Prolonged intensive care unit treatment (> 3 days) contributes to increased health costs and resource utilization. In order to devise strategies to limit intensive care unit stay, and provide cost-effective medical care, it is necessary to identify the pre- and perioperative risk factors of prolonged treatment. Over 100 potential risk variables were collected prospectively in 889 consecutive patients undergoing isolated coronary bypass surgery between 1990 and 1992. The incidence of intensive care unit therapy lasting > 3 days was 6.8%. Univariate statistics identified 23 pre- and perioperative variables that were potential contributors to prolonged intensive care unit therapy. However, multivariate analysis of preoperative risk variables identified only recent myocardial infarction (within 30 days of surgery) and continued preoperative smoking (within 30 days of surgery) to be independent risk factors. Only 6.3% of patients without preoperative myocardial infarction and 6.1% of non-smokers required prolonged intensive care unit treatment, compared with 14.8% of patients with preoperative myocardial infarction (P = 0.01) and 10.1% of smokers (P = 0.07). When multivariate analysis was repeated with both pre- and perioperative variables, only ischemic morbidity (inotropes, myocardial infarction and low-output syndrome; 138 patients) and non-ischemic morbidity (infection, stroke or bleeding; 37 patients) predicted prolonged intensive care unit treatment. Intensive care unit treatment for > 3 days occurred in 26.8% of patients with ischemic morbidity compared with 3.2% of patients without ischemic morbidity (P = 0.001). Prolonged intensive care stay occurred in 32.4% of patients who suffered non-ischemic complications compared with 5.7% of patients who did not suffer these complications. The multiple logistic regression analysis odds ratio for ischemic morbidity was 7.4 (95% c.i. 4.0-13.4) compared with 4.8 (95% c.i. 1.9-10.1) for non-ischemic morbidity. Strategies designed to reduce the incidence of prolonged intensive care unit treatment should include prevention of stroke, infection and bleeding. However, the greatest reduction of intensive care unit utilization would be mediated by prevention of ventricular dysfunction secondary to myocardial ischemia or inadequate myocardial preservation.