Medicaid Coverage and Health Outcomes: Evidence from Tennessee

Medicaid provides health insurance coverage to over 65 million Americans, mainly insuring low income individuals, pregnant women and children. Despite its important role as a safety net program, Medicaid’s impact on health outcomes is hotly debated. Previous research has found that state Medicaid expansions during the late 1990s and early 2000s reduced mortality among the non-elderly. However, more recent evidence from the Oregon Experiment provides mixed results with improvements identified in self-rated physical health and depression but not in clinical measures such as blood pressure or cholesterol levels. Understanding the impact of Medicaid on health is important given the expansion of the program under the 2010 Affordable Care Act (ACA) and the Supreme Court’s decision to allow states to choose whether or not to expand Medicaid.

In this study, I provide evidence on the impact of Medicaid coverage on health outcomes using a policy change in 2005 which led to the disenrollment of approximately 170,000 individuals from Tennessee’s Medicaid program. Previous research on the impact of this policy change has identified significant increases in employment primarily to secure private insurance coverage and increases in uninsured inpatient stays primarily originating in the emergency department. However, to the best of my knowledge, its impact on health outcomes has not been studied. This study examines the extent to which loss of Medicaid coverage due to the 2005 policy change affected self-rated physical and mental health and mortality. Data on self-rated health is obtained from the Behavioral Risk Factor Surveillance System (BRFSS) while information on mortality is obtained from the Compressed Mortality Files provided by the National Center for Health Statistics (NCHS).

To identify causal effects, I use a difference-in-differences approach that compares changes in health outcomes before and after the policy change in 2005 between Tennessee and other comparison states. Following prior work on this policy, I use multiple approaches to constructing the comparison group. I first include all states in the US and then restrict the comparison group to southern states. Next, I use a synthetic control method that uses a weighted average of comparison states to address differences between Tennessee and the control group. Finally, I use a triple difference approach that compares non-elderly individuals who were primarily affected by the policy change to elderly persons who were not. Preliminary findings provide some evidence that health status worsened due to the loss of coverage. I discuss the implications of my findings for the populations affected by the Medicaid expansions under the ACA.