We have developed a risk stratification score to identify cardiac surgical patients at higher risk of severe postoperative bleeding to aid a decision of whether to use a specific intervention preoperatively. We prospectively created a database of 11592 consecutive patients, who underwent cardiac surgery with cardiopulmonary bypass. An adverse outcome was formally defined as a mean blood loss exceeding 2 ml kg(-1)h(-1) measured between arrival in the intensive care unit (ICU) and the earliest of the elapse of 3h; the start of transfusion of any one of fresh-frozen plasma, platelets or cryoprecipitate; return to theatre or death. Univariate and multivariate associations of severe postoperative bleeding with patient characteristics, clinical features and procedure details were analysed on a development set. The final risk stratification scheme was then evaluated on a test set. Severe postoperative bleeding was associated with urgent or emergency surgery, surgery that was not coronary artery bypass grafting or single valve surgery, presence of aortic valve disease, low body mass index and older age. A risk stratification score was constructed from the above variables to define preoperative categories that demonstrated high, medium and low risk of severe postoperative bleeding. Patients deemed to be at high, medium and low risk by our preoperative scoring had a 21% (95% confidence interval: 18-24%), 8% (7-10%) and 3% (2-4%) rate of severe postoperative bleeding, respectively, within the test set. We have developed a simple risk stratification score that can separate, preoperatively, patients into risk groups with markedly different rates of severe postoperative bleeding.