Design, Implementation, and Clinical Impact of a Machine Learning–Assisted Intervention Bundle to Improve Opioid Prescribing

SummaryOpioid-prescribing patterns, including those after surgery, have been implicated as a significant contributor to the ongoing U.S. opioid crisis. Personalized postoperative opioid prescribing, tailored to a patient's unique characteristics, is now feasible using modern machine-learning techniques and large data sets. However, there are many challenges facing machine learning–based interventions, including data set shift and barriers to converting model predictions into action that delivers clinical impact. A multidisciplinary group at the Beth Israel Deaconess Medical Center and Harvard Medical School built, deployed, and evaluated the impact of a machine learning–assisted intervention aimed at improving the postsurgical opioid prescribing by providers. The intervention consisted of: (1) a machine learning–augmented point-of-care dashboard to guide personalized opioid prescribing, (2) a system for continual automated collection of patient opioid consumption data to adjust for temporal changes in opioid use and to update the machine-learning model, and (3) an automated near–real-time feedback system to promote durable behavior change in surgical opioid prescribers. The authors describe how their group built the infrastructure to automate rapid feedback and data collection and how the group was able to evaluate the effect of an intervention that existed in a context of rapidly changing policy norms and a pandemic that dramatically reduced surgical volumes. Over the course of 2 years, the intervention was associated with a reduced size and reduced frequency of inappropriately large opioid prescriptions, prescribing practices evolved to be closer to typical patient opioid consumption, and the volume of excess pills prescribed dropped significantly. Beth Israel continues to update the intervention with the most recently collected data and user feedback and iterate on the design to properly target surgical opioid prescribers in the health system.