The need for therapeutic endoscopy in patients with upper GI hemorrhage is important in determining the risk and disposition of these patients. Pre-endoscopic risk scores may be helpful in predicting this need.

To test the Blatchford and pre-endoscopic Rockall scores with the need for therapeutic endoscopy as the primary outcome.

Prospective validation study.

Tertiary-care university-affiliated hospital.

Between January 1, 2006 and February 28, 2007, 1087 patients with upper GI hemorrhage who had undergone an inpatient EGD within 24 hours were entered in the study.

Blatchford and pre-endoscopic Rockall scores were prospectively calculated for all patients, and the need for therapeutic endoscopy was determined during the EGD.

Of the 1087 patients, 297 (27.3%) needed therapeutic endoscopy. The mean Blatchford score for those who needed therapeutic endoscopy was significantly higher (mean [standard deviation]: 10.3 [3.5] vs 7.0 [4.4], P < .001). The area under a receiver-operating characteristic curve was 0.72 (95% CI, 0.68-0.75). A threshold of 0 (low risk) predicted the need for therapeutic endoscopy with 100% sensitivity and 6.3% specificity. Fifty (4.6%) patients were identified as low risk. The pre-endoscopic Rockall score was unable to predict this need.

The decision to perform therapeutic endoscopy is a subjective one, although endoscopists are trained to follow international consensus guidelines.

The Blatchford score is more useful for predicting low-risk patients who do not need therapeutic endoscopy and who may be suitable for outpatient management. A threshold of 0 for low risk should be used. The Rockall score is not helpful in predicting the presence of low-risk lesions.