Estimating hospital choices when the true choice set is unknown

Random utility models are frequently used in the health care economics literature to estimate a patient's probability of choosing a hospital. A key assumption of such models is that individuals choose the hospital that would yield the highest utility, given a known set of alternatives. However, managed care networks sometimes do not include all hospitals in a local area, so an individual's choice may reflect both the utility of the hospital and the unobserved limits on the choice set. This paper describes a method of estimating patient choice based on random utility when each hospital in the observed choice set has some probability of being unavailable. The model is tested using both simulated data and data hospital choices in several markets in Florida.