Data science-driven analysis of opioid use trajectories among opioid-naïve individuals following arthroplasty

The National Institute on Drug Abuse awarded The Trustees of the University of Pennsylvania $100,228 on January 14, 2027, under the Drug Use and Addiction Research Programs (CFDA 93.279) to conduct data science-driven analysis of opioid use trajectories among opioid-naive individuals following arthroplasty.

The recipient will develop a prediction model to identify patients at risk for new persistent opioid use (NPOU)—prolonged opioid exposure beyond the typical three-month post-operative recovery period—and determine the extent to which NPOU reflects unhealthy opioid use including misuse, abuse, and addiction. The analysis will examine electronic health records from a large academic health system serving diverse sociodemographic populations using machine learning and natural language processing methods to characterize post-operative opioid use trajectories and identify patient characteristics associated with unhealthy use. The research aims to generate insights for preventing transition from acute to chronic opioid use and inform safe perioperative opioid practices among opioid-naive patients. Work is performed in Philadelphia, Pennsylvania, with performance ending September 30, 2028. The award funds structured training activities and an interdisciplinary mentorship team consisting of experts in medicine, nursing, health informatics, and computer science.

Abstract

Optimizing safe perioperative opioid use is a critical priority. Despite being intended for short-term use, a significant number of patients who are first exposed to opioids (opioid-naïve) following surgery continue using them longer than the typical recovery period (i.e., 3MONTHS). This prolonged exposure is termed new persistent opioid use (npou). To date, the existing literature has heavily focused on identifying pre-existing risk factors. However, to better inform practice and facilitate actionable interventions, there is a critical need to discern which patients are at higher risk and characterize trajectories of post-operative opioid use that develop into unhealthy opioid use (i.e., misuse, abuse, and addiction). Thus, the proposed work has two specific aims: (1) to develop an optimal prediction model to identify patients at risk for npou, and (2) to determine the extent to which npou reflects unhealthy opioid use and identify patient characteristics associated with such use. To achieve these aims, we will analyze electronic health records from a large academic health system that serves diverse sociodemographic populations using state-of-the-art analytic methods, including machine learning and natural language processing. Complementing the proposed training plan, the overall objective is to promote the safe perioperative use of opioids among opioid-naïve patients by generating insights to prevent the transition from acute to chronic opioid use, and to discern chronic opioid use that reflects unhealthy use. This proposed research and training will be strongly supported by structured training activities, an interdisciplinary mentorship team consisting of experts in medicine, nursing, health informatics, and computer science, as well as robust research resources from the university of pennsylvania. Supporting nida's strategic priority of preventing prolonged drug use, addiction, and related adverse consequences through data science, the proposed research training proposal will provide the applicant an essential foundation for a sustained research career focused on identifying mechanisms underlying the continuum from initial opioid use to addiction, informing the development of novel, effective, and timely interventions to mitigate opioid-related harms, specifically among the large and growing population of opioid-naïve patients undergoing arthroplasty.