Related Experiment Video
Updated: Oct 4, 2025
14:56
An Experimental Paradigm for the Prediction of Post-Operative Pain PPOP
Published on: January 27, 2010
21.5K
Machine learning approach to predicting persistent opioid use following lower extremity joint arthroplasty
Rodney A Gabriel1, Bhavya Harjai2, Rupa S Prasad2
1Anesthesiology, University of California San Diego, La Jolla, California, USA ragabriel@health.ucsd.edu.
Regional Anesthesia and Pain Medicine
|February 4, 2022
Summary
Predictive models using ensemble learning and oversampling significantly improve the identification of patients at risk for persistent opioid use after joint arthroplasty. Early detection enables personalized interventions to manage postoperative pain effectively.
Area of Science:
- Orthopedic Surgery
- Data Science
- Pain Management
Background:
- Persistent opioid use after lower extremity joint arthroplasty is a significant clinical challenge.
- Developing accurate predictive models is crucial for identifying at-risk patients.
- Ensemble learning and oversampling techniques may enhance predictive model performance.
Purpose of the Study:
- To develop and compare predictive models for persistent opioid use following joint arthroplasty.
- To evaluate the impact of ensemble learning and Synthetic Minority Oversampling Technique (SMOTE) on model performance.
Main Methods:
- Compared six classification models: logistic regression, random forest, neural network, balanced random forest, balanced bagging, and support vector classifier.
- Utilized preoperative, intraoperative, and postoperative data.
- Employed repeated stratified k-fold cross-validation, calculating F1-scores and AUC, with and without SMOTE.
Main Results:
- 1042 patients were analyzed; 23.2% reported persistent opioid use.
- Ensemble methods outperformed logistic regression; balanced bagging achieved an F1 score of 0.80 and AUC of 0.94.
- SMOTE further improved performance, with balanced bagging reaching an F1 score of 0.84 and AUC of 0.96.
Conclusions:
- Ensemble learning significantly enhances predictive models for persistent opioid use.
- Accurate early identification of high-risk patients facilitates clinical decision-making and personalized interventions.
- Key predictors include early postoperative opioid use, BMI, age, preoperative opioid use, discharge opioid prescription, and hospital stay.
