Abstract: Objectives To evaluate the accuracy of four existing risk stratification models including the Society of Thoracic Surgeons(STS) 2008 Cardiac Surgery Risk Models for Coronary Artery Bypass Grafting (CABG), the European System for Cardiac Operative Risk Evaluation (EuroSCORE), the American College of Cardiology/American Heart Association (ACC/AHA) model, and the initial Parsonnet’s score in predicting early deaths of Chinese patients after CABG procedure. Methods We collected clinical records of 1 559 consecutive patients who had undergone isolated CABG in the Fu WaiHospital from November 2006 to December 2007. There were 264 females (16.93%) and 1 295 males (83.06%) with an average age of 60.87±9.06 years. Early death was defined as death inhospital or within 30 days after CABG. Calibration was assessed by the Hosmer-Lemeshow (H-L) test, and discrimination was assessed by the receiveroperatingcharacteristic (ROC) curve. The endpoint was early death. Results Sixteen patients(1.03%) died early after the operation. STS and ACC/AHA models had a good calibration in predicting the number of early deaths for the whole group(STS: 12.06 deaths, 95% confidence interval(CI) 5.28 to 18.85; ACC/[CM(159mm]AHA:20.67deaths, 95%CI 11.82 to 29.52 ), While EuroSCORE and Parsonnet models overestimated the number of early deaths for the whole group(EuroSCORE：36.44 deaths，95%CI 24.75 to 48.14；Parsonnet：43.87 deaths，95%CI 31.07 to 56.67). For the divided groups, STS model had a good calibration of prediction(χ2=11.46, Pgt;0.1)，while the other 3 models showed poor calibration（EuroSCORE：χ2=22.07，Plt;0.005；ACC/AHA：χ2=28.85，Plt;0.005；Parsonnet：χ2=26.74，Plt;0.005）.All the four models showed poor discrimination with area under the ROC curve lower than 0.8. Conclusion The STS model may be a potential appropriate choice for Chinese patients undergoing isolated CABG procedure.