Machine learning models to predict in-hospital mortality in patients with rhabdomyolysis combined with acute kidney injury

Background. Rhabdomyolysis-associated acute kidney injury (RI-AKI) is a serious complication in critically ill patients and is associated with increased in-hospital mortality. However, limited research has focused on predictive modeling of in-hospital mortality among this population. Objective. To develop and evaluate machine learning (ML) models for predicting in-hospital mortality in critically ill patients with RI-AKI. Methods. Data were extracted from the MIMIC-IV and eICU Collaborative Research Databases. Patients with RI-AKI were identified, and relevant clinical variables—including demographics, vital signs, laboratory indicators, comorbidities/complications, and treatments within the first 24 hours of ICU admission—were collected. The combined dataset was randomly divided into training and testing sets in an 8:2 ratio. Feature selection was performed using the least absolute shrinkage and selection operator (LASSO) and random forest (RF). ML models were constructed using Extreme Gradient Boosting (XGBoost), RF, and logistic regression (LR). Model performance was assessed by area under the receiver operating characteristic curve (AUC), Brier score, sensitivity, specificity, and calibration. Results. Ten key predictors, including age, sodium, phosphorus, and coagulation markers, were identified. In the training set, the XGBoost model achieved the highest AUC (0.889; 95% CI: 0.872–0.908), outperforming RF (0.797) and LR (0.740). Brier scores were 0.122, 0.185, and 0.203, respectively. Similar results were observed in the testing set. Conclusions. The XGBoost model demonstrated superior performance in predicting in-hospital mortality among critically ill RI-AKI patients, indicating its potential value in clinical risk stratification. Further external validation is warranted.