PHYSICIAN PEER NETWORKS AND PATTERNS OF OPIOID-RELATED BEHAVIOR

Abstract

Physicians are known to learn prescribing behavior from peers, although the extent and magnitude of peer influence on opioid prescribing is not well understood. Identifying the role peer networks play on influencing opioid prescribing, or opioid-related behavior, could elicit new understandings on how information in healthcare is spread and, in turn, lead to policy solutions and interventions to modify physician behavior in the direction of evidence-based medicine. The goal of this dissertation was to evaluate physicians prescribing opioids to patients in Medicare, or physicians receiving opioid industry payments, in order to determine if network-level characteristics are associated with patterns in opioid prescribing. This dissertation has three aims: (1) to determine whether patterns in opioid prescribing exist across physician networks and association with specialties, (2) to empirically demonstrate influence industry can have on clinical decision-making via targeted marketing within provider networks, and (3) to attempt to parse whether certain physicians with greater peer influence result in similar opioid prescribing among network peers. There are several findings and important implications related to this work. First, I find that primary care physicians who have more peer connections and more peers within a pain management specialty or surgery are more likely to have a higher median opioid prescribing rate across patient-sharing, hospital, and shared group clinic networks. Second, I find physicians who have any opioid payments are associated with three times the likelihood of at least one peer also having an opioid payment compared to physicians who did not have a similar payment. These physicians are more likely to belong to smaller and more interconnected patient-sharing networks. Finally, I perform a novel identification analysis of potential peer influencers to find certain provider-level characteristics that may shape peer prescribing behavior. The implications of this dissertation reveal that peer influence may serve as a potential mechanism for altering prescribing behavior and may be a lower-cost and efficacious way to increase adherence to evidence-based medicine.