Summary Background Several risk scoring systems exist for upper gastrointestinal bleed (UGIB). We hypothesised that a modified Glasgow Blatchford Score (mGBS) that eliminates the subjective components of the GBS might perform as well as current scoring systems. Aim To compare the performance of the mGBS to the most widely reported scoring systems for prediction of clinical outcomes in patients presenting with UGIB. Methods Prospective cohort study from 9/2010 to 9/2011. Accuracy of the mGBS was compared with the full GBS, full Rockall Score (RS) and clinical RS using area under the receiver operating characterstics‐curve (AUC). Primary outcome was need for clinical intervention: blood transfusion, endoscopic, radiological or surgical intervention. Secondary outcome was repeat bleeding or mortality. Results One hundred and ninety‐nine patients were included. Median age was 56 with 40% women. Thirty‐two per cent patients required blood transfusion, 24% endoscopic interventions, 0.5% radiological intervention, 0 surgical interventions, 5% had repeat bleeding and 0.5% mortality. Primary outcome: the mGBS (AUC 0.85) performed as well as the GBS (AUC = 0.86, P = 0.81), and outperformed the full RS (AUC 0.75, P = 0.005) and clinical RS (AUC 0.66, P < 0.0001). Secondary outcome: the mGBS (AUC 0.83) performed as well as the GBS (AUC 0.81, P = 0.38) and full RS (AUC 0.69, and outperformed the clinical RS (AUC 0.59, P = 0.0007). Conclusions The modified Glasgow Blatchford Score performed as well as the full Glasgow Blatchford Score while outperforming both Rockall Scores for prediction of clinical outcomes in American patients with upper gastrointestinal bleed. By eliminating the subjective components of the Glasgow Blatchford Score, the modified Glasgow Blatchford Score may be easier to use and therefore more easily implemented into routine clinical practice.