Universal Health Insurance And Mortality: Evidence From Massachusetts Charles Courtemanche Georgia State University and NBER Daniela Zapata IMPAQ International
Chapter 58 and the ACA follow the same Three-legged stool approach (Gruber, 2010): Massachusetts Reform (Chapter 58): Patient Protection and Affordable Care Act (ACA): 1. Reforms to non-group insurance market 1. Reforms to non-group insurance market Guarantee issue Guarantee issue Community rating Community rating Health insurance exchange Health insurance exchange Dependent care up to age 26 Dependent care up to age 26 2. Individual and employer mandate 2. Individual and employer mandate 3. Subsidies: Medicaid expansions Subsidized coverage on sliding scale 3. Subsidies: Medicaid expansions Subsidized coverage on sliding scale
Impacts of the Massachusetts Reform End results: Uninsured rate lowered to 2% in Massachusetts by 2010 (Massachusetts Division of Health Care, Finance, and Policy, 2010) Estimated annual cost : $707 million (Raymond, 2009) Negatives Delays from waiting to find a provider (Long and Stockely, 2011) Crowd out of private insurance (Yelowitz and Cannon, 2010) Increase in employer-sponsored insurance premiums (Cogan et al., 2010)
Positives Impacts of the Massachusetts Reform Decrease in unmet medical needs because of cost (Long and Stockely, 2011) Reduction in emergency room utilization (Kolstad and Kowalski, 2010; Miller, 2011a) Shorter length of hospital stays (Kolstad and Kowalski, 2010) Increased utilization of preventive services (Kolstad and Kowalski, 2010) Improved self-reported child health (Miller, 2011) Improved self-reported adult health and preventive care use (Wees, Zaslavsky, Ayanian, 2013) Improved self-reported adult health (physical and mental), joint disorders, d activity it limitations, it ti BMI (Courtemanche and Zapata, 2014)
What do we do and why is it important: What is the effect of the Massachusetts Health Care Reform on mortality? Previous evidence shows improvements on self-reported measures of health. Self-reports can have a degree of subjectivity. The Massachusetts health care reform is the most similar intervention to date to the ACA.
Data Annual mortality rates stratified by age, sex, and race from the CDC Mortality File. The sample is restricted to adults 20 to 64 years old State level information Years: 1986-2010
Data State level information on: Unemployment (Bureau of Labor Statistics) Median income (Census) Poverty rate (Census) Percent of the population > 25 with a High School Diploma (Census) Percentage of the state population that is married (Census) Percentage of the state population that is Hispanic (National Cancer Institute)
Methods Linear difference-in-differences model with mortality rate as the dependent variable Before and After Periods 1986 Apr-06 Oct-06 Jan-07 Jul-07 2010 Reform is signed into Law Subsidies for people <100% FPL start Subsidies for people <300% FPL start Individual mandate takes effect Before: During: After: 1986-2005 2006-2007 2008 2010
Methods All-Cause Mortality per 100,000 Adults, Massachusetts versus the rest of the country (1986-2010)
Synthetic Control Method (Abadie, et al. 2003, 2010, 2012) Data driven procedure to find a control group using a linear combination of non-treated units. It can be used when you only have one treated unit and a pool of potential control units. The synthetic control group is formed by a linear combination of units that best replicate pre-treatment trends of the outcome of interest. t If the number of pre-intervention periods is large enough matching in pre-treatment covariates (X) and outcomes (Y) reduces bias due to unobserved factors (Abadie et al. 2010).
In our case: Treated unit: MA Synthetic Control Method (Abadie, et al. 2003, 2010, 2012) Donor pool: 46 states (minus: CA, ME, VT, OR) Mortality predictors used to find the synthetic control group: Unemployment Median income Poverty rate Education Marital status Hispanic Population - Synthetic control group: Linear combination of CO, CT, NY, RI, UT.
Methods All-Cause Mortality per 100,000 Adults, Massachusetts versus Synthetic Massachusetts (1986-2010)
DID Results Estimated t change in All Cause Mortality Rate after the MA Health Care Reform Among Adults 20 to 64 (1986-2010) Average mortality rate in Estimated change in Percentage change in Massachusetts per 100,000 adults before the reform mortality per 100,000 adults after the reform mortality after the reform (2007-2010) (1986-2005) (2007-2010) 2010) 298-19 (8.25)** 6% N=3,368 368 ** Statistically significative at the 5% level. Robust standard errors clustered at the state level are shown in parentheses Regression controls for: State: unemployment, median income, poverty, education, percentage married, percentage Hispanic. Year fixed effects State*age*sex*race*year group fixed effects
Policy Implications The MA health care reform shares many features with the ACA suggesting that the mortality effects could be broadly similar. However, the ACA included additional cost-cutting measures that could potentially mitigate the mortality gains. Supreme Court ruling allows states to opt-out of Medicaid expansions. States that decide not to expand could see smaller improvements in mortality rates. On the other hand, baseline uninsured rates were unusually low in MA, so the coverage expansions andand corresponding mortality improvements from the ACA could potentially be greater
Future steps: Sommers, Long, and Baicker (May, 2014): Use county-level mortality rates and a DID estimation strategy and find that: MA reforms was associated with a 3 percent decrease in all- cause mortality compared with the control group. The decrease was explained mainly by a reduction in deaths amenable to health care (e.g. congestive heart failure, infection diseases, etc.) Changes were larger in counties with lower household incomes and higher pre-reform reform uninsured rates. Suggestions are welcome!