Application of Machine Learning Algorithms in the Prediction of Prolonged Postoperative Opioid Use in Patients Undergoing Elective Shoulder Arthroscopy (215). Issue 10 (31st October 2021)

• Record Type: Journal Article
• Authors: Lu, Yining; Lavoie-Gagne, Ophelie; Karhade, Aditya; Forlenza, Enrico; Schwab, Joseph; Cole, Brian; Verma, Nikhil; Forsythe, Brian
• Is Part Of: Orthopaedic journal of sports medicine. Volume 9:Issue 10(2021)Supplement 5
• Publication Date: 2021-10-31
• Subjects: Sports medicine -- Periodicals; Orthopedics -- Periodicals; Arthroscopy -- Periodicals; Arthroplasty -- Periodicals; Knee -- Surgery -- Periodicals; 616.7
• DOI: 10.1177/2325967121S00323
• Languages: English
• ISSNs: 2325-9671

Abstract

Objectives: Recovery after shoulder arthroscopy can be impaired by sustained postoperative opioid use, yet there are few validated risk calculators for this outcome. The purpose of this study is to develop and validate a machine learning algorithm that can reliably and effectively predict sustained opioid use in patients following elective shoulder arthroscopy. Methods: A retrospective review of an institutional outcomes database was performed at a tertiary academic medical center to identify adult patients who underwent shoulder arthroscopy between January 1, 2014 and October 1, 2019. Extended postoperative opioid consumption was defined as opioid consumption at least 150 days following surgery. Five machine learning algorithms were developed to predict this outcome. Performance of the algorithms were assessed through discrimination, calibration, and decision curve analysis. Results: Overall, of the 1504 patients included, 132 (8.8%) demonstrated sustained postoperative opioid consumption. The factors determined for prediction of prolonged postoperative opioid prescriptions were preoperative opioid use, age, median percentage of population in patient zip code living at the federal poverty line, preoperative pain score, duration of symptoms, BMI. The random forest model achieved the best performance based on discrimination (AUC = 0.78), calibration, and decision curve analysis. This model was integrated into a web-based open-access application able to provide both predictions and explanations. Conclusions: If externally validated in independent populations, the algorithm developed presently could effectively guide preoperative screening in patients at high risk for extended postoperative opioid prescriptions. Early identification and interdisciplinary counseling in high-risk cases can optimize both resource allocation and surgical outcomes.