Predictors for in-hospital mortality and need for clinical intervention in upper GI bleeding: a 5-year observational study.

BACKGROUND Upper GI bleeding (UGIB) is a potentially life threatening gastrointestinal emergency whose effective management depends on early risk stratification. METHODS We retrospectively studied 151 patients admitted to our unit with UGIB between 1st January 2007 and 31st December 2011 and in whom we calculated the clinical and complete Rockall, the Glasgow-Blatchford and modified Glasgow-Blatchford risk scores. We performed an analysis of the predictive value of these scores for in-hospital mortality and need for clinical intervention. RESULTS Of the 151 patients enrolled, 68.87% were male, and the mean age was 59.48 years. One in three patients had a history of chronic liver disease and one in eight had a previous episode of UGIB. Clinically, 58.3% of the patients presented with melena, 18.5% with hematemesis and 23.1% with both hematemesis and melena. 22% of cases were variceal hemorrhages and the other non-variceal. 16 patients died during hospitalization. The prognostic accuracy of all four scores for in-hospital death and need for clinical intervention was good, the complete Rockall score having the best performance (AUROC 0.849 and 0.653 respectively). CONCLUSIONS The Rockall and Blatchford scores were good predictors of mortality and need for clinical intervention in our study. The good predictive performance of these scores highlight the need for their use in day-to-day practice to select patients with likelihood of poor clinical outcome.