Predictive Value of Machine Learning for Mortality Risk in Acute Pancreatitis

Background: Acute pancreatitis (AP), a common acute abdominal disease, has a high mortality rate in severe cases. Accurate mortality prediction is crucial for clinical decision-making. Machine learning (ML) models have shown potential in predicting AP mortality, aiding clinicians in understanding prediction mechanisms and formulating personalized treatment plans. Objective: This study evaluates and compares the performance of ML models in predicting early mortality in AP patients to provide evidence for mortality prediction and guide the development of clinical prediction tools. Methods: A comprehensive search of PubMed, Web of Science, Cochrane Library, and Embase databases was conducted for literature published between January 1, 2012, and April 25, 2025. Effect sizes were synthesized using a random-effects model, and subgroup analyses were performed based on model characteristics to explore result heterogeneity. Results: Twenty-four studies were included. ML models demonstrated high predictive performance for AP mortality risk. In training sets, the pooled C-index was 0.84 (95% CI: 0.81-0.86), with 0.841 (95% CI: 0.806-0.877) for in-hospital mortality. External validation sets showed a pooled C-index of 0.84 (95% CI: 0.82-0.86) and in-hospital mortality prediction of 0.826 (95% CI: 0.798-0.855). ML models outperformed traditional scoring tools (pooled C-index: 0.754, 95% CI: 0.734-0.775 for standard systems). Common predictors included age, blood urea nitrogen, total bilirubin, white blood cells, hemoglobin, blood pressure, and respiratory rate. Conclusions: Machine learning demonstrates excellent accuracy in predicting the mortality of AP. This offers a reference for updating or creating a simple clinical prediction tool.