A Machine Learning-Based Model for Predicting Rhabdomyolysis in Patients With Sepsis

Background Sepsis is a common condition in the intensive care unit (ICU) and is frequently complicated by rhabdomyolysis, which can lead to serious consequences such as acute kidney injury. Currently, there is a lack of effective tools for the early prediction of sepsis-associated rhabdomyolysis. Methods This retrospective study analyzed 3,782 sepsis patients from the MIMIC-IV database. Three feature selection methods (multivariate analysis, LASSO regression, and the Boruta algorithm) were used to identify predictors. Six machine learning models (logistic regression, decision tree, random forest, XGBoost, support vector machine, and artificial neural network) were developed to predict the occurrence of rhabdomyolysis. Results Ten predictive features were ultimately identified. The XGBoost model demonstrated the best performance, achieving an area under the receiver operating characteristic curve (AUC) of 0.924 (95% CI: 0.907–0.941) in the validation set. SHAP analysis revealed that the CK/CKMB ratio, aspartate aminotransferase (AST), and invasive mechanical ventilation were the most important predictors. Conclusions The machine learning-based prediction model accurately identifies the risk of rhabdomyolysis in sepsis patients, facilitating the early recognition of high-risk individuals and enabling timely interventions.