Risk stratification scores for predicting mortality in coronary artery bypass surgery

Background: Four risk-stratification scores (RSSs - Euro, French, CCS/Higgins, Parsonnet) were tested as predictors of mortality in coronary artery bypass grafting (CABG) surgery. Methods: From March to April 2000, the perioperative courses of 245 consecutive CABG patients were compared to the predictions according to the RSSs. Sensitivity and specificity were determined with receiver operating characteristics (ROC) curves. Results: CCS/Higgins uses the most easily acquired patient data, and rates emergency conditions as high-risk. Euro, focuses on advanced age and septal rupture. French uses the smallest number of patient parameters and rates rare critical situations as high-risk. Parsonnet is partially based on the physician's subjective assessment of a "catastrophic state," making the scoring arbitrary. All RSSs gave similar (not significant) areas under the ROC curves regarding mortality (Euro 0.826 +/- 0.080, French 0.783 +/- 0.094, CCS/Higgins 0.820 +/- 0.060, Parsonnet 0.831 +/- 0.042). Predicted risk levels for the 11 patients who died differed between the RSSs - Higgins placed these patients in 3 of 5 risk levels with ascending distribution. The other RSSs placed these patients in the highest risk level except for one and two patients, respectively, who were placed in the lowest Euro and French risk level. Euro and Parsonnet placed about half of all patients with non-lethal outcome in the highest risk level. Conclusions: All RSSs satisfactorily estimated the group risk for mortality. No RSS expressed sufficient validity to predict individuals with lethal outcome. In clinical use, CCS/Higgins proved the most practicable.