Construction and internal validation of a machine learning model for predicting 30-day readmission after hip surgery

Abstract

Objective To develop and internally validate a machine learning (ML)-based model for predicting 30-day postoperative readmission after hip surgery using multidimensional perioperative data, and to evaluate its potential clinical utility. Methods This single-center retrospective cohort study included 720 patients who underwent hip-related surgery at Guangxi Zhuang Autonomous Region People’s Hospital between 2022 and 2025. Patients were randomly divided into training and test sets at a 7:3 ratio. Demographic characteristics, comorbidities, laboratory variables, anesthesia and temperature-management variables, surgical characteristics, and transfusion-related variables were extracted. Candidate predictors were first selected in the training set using the Boruta algorithm. Eleven base models and one stacking ensemble model were then developed using the selected features. Hyperparameters were optimized using 5-fold cross-validation combined with grid search. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC) and the area under the precision-recall curve (PRAUC), while decision curve analysis (DCA) and SHapley Additive exPlanations (SHAP) were used to assess clinical net benefit and model interpretability. Results The overall 30-day readmission rate was 4.3%. In the test set, the XGBoost model achieved an AUC of 0.88 and a PRAUC of 0.53. Across commonly used threshold probabilities, XGBoost provided a higher net benefit than treat-all or treat-none strategies. SHAP analysis identified preoperative albumin, post-anesthesia care unit (PACU) temperature, warming duration, age, and heart failure as the leading predictors of readmission risk. Conclusion The XGBoost-based ML model showed potential for predicting 30-day readmission after hip surgery and may provide supportive information for perioperative risk stratification and modifiable management strategies, particularly nutritional optimization and temperature management.