Betting on Your Health: How Uncertainty Affects Physician Choice

Much of the economics literature on physician decision making, particularly the empirical work, focuses on the expected (or marginal expected) outcomes for patients as the object of interest in understanding physician choices. This literature mirrors the approach taken in cost­‐effectiveness analyses, in which mean incremental outcomes are compared to mean incremental costs to determine whether a therapy is cost­‐effective. In this paper, we build a model of decision-­making that incorporates the variance of patient outcomes, as well as the mean. The approach allows us to test for the role of uncertainty in treatment decisions in the labor and delivery setting. Specifically, we ask whether variance in expected outcomes explains observed treatment choices, controlling for mean expected outcomes. We study length of stay, total spending and readmissions for newborns as outcomes that enter into physician and patient objectives. We estimate a structural treatment choice model that incorporates mean and variance in expected outcomes between vaginal and C-­section delivery. This allows us to estimate key parameters of choice, including the coefficient of risk aversion. Our results help explain observed treatment variation. Risk preferences explain more intensive treatment than the average outcome would predict; precisely the observation that motivates many policy proposals. We then estimate counterfactual models incorporating agency and alternative policy approaches to consider the impact on treatments, cost and outcomes of different policies.