Physician-Hospital Stability and Mortality in Cardiology

Cardiologists often work in multiple hospitals. We find that roughly 20% of physicians performing percutaneous coronary interventions (PCI) in the United States work in more than one hospital location each month. Little is known about the impact of this behavior at the hospital level despite the growing recognition that the hospital’s process of care delivery is instrumental in achieving better patient outcomes. This study develops a set of measures on physician-hospital stability and evaluates whether hospitals with cardiologists that work in fewer locations are associated with lower risk adjusted mortality rates (RAMR) following PCI. 

We identify 100% of FFS Medicare beneficiaries undergoing a PCI from 2001 through 2004 as well as the physician performing the procedure and where each PCI was delivered, thereby creating physician-hospital pairings for all PCIs delivered over the period. We incorporate patient level characteristics from the Medicare Beneficiary Summary files, physician level characteristics from the American Medical Associations (AMA) Masterfile, hospital level characteristics from the American Hospital Association (AHA) Annual Survey, and geographic-market level characteristics from the US Census Bureau’s Tiger shape files and the Area Health Resource files. Mortality is identified at the beneficiary level using the Medicare Vital Status files and is risk-adjusted using beneficiary demographics and co-morbidities.

Using the physician-hospital pairings identified in our analytical data, we create a Physician-Hospital Stability Index (PHSI) that captures how concentrated a physician’s PCI volume is across hospitals in a given year. The index is constructed by calculating the share of PCIs performed by a physician in each of their hospitals, then summing the squared shares from each hospital in a given year. The PHSI goes from values approaching 0 to 1. A physician with a PHSI of 1 does all of their PCI work in one hospital location. A physician with a PHSI less than 1 delivers PCIs in multiple hospitals, with increasingly less concentration in any one hospital as the PHSI approaches 0.

We summarize PHSI values at the hospital level, describing how concentrated a hospital’s set of physicians are within a given year. Preliminary findings indicate that higher PHSI values at the hospital level hold a significant inverse relationship with RAMRs. Forthcoming models will estimate PHSI using exogenous variation in physician-hospital distances to minimize the endogenous relationship between physician volume and patient outcomes.

Contractual affiliation between physicians and hospitals is a growing trend in medical markets, incentivizing greater stability between physicians and where they work. The impact of greater stability on health outcomes is not well understood. Some research indicates that physicians sort patients to the most appropriate medical facility. Other research indicates that operating in multiple locations is a threat to patient safety. This research will offer empirical evidence on the health impacts of greater physician-hospital stability.