New information technology systems and a Bayesian hierarchical bivariate probit model for profiling surgeon quality at a large hospital

New information technology systems at hospitals and medical centers provide administrators and policymakers with greater information than ever before on both the characteristics of patients and on the outcomes associated with individual surgeons. This paper develops a Bayesian hierarchical bivariate probit model describing surgeon performance in terms of the 30-day mortality and 30-day morbidity of their patients. We apply the model to a sample of 2,578 patients who received care from one of 36 surgeons at a large hospital. The model is estimated using Markov Chain Monte Carlo (MCMC) simulation methods. After accounting for observed differences in the health status of patients prior to surgery and the complexity of the procedure performed, we construct quality indices measuring surgeon performance in terms of morbidity and mortality. These indices are used to evaluate surgeon performance against absolute standards for morbidity and mortality rates, and then to conduct "head to head" comparisons of individual surgeons within subspecialty surgery departments at the hospital. Our approach highlights the potential benefits of new information technologies for monitoring surgeon quality.