POSSUM、P—POSSUM及Thoracoscore预测胸外科患者术后病死率的评价比较

Comparison of POSSUM, P-POSSUM and Thoracoscore in forecast of the postoperative mortality in thoracic surgery patients

目的评价比较POSSUM、P—POSSUM及Thoracoscore3种风险评分系统预测胸外科患者术后病死率的适用性与准确性。方法收集2012年3月至2013年8月中南大学湘雅二医院1450例胸外科手术患者的临床资料，分别按照3种风险评分系统计算患者的预测病死率，使用H—L检验和ROC曲线下面积比较评价三者预测胸外科患者术后病死率的符合程度与鉴别效度。结果1450例患者术后30天内死亡20例（1．4％），POSSUM高估了胸外科患者术后病死率（3．84％，P〈0．001），P—POSSUM（1．05％／1．10％）和Thoracoscore（0．90％）低估了胸外科患者术后病死率，P〉0．05。3种评分系统预测所有患者术后病死率的ROC曲线下面积分别为0．774、0．777、0．777和0．831，P〈0．05；预测肺部手术患者术后病死率的ROC曲线下面积分别为0．755、0．771、0．771、0．849，P〈0．05；预测食管手术患者术后病死率的ROC曲线下面积分别为0．640、0．650、0．650、0．764，仅ThoracoscoreP〈0．05；预测纵隔及其他手术患者术后病死率的ROC曲线下面积分别为1．000、1．000、1．000、0．854，仅ThoracoscoreP〈0．05。结论Thoracoscore评分系统最适合预测胸外科患者术后病死率，其预测准确性高且鉴别效度良好。

Objective To evaluate and compare the applicability and accuracy of POSSUM, P-POSSUM and Thora- coscore in predicting the postoperative mortality of thoracic surgery patients. Methods We collected the clinical data of 1 450 thoracic surgery patients during March 2012 to August 2013 in the Second Xiangya Hospital. We used the 3 kinds of risk sco- ring system to calculate the predicted postoperative mortality in these patients. Then we used H-L test and the area under the ROC curve to evaluate and compare their degrees of compliance and validities of identification independently. Results Within 1 450 cases,20 patients （ 1.4% ） died in 30 days after surgery, POSSUM overestimated the mortality （3.84% ,P 〈 0. 001 ） , while the P-POSSUM （1.05%/1.10%） and Thoracoscore （0.90%） underestimated the mortality, P 〉 0.05. The area under the ROC curve of POSSUM, P-POSSUM （ two kinds of death prediction formula） and Thoracoscore to predict postoperative mortality in all thoracic surgery patients was 0. 774,0. 777,0. 777 and 0. 831 independently, P 〈 0.05; the area under the ROC curve to predict postoperative mortality in patients after lung surgery was 0. 755,0. 771,0. 771,0. 849 independently, P 〈 0.05 ; the area under the ROC curve to predict postoperative mortality of patients with esophageal surgery was 0. 640,0. 650,0. 650,0. 764 independently, only Thoracoscore P 〈 0.05 ; the area under the ROC curve to predict postoperative mortality of pa- tients with mediastinal and other surgical was 1. 000,1. 000,1. 000,0. 854 independently, only Thoracoscore P 〈 0.05. Con- clusion Thoracoscore scoring system is the most suitable risk scoring system to predict postoperative mortality in thoracic sur- gery patients with high predictive accuracy and good identification validity.