Developing and validating machine learning models to predict length of hospitalization before obese patients undergo elective arthroplasty

Introduction: Obese patients undergoing Total Joint Arthroplasty (TJA) have been associated with increased length of hospital stay (LOS) and in-hospital resource utilization. This poses challenges for institutions participating in bundled-payment programs. We investigated whether pre-operative information could predict prolonged hospital stays in obese patients undergoing TJA. Methods: Using the arthroplasty registry of a single large institution, we identified 4563 obese patients who underwent unilateral THA or TKA between 2020 and 2021. No data was missing or incomplete. A total of 31 pre-surgical parameters were included as potential predictor variables. The data was partitioned into training (80 %) and test (20 %) set. For binary modelling, patients were categorized by a LOS less than 2 nights (41.4 %) and a LOS of 2 nights or more (58.6 %). Binary classification model performance was evaluated using area under the receiver operating characteristic curve (AUC-ROC), Matthews correlation coefficient (MCC), F1 score, accuracy, sensitivity, precision, and the Brier score. Regression models were evaluated utilizing Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE). 95 % confidence intervals (95 % CI) were calculated via bootstrapping. Results: The explainable boosted machine (EBM) model was the best performing binary model: F1-Score 0.75 (95 % CI 0.72, 0.78); accuracy 0.69 (0.67, 0.72); sensitivity 0.71 (0.67, 0.75); precision 0.74 (0.7, 0.78), AUC-ROC 0.75 (0.72, 0.79), MCC 0.38 (0.32, 0.44) and Brier Score 0.199. The QRF was the best performing regression model with MAE of 21.5 (20.0, 23.2) and RMSE of 32.3 (28.8, 40.0) hours. Male sex, older age, and not being married were the most important predictors across both models. Correlation of QRF predicted and actual LOS revealed a Spearman Correlation Coefficient of 0.43 (0.36, 0.47). Conclusion: All models demonstrated strong predictive capability, underscoring their clinical relevance. These insights can guide preoperative planning, patient counseling, and resource allocation, helping optimize care and discharge strategies.