Performance of the Glasgow-Blatchford score in predicting clinical outcomes and intervention in hospitalized patients with upper GI bleeding

Background: Data regarding the utility of the Glasgow-Blatchford bleeding score (GBS) in hospitalized patients with upper GI hemorrhage are limited. Objective: To evaluate the performance of the GBS in predicting clinical outcomes and the need for interventions in patients with upper GI hemorrhage. Design: Prospective observational study. Setting: Single, tertiary-care endoscopic center. Patients: Between July 2010 and July 2012, 888 consecutive hospitalized patients managed for upper GI hemorrhage were entered into the study. Intervention: GBS and Rockall scores. Main Outcome Measurements: GBS and Rockall scores were prospectively calculated. The performance of these scores to predict the need for interventions and outcomes was assessed by using a receiver operating characteristic curve. Results: Endoscopy was performed in 708 patients (80%). A total of 286 patients (40.3%) required endoscopic therapy, and 29 patients (3.8%) underwent surgery. GBS and post-endoscopy Rockall scores (post-E RS) were superior to pre-endoscopy Rockall scores in predicting the need for endoscopic therapy (area under the curve [AUC] 0.76 vs 0.76 vs 0.66, respectively) and rebleeding (AUC 0.71 vs 0.64 vs 0.57). The GBS was superior to Rock-all scores in predicting the need for blood transfusion (AUC 0.81 vs 0.70 vs 0.68) and surgery (AUC 0.71 vs 0.64 vs 0.51). Patients with GBS scores <= 3 did not require intervention. Limitations: Subjective decision making as to need for endoscopic therapy and blood transfusion. Conclusion: Compared with post-E RS, the GBS was superior in predicting the need for blood transfusion and surgery in hospitalized patients with upper GI hemorrhage and was equivalent in predicting the need for endoscopic therapy, rebleeding, and death. There are potential cutoff GBS scores that allow risk stratification for upper GI hemorrhage, which warrant further evaluation.