Over a 2-year period from January 1, 1992, to December 31, 1993, of 2,221 patients undergoing cardiac operations in our unit, 85 (3.8%) were reopened for the control of bleeding (9 patients more than once). The incidence of resternotomy in coronary cases was 2.3%, but resternotomy was more than three times as likely in valve cases (odds ratio, 3.4; 95% confidence interval, 2.1 to 5.4). Previous cardiac operation was more common among resternotomy patients than among the remainder (18% versus 9%, respectively; p = 0.018). An identifiable source of bleeding was found in 57 of the 85 patients (67%), but a concurrent coagulopathy was common (45 patients). Resternotomy patients, as a group, had higher preoperative risk scores (Parsonnet) than did the other patients (p < 0.0001), stayed longer in the intensive care unit (p < 0.0001), and had greater requirements for intraaortic balloon counterpulsation (14% versus 3%) and hemofiltration (9% versus 3%) (p < 0.0001 and p < 0.01, respectively). Nineteen resternotomy patients (22%) died in the hospital, a proportion significantly greater than the risk assigned to this group of patients preoperatively (12.8%) (p = 0.008). In contrast, the observed mortality for the other 2,136 patients (5.5%) was significantly less (8.3%) (p < 0.00006). Multiple forward stepwise logistic-regression analysis confirmed resternotomy for excessive bleeding after cardiac operation to be a significant independent predictor of a prolonged stay in the intensive care unit (p < 0.0001), the need for intraaortic balloon counterpulsation (p < 0.0001), and death (p < 0.0001).