新型冠状病毒肺炎患者重症风险模型建立及评价  

Establishment and evaluation of severe risk model of novel coronavirus pneumonia

作　　者：韩晶[1,2] 黄淑萍 谢祎[1] 刘玲梅[1] 张永进[1] 史丽霞[1,2] HAN Jing;HUANG Shu-ping;XIE Yi;LIU Ling-mei;ZHANG Yong-jin;SHI Li-xia(Tianjin Haihe Hospital,Tianjin University,Tianjin 300350,China)

机构地区：[1]天津市海河医院,天津300350 [2]天津市呼吸疾病研究所,天津300350

出　　处：《现代预防医学》2020年第24期4439-4442,共4页Modern Preventive Medicine

基　　金：天津市卫生健康新冠肺炎防治科技项目(2020xkc03)。

摘　　要：目的探究一种预测新型冠状病毒肺炎患者的重症风险模型的构建。方法对天津市收治的168例新型冠状病毒肺炎患者进行回顾研究分析,将研究人群以是否重症进行分层,并按7∶3比例随机分为建模组和验证组,其中建模组共117例用作模型的建立,验证组共51例用于检验模型应用的效度及评价。采用受试者工作特征(ROC)曲线评价各评分系统对患者重症风险的预测能力。结果研究人群的新型冠状病毒肺炎患者重症率37.5%,logistic回归分析结果表明,年龄(OR=1.091,95%CI:1.007~1.181,P=0.032)、糖尿病(OR=28.549,95%CI:1.753~465.039,P=0.019)、冠心病(OR=125.649,95%CI:2.728~5787.957,P=0.013)、淋巴细胞计数(OR=0.007,95%CI:0.001~0.160,P=0.007)、肌红蛋白MB(OR=1.087,95%CI:1.036~1.140,P=0.001)、Pa O2/Fi O2(OR=0.974,95%CI:0.957~0.991,P=0.004)为患者重症的独立危险因素,预测模型:P=1/[1+e-(0.087×年龄+3.352×糖尿病+4.833×冠心病-5.003×LY+0.083×肌红蛋白MB-0.027×Pa O2/Fi O2+6.540)],其ROC曲线下面积(AUC)为0.914(95%CI:0.836~0.993),logistic回归模型的预测敏感度为89.47%、特异度为87.5%。结论logistic回归模型能较好预测新型冠状病毒肺炎患者的重症风险,可提高对重症高危患者的早期识别,以期早期采取干预等治疗策略降低重症风险。Objective To explore novel coronavirus pneumonia prediction model for severe risk.Methods 168 novel coronavirus pneumonia patients in Tianjin City were retrospectively analyzed.The study population was divided into modeling group and validation group according to the proportion of 7:3.117 cases in the modeling group were used to establish the model,and 51 cases in the validation group were used to test the validity and evaluation of the model application.ROC curve was used to evaluate the ability of each scoring system to predict the severe risk of patients.Results The severe rate of novel coronavirus pneumonia in the study population was 37.5%.Logistic regression analysis showed that age(OR=1.091,95%CI:1.007-1.181,P=0.032),diabetes mellitus(OR=28.549,95%CI:1.753-465.039,P=0.019),coronary heart disease(OR=125.649,95%CI:2.728-5787.957,P=0.013),LY(OR=0.007,95%CI:0.001-0.160,P=0.007),MB(OR=1.087,95%CI:1.036-1.140,P=0.001)and PaO2/FiO2(OR=0.974,95%CI:0.957-0.991,P=0.004)were independent risk factors for severe patients.Prediction model:P=1/[1+e-(0.087×age+3.352×diabetes mellitus+4.833×coronary heart disease-5.003×LY+0.083×MB-0.027×Pa O2/Fi O2+6.540)].The area under ROC curve(AUC)was 0.914(95%CI:0.836-0.993),and the predictive sensitivity and specificity of logistic regression model were 89.47%and 87.5%,respectively.Conclusion Novel coronavirus pneumonia patients can be predicted by logistic regression model,and early recognition of high-risk patients can be improved.

关 键 词：新型冠状病毒肺炎 重症 预测模型 LOGISTIC回归 

分 类 号：R181.2[医药卫生—流行病学]