION AND MOTIVATION Americans living in poverty face the unique health care challenge of facing high rates of mental illness. Medicaid benefits are tailored to meet the unique needs of its target population, and this includes coverage for mental health care. Currently, Medicaid is the single largest payer for mental health services in the United States and reimburses a significant proportion of substance abuse services. 1 The Affordable Care Act (ACA) sought to expand access to health care coverage to lowincome Americans by expanding Medicaid to all individuals up to 133% of the federal poverty level (FPL) beginning in The ACA offered states the opportunity to expand Medicaid to any level up to 133% FPL while receiving federal matching funds prior to Six states, among them Connecticut and Minnesota, chose to participate in this early expansion in 2010 and 2011 respectively. 2 Outcomes among these early expansion states have the potential to give insight as to what can be expected from the 2014 expansion. An important policy question to consider is does Medicaid expansion impact mental health outcomes? It is imperative to determine whether we are receiving high quality health outcomes from Medicaid, as the United States spent roughly $450 billion on this public program in This thesis will analyze what effect early Medicaid expansion in Connecticut and Minnesota had on mental health. This will be done using a difference-in-difference approach to compare how mental health outcomes changed in these two states versus control states that did not implement expansion. By examining the existing literature, it is clear that Medicaid has a 1

6 history of improving various health outcomes for low-income populations and that Medicaid plays a significant role in the delivery of mental health services. By providing coverage for lowincome Americans and reducing the cost of care, Medicaid provides access to necessary care and improves the health of beneficiaries. I would expect that early Medicaid expansion in Connecticut and Minnesota in 2010 and 2011 would result in improvements in mental health due to increased access to treatment. 2

7 LITERATURE REVIEW AND INSTITUTIONAL BACKGROUND Institutional Background: Medicaid serves as the public health insurance program for low-income Americans. It is the largest source of public coverage, with 68 million enrollees. Medicaid is jointly financed between state and federal governments, with the federal government mandating requirements and the states designing and administering their programs. Because of this, no two states Medicaid programs are exactly the same. The federal government matches state spending at a rate ranging from 50% to 73.6%. 4 Medicaid is the main payer for safety-net hospitals and health centers as well as for nursing home and community-based long-term care. In total, Medicaid finances 16% of total personal health spending in the United States and half of long-term care spending. In FY 2013, Medicaid cost a total of $ billion, with 0.8% of spending going towards mental health. 5 The ACA made some major changes to the Medicaid program. It expanded eligibility to all non-elderly adults at or below 133% of the federal poverty level beginning in January Furthermore, the federal government will finance 100% of this expansion through 2016, declining incrementally to 90% in year 2020 and on. However, a 2012 Supreme Court ruling made Medicaid expansion optional for the states. Currently, 32 states including the District of Columbia have expanded Medicaid. 6 Prior to the ACA, federal law provided funding for Medicaid coverage for only some groups of low-income individuals, including children, pregnant women, parents of dependent children, the disabled and the elderly. States are required to cover individuals in these groups up 3

8 to a minimum income threshold, but many extended coverage to higher levels of income. Before expansion under the ACA, adults were largely excluded from Medicaid, particularly childless adults. While Medicaid eligibility levels are higher for children, adults have a relatively lowincome threshold. In 2013, the median state income eligibility level was 61% of the federal poverty level for working parents, and in the vast majority of states, childless adults were ineligible for Medicaid at any level of income. Eligibility thresholds tend to be higher for adults with disabilities and the elderly. In FY 2011, children accounted for 48% of Medicaid beneficiaries, adults made up 27%, the elderly were 9% and the disabled were 15%. Furthermore, in FY 2010, 14% of Medicaid beneficiaries were dually eligible for Medicaid and Medicare. 7 Medicaid covers a broad range of services tailored to meet the needs of its population, often offering services that private insurance does not offer or strictly limits. States are required to provide mandatory services, and may also choose to cover optional services. Mandatory services include: inpatient and outpatient hospital services; physician, midwife and nurse practitioner services; early and periodic screening, diagnosis, and treatment (EPSDT) for children up to age 21; laboratory and x-ray services, family planning services and supplies; federally qualified health center (FQHC) and rural health clinic (RHC) services; freestanding birth center services; nursing facility (NF) services for individuals age 21+; home health services for individuals entitled to NF care; tobacco cessation counseling and pharmacotherapy for pregnant women; and non-emergency transportation to medical care. Popular optional services include: prescription drugs; dental care; durable medical equipment; personal care services and home and community-based services (HCBS). 8 4

9 Every state Medicaid program offers some mental health services and some offer substance use disorder services to beneficiaries. These services include counseling, therapy, medication management, social work services, peer supports, and substance use disorder treatment. 9 A Kaiser Family Foundation study shows that mental health services covered by Medicaid are generally more comprehensive than services covered by private plans on government-run exchanges. Medicaid plans typically cover psychiatric hospital visits, case management, day treatment, psycho-social rehabilitation, psychiatric evaluation and testing, medication management, individual, group and family therapy, inpatient detoxification, methadone maintenance and smoking and tobacco cessation services. 10 Furthermore, coverage offered to the adult expansion population is required to include essential health benefits, which includes mental health and substance abuse disorder benefits. This coverage must also meet mental health and substance abuse parity requirements under the Mental Health Parity and Addiction Equity Act of 2008 (MHPAEA). MHPAEA requirements state that coverage for mental health and substance abuse services cannot be more restrictive than coverage for physical health. 11 Premiums are prohibited for individuals with income less than 150% of the federal poverty level and cost-sharing for individuals below 100% of the federal poverty level is limited to nominal amounts. Combined Medicaid premiums and cost sharing for a family may not be more than 5% of family income. Partially due to these restrictions against financial barriers, both child and non-elderly adult beneficiaries are more likely to have a usual source of care and more likely to see a doctor compared to the uninsured. Furthermore, Medicaid beneficiaries have 5

10 access to care and use primary and preventive care at similar rates to their counterparts with employer-sponsored coverage. 12 Prior to Medicaid expansion under the ACA, states could only extend coverage to lowincome adults through a Section 1115 waiver of federal Medicaid rules or through a solely statefunded program. Furthermore, under a Section 1115 waiver, a state could not receive additional federal Medicaid funds for covering these individuals. Instead, it would have to reorganize existing federal funds. 13 In April 2010, the ACA allowed states to begin expanding Medicaid early (rather than waiting until January 2014) to adults with incomes up to 133% of the federal poverty level while receiving federal matching funds. Moreover, the ACA allowed states to expand coverage through a Section 1115 waiver instead of the ACA option and could still receive federal matching funds. States may have chosen this option in order to provide coverage in ways that do not meet other federal Medicaid rules, pending federal approval. 14 Between April 2010 and April 2012, six states, including the District of Columbia, chose to expand Medicaid early, covering roughly 600,000 additional adults. All of these states already covered some low-income adults through state or county funded programs, so this transition to Medicaid allowed states to preserve or expand coverage for low-income adults. 15 California expanded coverage up to 200% of the federal poverty level through a waiver in November of Connecticut expanded coverage up to 56% of the federal poverty level through the ACA option in April of The District of Columbia expanded coverage up to 133% of the federal poverty level through the ACA option in July 2010 and then further expanded up to 200% of the federal poverty level through a waiver in December

11 Minnesota expanded coverage up to 75% of the federal poverty level through the ACA option in March 2010 and then further expanded coverage up to 250% of the federal poverty level through a waiver in August New Jersey expanded coverage up to 23% of the federal poverty level through a waiver in April Washington expanded coverage up to 133% of the federal poverty level through a waiver in January Although each of these states technically expanded Medicaid early under the ACA, not all expansions actually resulted in getting more individuals enrolled in an insurance program. In all states, thousands of individuals were transferred from previously existing state programs onto Medicaid. In Washington and New Jersey, there was no new enrollment. However, California enrolled 440,000 new beneficiaries, Connecticut enrolled 36,000 new beneficiaries, the District of Columbia enrolled 10,000 new beneficiaries and Minnesota enrolled 7,000 new beneficiaries. 17 Literature Review: Mental illness tends to be associated with lower incomes. Sareen and colleagues conducted a national longitudinal study that found the presence of most lifetime Axis I and Axis II mental disorders were associated with lower levels of income. Individuals with household incomes of less than $20,000/ year had an increased risk of incident of mood disorders in a 3- year follow-up period than those with a household income of $70,000 or more. Furthermore, a decrease in household income over the study period was associated with an increased risk of incident mood, anxiety, or substance use disorders. This increased risk for mental disorders is a unique health characteristic of low-income populations. 18 7

12 Considerable evidence exists that gains in health insurance coverage among low-income populations can positively affect health outcomes. For example, Sommers et al. found that Medicaid expansions in the early 2000s led to a significant reduction in mortality. The purpose of this paper was to examine three states (New York, Arizona and Maine) that had expanded Medicaid since The authors compared these states with neighboring states that did not implement expansions, with the sample consisting of adults, ages The authors used a differences-in-differences quasi-experimental design. Controls were neighboring states without Medicaid expansion that had similar population and demographic characteristics. New Hampshire (for Maine), Pennsylvania (for New York) and Nevada and New Mexico (for Arizona) served as controls. The experimental design included data before and after Medicaid expansion in both expansion and control states, starting 5 years before expansion and ending 5 years after each state s expansion. 20 Sommers et al. found that expansions were associated with a significant reduction in adjusted all-cause mortality, yielding a relative reduction of 6.1%. Furthermore, expansions increased Medicaid coverage by 2.2 percentage points, decreased rates of uninsurance by 3.2 percentage points, decreased rates of delayed care due to costs by 2.9 percentage points, and increased rates of self-reported health status of excellent or very good by 2.2 percentage points. 21 This study design serves as a strong basis for evaluating how Medicaid expansion affects health outcomes. Moreover, Baicker et al. found that receiving Medicaid coverage in the Oregon lottery decreased the probability of a positive depression screening. This paper studied the 2008 Medicaid expansion in Oregon, which was based on a lottery drawing from a waiting list, 8

13 offering a scenario to analyze the differences in clinical outcomes between those who randomly won the lottery and received Medicaid coverage and those who signed up for the waitlist yet were not chosen. 22 These researchers collected data using an in person survey roughly 2 years after the lottery began. Measures of interest in the analysis included blood pressure, cholesterol, and glycated hemoglobin levels, screening for depression, medication inventories, and self-reported diagnoses, health status, health care utilization and out-of-pocket spending. 23 The authors found that there was no significant effect of Medicaid coverage on prevalence or diagnosis of hypertension or high cholesterol or the use of related medications. However, they did find that Medicaid coverage significantly increased the probability of a diabetes diagnosis and the use of medication to treat this condition, but there was no significant difference in the average glycated hemoglobin levels. They found that Medicaid coverage increased the use of preventative services and was associated with a large reduction in catastrophic out-of-pocket medical expenditures, almost eliminating them. Most importantly for the current context, Baicker and her colleagues found that Medicaid coverage significantly decreased the probability of a positive screening for depression by 9.15 percentage points. The eight-question version of the Patient Health Questionnaire was used to assess depression. 24 Furthermore, this study found that Medicaid coverage was associated with a significant increase in the proportion of respondents who reported their health was the same or better as compared to their health one year before. The health-related quality of life measure is based on a physical and a mental component. The results showed that Medicaid coverage was significantly associated with and increase of 1.95 points in the average score of the mental component, but 9

14 there were no significant differences in the quality of physical health or in self-reported pain or happiness. 25 The results showing a significant decrease in a positive screening for depression and a significant increase in score for the mental component of the health-related quality of life measure suggest that Medicaid coverage may have positive effects on mental health outcomes. Garfield et al. examined how the implementation of the ACA would affect adults with severe mental disorders. This study used data from the Medical Expenditure Survey Panel (MEPS), a survey conducted by the Agency for Healthcare Research and Quality (AHRQ). 26 The authors found a prevalence of severe mental disorders in the sample of 9.2%. Adults with severe mental illness were more than twice as likely to have incomes under 133% of the federal poverty level (the cutoff for Medicaid eligibility under the ACA) as those without severe mental illness. Furthermore, across all income groups, those with severe mental disorders had a higher probability of not having insurance for the full year than individuals without severe mental disorders (21% vs. 16.5%). 27 Adults with severe mental illness were also found to be significantly more likely to be enrolled in public programs (Medicare of Medicaid). Within the low-income population, people with severe mental disorders were more likely to have full-year Medicaid coverage than those without severe mental disorders (25.2% vs. 16.2%). Rates of full-year uninsurance within the low-income population were lower for those with severe mental disorders than for those without (29.4% vs. 38.1%)

15 These researchers found that only 21.5% of individuals with severe mental disorders who did not have full-year coverage had any mental health services use in the period while 48.4% of those with Medicaid coverage accessed mental health services (compared to 63.7% of those with Medicare and 38.2% of those with private insurance). 29 The authors estimated that at full implementation of the ACA in 2019, 3.7 million individuals with severe mental disorders who, in the period of analysis, were uninsured part year (1.6 million) or full year (2.1 million) would gain coverage. It is important to note that at the time of this study, the authors assumed that all states would be required to expand Medicaid to 133% of the federal poverty level. Since then, Supreme Court ruled that this expansion was optional, so these estimates are overstated. Furthermore, they estimated that 31.2% of the uninsured individuals with severe mental disorders would be covered under the Medicaid expansion. They estimated that in 2019, 24.5% of the population of individuals with severe mental disorders would be covered by Medicaid, compared with 12.8% of this population in In addition to changes in coverage, the authors estimated significant changes in use due to reform. They estimated a net increase of 1.15 million mental health care users, including 0.45 million with a severe mental disorder, because uninsured individuals will be more likely to use services when they become insured. Moreover, the authors estimated that reform would result in an increase of 2.3 million users of mental health services in Medicaid

16 METHODOLOGICAL APPROACH AND UNDERLYING MODEL The literature shows that receiving health insurance positively impacts access to health care and health outcomes. When consumers have health insurance, they purchase more health care because the price of these services is effectively lower. This concept is known as moral hazard. Generally, possessing health insurance leads to greater utilization of health care services, and this leads to better health outcomes. It is reasonable to hypothesize that low-income individuals who receive health insurance coverage through Medicaid expansion would have access to treatments for mental illness and would exhibit lower levels of mental illness. My study tests to see whether early Medicaid expansion in Minnesota and Connecticut decreased the prevalence of adverse mental health outcomes compared to states that did not implement Medicaid expansion. I compared results for outcomes over time for both expansion and control states, roughly 3 years before and after expansion. Outcomes related to mental health include the number of days in which mental health was not good, as well as other outcomes about access to and use of care. 12

17 EMPIRICAL MODEL AND ESTIMATION STRATEGY Similarly to Sommers et al., I used a difference-in-difference approach where the independent variable of interest is the interaction between timing after Medicaid expansion and expansion state, which compares the difference in mental health outcomes between expansion and control states in the period before expansion with the period after expansion. Connecticut and Minnesota serve as the treatment states while New Hampshire and Rhode Island serve as a control group for Connecticut, and Ohio serve as a control for Minnesota serves as the preintervention period while 2013 serves as the post-intervention period. Here is a simple version of the regression equation using difference-in-difference estimation: Y= B0 + B1Post-Expansion-Year+ B2Expansion-State i + B3Post-Expansion- Year*Expansion-State i + B4x+ u 13

18 DESCRIPTION OF DATA For my analysis I used the Behavioral Risk Factor Surveillance System for years 2007 and This dataset is a nationally representative household survey. The sample includes adults between the ages of 18 and 64 from Connecticut, Minnesota, New Hampshire, Ohio and Rhode Island. Furthermore, the sample was limited to respondents that had a household income of less than $20,000 in order to most closely estimate the effect on the Medicaid population. The choice in age range reflects the need to exclude the elderly population that receives health insurance coverage through Medicare, and thus was not effected by the Medicaid expansion taking place. The dataset consists of 7,040 observations. Key variables used in the analysis as outcomes include self reported overall health and number of days in the past 30 days that mental health was not good. Other outcomes related to access to care include the rate of uninsurance, the rate of not being able to see a doctor due to cost in the past 12 months and the length of time since the last routine checkup. Factors to control for include age, race, sex, income, employment status, education, marital status, number of children, year and state of residence. 14

19 RESULTS Descriptive Statistics: Table 1: Descriptive Statistics for Controls State Connecticut 18.28% 14.86% Minnesota 9.78% 25.74% New Hampshire 16.01% 10.87% Ohio 41.72% 32.80% Rhode Island 14.21% 15.74% Marital Status Married 18.44% 16.98% Divorced 31.10% 26.76% Widowed 8.30% 5.84% Separated 6.75% 5.55% Never Married 31.30% 39.66% A member of an unmarried 3.75% 4.40% couple Refused 0.36% 0.89% Education Never attended school or only 0.24% 0.13% kindergarten Grades 1 through % 3.79% (Elementary) Grades 9 through 11 (Some 13.69% 11.60% high school) Grade 12 or GED (High 43.87% 38.30% school graduate) College 1 year to 3 years 25.59% 30.80% (Some college or technical school) College 4 years or more 11.94% 15.22% (college graduate) Refused 0.24% 0.16% Employment Status Employed for wages 29.10% 28.00% Self-employed 6.07% 6.10% Out of work for 1 year or more 8.66% 11.06% 15

20 Out of work for less than % 7.90% year A homemaker 6.23% 4.46% A student 4.55% 6.69% Retired 6.79% 6.33% Unable to work 31.66% 29.08% Refused 0.48% 0.38% Income Less than $10, % 32.58% Less than $15,000 ($10, % 29.16% to less than $15,000) Less than $20,000 ($15, % 38.26% to less than $20,000 Sex Male 32.61% 38.55% Female 67.39% 61.45% Race/ Ethnicity White only, non-hispanic 69.38% 67.03% Black only, non-hispanic 12.97% 12.87% Asian only, non-hispanic 0.64% 1.45% Native Hawaiian or other 0.16% 0.26% Pacific Islander only, non- Hispanic American Indian or Alaska 1.72% 1.72% Native only, non-hispanic Other race only, non- 1.32% 2.09% Hispanic Multiracial, non-hispanic 2.20% 2.09% Hispanic 10.42% 10.47% Don t know/not sure/refused 1.20% 2.01% Age % 11.64% % 16.00% % 14.53% % 24.75% % 33.08% Children Mean Std. Dev Min 0 0 Max

21 Table 2: Descriptive Statistics for Primary Outcomes Do you have health insurance? Yes 69.70% 70.07% No 30.10% 29.47% Don t know/not sure 0.20% 0.42% Refused % Could not see doctor because of costs Yes 30.86% 30.88% No 68.86% 68.79% Don t know/not sure 0.28% 0.20% Refused % Time since last checkup Within past year (anytime 65.67% 64.76% less than 12 months ago) Within past 2 years (1 year 12.81% 13.33% but less than 2 years ago) Within past 5 years (2 years 8.38% 9.37% but less than 5 years ago) 5 or more years ago 10.54% 9.85% Don t know/not sure 1.24% 1.21% Never 1.36% 1.32% Refused % Table 3: Descriptive Statistics for Secondary Outcomes General Health Excellent 9.22% 10.67% Very Good 18.88% 20.06% Good 29.92% 31.10% Fair 26.59% 25.21% Poor 14.33% 12.54% Don t Know/Not Sure 0.20% 0.37% Refused 0.96% 0.04% How many days in past days mental health not good? Mean Std. Dev Min 0 0 Max

22 Connecticut Analysis: Primary Outcomes: Difference-in-difference models were used throughout the Connecticut analysis to compare Connecticut to the control group states with no Medicaid expansion (New Hampshire and Rhode Island). I found that Medicaid expansion had no impact on the probability of having health insurance. However, the coefficient on the Medicaid expansion variable is positive, which is what we would expect. Table 4 shows the marginal effects of Medicaid expansion in Connecticut (as shown by the interaction between Connecticut and the post intervention year, 2013) on the probability of having health insurance and not being able to see a doctor due to cost. Additionally, the models examining the probability of not being able to see a doctor due to costs and the time since last checkup found that all else equal, the Connecticut Medicaid expansion had no significant effect on these outcomes. Table 5 shows the marginal effects of Medicaid expansion in Connecticut on the time since last checkup. Again, although these impacts were not statistically significant, the impacts were moving in the correct direction with Medicaid expansion having a negative effect on the probability of not being able to see a doctor due to costs and a positive impact on the probability of having a checkup within the past year. 18

23 Table 4: Primary Outcomes, Connecticut Expansion Marginal Effect (Std. Err.) Health Insurance Could Not See Doctor Due to Cost Connecticut.06229** (.0265) (.0282) Year * (.0201).0205 (.0214) Connecticut*Year (.0331) (.0353) *.05<p<.10, **.01<p<.05, ***p<.01 Table 5: Primary Outcomes: Time Since Last Checkup, Connecticut Expansion Marginal Effect (Std. Err.) Within past Within past Within past 5 years or Never year 2 years 5 years more Connecticut (.0270).0082 (.0078).0079 (.0075).0107 (.0101).0018 (.0017) Year (.0209).0069 (.0060).0066 (.0058).0090 (.0078).0015 (.0013) Connecticut*Year (.0338) (.0097) (.0094) (.0126) (.0021) *.05<p<.10, **.01<p<.05, ***p<.01 + Within past year= anytime less than 12 months ago; Within past 2 years= greater than or equal to 1 year but less than 2 years ago; Within past 5 years= greater than or equal to 2 years but less than 5 years ago Secondary Outcomes: Continuing use of difference-in-difference estimation, I examined the effect of Medicaid expansion in Connecticut on health outcomes. Table 6 shows the effect of Medicaid expansion on all secondary outcomes. Although I hypothesized that Medicaid expansion would increase insurance coverage, and that this increase in access to health care would significantly improve outcomes, the data did not support this. All else equal, Medicaid expansion had no significant effect on self-reported general health at any level (Excellent, Very Good, Good, Fair, or Poor). However, the direction of these impacts are what we would have expected, increasing the 19

24 probability of reporting excellent, very good or good general health and decreasing the probability of reporting fair or poor general health. Finally, I analyzed the effect of Medicaid expansion in Connecticut on the number of days in the past 30 days that respondents reported that mental health was not good. First, I calculated an unadjusted difference-in-difference estimate to represent the difference in the effect of the intervention between the treatment group (Connecticut) and the control group (New Hampshire and Rhode Island). Although I expected Medicaid expansion in Connecticut to cause a decrease in the average number of days in the past 30 days that mental was not good, I found the opposite to be true. In fact, this estimate suggested that the intervention increased the number of days by days. These calculations are shown in Table 7. Using difference-in-difference regression analysis, the direction of these results were upheld, but all else equal, I found that Medicaid expansion did not have any significant effect on the number if days in the past 30 days that mental health was not good. 20

25 Table 6: Secondary Outcomes, Connecticut Expansion Marginal Effect (Std. Err.) General Health Number of Days in Past 30 Days Mental Health Not Good Connecticut Excellent:.0001 (.0108) * (.6757) Very Good:.0001 (.0087) Good:.0000 (.0022) Fair: (.0096) Poor: (.0121) Year 2013 Excellent: (.0083) (.4729) Very Good: (.0067) Good: (.0017) Fair:.0079 (.0074) Poor:.0100 (.0093) Connecticut*Year 2013 Excellent:.0088 (.0136) Very Good:.0070 (.0109) Good:.0018 (.0028) Fair: (.0121) Poor: (.0153) (.8668) *.05<p<.10, **.01<p<.05, ***p<.01 Table 7: Unadjusted Difference-in-Difference, Number of Days in Past 30 Days Mental Health Not Good, Connecticut Expansion Year Connecticut New Hampshire/ Rhode Island Difference (Connecticut-New Hampshire/ Rhode Island) Difference ( ) Difference-in- Difference: Other covariates in this model provided significant results. I found that all else equal, those who were out of work for one year or more experienced more days in which mental health was not good in the past 30 days compared to respondents who were employed for wages. Additionally, all else equal, those who were out of work for less than one year reported more days in which mental health was not good in the past 30 days compared to those who were 21

26 employed for wages. Also, all else equal, those who were unable to work reported more days in which mental health was not good in the past 30 days compared to those who were employed for wages. All of these results were highly statistically significant. Furthermore, all else equal, respondents who had incomes between $15,000 and $20,000 reported fewer days in which mental health was not good in the past 30 days compared to respondents with incomes less than $10,000. This result was highly statistically significant. Interestingly, all else equal, males experienced fewer days where mental health was not good in the past 30 days compared to females. This result was highly statistically significant. Moreover, there were some interesting distinctions between racial and ethnic groups. All else equal, Black non-hispanics experienced fewer days in the past 30 days in which mental health was not good compared to White non-hispanics, Asian non-hispanics experienced fewer days compared to White non-hispanics and Hispanics, and Hispanics experienced fewer days compared to White non-hispanics. All of these results were highly statistically significant. Marginal effects for all covariates in this model can be found in Appendix 1. Minnesota Analysis: Primary Outcomes: Difference-in-difference models were used throughout the Minnesota analysis as well to compare Minnesota to the control group state with no Medicaid expansion (Ohio). I found that all else equal, there was no significant change in the probability of having health insurance or the time since last checkup as a result of Medicaid expansion. However, the direction of these results 22

27 are what we would expect with Medicaid expansion having a positive effect on the probability of having health insurance and the probability of having a checkup within the last year. Medicaid expansion in Minnesota did have a negative and significant impact on the probability of not being able to see a doctor due to cost. All else equal, the probability of not being able to see a doctor due to costs decreased by 7.39 percentage points as a result of Medicaid expansion in Minnesota. Table 8 shows the marginal effects of Medicaid expansion in Connecticut on the probability of having health insurance and not being able to see a doctor due to costs. Table 9 shows the marginal effects of Medicaid expansion in Connecticut on the time since last checkup. Table 8: Primary Outcomes, Minnesota Expansion Marginal Effect (Std. Err.) Health Insurance Could Not See Doctor Due to Cost Minnesota.0913*** (.0302) (.0315) Year (.0166).0266 (.0180) Minnesota*Year (.0348) ** (.0362) *.05<p<.10, **.01<p<.05, ***p<.01 Table 9: Primary Outcomes: Time Since Last Checkup, Minnesota Expansion Marginal Effect (Std. Err.) Within past Within past Within past 5 years or Never year 2 years 5 years more Minnesota.0504 (.0307) (.0069) (.0083) (.0136) (.0020) Year (.0180) (.0041) (.0049) (.0080) (.0011) Minnesota*Year (.0355) (.0080) (.0096) (.0157) (.0022) *.05<p<.10, **.01<p<.05, ***p<.01 + Within past year= anytime less than 12 months ago; Within past 2 years= greater than or equal to 1 year but less than 2 years ago; Within past 5 years= greater than or equal to 2 years but less than 5 years ago 23

28 Secondary Outcomes: I also examined the effect of Medicaid expansion in Minnesota on health outcomes. I found that all else equal, Medicaid expansion had a marginally significant effect on self-reported general health. All else equal, Medicaid expansion resulted in an increased probability of reporting excellent, very good or good general health by 2.36 percentage points, 2.44 percentage points and.45 percentage points respectively. Furthermore, all else equal, Medicaid expansion resulted in a decreased probability of reporting fair or poor health by 2.34 percentage points and 2.91 percentage points respectively. These results were only marginally significant, however. I analyzed the effect of Medicaid expansion in Minnesota on the number of days in the past 30 days that respondents reported that mental health was not good as well. As I did in the Connecticut analysis, I started by calculating an unadjusted difference-in-difference estimate to represent the difference in the effect of the intervention between the treatment group (Minnesota) and the control group (Ohio). Again, I found the opposite of what I hypothesized to be true. This estimate implies that Medicaid expansion in Minnesota increased the number of days in which mental health was not good in the past 30 days by days. These calculations are shown in Table 11. Using difference-in-difference regression analysis, the direction of these results were smaller in magnitude and insignificant. These results are shown in table

29 Table 10: Secondary Outcomes, Minnesota Expansion Marginal Effect (Std. Err.) General Health Days in Past 30 Days Mental Health Not Good Minnesota Excellent: (.0111) ** (.8188) Very Good: (.0114) Good: (.0021) Fair:.0027 (.0110) Poor:.0036 (.0137) Year 2013 Excellent:.0010 (.0064) (.4151) Very Good:.0010 (.0066) Good:.0002 (.0012) Fair: (.0063) Poor: (.0079) Minnesota*Year 2013 Excellent:.0236* (.0128) Very Good:.0244* (.0131) Good:.0045* (.0025) Fair: * (.0126) Poor: * (.0157) (.9299) *.05<p<.10, **.01<p<.05, ***p<.01 Table 11: Unadjusted Difference-in-Difference, Number of Days in Past 30 Days Mental Health Not Good, Minnesota Expansion Year Minnesota Ohio Difference (Minnesota-Ohio) Difference ( ) Difference-in- Difference: Other covariates in this model proved to be significant. All else equal, respondents who were out of work for less than one year had more days in which mental health was not good in the past 30 days compared to those who were employed for wages. This result was highly statistically significant. Additionally, all else equal, homemakers experienced more days in the past 30 days in which mental health was not good compared to those employed 25

30 for wages, and those unable to work experienced more days compared to those employed for wages. These results were highly statistically significant. Income also significantly impacted this outcome. All else equal, respondents who reported household income between $10,000 and $20,000 experienced fewer days in which mental health was not good in the past 30 days compared to respondents who reported income less than $10,000 and respondents who reported household income between $15,000 and $20,000 experienced fewer days in which mental health was not good in the past 30 days compared to respondents who reported income less than $10,000. These results were highly statistically significant. Again, males experienced better mental health outcomes than females. All else equal, males had fewer days in which mental health was not good in the past 30 days compared to females. This result was also highly statistically significant. Race also played an important role. All else equal, Black non-hispanics had fewer days in which mental health was not good in the past 30 days compared to White non-hispanics, Asian non-hispanics experienced fewer days compared to White non-hispanics, Native Hawaiian or other Pacific Islander non-hispanics experienced fewer days compared to White non-hispanics, and those who were unsure of their race or refused to respond reported fewer days compared to White non-hispanics. All of these results were significant at conventional levels. Finally, age had an impact. All else equal, respondents age had more days in which mental health was not good in the past 30 days compared to respondents age This result was highly statistically significant. Additionally, all else equal, respondents age experienced more days compared to respondents age However, this result was only marginally significant. Marginal effects for all covariates in this model can be found in Appendix 2. 26

31 POLICY IMPLICATIONS, CAVEATS AND LIMITATIONS Using these outcomes from early Medicaid expansion as a prediction for what we can expect to see as a result of the 2014 Medicaid expansion under the ACA, we may expect to see positive impacts on low-income Americans access to the health care system. Although most results in this study were insignificant, results showed an increase in the probability of having health insurance, a decrease in the probability of not being able to see a doctor due to costs and an increase in the probability of having a checkup in the past year. In the Minnesota analysis, there was a statistically significant decline in the probability of not being able to see a doctor due to costs by 7.39 percentage points as a result of early Medicaid expansion in this state. These results are encouraging, showing that Medicaid expansion assists in bringing down the barriers to the health care system that many Americans face. Moreover, the goal of increasing access to care is ultimately to improve health outcomes. This study found that early Medicaid expansions had a positive, yet insignificant, impact on self reported general health in Connecticut and a positive and marginally significant impact on self reported health in Minnesota. However, the analysis in this study found that Medicaid expansions resulted in an increase in the number of days in the past 30 days in which mental health was not good, but these results were insignificant. We would of course like to see Medicaid expansion decreasing this number, especially since mental health is a particularly troubling issue for low-income individuals. Despite insignificant results in this study, we should be optimistic about the long-term effects of the 2014 Medicaid expansion under the ACA. First, these are only 2 states of the 32 27

32 states that have decided to expand Medicaid at this time, and differences across these states make it unlikely that the results are generalizable. Furthermore, the ACA requires states to expand Medicaid eligibility up to 133% FPL. In the early expansion being studied, Connecticut only expanded up to 56% FPL and Minnesota expanded up to 75% FPL. While Connecticut expanded Medicaid by 81,000 individuals, only 36,000 were new enrollees. In Minnesota, total expansion enrollment was 84,000 with only 7,000 new enrollees. 32 As of May 2015, 12 million new beneficiaries have been enrolled in Medicaid nationwide since the implementation of the ACA. 33 Compared to the enormous growth that has taken place since 2014, it would be premature to make conclusions about insignificant results based off of these small increases in enrollment. Other limitations exist in this study as well. First, the data reported here is self-reported survey data. The subjective nature of this data makes it difficult to determine objective impacts of legislation. Moreover, it is possible that increased access to the health care system in some circumstances could result in respondents becoming more aware of their general health or mental health status, and they may overestimate the number of days they felt their mental health was not good. 28

33 CONCLUSIONS This study found that early Medicaid expansion in Connecticut and Minnesota were associated with positive impacts on access to the health care system for low-income populations, with the probability of not being able to see a doctor due to costs significantly decreasing in Minnesota. Moreover, early Medicaid expansions were associated with positive impacts on general health, and these results were marginally significant in Minnesota. However, results did not show that Medicaid had a positive or significant impact on mental health outcomes. Due to the limitations discussed in the previous section, there are reasons to believe that the much larger Medicaid expansion implemented in 2014 may provide stronger results in expected directions. Future research should focus on the 2014 Medicaid expansion to analyze whether Medicaid coverage results in improved mental health outcomes for a larger population nationwide. Furthermore, future research should rely less on self-reported data in favor of more objective alternatives. Medicaid is the largest health insurance program in the United States, and due to its expansion under the ACA, its reach is the largest in our nation s history. The ACA has made tremendous strides in expanding insurance coverage to the nation s most vulnerable citizens. Medicaid has a long history of providing access to health care for low-income parents and children and consequently has improved health outcomes. Now, childless adults are finally receiving access to this necessary care. It is imperative to continue to study and improve this program to protect the health of low-income Americans. 29

34 REFERENCES 1. Behavioral Health Services, Medicaid.gov States Getting a Jump Start on Health Reform s Medicaid Expansion, Kaiser Family Foundation, April 2, National Health Expenditure Fact Sheet, Centers for Medicare and Medicaid Services, December 3, Paradise, J. Medicaid Moving Forward, Kaiser Family Foundation, May 9, ibid. 6. ibid. 7. ibid. 8. ibid. 9. Health Insurance and Mental Health Services, MentalHealth.gov Canon et al. Adult Behavioral Health Benefits in Medicaid and the Marketplace, Kaiser Family Foundation, June Health Insurance and Mental Health Services, MentalHealth.gov. 30

35 12. Paradise, J. Medicaid Moving Forward, Kaiser Family Foundation, May 9, States Getting a Jump Start on Health Reform s Medicaid Expansion, Kaiser Family Foundation, April 2, ibid. 15. ibid. 16. ibid. 17. Sommers et al. Lessons from Early Medicaid Expansions Under Health Reform: Interviews with Medicaid Officials, Medicare & Medicaid Research Review 3, no. 4 (2013) Sareen et al. Relationship Between Household Income and Mental Disorders: Findings From a Population-Based Longitudinal Study, JAMA Psychiatry 68, no. 4 (April 2011): Sommers et al. Mortality and Access to Care among Adults after State Medicaid Expansions, N Engl J Med 367 (September 2012): ibid. 21. ibid. 22. Baicker et al. The Oregon Experiment Effects of Medicaid on Clinical Outcomes, N Engl J Med 368 (May 2013):

36 23. ibid. 24. ibid. 25. ibid. 26. Garfield et al. The Impact of National Health Care Reform on Adults With Severe Mental Disorders, American Journal of Psychiatry 165, no. 5 (May 2011): ibid. 28. ibid. 29. ibid. 30. ibid. 31. ibid. 32. Sommers et al. Lessons from Early Medicaid Expansions Under Health Reform: Interviews with Medicaid Officials, Medicare & Medicaid Research Review 3, no. 4 (2013) Pradhan, R. Skyrocketing Medicaid signups stir Obamacare fights, Politico, May 18,

37 APPENDICES Appendix 1: Secondary Outcomes: Number of Days in Past 30 Days Mental Health Not Good, Connecticut Expansion Marginal Effect (Std. Err) Number of Days in Past 30 Days Mental Health Not Good Connecticut * (.6757) Year (.4729) Connecticut*Year (.8668) Marital Status Education+ Divorced Widowed Separated Never Married A member of an unmarried couple Refused Elementary Some high school High school graduate Some college or technical school College 4 years or more Refused Employment Status Self-employed Out of work for 1 year of more Out of work for less than 1 year Homemaker Student Retired Unable to work Refused Income++ Less than $15,000 Less than $20, (.6552) (.9973) (.9586) (.6367) (1.1087) ** (1.9028) * (3.9968) (4.0136) (4.0010) (4.0111) (4.0338) (5.9608) * (.7979) *** (.7256) *** (.7782).9874 (.8873).0474 (.8191) (.8360) *** (.5638) (4.6555).1056 (.4922) ** (.4877) Sex *** (.4305) Race Black only, non-hispanic *** (.6620) Asian only, non-hispanic *** (1.2845) Native Hawaiian or other Pacific Islander (3.9897) only, non-hispanic American Indian or Alaska Native only,.1512 (1.4725) non-hispanic 33