Physician Trainees’ Interactions with the Pharmaceutical Industry

Abstract

The Authors’ Reply—The differences Ramachandran et al. note between our results1 and two other recent publications2,3 may be explained by both the analytic approaches used in our different studies and the outcomes we examined. In their important papers, both King et al. and Epstein et al. analyzed prescribing patterns for psychoactive drugs, using difference-in-differences models that dichotomized medical schools or residency training programs based on their conflict of interest policies alone, and found that those policies were important predictors of subsequent actual drug prescribing decisions. By contrast, we examined the relationship between students’ self-reported receipt of gifts and two characteristics of their medical schools—conflict of interest policies represented by the American Medical Student Association (AMSA) PharmFree Scorecard score and National Institutes of Health (NIH) funding. We found that students’ receipt of gifts correlates strongly with lower levels of NIH funding at their schools, but less so with the schools’ AMSA score after controlling for NIH funding level, suggesting that a higher AMSA score is closely correlated with a school’s receiving above-average NIH funding. As a result, the score may add little further power to account for this behavior beyond the research intensity of the institution. Our findings should not be interpreted to reduce the importance of institutional policies to control conflict of interest issues for trainees. This is still a young area of investigation, and more studies such as these are needed to clarify the best way to implement conflict of interest policies, and the effect of doing so on improving clinical decision-making. These policies emerged in the King et al. and Epstein et al. studies as valid predictors of subsequent rational prescribing behavior on the part of trainees. Despite these salutary downstream effects, our findings also suggest that the reported policies may not completely reflect the day-to-day reality of the medical learning environment that shape trainees. We agree with Ramachandran et al. that meaningful enforcement of conflict of interest policies is essential. Further research can identify better ways to measure implementation of these policies to ensure that resources are optimally directed to interventions that will change trainees’ attitudes and behaviors to benefit patients. Finally, we would like to direct readers’ attention to the revised Figure 1 in an accompanying Erratum. The original Figure contained clerical errors in the regression models.