Physician Admitting Patterns and Integration with Hospitals

Spurred by a variety of policy changes and market forces, relationships between physicians and hospitals are increasingly important for physician and hospital reimbursement, pricing, and quality. The nature of these relationships, however, is not well understood. At one extreme, physician practices or groups may be formally acquired by a hospital or hospital system. Indeed, recent years have seen a new wave of this form of hospital-physician integration in the U.S., with the fraction of physicians working in practices owned by hospital systems increasing from 7% in 2009 to 25% in 2015. At the other extreme, physicians may remain fully independent but nonetheless choose to operate almost exclusively with a given hospital. We refer to this latter relationship as "physician loyalty" and the former relationship as integration. We refer to each of these terms collectively as "alignment." In this sense, alignment reflects a physician's observed referral or admitting patterns across hospitals, while loyalty and integration reflect different underlying mechanisms for these observed patterns.

In this study, we first examine the role of vertical integration in determining alignment between physicians and hospitals observed in the data. Essentially, does vertical integration serve to increase hospital-physician alignment, or do hospitals simply integrate with physician practices that are already loyal to that hospital? We answer this question by estimating the effects of formal integration on overall hospital-physician alignment. Here, in addition to linear fixed effects estimates allowing for unobserved, time-invariant physician factors, we identify the causal effects of integration on alignment by estimating the predicted probability of vertical integration at the physician-level from a first-stage regression estimated using plausibly exogenous physician characteristics before the practice is acquired by a hospital system. We then use this predicted probability of vertical integration as an instrument for the observed vertical integration in a traditional two-stage least squares (2SLS) estimator. As an initial summary finding, this first-stage regression also identifies the types of physicians and hospitals most likely to become formally integrated.

We then examine the effect of hospital-physician alignment on treatment intensity, hospital prices, and quality of care, where we differentiate between effects due to formal integration versus physician loyalty. Our analysis is based on a panel of all U.S. general medical and surgical hospitals from 2009 to 2015 and derives from various data sources. We employ the SK&A physician database to identify formal physician-hospital integration, and we quantify hospital-physician alignment using observed patient flows from 100% of the Medicare fee-for-service inpatient claims over the same time period. The Healthcare Cost Report Information System (HCRIS) provides measures of hospital prices, the Hospital Compare database provides quality indicators, and American Hospital Association (AHA) annual surveys provide additional hospital characteristics.

Preliminary results show that increased hospital-physician alignment may improve quality and efficiency, with no significant increase on price, when alignment is driven by physician loyalty. If instead we observe increased alignment due to formal integration, we see no improvement in quality or efficiency with a significant increase in price.