Development of a prediction method for severe pancreatitis using a nomogram

The identification of severe acute pancreatitis (SAP) is paramount for effective patient management, but the gold standard for diagnosing SAP requires over 48 h of organ failure, which may delay timely treatment. Hence, this study aimed to develop a prediction model for SAP using clinical characteristics, laboratory examinations, and non-contrast computed tomography (CT) signs.

This retrospective study included patients admitted for acute pancreatitis between November 2019 and December 2025. The patients were randomized 7:3 to the training and internal validation sets. Patients from another hospital were included as the external validation set. The patients were grouped according to severity by the Atlanta classification. Selected features were used for the development and evaluation of a nomogram model for SAP prediction. The ROC curve, decision curve, and calibration curve were used for model performance evaluation.

A total of 1,128 patients (703, 300, and 125 patients in training, internal validation, and external validation sets) were included for analysis. The multivariable analysis revealed that age, diabetes, D-dimer, creatinine, serum calcium, WBC, VFR, CRP, decreased SpO₂, ascites, and anterior renal fascia thickening were independently associated with SAP. The nomogram developed from the multivariable analysis demonstrated good performance for SAP prediction (AUC = 0.88, 95% confidence interval: 0.84-0.92) and exhibited excellent calibration abilities. Subgroup analysis among patients with hyperlipidemic or non-hyperlipidemic acute pancreatitis yielded similar results.

A nomogram-based SAP prediction model may improve the capacity for risk stratification of pancreatitis severity. Exploratory subgroup analysis suggested the potential applications in patients with different causes.