Physician Characteristics and De-Implementation of Ineffective or Unsafe Treatments

Tuesday, June 12, 2018: 3:30 PM
Azalea - Garden Level (Emory Conference Center Hotel)

Presenter: Laura Smith

Co-Authors: Lucas Higuera; Nihar Desai; Bryan Dowd; Alexander Everhart; Jeph Herrin; Anupam Jena; Joseph Ross; Nilay Shah; Aaron Spaulding; Pinar Karaca-Mandic

Discussant: Kelsey Drewry


What influences a physician to discontinue a previously-approved therapy that was subsequently proven ineffective or unsafe? It is not clear how providers modify their prescribing when the clinical literature introduces new evidence on the effectiveness or safety of a treatment. This paper analyzes the physician characteristics associated with the de-implementation of ineffective or unsafe drugs. Using data from the OptumLabs® Data Warehouse (OLDW), a comprehensive, longitudinal, real-world data asset with de-identified claims and clinical information for 2007-2015, we identified cohorts of patients (Medicare Advantage and commercially insured) taking drugs considered ineffective or unsafe focusing on two case studies. First, we considered physicians’ de-implementation of dronedarone use for patients with heart failure or permanent atrial fibrillation. In December 2011, the PALLAS trial provided evidence of safety concerns associated with dronedarone use, which were subsequently followed up by black box safety warnings and REMS by the FDA. Second, we considered de-implementation of fibrate use among type 2 diabetes patients. In April 2010, the ACCORD lipid trial provided evidence that using fibrates in combination with statins was no more effective in reducing cardiovascular events compared to using statins exclusively. In order to identify the physicians responsible for de-implementation following these new bodies of evidence, we used a claims-based attribution algorithm to retrospectively link patients to the physicians most likely responsible for the decision to administer the treatment under consideration. We then obtained characteristics of the attributed physicians from Doximity®, an online social networking site for healthcare professionals. Doximity® provides information on physicians’ sex, their specialty, the number of years since they were in residency, where they went to medical school, and where they went to residency. We estimate interrupted time series models controlling for individual characteristics (age, gender, insurance type, and comorbidities) and the physician characteristics from Doximity® to understand which physicians’ characteristics are associated with de-implementation. In particular, we analyze how trends in drug utilization differ by physician characteristics after the new evidence of unsafety or inefficiency was introduced. Our preliminary findings show differences in de-adoption between physician gender and specialty. Moreover, the differences in de-adoption were heterogeneous between the two case studies, suggesting that physicians react differently to evidence about unsafe medicine compared to evidence about ineffective medicine.