Severe perioperative bleeding after coronary artery bypass grafting (CABG) is associated with poor outcome. An additive score for prediction of severe bleeding was derived (n=2494) and validated (n=1250) in patients from the E-CABG registry. Severe bleeding was defined as E-CABG bleeding grades 2-3 (transfusion of >4 units of red blood cells or reoperation for bleeding). The overall incidence of severe bleeding was 6.4 %. Preoperative anaemia (3 points), female gender (2 points), eGFR <45 ml/min/1.73 m² (3 points), potent antiplatelet drugs discontinued less than five days (2 points), critical preoperative state (5 points), acute coronary syndrome (2 points), use of low-molecular-weight heparin/fondaparinux/unfractionated heparin (1 point) were independent predictors of severe bleeding. The WILL-BLEED score was associated with increasing rates of severe bleeding in both the derivation and validation cohorts (scores 0-3: 2.9 % vs 3.4 %; scores 4-6: 6.8 % vs 7.5 %; scores>6: 24.6 % vs 24.2 %, both p<0.0001). The WILL-BLEED score had a better discriminatory ability (AUC 0.725) for prediction of severe bleeding compared to the ACTION (AUC 0.671), CRUSADE (AUC 0.642), Papworth (AUC 0.605), TRUST (AUC 0.660) and TRACK (AUC 0.640) bleeding scores. The net reclassification index and integrated discrimination improvement using the WILL-BLEED score as opposed to the other bleeding scores were significant (p<0.0001). The decision curve analysis demonstrated a net benefit with the WILL-BLEED score compared to the other bleeding scores. In conclusion, the WILL-BLEED risk score is a simple risk stratification method which allows the identification of patients at high risk of severe bleeding after CABG.