Clinical and endoscopic criteria are already being used to determine GI bleeding outcomes and need for endoscopic therapy. These criteria apply only to upper GI bleeding and require an EGD. We have previously examined, in a retrospective study, objective clinical criteria for an ICU risk determination to be used to predict in-hospital complications for acute upper (UGIB) and lower (LGIB) gastrointestinal bleeding (Crit Care Med 1995;23:1048). METHODS: Consecutive patients with acute UGIB or LGIB from two hospital emergency departments were stratified into high or low risk for negative outcomes, using our BLEED classification (ongoing Bleeding, Low systolic BP, Elevated prothrombin time, Erratic mental status, presence of an unstable comorbid Disease). The presence of one or more BLEED criteria assigned the patient to a "high-risk" categorization. All other patients were classified as low risk. The ER triage physicians were not aware of the BLEED criteria. All patients were followed for outcomes. RESULTS: A total of 465 patients were evaluated over one year (mean age 66.3). The source of the bleeding was upper in 286 patients (61%), lower in 167 patients (36%), small bowel in 3 patients (1%) and indeterminant in 9 (2%). All five elements of the BLEED classification were significantly different between patients having in-hospital complications and those without complications (p≤0.014). High-risk patients (N=209) had significantly greater rates of recurrent GI hemorrhage (Relative RR 4.21; 95% CI, 2.49 to 7.12), surgery to control the source of hemorrhage (RR, 3.37; 95% CI, 1.08 to 10.49) and hospital mortality (RR 14.70; 95% CI, 1.93 to 112.11) compared to low risk patients (N=256). ICU admission (N=126), repeat ICU admission (N=11) and transfused units of RBCs (3.7 ± 3.7) were significantly greater in high-risk group (p=0.001). The BLEED classification also identified a greater incidence of ICU admissions for low-risk patients (RR, 4.21; 95% CI, 2.24 to 7.89) in one of the study hospitals despite no beneficial effect on patient outcomes. CONCLUSIONS: The BLEED classification, applied in the emergency department prior to hospital admission, predicts hospital outcomes for patients with acute upper and lower GI bleeding. This prediction tool may also be used to identify variations in ICU Utilization.