The aim of this study was validate the Glasgow-Blatchford score and the pre-endoscopic Rockall score to assess their ability to predict the presence of active bleeding in emergency department patients with suspected upper gastrointestinal bleeding. We reviewed and extracted data from electronic medical record on patients presenting with a suspicion of acute upper gastrointestinal bleeding at our emergency from January 1, 2012 to December 31, 2012. For each patient we calculated the Glasgow-Blatchford score and the pre-endoscopic Rockall score. Discriminative ability of the scoring systems for predicting active bleeding was evaluated by receiver operator characteristic (ROC) curve analysis. We identified 636 patients with upper gastrointestinal bleeding. There were 118 (18.6%) patients with active bleeding and 520 (81.8%) patients with the need for intervention. The ROC curve analysis showed the poor discriminative ability of the Glasgow-Blatchford score and the pre-endoscopic Rockall score for determining the presence of active bleeding (area under the curve (AUC) = 0.546, 95% confidence interval (CI) 0.490-0.602 vs. 0.576, 95% CI 0.523-0.630, p=0.34). The sensitivity and the specificity of two scoring systems were suboptimal. However, the Glasgow-Blatchford score outperformed the pre-endoscopic Rockall score in predicting the need for clinical intervention (AUC = 0.867, 95% CI 0.831-0.903 vs. 0.698, 95% CI 0.643-0.754; p<0.001). The Glasgow-Blatchford score was superior in predicting the need for intervention in emergency department patients with suspected gastrointestinal hemorrhage. However, these clinical decision rules may be insufficient to predict the presence of active bleeding.