Background: Prediction of perioperative cardiac complications is important in the medical management of patients undergoing noncardiac surgery. Several indices have been developed to aid prediction, but their performance has not been systematically compared.
Objective: To compare four existing methods for predicting perioperative cardiac risk.
Design: Prospective cohort study.
Setting: Two teaching hospitals in London, Ontario, Canada.
Patients: 2,035 patients referred for medical consultation before elective or urgent noncardiac surgery.
Measurements: Myocardial infarction, unstable angina, acute pulmonary edema, or death. The indices were compared by examining the areas under their respective receiver-operating characteristic (ROC) curves.
Results: Cardiac complications occurred in 6.4% of patients. The area under the ROC curve was 0.625 (95% CI, 0.575 to 0.676) for the American Society of Anesthesiologists index, 0.642 (CI, 0.588 to 0.695) for the Goldman index, 0.601 (CI, 0.544 to 0.657) for the modified Detsky index, and 0.654 (0.601 to 0.708) for the Canadian Cardiovascular Society index. These values did not significantly differ.
Conclusions: Existing indices for prediction of cardiac complications perform better than chance, but no index is significantly superior. There is room for improvement in our ability to predict such complications.