The Patient Protection and Affordable Care Act (PPACA) came into law in 2010. Originally, according to the Act, a state would lose its federal Medicaid funding if it did not expand its Medicaid eligibility to include all persons earning below 138% of the federal poverty level on January 1, 2014. However, in a Supreme Court Case in 2012 this was ruled as unconstitutional and Medicaid expansion in 2014 was made optional. Twenty four states and the District of Columbia opted to expand their Medicaid programs on January 1, 2014 and the remaining 26 states opting against it. Section 1115 of the Social Security Act allows states to alter the federal Medicaid requirements to promote the overall state Medicaid program. Among those states that expanded Medicaid Arkansas, Arizona and Iowa adopted approved Section 115 Waivers to expand their Medicaid programs. The variability in the states' decisions regarding Medicaid expansion presented researchers with the opportunity to study the impacts of Medicaid expansion on various facets of health care.

There is a growing body of evidence suggesting that implementation of the coverage expansions under the PPACA and Medicaid expansion led to significant decreases in rate of uninsured persons, increase in access to health care and improvements in affordability of healthcare. Along with improving access and affordability of health care, the PPACA aimed at reducing the growth rate of health care expenditures by reducing wasteful use of resources such as preventable inpatient and emergency department (ED) visits. According to previous research, access to primary care and insurance coverage are significantly and negatively associated with experiencing preventable inpatient and ED visits. Historically, racial/ethnic minorities have had lower rates of access to primary care and insurance and higher rates of preventable inpatient and ED visits which might change with implementation of PPACA. Within states that have expanded Medicaid, adopting different methods of expansion may also impact patterns of inpatient and ED utilization and disparities in those. In the current political scenario and looming uncertainty over the future of PPACA and the possibilities of modifying the PPACA it might benefit policy makers to gain knowledge on the early impact of Medicaid expansions and different approaches to expanding Medicaid under the PPACA. This study seeks to determine the impact of Medicaid expansion and different types of Medicaid expansion on overall and preventable inpatient and ED utilization and disparities in those through a three-state comparison between Kentucky, Arkansas and Florida.

Another major reform under PPACA was in the area of substance use disorder treatment. Despite the high societal burden exerted by patients with substance use disorders treatment rates among them have been low. The most common reasons cited for the poor access to treatment have been lack of insurance coverage. The PPACA required all insurance plans sold after January 1, 2014 to cover substance use disorder treatments. Additionally, plans were required to cover screening, brief intervention and referral to treatment for substance use disorders. This might potentially lead to changes in treatment rates and sources of payments for substance use disorder treatment. Further, the promotion of integration between substance use disorder treatment and primary care might lead to increased referrals by healthcare professionals to substance use treatment. Thus, in this study we also seek to assess the impact of Medicaid expansion on admission to substance use disorder treatment facilities and changes in sources of payments and rate of health care referrals to those treatment facilities.

Background The Patient Protection and Affordable Care Act (PPACA) was signed into law by President Obama on March 23, 2010. Two of the most contentious clauses under the PPACA were penalizing all persons who lacked insurance after January 1, 2014 and expansion of Medicaid insurance to all adults with incomes at or below 138% of the federal poverty level. The constitutionality of the act came under scrutiny and led to the "National Federation of Independent Businesses vs Sebelius" case wherein the Supreme Court upheld most of the provisions of the PPACA but decided to give states the option of not expanding Medicaid and at same time retain their Federal funding for the program. This Supreme Court decision resulted in Medicaid expansion becoming optional for states. As of January 1, 2014 24 states and the District of Columbia elected to expand Medicaid coverage as per the PPACA requirements. This variation in Medicaid expansion provided natural experiments to investigate the impact of expansion decisions.

Inpatient care accounts for the largest share of national health care expenditures in the United States. There are several provisions in the PPACA which encourage coordination of care between providers through use of patient centered medical homes and base provider reimbursements on patient outcomes as opposed to volume of care delivered. Through such reforms the PPACA aimed at reducing wasteful use of resources such as preventable hospitalizations. Previous research has shown that inpatient use, specifically preventable hospitalizations can be sensitive to coverage gains and can serve as an indicator of access to primary care. Thus, assessing how insurance expansion under the PPACA has impacted overall inpatient utilization and rate of preventable hospitalizations may provide insight on how successful PPACA has been in replacing high cost wasteful services with lower cost primary care services. Wherry et al, using data from National Health Interview Survey, determined a positive association between the PPACA Medicaid expansion and rate of overnight hospitalizations, whereas Sommers et al, using data from a survey fielded in Arkansas, Kentucky and Texas found no significant association between overnight hospitalizations and Medicaid Expansion. They did not detect any differences in probability of overnight hospitalization and method of expansion (Arkansas vs Kentucky). However, their study was limited by a small sample size and based on results of a survey. Additionally, due to lack of administrative data they were unable to assess outcomes which might be more sensitive to increased access to primary care such as preventable hospitalizations. To the best of our knowledge only one study conducted thus far, assessed the impact of Medicaid expansion under PPACA on overall inpatient utilization and rates of preventable hospitalization. However, their study was limited to California, additionally different counties had different levels of expansions, with some expanding coverage up to 133% FPL, some up to 200% FPL and some were below 100% FPL. None of the studies thus far have determined the impact of different methods of Medicaid expansion under the PPACA on rates of all-cause and preventable hospitalizations using administrative data. One of the goals of the PPACA was to curb health care costs by reducing the volume of care delivered at ED. Certain measures incorporated in the PPACA such as increasing the number of insured persons and hence increasing access to care outside the ED and integration of health care delivery may contribute to reducing ED utilization. Two studies assessing the impact of Medicaid expansion on ED use also analyzed rates of ambulatory care sensitive conditions (ACSC), emergent but primary care treatable conditions and non-emergent ED visits. However, one study was limited to only the state of Maryland lacking a comparison state which did not expand Medicaid and the other was limited to only a single for-profit investor owned chain of hospitals in 6 states which expanded Medicaid and 14 which didn't. Furthermore, none of the studies have assessed how different methods of Medicaid expansion under the PPACA impact rates of overall and non-emergent ED use. Racial/ethnic disparities in various facets of health care such as insurance coverage and overall access to health care is well documented in literature. Combined with the coverage provisions, additional reforms made by the PPACA such as, elevating the National Center on Minority Health and Health Disparities at the National Institutes of Health from a Center to a full Institute, might lead to narrowing of these disparities. The studies that have assessed the impact of Medicaid expansion under PPACA on disparities in inpatient use or ED use have relied primarily on self-reported data which is subject to cognitive, non-response, recall and other biases. Furthermore, granular measures of inpatient or ED use such as preventable hospitalizations and preventable ED visits cannot be accurately obtained from self-reported data. None of the studies thus far have determined the impact of Medicaid expansion or different methods expansions under PPACA on disparities in types of inpatient or emergency department utilization.

Persons with substance use disorders (SUD) have significantly higher health-care costs, higher rates of suicide attempts and disabilities compared to the general population thus posing a burden to society overall. Despite this, in 2012, nearly 90% of the persons aged 12 and older who required treatment for SUDs did not receive adequate treatment. The lack of insurance or insurance not covering SUD treatment services leading to the patient's inability to pay for them are some of the most commonly cited barriers to treatment. The coverage expansions under the PPACA have led to nearly 1.6 million Americans suffering from SUDs in Medicaid expansion states gaining health insurance coverage. Further, under the PPACA, SUD treatment is one of the ten essential health benefits that all plans must offer beginning in 2014. The PPACA also requires all plans to adhere to the federal Mental Health Parity and Addiction Equity Act of 2008 (MHPAEA). Under this act all plans which offer mental health and SUD treatment benefits to their beneficiaries must make the benefits no more restrictive than medical benefits. The PPACA mandates all plans to cover screening, brief intervention and referral to treatment (SBIRT) for SUDs. Along with mandating SBIRT the PPACA also encourages Accountable Care Organizations and Patient Centered Medical Homes which might increase coordinated care for patients suffering from SUDs. The efforts for increasing care coordination and incorporating SBIRT for SUD in primary might translate to increase in number of health care referrals for SUD treatment which has historically, predominantly been through law enforcement agencies. Only one study has examined the impact of 2014 coverage expansions on the eligible adult population (18-64 year olds). The authors used the National Survey on Drug Use and Health data and found a significant increase in mental health treatment utilization without any significant changes in treatment of SUDs. However, they did not draw comparisons between states that expanded Medicaid vs states that did not. The study also did not examine changes in sources of payment for the SUD treatment following Medicaid expansion.

Arkansas, Kentucky and Florida responses. Arkansas was the first state to secure an approval for its "Private Option" demonstration project to implement the Medicaid expansion under the PPACA. Arkansas adopted a premium assistance strategy, which involved using federal funds to provide individual commercial health insurance for all the newly Medicaid eligible persons earning up to 138% of the FPL by placing them into one of the federally qualified health plans. On May 9, 2013 Governor Beshear declared that Kentucky would go ahead with Medicaid expansion as proposed under the PPACA. Kentucky decided to carry out the Medicaid expansion by placing the newly eligible persons in this pre-existing managed Medicaid program. Despite the state's vehement opposition to the PPACA, Governor Rick Scott expressed support for a "limited Medicaid expansion" through a federally funded and privately administered managed care plan. However, Florida opted against the Medicaid expansion. This resulted in almost 764,000 individuals not having any affordable coverage options by virtue of not being eligible for Medicaid coverage or Marketplace subsidies.

Comparing inpatient and emergency room utilization between these three states will highlight the impact Medicaid expansion and different approaches to Medicaid expansion might have on health service utilization. Further, using national level data, comparing the change in rate of SUD treatment admissions after Medicaid expansion in states that expanded Medicaid vs states that did not will determine the impact of Medicaid expansion on SUD treatment utilization.

Objectives and Specific Aims

1. To determine the impact of Medicaid expansion and type of Medicaid expansion (purchase commercial insurance vs traditional Medicaid expansion) on all cause inpatient utilization and preventable hospitalizations and if effect of Medicaid expansion on inpatient utilization differs by race/ethnicity.

   1. Using a differences in differences approach we will compare the rate of all-cause hospitalizations and preventable hospitalizations for adults aged 19-64 between 2013 and 2014 in the two states that expanded Medicaid vs Florida. A sub-group analysis comparing Arkansas and Kentucky will reveal if different approaches to expansion influenced inpatient utilization.
   2. Using a differences in differences in differences approach we will determine whether the impact of Medicaid expansion on inpatient utilization for adults aged 19-64 differed by racial/ethnic groups (Hispanics, Non-Hispanic Whites and Non-Hispanic Blacks) between the three states.

2. To determine the impact of Medicaid expansion and type of Medicaid expansion on ED utilization and non-emergent ED use and if the effect of Medicaid expansion on ED utilization differs by race/ethnicity.

   1. Using a differences in differences approach we will compare the rate of ED visits and non-emergent ED visits for adults aged 19-64 between 2013 and 2014 in the two states that expanded Medicaid vs Florida. A sub-group analysis comparing Arkansas and Kentucky will reveal if different approaches to expansion influenced ED utilization.
   2. Using a differences in differences in differences approach we will determine whether the impact of Medicaid expansion on emergency department utilization, for adults aged 19-64, differed by racial/ethnic sub-groups (Hispanics, Non-Hispanic Whites and Non-Hispanic Blacks) between the three states.

3. To determine the impact of Medicaid expansion on the rate of SUD treatment admissions in facilities receiving some public support.

   1. Using a differences in differences in differences approach we will compare the rate of SUD treatment admissions from 2010-2014 for persons aged 18-54 vs those aged 12-17 in the states that expanded Medicaid in 2014 vs states that did not expand Medicaid in 2014.
   2. Using a differences in differences in differences approach we will compare the changes in sources of payment for the SUD treatment admissions from 2010-2014 for persons aged 18-54 vs those aged 12-17 in the states that expanded Medicaid in 2014 vs states that did not expand Medicaid in 2014.
   3. Using a differences in differences in differences approach we will compare the rate of health care referral for the SUD treatment admissions from 2010-2014 for persons aged 18-54 vs those aged 12-17 in the states that expanded Medicaid in 2014 vs states that did not expand Medicaid in 2014.

Methods Impact of Private Option, Traditional Medicaid Expansion and Opting Against Medicaid Expansion on Inpatient Utilization: A Three State Comparison The investigators will use a longitudinal cross-sectional quasi-experimental difference in difference study design to assess the impact of Medicaid expansion and method of expansion on our inpatient utilization measures. We will determine the change in rates of all-cause and preventable hospitalization from 2013 - 2014 and contrast the change between states based on expansion/method of expansion status. The investigators will use a longitudinal cross-sectional quasi-experimental difference in difference in difference study design to assess the impact of Medicaid expansion and method of expansion on racial/ethnic disparities on our inpatient utilization measures. The investigators will determine the change in rates of all-cause and preventable hospitalization from 2013 - 2014 for each race/ethnicity and contrast the change between states based on expansion/method of expansion status.

Data CDC - Wonder Bridged-Race Population Estimates The national center for health statistics (NCHS) combines 31 race categories used in the Census to four race categories: Asian or Pacific Islander, Black or African American, American Indian or Alaska Native, White and provides age, gender and ethnicity (Hispanic/Non-Hispanic) specific population counts for each county of the United States. These files are available for public use on CDC Wonder website. We will use these files to obtain the total population counts that will serve as denominators for our utilization metrics for each county of Arkansas, Florida and Kentucky for 2013 and 2014.

HCUP SID For our inpatient utilization measures, the investigators will acquire the Healthcare Cost and Utilization Project (HCUP) State Inpatient Database (SID) from Arkansas, Kentucky and Florida respectively. The SIDs capture the particular state's hospital inpatient discharge records for the given year and contain them in a uniform format to allow inter-state comparisons. The unit of analysis in the dataset is a hospital discharge. In addition to information each hospitalization such as length of stay, charges, admit and discharge dates, the SIDs also contain clinical information such as the diagnosis recorded and procedures performed during each hospitalization and demographics of the hospitalized patient such as birth date, zip code of residence, gender, race, ethnicity, payer type etc.

Analysis The investigators will depict the rates of outcomes for each state be quarter and determine if there is any change in trends between 2013 and 2014 or if there is an immediate change between 2013 and 2014 by using interrupted time series analysis. The investigators will calculate the crude rates of all-cause and preventable hospitalization for each state for each year. Using the US 2014 population as the reference population we will carry out direct standardization to compute the age and sex adjusted rates of all-cause and preventable hospitalizations for each state for each year. The investigators will also carry out direct standardization for race/ethnicity subgroups. Next, we will perform multivariate regression analysis to estimate the excess change in inpatient utilization due to Medicaid expansion. The investigators will use a Poisson distribution for our outcome variables. The population at risk will be the total population count of each the cohort as obtained from NCHS data. Standard errors will be clustered at the emergency management region level to account for nesting of each cohort within a emergency management region. The model will include emergency management region level fixed effects, to account for time invariant emergency management region level factors, dummies for age-group, gender race/ethnicity, and year. The investigators will create indicators for period (2013 and 2014) and expansion status (expanded Medicaid or not). The investigators will interact period indicator with expansion status indicator to determine the excess change in the outcome measures for states which expanded Medicaid (AR and KY) vs state that did not expand Medicaid in 2014 (FL). The same model specification will be used to determine the impact of method of Medicaid expansion on rates of all-cause and preventable hospitalizations. For this analysis, the investigators will compare the 2013 and 2014 data from Arkansas and Kentucky. To assess if the impact of Medicaid expansion on our outcome measures differed by racial/ethnic groups in the three states we will use a multivariate difference in difference in difference model. The data will be restricted to Non-Hispanic White, Hispanic and Non-Hispanic Black cohorts. In addition to the terms in the prior models, we will include interactions between race/ethnicity and state indicators to control for baseline disparities in the outcome in the states and interactions between race/ethnicity and time to control for longitudinal trends in disparities overall. The indicators for Non-Hispanic Blacks and Hispanics will be interacted with period indicators and state indicators. The coefficients of these interactions will provide us with the excess change in the outcome measures between Hispanic vs Non-Hispanic Whites and Non-Hispanic Blacks vs Non-Hispanic Whites in states which expanded Medicaid (AR and KY) vs state that did not expand Medicaid in 2014 (Referent = FL).

Sensitivity Analysis The investigators will use a negative binomial distribution for our outcome measures in place of a Poisson distribution. In the second sensitivity analysis, for our measure of all-cause hospitalizations, the investigators will exclude all maternal discharges. In a third sensitivity analysis, the investigators will include only hospitalizations for adults aged 27-64 to exclude all young adults who might have experienced coverage gains prior to 2014 due to the "young-adult mandate". In a fourth sensitivity analysis, the investigators will include only those hospitalizations that took place in community hospitals. In a fifth sensitivity analysis, the investigators will collapse the racial/ethnic groups and define each cohort based on year, state, emergency management region, age-group and sex. The investigators will do this in order to incorporate those admissions in the count of the outcome measures where the patient's race/ethnicity is missing. Sixth, instead of grouping the data annually the investigators will group it by quarter and perform a segmented regression analysis. Lastly, the investigators will calculate all the outcome variables as rates per 10,000 persons and use weighted ordinary least squares regression analysis to determine if our interpretations change.

Impact of Private Option, Traditional Medicaid Expansion and Opting Against Medicaid Expansion on Emergency Department Utilization: A Three State Comparison The investigators will use a longitudinal cross-sectional quasi-experimental difference in difference study design to assess the impact of Medicaid expansion and method of expansion on the inpatient utilization measures. The investigators will determine the change in rates of overall and types of ED visits from 2013 - 2014 and contrast the change between states based on expansion/method of expansion status. The investigators will use a longitudinal cross-sectional quasi-experimental difference in difference in difference study design to assess the impact of Medicaid expansion and method of expansion on racial/ethnic disparities on our ED utilization measures. The investigators will determine the change in rates of overall and types of ED visits from 2013 - 2014 for each race/ethnicity and contrast the change between states based on expansion/method of expansion status.

Data CDC - Wonder Bridged-Race Population Estimates HCUP SEDD For our outcome measures the investigators will acquire the Healthcare Cost and Utilization Project (HCUP) State Emergency Department Database (SEDD) for Arkansas, Kentucky and Florida respectively. The SEDDs capture all the ED visits for the particular year which take place in EDs that are affiliated with hospitals and that do not result in hospitalizations. In addition to information on each ED visit such as charges and admit dates, the SEDDs also contain clinical information such as the diagnosis recorded and procedures performed during each visit and demographics of the patient such as birth date, zip code of residence, gender, race, ethnicity, payer type etc.

Outcome Measures

1. ED visit.

2. Type of ED visit:

   1. Preventable/Avoidable.
   2. Emergent. Analytic data structure Similar to previous aim, only difference would be in terms of the outcomes. Analysis Similar to previous aim, only difference would be in terms of the outcomes. Sensitivity Analysis The investigators will use a negative binomial distribution for our outcome measures in place of a Poisson distribution. In a second sensitivity analysis, we will include ED visits for adults aged 27-64 to exclude all young adults who might have experienced coverage gains prior to 2014 due to the "young-adult mandate". In a third sensitivity analysis the investigators will change our definition of preventable visits and emergent visits. The investigators will define a preventable visit as any visit whose probability of being preventable is ≥ 0.5 and any visit whose probability of being emergent is ≥ 0.5 will be defined as such. In a fourth sensitivity analysis the investigators will include only those ED visits that took place in community hospitals. In a fifth sensitivity analysis we will collapse the racial/ethnic groups and define each cohort based on year, state, emergency management region, age-group and sex. The investigators will do this in order to incorporate those ED visits in the count of the outcome measures where the patient's race/ethnicity is missing. Sixth, instead of grouping the data annually we will group it by quarter and perform a interrupted time series analysis. Seventh, the probability of ED visit being preventable will be defined as sum of probability of ED visit being non-emergent and emergent but primary care treatable and the probability of ED visit being emergent will be defined as sum of probability of ED visit being emergent and ED care needed. Lastly, the investigators will calculate all the outcome variables as rates per 10,000 persons and use weighted ordinary least squares regression analysis to determine if our interpretations change.

Early Impact of Medicaid Expansion on Public Sector Substance Abuse Treatment: Evidence from the Affordable Care Act We will use a longitudinal cross-sectional quasi-experimental difference in difference in difference study design. We will determine the changes in rates of admissions to SUD treatment facilities, changes in the sources of payment and changes in the rate of health care referrals from 2010 - 2014 for persons aged 18-54 vs 12-17 and compare the changes between them based on states' Medicaid expansion status.

Data Treatment Episode Data Set - Admissions (TEDS-A) TEDS-A is a nationwide administrative database of admissions to specialty SUD treatment facilities. It consists of facilities that receive some public source of funding for providing SUD treatment. Some of the states collect data only on admissions that are publically funded whereas others also collect privately funded admissions from facilities that receive some public funding. TEDS-A includes data on the demographics of the patient including age-group, gender, race/ethnicity, state of residence, substances abused, type of facility, payment source for the admission for a sub-group of states and source of referral to the treatment. It covers nearly 80% of SUD treatment admissions in the US.

Current Population Survey (CPS) The CPS reports monthly information on population demographics including age, race/ethnicity, gender, state of residence etc. We will use this data to get information on total population counts for our cohorts developed on the basis of age-group (12-14, 15-17, 18-29, 30-39, 40-54), gender, race/ethnicity (Non-Hispanic White, Hispanic, Non-Hispanic Blacks, Others), state and year and state-level covariates. The investigators will exclude New Hampshire and Michigan from our data to avoid ambiguity since these states expanded later in 2014. To assess impact of Medicaid expansion on changes in payment source for SUD admissions the investigators will include only those states which provide primary payer information in the TEDS files for at least 85% of the admissions in each of the years.

Outcome Measures

1. Treatment admissions.

2. Source of payment

   1. Treatment admissions privately funded.
   2. Treatment admissions funded by Medicaid.
   3. Treatment admissions self-funded.
   4. Treatment admissions that are free/ or funded by other government sources.

3. Treatment admissions through a health care source of referral Analytic Data Structure We will sum all the outcome measures from the TEDS-A data by state, year, age, gender, and race/ethnicity. The investigators will ink the CPS data and the aggregate level TEDS file files. The investigators will gather the covariate information from the CPS data at the state-year level which will give us our final analytic dataset to assess the impact of Medicaid expansion on SUD treatment in publically funded facilities.

Analysis The investigators will compare the demographics of the admissions and covariates between states that expanded Medicaid vs states that did not expand Medicaid for each year of the study period. The investigators will conduct t-tests/chi-square to determine significant differences for continuous and categorical variables respectively. The investigators will sum up all the outcome measures for each state and plot them year wise on separate graphs to visually examine the trends. The investigators will perform multivariate regression analysis to estimate the excess change in our outcome measures attributable to Medicaid expansion. The investigators will use a Poisson distribution for outcome variables. The population at-risk will be the population count of each cohort. The model will include state level fixed effects, to account for time invariant state level factors, and dummies for age-group, gender, period indicator (pre-expansion vs post-expansion), race/ethnicity, state-year level covariates (marital status, educational attainment, and proportion unemployed), and time varying state-policies (eg. legalizing medical marijuana). The investigators will create indicators for period before and after Medicaid expansion (2014 and 2010-2013) and expansion status of state (expanded Medicaid or not) and an indicator for treatment group (12-17 vs 18-54). The investigators will interact period indicator with expansion status indicator to control for overall trend in outcome in states that expanded vs states that did not expand. The investigators will interact the treatment indicator with period indicator to control for overall trend in outcome among treatment group (18-54 year olds) vs non-treatment group (12-17 year olds). The investigators will also control for baseline differences in the outcome among treatment group (18-54 year olds) vs non-treatment group (12-17 year olds) in states that expanded vs states that did not expand by interacting expansion status with treatment group indicator. The model will include a difference in difference in difference indicator to assess the effect of Medicaid Expansion on the outcome measure among treatment group (18-54 year olds) vs non-treatment group (12-17 year olds). The difference in difference in difference indicator will be specified by interacting the expansion, treatment and period indicators.

Sensitivity Analysis The investigators will conduct a number of sensitivity analyses to test the robustness of our model and study subject selection criteria. First, The investigators will use a negative binomial distribution for the outcome counts instead of Poisson distribution. Second, The investigators will exclude persons aged 18-29 to eliminate the population that had an increase in access to insurance in 2010 due to the "young adult mandate". The age-group that was impacted by the mandate was 18-26. However, due to the availability of age-groups in the TEDS-A data (18-20, 21-25 and 26-29) The investigators decided to include the entire 18-29 age-group. Third, consistent with previous literature The investigators will exclude admissions only for detoxification since these are considered to be forerunners for treatments rather than treatment itself. Fourth, The investigators will determine the impact of Medicaid expansion only on rate of first time admission. Finally, The investigators will conduct a weighted least squares regression for each of our outcome measures. Here the weights will be the population counts for each cohort. The outcomes will be as follows: a. rate of treatment admissions, privately funded admissions, Medicaid funded admissions, self-funded admissions, admissions payed for by charity or other government organizations, admissions through health care referrals per 10,000 persons.

Limitations Proposed studies on the impact of Medicaid expansion on inpatient and ED utilization have several limitations. The investigators analyze data from only three states, hence our findings might not be generalizable to other states of the US. Although Florida is a southern state, it is significantly different in terms of demographic composition from Arkansas and Kentucky. However, since the rate of uninsured in the three states was similar prior to Medicaid expansion we believe that it would serve as a reasonable control. Additionally, The investigators do not have access to data from other southern states which did not expand Medicaid in 2014 and which might be better controls such as Texas, Tennessee, Mississippi, Louisiana or Georgia. The investigators have access to only one year of post-reform data hence The investigators might not be able to capture the full impact of the health care reform. Since The investigators have only one year of pre-reform data we might not be able to fully account for differences in trends of the outcomes pre-reform in our three states. Our third study is subject to limitations as well. TEDS-A data does not include all the substance abuse treatment centers. Thus, The investigators may potentially miss some of the admissions. Our findings might be biased by showing higher rates of admissions in states have greater public funding by virtue of them contributing more data. Further since the funding may fluctuate annually, The investigators cannot ascertain whether the changes in treatment rates are a function of the fluctuations in reporting or due to Medicaid expansion. The investigators mitigate this limitation to some extent by including the 12-17 age group as a control group which will not be impacted by the Medicaid expansion but if there are changes in reporting rates for states The investigators expect them to have equal effect on 12-17 years olds as on 18-54 year olds. Not all states provide information on source of funding, thus limiting the generalizability of our findings about the changes in source of payment. However, previous research has shown no differences in states that provide data on sources of funding vs states that do not.

Inclusion Criteria:

• Age 19-64 for aims 1 and 2
• Age 12-54 for aim 3

Exclusion Criteria:

• None

For Aim 1 and 2. The population will be obtained from NCHS CDC Wonder based on county of residence, gender, age-group (19-26, 27-39, 40-49 and 50-64), race/ethnicity (Non-Hispanic Whites, Non-Hispanic Black, Hispanics and Others) for each of the AR, FL and KY for 2013 and 2014.

For Aim 3.