Severity scores are used to predict the outcome of acute pancreatitis (AP). Several scores are used in adult patients, but none has been thoroughly validated for specific use in paediatric patients. We retrospectively collected data from 48 children with AP (13 severe and 35 mild). The main causes were trauma (23%), idiopathic (23%), lithiasis (12.5%), and virus (10.5%). We evaluated 3 clinical scores (Ranson, Glasgow modified, and DeBanto) and Balthazar computed tomography severity index. The clinical scores had a good specificity (approximately 85%) but a low sensitivity (approximately 55%) in predicting the severity of paediatric AP. The radiological score is better (sensitivity 80%, specificity 86%). The area under the receiver operator characteristic curve was 0.699 (95% CI 0.508%-0.891%, P = 0.054) for the DeBanto score, 0.846 (95% CI 0.69%-1%, P = 0.001) for the Ranson score, and 0.774 (95% CI 0.584%-0.964%, P = 0.008) for the Glasgow and 0.898 (95% CI 0.73%-1%, P = 0.011) for the Balthazar computed tomography severity index score. In our paediatric cohort, the severity of AP was best predicted by Balthazar computed tomography-based scoring scale. Our results confirm previously reported low sensitivity of adult-based clinical scoring scales.