Fig. 2: Feature importance plots on predicting patient dissatisfaction at 2-year follow-up.
From: Identifying who are unlikely to benefit from total knee arthroplasty using machine learning models
SHAP explanation of clinical-data only model for predicting each PROM MCID: a Not achieve KSS MCID. b Not achieve SF-PCS MCID. c Not achieve SF-MCS MCID. d Not achieve OKS MCID. KSS Knee Society knee and function scores, MCID minimal clinically important difference, MCS mental component score, OKS Oxford Knee Score, PCS physical component score, PREOP preoperative, rome terminal flexion, roms terminal extension.
