The diffusion of a medical innovation: is success in the stars? Further evidence.

• Author: Burke, Mary A.

1. Introduction

   In Burke, Fournier, and Prasad (2007) we examined the role that local interaction among physicians played in the adoption and utilization of stents following their introduction in the mid-1990s. Our interest was in the mechanics of social learning in medical practice, and the key hypothesis was that of asymmetric effects, whereby high-status physicians (stars) exert greater influence than others. Given the limitations of data, this is a challenging question to resolve empirically. Our paper was grounded in the belief that significant progress could, nevertheless, be made. The Comment (Huesch 2009) raises important questions, and, to a considerable extent, we are sympathetic to its concerns. In response, we have performed some of the suggested tests, to the extent feasible with the data. The new results are found to strengthen our previous conclusions. On balance, we think the evidence continues to point to star effects in the diffusion of stents. Of course, much work remains to be done.

   Huesch (2009) points to four difficulties with our empirical approach: (i) our definition of stars based on residency schools, (ii) inadequate controls for time-varying hospital effects, (iii) our focus on the operating physician, as opposed to the attending physician, and (iv) inclusion of nonadopting stars in the model. Among these concerns, the definition of stars is the most fundamental concern, and also the hardest to address. We would like a definition that identifies the true opinion leaders, or most influential practitioners (Young et al. 2003). A physician's status within the local, professional community is likely to determine the influence he or she exerts, but this status is difficult to observe in the available data. The best option, in our view, is to consider a number of plausible alternatives and hope for robust results.

   With regard to (i), Huesch (2009) suggests incorporating cumulative angioplasty volume, age, and periodic performance (mortality and morbidity rates) in the definition of stars. In the original paper, we chose not to define star status based on a physician's patient volume as a result of our desire to identify the influence of specific types of peers, rather than just a general peer effect. For example, physicians with higher cumulative volumes in angioplasty may be more likely to adopt stents than lower volume physicians because fixed adoption costs (such as human capital investment) would be recovered more quickly with a higher patient volume (assuming past volume predicts future volume). If this were true, the number of colleagues with high volume could proxy for the overall peer stent adoption frequency, and we would not be able to determine whether the number of high-volume peers raised adoption risk because high-volume peers in particular were influential or because peer adoption in general created a spillover. The stars definition we employed had the advantage (somewhat counterintuitively) of not being a good predictor of early adoption of stents, and we stand behind this logic.

   We do accept the proposition that older stars may be less influential than younger stars, and in this response we divide our stars group into two (based on a residency vintage cutoff) and test for differential influence between younger and older graduates of top residency programs. We are also intrigued by the idea of using...