Development of an Explainable Machine Learning Model to Predict Mortality Risk in Sepsis Patients: Insights From a Real-World Clinical Data.

BACKGROUND

Sepsis is a life-threatening dysregulated host response to infection; early risk stratification is essential to guide intensive care.

OBJECTIVE

To develop and interpret a machine learning model for predicting in-hospital mortality among intensive care unit (ICU) patients with sepsis.

METHODS

We retrospectively analyzed clinical data from sepsis patients admitted to the ICU of the First Affiliated Hospital of Xinjiang Medical University. Missing values were imputed using the mice package, and class imbalance was addressed with the Synthetic Minority Oversampling Technique. Seven models-Random Forest (RF), k -Nearest Neighbors, Support Vector Machine, logistic regression, eXtreme Gradient Boosting, least absolute shrinkage and selection operator-logistic regression, and light gradient boosting machine-were trained and evaluated using area under the receiver operating characteristic curve (AUC), precision-recall curves, and decision curve analysis. Shapley Additive Explanations (SHAP) were applied for global and local interpretation.

RESULTS

RF achieved the best performance (AUC = 0.9816) and provided the highest net clinical benefit. SHAP identified key prognostic variables and illustrated patient-specific risk contributions.

CONCLUSIONS

An RF-based, SHAP-interpretable model offers accurate and explainable mortality risk prediction for septic ICU patients, supporting individualized clinical decision-making.