Is blood thicker than water? Peer effects in stent utilization among Floridian cardiologists

Variations in physician practice are pervasive and costly, and may be harmful. The objective of much policy in the West is to increase the interconnectedness of physicians, furthering the transfer of information and thus reducing variation.This study tests whether physicians are influenced by the practice of peers, or if propensity, mere context or sorting of like-minded physicians better explain similarities and differences in practice. We study US cardiologists who place coronary stents into patients with blocked arteries around the heart. Organized in locally competing physician groups and also as solo practitioners, they see patients in offices, but insert the stents at a shared production facility – the cath lab.We examine their use of the popular drug-eluting coronary stents between their launch and rapid adoption in early 2003, and through the period of late 2006 in which private and public reports of serious late side-effects eventually led to reductions in use. Our analyses use administrative claims data on nearly 1000 cardiologists and their patients in Florida, USA, merged with Florida physician licensure data. Collectively these physicians used these stents nearly a quarter of a million times in the 4 year period reviewed. Pooled and panel linear regressions for device utilization by a physicians are estimated using measures of peer utilization, physician characteristics and controls for unobservable physician characteristics, common shocks and selection effects.We find strong evidence for intra-group but against inter-group practice spillovers. Even when sharing the same lab, competing cardiologists did not appear to correlate practices. Our results are consistent with a view that policies aimed at increasing the interconnectedness of physicians must first consider the organizational barriers and competitive forces that can stymie knowledge transfer even among physicians working closely together.

► Increasing the flow of information between physicians could reduce pervasive, costly and harmful practice variations.
► However, such physician peer effects could be moderated by organizational barriers, competition between physicians and physical isolation.
► Such effects can be confounded by non-random selection into practice groups and local shocks and designs must be robust.
► We found Florida coronary stent use was correlated within groups but not across competitors, even when these practiced at the same lab.
► Findings suggest that physician peer influences are small, and may be restricted to non-competing physicians, reducing their policy attractiveness.