Predictors of outcome in massive upper gastrointestinal hemorrhage

We reviewed 100 consecutive cases of massive upper gastrointestinal hemorrhage (UGIH). The criteria for inclusion were a decrease in hematocrit greater than or equal to 6%, unstable vital signs, and greater than or equal to units of blood transfused (16 +/- 18 units, mean +/- SD). A multiple regression analysis of 96 variables was employed to determine the most accurate predictors of outcome. The overall mortality was 35%. Hospital status (whether the patient was an inpatient or outpatient when the UGIH began) showed a striking association with mortality (70% for inpatients vs. 22% for outpatients, p less than 0.001). Nonsurvivors also had a greater number of life-threatening diseases than survivors (1.4 +/- 1.1 vs. 0.3 +/- 0.5, p less than 0.001) and greater transfusion requirements (27 +/- 20 units vs. 10 +/- 13 units, p less than 0.001). Age, the presence of cirrhosis, and recent excessive alcohol intake were not important risk factors. At presentation, the most reliable predictor of a fatal outcome was the brevity of the interval between the onset of bleeding and the initiation of a medical work-up. The primary predictor when considering the entire hospitalization was the number of life-threatening diagnoses. Our data indicate that stratification for hospital status and for other potentially predictive risk factors should be incorporated in future trials of therapy for UGIH.