Patient-Reported Outcomes for Function and Pain in Total Knee Arthroplasty Patients

Area of Science:

• Orthopedic surgery outcomes
• Health informatics
• Predictive analytics in healthcare

Background:

• Postoperative recovery after total knee arthroplasty (TKA) varies significantly among patients.
• Some patients experience challenges in functional recovery and pain management at home following TKA.

Purpose of the Study:

• To develop predictive models using traditional statistics and machine learning.
• To identify patients at higher risk for increased care needs concerning function and pain post-TKA.

Main Methods:

• Included 201 patients undergoing TKA.
• Utilized Oxford Knee Score (function and pain subcomponents) changes from baseline.
• Applied classification (random forest, stochastic gradient boosting) and regression (support vector machine, stochastic gradient boosting) modeling.
• Evaluated models using metrics like RMSE, MAE, R-squared, and AUC.

Main Results:

• Machine learning models, particularly random forest and stochastic gradient boosting, showed strong performance in classification.
• Support vector machine and stochastic gradient boosting excelled in regression modeling.
• Predictive accuracy was higher for functional challenges than for pain management challenges.

Conclusions:

• Routine clinical data contains valuable information for predicting TKA patient outcomes.
• Developed models show potential for identifying patients needing enhanced care for function and pain.
• Further improvements are necessary for clinical implementation of these predictive models.