Prediction of 30-day mortality risk in patients with acute pancreatitis in ICU: a study based on the MIMIC-IV database

Acute pancreatitis (AP) in the intensive care unit (ICU) exhibits rapid progression and high 30-day mortality, necessitating efficient early risk prediction methods. This study analyzed 724 ICU patients from the MIMIC-IV database. Key predictors were screened using LASSO regression, and a multivariate logistic regression model was constructed to evaluate 30-day mortality risk. Twelve key predictors–including blood urea nitrogen (BUN), red blood cell distribution width (RDW), pH, sepsis status, and vasopressor use—were identified. The model demonstrated strong discrimination, with area under the curve (AUC) values of 0.85 and 0.89 across sets, an overall accuracy of 77%, and significantly lower 30-day survival in high-risk groups (60% vs. 97%). The proposed model effectively stratifies 30-day mortality risk in ICU AP patients, offering valuable support for early clinical intervention and personalized management. Not applicable.