OBJECTIVETo assess the prognostic accuracy for operative mortality of 4 different risk stratification models in Chinese patients underwent (coronary artery bypass graft) CABG.

METHODSBetween 2002 and 2003, all patients undergoing CABG in our institution were prospectively scored for operative mortality using Parsonnet, EuroSCORE, Cleveland and OPR scoring systems and operative mortality was registered. Operative mortality is defined as postoperative death of any cause during hospitalization. Calibration was estimated using the Hosmer-Lemeshow goodness-of-fit test. Discrimination was evaluated using receiver operating characteristic (ROC) curves and area under a ROC curve (AUC).

RESULTSFollow-up was able to be completed in 2251 prospectively scored patients. Registered operative mortality was 1.87% (42/2251). The overall expected mortality calculated by Parsonnet, EuroSCORE, Cleveland and OPR scoring systems were 5.78%, 2.82%, 3.30% and 1.65%, respectively. The actual operative mortality was within the range of 95% confidence interval of OPR model and the other 3 predicted significantly higher operative mortality. Among the four risk scores, Calibration was good in OPR model (chi(2) = 4.842, P = 20.182) and poor in other 3 models (P < 0.001) while discrimination was acceptable in Parsonnet, Cleveland and OPR scoring systems (AUC: 0.711, 0.754, 0.757, respectively) and excellent in EuroSCORE scoring system (0.813).

CONCLUSIONFor Chinese patients undergoing CABG, OPR scoring system best predicted the operative mortality. All systems could be used to discriminate operative mortality for individual patient.