Observational Studies of the Effect of Medicaid on Health: Controls Are Not Enough

Abstract

Covariate-adjusted cross-sectional comparisons show that Medicaid patients have worse health outcomes than other patients. We evaluate the validity of this research design for estimating the causal effect of Medicaid on mortality. Even after controlling for common covariates, Medicaid patients have worse preoperative health and lower socioeconomic status than privately insured patients. Controlling for additional variables shrinks the mortality differences but still does not eliminate imbalance in other predetermined variables. These results can be explained by fairly weak assumptions about unmeasured confounders. We conclude that cross-sectional observational methods do not produce valid causal estimates of Medicaid’s mortality effects.