Development of Machine Learning Algorithms to Predict Patient Dissatisfaction After Primary Total Knee Arthroplasty

Area of Science:

• Orthopedic Surgery
• Machine Learning in Healthcare
• Patient Outcomes Research

Background:

• Postoperative dissatisfaction following total knee arthroplasty (TKA) leads to increased healthcare costs.
• Predicting dissatisfaction is crucial for optimizing patient care and resource allocation.

Purpose of the Study:

• To develop and validate machine learning algorithms for predicting patient dissatisfaction after primary TKA.
• To identify key preoperative factors associated with TKA dissatisfaction.

Main Methods:

• Retrospective review of 430 TKA patients (2014-2016) from academic and community hospitals.
• Utilized demographics, medical history, and preoperative scores (KSS, KSS-F, health state) as predictors.
• Developed and tested five supervised machine learning algorithms using 10-fold cross-validation and an independent testing set.

Main Results:

• 9.0% of patients reported dissatisfaction at 2-year follow-up.
• The random forest algorithm demonstrated strong predictive performance (c-statistic: 0.77).
• Significant predictors included age, comorbidities, drug allergies, and preoperative patient-reported scores.

Conclusions:

• Machine learning models can effectively predict TKA dissatisfaction using partially modifiable risk factors.
• The developed model offers good discriminative capacity to identify at-risk patients.
• Preoperative health optimization may reduce TKA dissatisfaction rates.