Clinical decision support system based on an artificial neural network for prediction of infected pancreatic necrosis

s / Pancreatology 15 (2015) S1eS141 S64 data. Only limited data is available regarding validation among different patient populations. Aims: To evaluate the accuracy of the BISAP in predicting severe clinical outcomes in AP. Patientsm creatinine 2mg/dl; systolic blood pressure 90mmHg or need for vasopressor), its persistence ( 48h), pancreatic necrosis and early ( 7d) and late mortality ( 8d). Accuracy was assessed separately according to etiology (lithiasic, alcoholic and non-lithiasic). Statistical analysis: SPSSv21.0. Results: 198 patients, 69.2% male, mean age 58.5±19.0 years. 56 patients (28.3%) were admitted to ICU and 19 patients (9.6%) died. BISAP was accurate in predicting pancreatic necrosis (AUC 0.74, p 2 (sensibility 78.9%; specificity 86.7%). Separately, only in lithiasic and non-lithiasic pancreatitis could BISAP predict, with fair accuracy, persistent organ dysfunction (AUC 0.725 and 0.745, p1⁄40.03 respectively). Conclusion: Overall, BISAP demonstrated high accuracy in all outcomes. This score is simple and easily allowing for early assessment of prognosis.