[Development and validation of a prognostic model for elderly patients with acute pancreatitis]

To identify independent predictors of in-hospital mortality in elderly patients with acute pancreatitis (AP) and to develop and validate a nomogram prediction model.

A retrospective cohort study was conducted, including patients aged ≥ 60 years admitted to the Affiliated Hospital of Zunyi Medical University with a diagnosis of AP from January 2015 to December 2024. Based on in-hospital outcomes, patients were divided into survival and death groups and were then randomly allocated to a training set and a validation set in a 7 : 3 ratio. Predictors were initially screened using Lasso regression and subsequently entered into multivariate Logistic regression analysis to identify independent risk factors for constructing the nomogram. Model discrimination, calibration, and clinical utility were evaluated using the receiver operator characteristic curve (ROC curve), calibration curve, and decision curve analysis (DCA), respectively.

A total of 2 569 elderly AP patients were enrolled, with 2 323 survivors and 246 deaths in the hospital. The training set contained 1 801 patients (177 deaths, 9.8%), and the validation set contained 768 patients (69 deaths, 9.0%). Lasso regression identified five candidate variables including age, Ranson score, aspartate aminotransferase (AST), acute respiratory distress syndrome (ARDS), and use of vasoactive agents. Multivariate Logistic regression showed that age [odds ratio (OR) = 1.076, 95% confidence interval (95%CI) was 1.054-1.099, P < 0.001], Ranson score (OR = 1.318, 95%CI was 1.215-1.429, P < 0.001), AST (OR = 1.001, 95%CI was 1.000-1.001, P < 0.001), ARDS (OR = 3.782, 95%CI was 2.495-5.732, P < 0.001), and use of vasoactive agents (OR = 4.850, 95%CI was 3.192-7.370, P < 0.001) were independent predictors of in-hospital mortality. The nomogram prediction model was constructed based on the above five factors, ROC curve analysis shows that, the area under the curve (AUC) was 0.817 (95%CI was 0.784-0.851) in the training set and 0.823 (95%CI was 0.775-0.871) in the validation set, indicating good discriminative ability. Calibration plots demonstrated good agreement between predicted and observed probabilities, and DCA showed favorable net clinical benefit across a wide range of threshold probabilities.

The nomogram incorporating five independent predictors-age, Ranson score, AST, ARDS, and use of vasoactive agents-demonstrated good predictive performance for in-hospital mortality among elderly AP patients. This model provides a practical tool for individualized prognostic assessment and for informing clinical decision-making in this population.