Explainable machine learning model predicts mortality in infected pancreatic necrosis

An explainable machine learning model for predicting 90-day mortality in patients with infected pancreatic necrosis (IPN) has shown promising results. The final model, developed from a cohort of 344 patients, achieved a c-index of 0.863 in internal validation and 0.857 in external validation, outperforming nine other models. Key mortality predictors included multiple organ failure and acute physiology scores. The model’s utility is enhanced by two web-based applications, improving clinical decision-making and patient outcomes.