Machine learning prediction of sepsis in paralytic ileus using interpretable clinical models - PubMed
5 hours ago
- #clinical models
- #machine learning
- #sepsis prediction
- The study developed a machine learning model to predict sepsis in paralytic ileus patients, using clinical data from the MIMIC-IV database.
- Seven predictors were identified: pneumonia, red cell distribution width, heart failure, blood urea nitrogen, atrial fibrillation, serum chloride, and white blood cell count.
- Logistic regression was selected as the final model, showing good performance with AUCs of 0.687 in internal validation and 0.715 in external validation.
- The model demonstrated good calibration and clinical net benefit, with SHAP analysis highlighting pneumonia and RDW as the most influential predictors.
- This interpretable model aims to support early risk stratification and preventive interventions for sepsis in paralytic ileus patients.