SinoSCORE对心脏瓣膜术后院内死亡及并发症的预测价值

Performance of SinoSCORE in predicting in-hospital mortality and postoperative complications after cardiac valve surgery

目的评价中国冠状动脉旁路移植手术风险评估系统(SinoSCORE)对心脏瓣膜术后院内死亡及术后并发症的预测价值。方法回顾性收集2005年至2011年第二军医大学长海医院行心脏瓣膜手术患者的围手术期资料,采用Si-noSCORE模型计算全组患者院内预期死亡风险,并验证SinoSCORE与术后并发症[低心排血量综合征、肾衰竭、肺部感染、使用主动脉内气囊泵(IABP)、延迟拔除气管插管、ICU住院时间延长以及二次手术]发生率间的关系,采用ROC曲线下面积评价模型区分度,以Hosmer-Lemeshow拟合优度检验评价模型校准度,采用Youden指数确定SinoSCORE有较好预测价值的术后并发症的最佳诊断界值。结果入选3 407例患者,平均年龄(49.2±13.3)岁。术后院内死亡率ROC曲线下面积为0.754(95%可信区间为0.701～0.806),提示模型有较好区分度。总体术后院内死亡率为3.05%(104/3 407),SinoSCORE预测院内死亡率为(3.1±0.1)%,Hosmer-Lemeshow拟合优度检验χ2=9.545,P=0.490,提示模型有较好校准度。SinoSCORE对术后低心排血量综合征、肾衰竭及使用IABP有较好的预测价值(ROC曲线下面积分别为0.708、0.711和0.718),最佳诊断界值分别为5.5、7.5和6.0。结论SinoSCORE模型预测中国心脏瓣膜手术患者的术后院内死亡风险效果良好,对术后低心排血量综合征、肾衰竭及使用IABP也有较好的预测作用。

Objective To evaluate the performance of the Chinese system for cardiac operative risk evaluation （SinoSCORE） in predicting the in-hospital mortality and postoperative complications in patients undergoing cardiac valve surgery. Methods The clinical data of patients undergoing valve surgery in Changhai Hospital between 2005 and 2011 were retrospectively analyzed. SinoSCORE was used to calculate the in-hospital mortality risk. The relationship of SinoSCORE result with postoperative complications （low cardiac output syndrome, renal failure, lung infection, application of intra-aortic balloon pump [IABP] , prolonged ventilation, prolonged postoperative ICU stay and reoperation） was verified. Discrimination degree of the model was tested by determining the area under the receiver operating characteristic （ROC） curve, and calibration of the model was evaluated by Hosmer-Lemeshow goodness-of-fit test. The optimal cut-off points for postoperative complications, which could be well predicted by SinoSCORE, were obtained by Youden index. Results The mean age of the 3 407 enrolled patients was （49.2 ± 13.3） years. The area under ROC was 0. 754 （95CI： 0. 701-0. 806）, indicating good discrimination power of the model in predicting in-hospital mortality. The overall in-hospital mortality was 3.05 % （104/3 407）. The predicted in-hospital mortality by SinoSCORE was （3.1 ± 0. 1）%. Hosmer-Lemeshow calibration test yielded Z2 = 9. 454, P = 0. 490, suggesting a high calibration ability of the model. The areas under ROC of low cardiac output syndrome, renal failure, and application of IABP were 0. 708, 0. 711, and 0. 718, respectively, suggesting that SinoSCORE had a satisfactory performance in predicting post-operative low cardiac output syndrome, renal failure, and application of IABP. And the optimal cut-off points for the above three complications predicted by SinoSCORE were 5.5, 7.5, and 6.0, respectively. Conclusion SinoSCORE has a better performance in predicting the in-hospital mortality risk in Chinese patients undergoing valve surgery, and it can betterpredict post-operative low cardiac output syndrome, renal failure, and application of IABP.