重症新型冠状病毒肺炎的危险因素分析及预测模型构建

Developing a prediction model for severe coronavirus disease 2019 based on the analysis of early predictors

目的探讨重症新型冠状病毒肺炎(COVID-19)的危险因素。方法选取2020年1月10日—2月27日江苏省部分医院共计342例COVID-19确诊患者,将其分为非重症组307例和重症组35例,通过logistic回归分析及建立预测模型分析重症感染患者的独立危险因素。结果与非重症组相比,重症组患者中糖尿病患病率、年龄、空腹血糖(FBG)、体重指数(BMI)、中性粒细胞百分比/淋巴细胞百分比(NLR)、中性粒细胞百分比(NE%)更高,淋巴细胞百分比(LY%)、血小板计数(PLT)较低,差异具有统计学意义(P<0.05)。逐步logistic回归分析显示FBG、NLR、BMI是重症COVID-19的独立危险因素,其中,FBG的预测价值最高,AUC值为0.839。构建的预测模型中,logistic全模型预测效果最佳,AUC值为0.906。logistic全模型中筛选出重症COVID-19的危险因素为:合并糖尿病,LY%、NLR、PLT、总胆红素(TBIL)降低。结论FBG、BMI、NLR与COVID-19的严重程度呈正相关,是其独立危险因素。是否合并糖尿病、TBIL、LY%、PLT等指标对重症COVID-19的发生也有一定的预测价值,并且与NLR联合预测效果更好。

Objective To explore the relevant risk factors of severe coronavirus disease 2019(COVID-19).Methods A total of 342 patients confirmed with COVID-19 diagnosis were selected in selected hospitals across Jiangsu Province from January 10,2020 to February 27,2020.The patients were assigned to non-severe group(n=307)and severe group(n=35)based on universal diagnostic criteria.The independent risk factors for severe cases were identified by logistic regression analysis to construct a prediction model.Results The severe COVID-19 cases showed significantly older age,higher prevalence of diabetes mellitus,fasting blood glucose(FBG),body mass index(BMI),neutrophil to lymphocyte ratio(NLR),and neutrophil percentage,but significantly lower lymphocytes percentage(LY%)and platelet count(PLT)than the non-severe cases(P<0.05).Stepwise logistic regression analysis demonstrated that FBG,NLR,and BMI were independent risk factors for severe COVID-19 disease.FBG provided the highest predictive value,evidenced by an area under the receiver operating characteristic curve(AUC)of 0.839.Among the constructed prediction models,the full logistic model showed the best prediction result,evidenced by AUC of 0.906.Diabetes mellitus,LY%,NLR,PLT,and total bilirubin(TBIL)were the relevant risk factors in the logistic model for identifying severe COVID-19 cases.Conclusions FBG,BMI,and NLR levels are positively correlated with the severity of COVID-19.They are also independent risk factors for severe COVID-19.Diabetes mellitus,TBIL,LY%,PLT are also useful in predicting severe COVID-19.A prediction model performs better when combining these variables with NLR.