Artificial neural networks accurately predict mortality in patients with nonvariceal upper GI bleeding

• Published in: Gastrointestinal endoscopy 2011-02, Vol.73 (2), p.218-226.e2
• Main Authors: Rotondano, Gianluca, MD, FASGE, Cipolletta, Livio, MD, Grossi, Enzo, MD, Koch, Maurizio, MD, Intraligi, Marco, MD, Buscema, Massimo, PhD, Marmo, Riccardo, MD
• Format: Article
• Language: English
• Subjects: Aged; Biological and medical sciences; Digestive system. Abdomen; Endoscopy; Female; Follow-Up Studies; Gastroenterology and Hepatology; Gastroenterology. Liver. Pancreas. Abdomen; Gastrointestinal Hemorrhage - diagnosis; Gastrointestinal Hemorrhage - mortality; Humans; Investigative techniques, diagnostic techniques (general aspects); Italy - epidemiology; Male; Medical sciences; Neural Networks (Computer); Other diseases. Semiology; Prognosis; Prospective Studies; Risk Factors; Severity of Illness Index; Stomach. Duodenum. Small intestine. Colon. Rectum. Anus; Survival Rate
• ISSN: 0016-5107; 1097-6779
• DOI: 10.1016/j.gie.2010.10.006

Background Risk stratification systems that accurately identify patients with a high risk for bleeding through the use of clinical predictors of mortality before endoscopic examination are needed. Computerized (artificial) neural networks (ANNs) are adaptive tools that may improve prognostication. Objective To assess the capability of an ANN to predict mortality in patients with nonvariceal upper GI bleeding and compare the predictive performance of the ANN with that of the Rockall score. Design Prospective, multicenter study. Setting Academic and community hospitals. Patients This study involved 2380 patients with nonvariceal upper GI bleeding. Intervention Upper GI endoscopy. Main Outcome Measurements The primary outcome variable was 30-day mortality, defined as any death occurring within 30 days of the index bleeding episode. Other outcome variables were recurrent bleeding and need for surgery. Results We performed analysis of certified outcomes of 2380 patients with nonvariceal upper GI bleeding. The Rockall score was compared with a supervised ANN (TWIST system, Semeion), adopting the same result validation protocol with random allocation of the sample in training and testing subsets and subsequent crossover. Overall, death occurred in 112 cases (4.70%). Of 68 pre-endoscopic input variables, 17 were selected and used by the ANN versus 16 included in the Rockall score. The sensitivity of the ANN-based model was 83.8% (76.7-90.8) versus 71.4% (62.8-80.0) for the Rockall score. Specificity was 97.5 (96.8-98.2) and 52.0 (49.8 4.2), respectively. Accuracy was 96.8% (96.0-97.5) versus 52.9% (50.8-55.0) ( P < .001). The predictive performance of the ANN-based model for prediction of mortality was significantly superior to that of the complete Rockall score (area under the curve 0.95 [0.92-0.98] vs 0.67 [0.65-0.69]; P < .001). Limitations External validation on a subsequent independent population is needed, patients with variceal bleeding and obscure GI hemorrhage are excluded. Conclusion In patients with nonvariceal upper GI bleeding, ANNs are significantly superior to the Rockall score in predicting the risk of death.