Quality vs. Familiarity: An Equilibrium Analysis of Physician Referrals
Broadly speaking, the economic purpose of referrals is to address informational frictions. However it is unlikely that referring physicians are fully informed about all the available specialists in a market, even if they do serve as perfect agents for their patients. In this paper we evaluate the importance of specific, objective information about specialist quality, compared with other factors that may simply foster greater familiarity, in the context of referrals for heart surgery. This is a unique setting where relevant quality measures—patient mortality rates—are publicly available through “report cards” on individual surgeons. We wish to recover the apparent value that referring physicians place on these reported mortality rates, relative to factors such as shared time in medical school that could increase their familiarity with a surgeon without necessarily relating to quality.
We use a simple matching framework to model the interaction between referring physicians (cardiologists) and specialists (surgeons). Cardiologists are assumed to internalize patient preferences, so that their utility of a match (i.e., a completed referral) to a particular surgeon is based largely on beliefs about the likely outcomes for the patient. Risk-adjusted mortality rates from the surgeon report cards enter into these beliefs as an indicator of surgeon quality. However, because cardiologists do not have perfect information about the performance and capabilities of all surgeons in their market, we also include factors that would tend to increase their familiarity with a particular surgeon: practicing in the same hospital, physician group, or health system, training at the same institution, or having nearby offices.
Surgeon utility depends simply on the payment for performing a procedure and the marginal cost of providing this treatment. The marginal cost is specified to be increasing in the number of patients seen over a given time period. Importantly, this captures any capacity constraints that limit the quantity of services each surgeon may supply, implying that a patient is less likely to be seen by a surgeon who already faces high demand.
Estimation uses a conditional logit model, suitably augmented. This follows Bayer and Timmins (2007), which extends methods pioneered in Berry, Levinsohn, and Pakes (1995) to incorporate non-price spillovers in demand—in our case, the crowding out of a surgeon’s available capacity. The data come from four HRRs in Pennsylvania. Referrals are inferred from Medicare claims records: the cardiologist who provides a specific diagnostic procedure to the patient most recently before the surgery is treated as the referring physician.