The proportion of high-risk patients undergoing cardiac surgery has increased steadily over the last two decades. Many of those patients have a catastrophic postoperative course and use hospital resources in a proportion that largely outweighs their number. Consequently, the appropriateness of invasive and intensive interventions in those patients has been questioned. If futility of care were predictable preoperatively, cardiac surgery would probably be denied to many high-risk patients. Logistic regression has been used to develop many complex predictive models to identify high-risk patients and predict their outcome; however, those models do not provide much more discrimination than clinical judgment alone. Moreover, with continuous improvement in medical care all risk models lose their calibration over time. As a result, they often overestimate the probabilities of poor outcome in the individual patients. Many high-risk cardiac surgical patients require a prolonged stay in the intensive care unit (ICU). The analysis of small cohorts of patients who had a prolonged postoperative stay in the ICU shows that 50% and 40% of them are still alive at 1- and 2-year follow-up, respectively; and most survivors report a good quality of life. Considering the limitations of predictive risk models and the satisfaction of cardiac surgical patients who survive after a prolonged ICU stay, it is reasonable to recognize that cardiac surgery should rarely be denied to high-risk patients unless technically unfeasible, and clinical predictions should have only a marginal role in the decision to operate on those patients.