EuroSCORE模型对心瓣膜手术患者死亡风险的预测

Validation of European System for Cardiac Operative Risk Evaluation in Chinese Heart Valve Surgery

目的评价欧洲心脏手术风险评估系统(European System for Cardiac Operative Risk Evaluation,EuroSCORE)模型预测行心脏瓣膜手术患者在院死亡率的准确性。方法收集1998年1月至2008年12月于第二军医大学长海医院因心脏瓣膜疾病行外科治疗4 155例患者的临床资料,其中男1 955例,女2 200例;年龄45.90±13.64岁。先按additive及logistic EuroSCORE两种方法评分,将患者分为低风险(n=981)、中风险(n=2 492)、高风险(n=682)3个亚组,比较全组及各亚组患者的实际与预测死亡率。模型预测的校准度用Hosmer-Lemeshow卡方检验,预测的鉴别度采用受试者工作特征(receiver operating characteristic,ROC)曲线下面积检验。结果 4 155例患者在院死亡205例,实际在院死亡率4.93%;additive EuroSCORE预测死亡率为3.80%,而logisticEuroSCORE为3.30%;提示两种评分方法均低估了实际在院死亡率(χ2=11.13,44.34,P<0.05)。additiveEuroSCORE对高风险亚组在院死亡预测校准度较高(χ2=3.61,P=0.31),但对低风险亚组(χ2=0.00,P<0.01)及中风险亚组(χ2=14.72,P<0.01)较低;而logistic EuroSCORE对低风险亚组(χ2=1.66,P=0.88)及高风险亚组(χ2=11.71,P=0.11)在院死亡预测准确性均较高,却低估了中风险亚组(χ2=17.48,P<0.01)的实际在院死亡率。两种评分方法对全组患者在院死亡预测的鉴别度均较差(ROC曲线下面积分别为0.676和0.677)。结论 EuroSCORE模型对本中心心瓣膜手术患者死亡风险预测的准确性较差,不适合本中心心瓣膜手术的风险预测,在今后的临床实践中应慎重使用。

Objective To assess the accuracy of the European System for Cardiac Operative Risk Evaluation（EuroSCORE） model in predicting the in-hospital mortality of Chinese patients undergoing heart valve surgery in Changhai Hospital. Methods We collected the clinical data of 4 155 consecutive patients who underwent heart valve surgery at our center from January 1998 to December 2008.Among them,there were 1 955 males and 2 200 females with an average age of 45.90±13.64 years.According to the score for additive and logistic EuroSCORE models,the patients were divided into three risk subgroups including low risk subgroup（n=981）,moderate risk subgroup（n=2 492）,and high risk subgroup（n=682）.The actual and predicted mortality of whole cohort and each risk subgroup were studied and compared.Calibration of the EuroSCORE model was assessed by the Hosmer-Lemeshow（H-L） test.Discrimination was tested by calculating the area under the receiver operating characteristic（ROC） curve. Results A total of 205 patients died among the 4 155 patients with an actual mortality of 4.93%.The predicted mortality by additive EuroSCORE model and logistic EuroSCORE model was 3.80% and 3.30% respectively,which suggested that both models underpredicted the mortality（χ2=11.13,44.34,P0.05） for the whole cohort.The additive EuroSCORE model had a good calibration in predicting in-hospital mortality for the high-risk subgroup（χ2=3.61,P=0.31）,a poor calibration for the low-risk subgroup（χ2=0.00,P0.01）,and the moderate risk subgroup（χ2=14.72,P0.01）.While the logistic EuroSCORE model had a good calibration in predicting in-hospital mortality for the low risk subroup（χ2=1.66,P=0.88） and high risk subgroup（χ2=11.71,P=0.11）,but underpredicted the mortality for the moderate risk subgroup（χ2=17.48,P0.01）.The discriminative power of both models for the whole cohort was poor with area under the ROC curve at 0.676 by additive model and 0.677 by logistic model,respectively.Conclusion The additive and logistic EuroSCORE models can＇t give an imprecise prediction for individual operative risk in heart valve surgery patients in our center.Clinical use of these models should be considered cautiously.