In an effort to identify areas of high cost for coronary artery bypass (CAB) surgery, we used logistic regression methods to predict patients at risk for postoperative morbidity. The predictive model was derived from 1567 consecutive isolated CAB cases performed between July 1, 1991 and December 31, 1992. The patients were categorized into 3 groups: Group 1 (N = 756,48%) no complications, Group 2 (N = 560, 36%) minor complications, and Group 3 (N = 251, 16%) major complications with extended stay and/or death. Estimated hospital costs were 3.5 times higher for Group 3 than Group 1, and 2 times higher for Group 2 than Group 1. The following patient factors were found to be significant independent predictors of morbidity, listed in decreasing order of importance: preoperative cardiogenic shock, emergency surgery, severe left ventricular dysfunction, cardiac laboratory induced cardiac collapse, increasing age, diabetes mellitus, high serum creatinine levels, peripheral vascular disease, cardiomegaly, female gender, reoperation, serum albumin levels of less than 4 mg/dl, low body mass index, history of congestive heart failure, atrial arrhythmia, cerebral vascular disease, chronic pulmonary disease, anemia. and blood urea nitrogen levels greater than 29 mg/dl. A cumulative clinical risk score (CRS) was developed by assigning each factor a score of 1 to 7 based on the predictive power. The mean CRS for Groups 1, 2 and 3 were 1.5 ± 4.9, 5.3 ± 3.2 and 9.0 ± 4.5. respectively (p = 0.0001). The model was prospectively validated on 1235 CAB surgery patients from January 1. 1993 to April 30. 1994. Postoperative morbidity was stratified by CRS points as low (0–2). moderate (3–5), high (681 and extremely high (≥9). The validation group morbidity rates fell within the 95% confidence interval of the predicted rates in each category of risk. Preoperative patient variables can predict postoperative morbidity for CAB patients. Significant cost reduction can only be accomplished by reducing postoperative morbidity. The ability to predict morbidity risk provides data for improved patient selection and for development of targeted protocols to reduce specific complications.