Physician treatment preference formation and diffusion: The case of specialty referrals
Date:
Pany, M., Weinreb, G. G., McWilliams, J. M. Download presentation
Abstract: Variation in physician practice patterns is a key determinant of health care spending and care quality. Practice variation at the regional and local levels has been linked to variation in local health care capital and technology (e.g., the number of available hospital beds or advanced imaging), patient preferences, and the diffusion of knowledge, skills, and preferences among physicians. While existing empirical evidence highlights the key influence of physician preferences on practice patterns, evidence on how these preferences form and diffuse is lacking.
In this study, we investigate how practice preferences form and diffuse among communities of physicians. We consider the case of primary care physicians’ (PCPs’) specialty referral decisions, an important component of practice variation where PCPs have significant discretion and their choices have downstream consequences for patients’ care cascades. Specifically, we examine whether PCPs who work together during quasi-randomly assigned overlapping shifts at one high-volume primary care practice influence each other’s specialty referrals. The practice comprises a mix of experienced attending physicians and less experienced resident physician trainees supervised by the attendings, allowing us to investigate preference diffusion across horizontal (i.e., attending-attending) and vertical (i.e., attending-trainee) physician relationships. Using records of PCPs’ specialist choices in the practice’s electronic health record (EHR), we find that PCPs who share shifts are more likely to refer to the same specialists than those who do not. The effect exhibits a dose response, increasing in magnitude with the number of overlapping shifts. We also find evidence suggestive of heterogeneity across horizontal versus vertical physician relationships.
Given the outsize role of providers in influencing patient treatment patterns, the dynamics of preference formation among physicians and other providers are critical to understand, especially because provider preferences appear to ossify over the course of a career. For example, it may be the case that preference convergence among PCPs promotes patient welfare by channeling patients to the highest quality specialists, either in terms of average specialist quality or patient-specialist match quality. At the other extreme, it could be that specialist quality is unknown to PCPs and referral patterns within a practice reflect new PCPs uncritically adopting the preferences of established PCPs. Our findings suggest that informing preference formation of early-career providers with the information contained in experienced providers’ preferences could help establish more streamlined practice workflows, potentially improving care quality. At the system level, for example, ongoing efforts to make PCPs accountable for total medical expenditures could give established PCPs incentive to coach the referral behaviors of junior PCPs in the same practice. Within a health system, practice managers can implement EHR workflows that surface the preferred specialists of established PCPs at the point of referral, giving junior PCPs visibility into their colleagues’ private knowledge of specialist quality.