Current risk-stratification systems for patients with acute upper-gastrointestinal bleeding discriminate between patients at high or low risks of dying or rebleeding. We therefore developed and prospectively validated a risk score to identify a patient's need for treatment. Our first study used data from 1748 patients admitted for upper-gastrointestinal haemorrhage. By logistic regression, we derived a risk score that predicts patients' risks of needing blood transfusion or intervention to control bleeding, rebleeding, or dying. From this score, we developed a simplified fast-track screen for use at initial presentation. In a second study, we prospectively validated this score using receiver operating characteristic (ROC) curves--a measure of the validity of a scoring system--and chi2 goodness-of-fit testing with data from 197 patients. We also validated the quicker screening tool. We calculated risk scores from patients' admission haemoglobin, blood urea, pulse, and systolic blood pressure, as well as presentation with syncope or melaena, and evidence of hepatic disease or cardiac failure. The score discriminated well with a ROC curve area of 0.92 (95% CI 0.88-0.95). The score was well calibrated for patients needing treatment (p=0.84). Our score identified patients at low or high risk of needing treatment to manage their bleeding. This score should assist the clinical management of patients presenting with upper-gastrointestinal haemorrhage, but requires external validation.