This presentation draws upon the experience of the O.M.G.E. Multi-national Upper G.I. Bleeding Survey, using data collected during 1980-1982 by 185 clinicians from 44 centres in 21 countries to discuss two questions. First, can prognostic factors be identified in patients presenting to hospital with upper G.I. bleeding, and if so what are they? Second, is it possible - by combining the two technologies of endoscopy and computers - to provide an individual patient with a short-term prognostic prediction sufficiently accurate to affect patient management. Amongst 4,010 patients, a number of clinical factors were found to affect short-term prognosis. These included patient age, previous history of heart or liver disease, confusion and dehydration on admission, jaundice, and ascites. Identification of the bleeding source via endoscopy was shown to aid short-term prognosis - especially in the period of the 2nd to 10th days post-admission. Use of computer analysis enabled "high risk" patients to be defined (of whom 63.8% suffered further bleeding and 30.0% died), and also a comparable "low risk" group (of whom only 4% suffered further bleeding and none died). Finally, "time-dependence" studies have been used to identify a group of patients who (by the 2nd day post-admission) have a residual risk of further bleeding sufficiently low (well under 1%) to suggest that considerable resources can be saved by the judicious use of endoscopy and computer science.