甲型流感社区获得性肺炎死亡风险预测模型的建立

A predictive tool for mortality of influenza A community-acquired pneumonia

目的探索一种预测甲型流感社区获得性肺炎30天死亡风险的模型。方法进行多中心回顾研究分析178例甲型流感社区获得性肺炎,包括存活组144例和死亡组34例的病例资料。采用受试者工作特征(ROC)曲线评价各评分系统对甲型流感社区获得性肺炎30天死亡风险的预测能力。结果甲型流感社区获得性肺炎30天病死率19.1%。肺炎严重度指数(PSI)Ⅰ～Ⅱ和社区获得性肺炎CURB-65评分0～1分的实际病死率分别为14.5%和15.7%,远高于预测病死率。Logistic回归分析证实血尿素氮>7 mmol/L、白蛋白<35 g/L和外周血淋巴细胞<0.7×10~9/L是30天死亡的独立危险因素。以血尿素氮>7 mmol/L(U)+白蛋白<35 g/L(A)+外周血淋巴细胞<0.7×10~9/L(L)组合的UAL作为模型预测甲型流感社区获得性肺炎30天死亡风险,其ROC曲线下面积(AUC)为0.891,高于CURB-65评分(AUC=0.777,P=0.008 3)、CRB-65评分(AUC=0.590,P<0.000 1)、PSI分级(AUC=0.568,P=0.000 1)。结论 UAL可作为一种可靠的模型用以预测甲型流感社区获得性肺炎30天死亡风险。

Objectives To explore a reliable and simple predictive tool for 30-day mortality of influenza A community-acquired pneumonia （CAP）. Methods A multicenter retrospective study was conducted on 178 patients hospitalized with influenza A CAP, including 144 alive patients and 34 dead patients. Receiver operating characteristic （ROC） curves were performed to verify the accuracy of severity scores as 30-day mortality predictors in the study patients. Results The 30-day mortality of influenza A CAP was 19.1%. The actual mortality of PSI risk class Ⅰ- Ⅱ and CURB-65 score 0-1 were 14.5% and 15.7%, respectively, which were much higher than the predicted mortality. Logistic regression confirmed blood urea nitrogen 〉7 mmol/L （U）, albumin 〈35 g/L （A） and peripheral blood lymphocyte count 〈 0.7×109/L （L） were independent risk factors for 30-day mortality of influenza A CAP. The area under the ROC curve （AUC） of UAL （blood urea nitrogen 〉7 mmol/L＋ albumin 〈35 g/L＋ peripheral blood lymphocyte count 〈0.Tx109/L） was 0.891, which was higher than CURB-65 score （AUC=0.777, P=0.008 3）, CRB-65 score （AUC=0.590, P〈0.000 1）, and PSI risk class （AUC=0.568, P=0.000 1）. Conclusion UAL is a reliable and simple predictive tool for 30-day mortality of influenza A CAP,