Multivariate Model for the Prediction of Severity of Acute Pancreatitis in Children

Abstract

This study aimed to develop a severity prediction system for pediatric patients with acute pancreatitis (AP) based on clinical and laboratory parameters recorded at disease onset. A retrospective cohort study including 130 patients with AP, aged 0 to 18 years, was conducted. Correlations between severe AP (SAP) and clinical and laboratory data were established. Parameters with a significant statistical correlation (P ≤ 0.05) were incorporated in logistic regression models, and receiver operating characteristic curves were generated. The best-performance cutoff points were calculated to propose a severity prediction score, for which sensitivity and specificity were determined. Thirty-eight cases (29.2%) were consistent with SAP. A value of ≥1 point yielded a sensitivity of 81.5% and specificity of 64.1% for SAP prediction, when using a score including blood urea nitrogen ≥12.5 mg/dL (1 point) or hemoglobin <13 mg/dL (1 point) as variables. The proposed severity score showed good performance in predicting SAP.