The efficacy of prophylaxis against stress ulcers in preventing gastrointestinal bleeding in critically ill patients has led to its widespread use. The side effects and cost of prophylaxis, however, necessitate targeting preventive therapy to those patients most likely to benefit. We conducted a prospective multicenter cohort study in which we evaluated potential risk factors for stress ulceration in patients admitted to intensive care units and documented the occurrence of clinically important gastrointestinal bleeding (defined as overt bleeding in association with hemodynamic compromise or the need for blood transfusion). Of 2252 patients, 33 (1.5 percent; 95 percent confidence interval, 1.0 to 2.1 percent) had clinically important bleeding. Two strong independent risk factors for bleeding were identified: respiratory failure (odds ratio, 15.6) and coagulopathy (odds ratio, 4.3). Of 847 patients who had one or both of these risk factors, 31 (3.7 percent; 95 percent confidence interval, 2.5 to 5.2 percent) had clinically important bleeding. Of 1405 patients without these risk factors, 2 (0.1 percent; 95 percent confidence interval, 0.02 to 0.5 percent) had clinically important bleeding. The mortality rate was 48.5 percent in the group with bleeding and 9.1 percent in the group without bleeding (P < 0.001). Few critically ill patients have clinically important gastrointestinal bleeding, and therefore prophylaxis against stress ulcers can be safely withheld from critically ill patients unless they have coagulopathy or require mechanical ventilation.