Development and Validation of an Interpretable Machine Learning Model for Predicting Enteral Nutrition-Associated Diarrhea in ICU Patients:
ICU patients receiving enteral nutrition face elevated risks of diarrhea, potentially worsening clinical outcomes. This multicenter study developed and validated a machine learning model to predict such risk.
Enteral nutrition-associated diarrhea (ENAD) poses significant challenges in intensive care units, where critically ill patients depend on nutritional support for recovery. Early identification of at-risk individuals is crucial for mitigating adverse clinical outcomes and reducing mortality risks associated with this complication.
This retrospective multicenter study aimed to develop and externally validate an interpretable machine learning model capable of predicting ENAD risk in critically ill patients. The research team utilized data from the First Affiliated Hospital of Jinzhou Medical University between January 2024 and December 2025, supplemented by an independent external validation cohort from Shangrao People's Hospital during the same timeframe.
Twelve machine learning models were developed using R software to evaluate predictive performance. Model efficacy was assessed through multiple metrics including area under the curve (AUC), accuracy, precision, negative predictive value (NPV), recall, and F1 score. Calibration curves and decision curve analysis (DCA) further evaluated model calibration and clinical utility. SHAP analysis provided insights into feature importance across the models.
Among the twelve models tested, random forest demonstrated superior discriminative performance with an AUC of 0.811 (95% CI: 0.769–0.853) in the test set and 0.808 (95% CI: 0.759–0.856) in the external validation cohort. The externally validated random forest model showed acceptable discriminative performance and provided an interpretable assessment of the contributions of individual features.
The study concludes that the externally validated random forest model may support static early risk stratification of ENAD among critically ill patients receiving enteral nutrition. However, the available evidence remains insufficient to support its routine clinical implementation. Further validation in prospective studies, independent populations, and longitudinal clinical datasets is required before broader application.