Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

Public insurance for the poor (Medicaid) in the United States often results in poor continuity of coverage due to the phenomenon of "churning." Churning occurs when individuals lose and regain coverage in a short period of time. Gaps occur within the Medicaid program because of changes in family circumstances that make individuals ineligible for public insurance, including when individuals find employment (often in jobs that do not provide health benefits), and administrative complexities that result in failure to renew coverage. The purpose of this study is to evaluate the impact of insurance transitions on health care utilization among the publicly insured non-elderly.

We evaluated the impact of gaps in health insurance coverage on medical care utilization using the Medical Expenditure Panel Survey (MEPS). MEPS provides monthly insurance status and resource utilization. Additionally, MEPS contains detailed demographic, socioeconomic and health status information. We calculate the number of transitions for an individual in each round of the survey and use the total number of inpatient hospitalizations, emergency room visits, outpatient visits, and medication prescriptions as our measure of resource utilization. We control for comorbid conditions using the Charlson index. We use survey rounds to create a panel of observations on each individual and analyze these data using static and dynamic panel data models that allow us to control for unobserved heterogeneity and state dependence. In the dynamic panel setting, we deal with the problem of endogenous initial conditions using the method suggested in Wooldridge (2005). We run both Poisson and Negative Binomial models, where the latter help us deal with overdispersion of the dependent variable.

Our sample has 9883 individuals, of whom 2,730 had one transition into or out of Medicaid and 598 had more than one transition. Our results indicate that using a dynamic model is important, especially when there is likely to be state dependence, and that unobserved heterogeneity is not a big concern in this population once we control for self-reported health status. We find that individuals with multiple transitions tend to use the emergency room (ER) 20% more often than those with no transitions. These individuals also tend to be hospitalized 47% more often and have 5% more physician office visits, in addition to having 6% lower prescription drug usage. Moreover, the dynamic effects are important: a hospitalization or ER visit in the previous round is negatively correlated to the likelihood of having one in the current round, whereas either of these in the initial round of the survey is highly positively correlated with that in the current round. Thus initial conditions are informative and assuming them to be exogenous can lead to misleading results.

Transitions in insurance status among Medicaid enrollees are associated with significant increases in emergency room visits, hospitalizations and physician office visits, and with decreases in prescription drug utilization. Policymakers may achieve a reduction in health expenditures by designing insurance plans that reduce transitions among vulnerable populations.