OBJECTIVE Pediatric pancreatitis remains poorly understood despite increasing incidence and risk of morbidity and mortality. Present predictive scores for severe pediatric acute pancreatitis (AP) are either extrapolated from adults or difficult to use in practice. We aimed to identify laboratory parameters for early prediction of severity of the course of pediatric AP. METHODS A retrospective review of children with AP (January 2000-July 2011) was performed at 2 pediatric hospitals. Predictors of severe AP using laboratory parameters measured within 24 hours of presentation were derived in the cohort from one institution and validated in the other. RESULTS A total of 131 pancreatitis episodes, 73 (34% severe) and 58 (24% severe) in the derivation and validation cohorts respectively, were reviewed. In the derivation cohort, serum lipase was significantly higher in severe versus mild AP (median [interquartile range] 18.1 [9.2-39.1] vs 4.9 (3.2-13.3) × upper limit of normal [ULN]; P = 0.002). Logistic regression analysis in the derivation cohort showed serum lipase ≥7 × ULN to be predictive of severe AP. This finding was confirmed in the validation cohort. Based on the combined derivation and validation data, serum lipase ≥7 × ULN was associated with an odds ratio of 7.1 (95% confidence interval 2.5-20.5; P < 0.001) for developing severe AP. Sensitivity, specificity, positive and negative predictive values, and positive and negative likelihood ratios were 85%, 56%, 46%, 89%, 1.939, and 0.27, respectively. CONCLUSIONS Serum lipase ≥7 × ULN within 24 hours of presentation may be a simple clinical predictor of severe AP in children. Lipase levels below this threshold are strongly associated with a milder course.