Development and validation of machine learning algorithms for postoperative opioid prescriptions after TKA

Area of Science:

• Orthopedic Surgery
• Data Science
• Pharmacology

Background:

• Prolonged opioid prescriptions after total knee arthroplasty (TKA) represent a significant clinical challenge.
• Identifying patients at high risk for extended opioid use is crucial for effective pain management and mitigating opioid misuse.
• Current methods for predicting this outcome have limitations.

Purpose of the Study:

• To develop and evaluate machine learning algorithms for the preoperative prediction of prolonged opioid prescriptions following TKA.
• To identify key clinical and demographic variables that contribute to the prediction of this adverse outcome.

Main Methods:

• Development of five distinct machine learning algorithms to predict prolonged postoperative opioid prescriptions.
• Utilized a dataset including patient demographics, medical history, and preoperative medication information.

Main Results:

• The stochastic gradient boosting (SGB) model demonstrated superior performance in predicting prolonged opioid prescriptions.
• Significant predictive factors identified include age, prior opioid use, marital status, diabetes diagnosis, and specific preoperative medications.

Conclusions:

• The developed SGB algorithm offers a promising tool for the preoperative identification of TKA patients at elevated risk for prolonged postoperative opioid use.
• This predictive capability can inform targeted interventions and improve patient management strategies.