A prognostic tool to predict severe acute pancreatitis in pediatrics.

BACKGROUND/OBJECTIVES: Approximately 15-20% of pediatric patients with acute pancreatitis (AP) develop severe disease. Severity scoring tools were developed for adult patients, but have limitations when applied in children. We aimed to identify early predictors of severe acute pancreatitis (SAP) on hospital admission for early risk stratification of patients. METHODS: Retrospective review of AP admissions was conducted. The derivation cohort included cases at Cincinnati Children's Hospital Medical Center (CCHMC) between 2009 and 2013. Clinical data collected during the first 24 h of admission were analyzed and a predictive model was derived through statistical analysis. The performance of the model was evaluated in a validation cohort from 2 more institutions other than CCHMC. RESULTS: In the derivation cohort 19% of the 284 admissions were SAP. A generalized linear mixed effect model analysis revealed that lipase, albumin and white blood count (WBC) play a role in the development of SAP (area under the receiver operating curve (AUROC 0.76)). In the validation cohort of 165 AP cases, SAP ranged from 8 to 20% at the three institutions. Performance of the model in this cohort was comparable to the derivation model (AUROC 0.77). There were 369 encounters in the combined derivation and validation pool (AUROC 0.76). CONCLUSIONS: The prognostic severity tool with 3 variables (lipase, albumin, and WBC) obtained within 24 h of admission can be applied to predict SAP in pediatric patients.