Introduction: Infected pancreatic necrosis is associated with high morbidity and mortality and is mandatory for surgical or radiosurgical intervention.

Background: The aim of this study was to construct and validate a Data Mining (DM) to predict infected pancreatic necrosis (IPN).

Methods: All patients who presented with severe acute pancreatitis from January 1996 to December 2007 were reviewed. Presentation data on admission and at 48 hours were collected. Acute Physiology and Chronic Health Evaluation (APACHE) II and Glasgow severity (GS) score were calculated. A Data Mining (Artificial Neural Networks (ANN) and Support Vector Machine SVM)) was created and trained to predict development of IPN and mortality from AP; 25% of the data set was withheld from training and was used to evaluate the accuracy of the DM. Accuracy of the DM in predicting infected pancreatic necrosis was compared with APACHE II and GS scores.

Results: A total of 1460 patients with acute pancreatitis were identified of whom 320 (21.9%) fulfilled the clinical and radiological criteria for severe pancreatitis and 91 patients died (6.2%). Median APACHE II score at 48 hours was 6 (range, 0 to 23). DM (ANN and SVM) was more accurate than APACHE II or GS scoring systems at predicting infected pancreatic necrosis (P < .05 and P < .01, respectively).

Conclusion: A DM was able to predict development of infected necrotizing pancreatitis with considerable accuracy and outperformed other clinical risk scoring systems. Further studies are required to assess its utility in aiding management decisions in patients with severe acute pancreatitis.