Healthcare Competition based on Physician-Hospital Networks

The welfare consequences of competition in healthcare have been the subject of significant theoretical and empirical interest in the health economics literature. Given the question's theoretical ambiguity, this is ultimately an empirical issue for which researchers have focused almost exclusively on hospital-level competition. However, healthcare competition naturally extends beyond the walls of the hospital and should arguably include a hospital's broader patient reach through its network of affiliated physicians. U.S. healthcare market structure has also changed dramatically in recent decades. For example, hospitals now operate as part of much larger hospital systems, hospitals regularly enter into exclusive arrangements with physicians such as direct employment or joint ventures, and hospitals may even operate their own insurance plans. Assessing competition (and its subsequent effects on health outcomes, prices, and access to care) based purely on hospital market shares fundamentally ignores these important issues. In this paper, I extend the measure of hospital competition to encompass the broader physician-hospital network along with the strength of this network. I then compare the effects of such "healthcare" competition versus hospital competition on health outcomes and utilization.

The analysis is based on Medicare claims data from 2009 through 2011, where I have 100% of the inpatient and (institutional) outpatient claims along with the universe of physician office claims for a 20% sample of Medicare beneficiaries. I supplement these data with self-reported health measures from the Medicare Health Outcomes Survey as well as hospital characteristics from the American Hospital Association (AHA) annual surveys.

With these data, I first derive a new measure of healthcare market concentration incorporating both the hospital and the extended physician-hospital network (PHN). This is essentially a weighted average of concentration indices across different physicians, constructed in two stages. First, I form the Herfindahl-Hirschman Index (HHI) for each physician based on the share of referrals to different hospitals. Physicians with high HHIs therefore heavily concentrate their referrals to very few hospitals, while physicians with low HHIs refer patients more uniformly to several different hospitals. Second, I take a weighted average of each physician HHI relevant to each hospital, weighted by the physician's share of patients relative to the hospital's total patients admitted in the year. The resulting hospital-level index reflects the breadth and tightness of referrals from their extended physician network. I find that measures of concentration are significantly different based on PHN versus standard hospital market share calculations, with markets substantially more concentrated. Preliminary evidence suggests that these increases in concentration may positively effect patient outcomes with potential reductions in utilization; however, prices may ultimately increase.