: This dissertation proposal examines responses to cardiac surgery mortality report cards in three states-New Jersey, New York, and Pennsylvania-over the course of the 1990s by three sets of stakeholders-cardiac surgeons, referring cardiologists, and HMOs. Report card proponents contend that publicly releasing this information provides incentives for hospitals and surgeons to improve the quality of patient care and supplies consumers with previously unavailable information to assist in selecting a provider. Critics counter that the report cards are inherently inaccurate and surgeons can act to improve their performance artificially, thereby biasing the scores. The first part of this work focuses on the incentives surgeons have to avoid riskier cases, looking across states and time for evidence of whether surgeons treat less severe cases during periods of report card data collection, and within report card states over time for evidence that lower-volume and worse-rated surgeons engaged in more patient avoidance. These difference-in-difference and surgeon fixed-effect regression analyses will be conducted using hospital discharge data from all three report card states, Maryland, and the AHRQ HCUP National Inpatient Sample. Part Two studies whether Pennsylvania cardiologists changed their referral patterns in response to the May 1998 report card release. Leveraging the fact that cardiac surgery patients undergo cardiac catheterization prior to surgery, a novel algorithm has been developed to infer the identity of the referring cardiologist from inpatient and outpatient surgery data. Nested Iogit methods will be used to explore whether the magnitude of the referral responses varied by patient and cardiologist characteristics and to attempt to uncover the extent to which the report cards presented information not already known by cardiologists. The third part considers whether HMOs in the three report card states engaged in selective contracting for cardiac surgery and whether HMOs responded to report cards by increasing their business with better or worse rated providers. Difference-in-difference, multinomial Iogit, and market-share models will compare the HMO volume response with that of Medicare FFS patients, who are not constrained in their choices of provider.
