基于实验室指标的新型冠状病毒肺炎和甲型流感鉴别诊断模型的建立及其临床意义

Establishment of differential diagnostic model for COVID-19 and influenza A based on laboratory indicators and its clinical significance

目的:探讨新型冠状病毒肺炎(COVID-19)与甲型流感实验室指标检测结果的差异,建立2种疾病的鉴别诊断模型,阐明该模型对于鉴别2种疾病的临床意义。方法:共纳入56例COVID-19普通型患者和54例甲型流感患者,以及用于模型验证的24例COVID-19普通型患者和30例甲型流感患者;计算患者住院后5d实验室指标的平均值,采用弹性网络模型和逐步Logistic回归模型,筛选鉴别COVID-19和甲型流感的指标;弹性网络模型用于第1轮选择,并通过10折交叉验证选择lambda的最佳截断值。采用不同的随机种子,将该模型拟合200次,选取高频指标(频率>90%);第2轮筛选采用以AIC作为选择标准的Logistic回归模型,列线图用来表示最终的模型;使用独立数据集作为外部验证集,计算受试者工作特征(ROC)曲线下面积(AUC)来评估该模型的预测性能。结果:第1轮筛选后,有16个实验室指标被选为高频指标;经过第2轮筛选,确定白蛋白(ALB)/球蛋白GLB(A/G)、总胆红素(TBIL)和红细胞比容(HCT)为最终鉴别指标;该模型具有较好的预测性能,验证集的AUC为0.844(95%CI:0.747~0.941)。结论:成功建立基于实验室检测结果的COVID-19和甲型流感鉴别诊断模型,该模型有助于临床及时对2种疾病做出准确、快速的诊断。

Objective:To explore the differences in laboratory indicators test results of coronavirus disease 2019(COVID-19)and influenza A and to establish a differential diagnosis model for the two diseases,and to clarify the clinical significance of the model for distinguishing the two diseases.Methods:A total of 56 common COVID-19 patients and 54 influenza A patients were enrolled,and24 common COVID-19 patients and 30 influenza A patients were used for model validation.The average values of the laboratory indicators of the patients 5 d after admission were calculated,and the elastic network model and the stepwise Logistic regression model were used to screen the indicators for identifying COVID-19 and influenza A.Elastic network models were used for the first round of selection,in which the optimal cutoff of lambda was chosen by performing 10-fold cross validations.With different random seeds,the elastic net models were fit for 200 times to select the high-frequency indexes(frequency>90%).A Logistic regression model with AIC as the selection criterions was used in the second round of screening uses;a nomogram was used to represent the final model;an independent data were used as an external validation set,and the area under the curve(AUC)of the validation set were calculate to evaluate the predictive the performance of the model.Results:After the first round of screening,16 laboratory indicators were selected as the high-frequency indicators.After the second round of screening,albumin/globulin(A/G),total bilirubin(TBIL)and erythrocyte volume(HCT)were identified as the final indicators.The model had good predictive performance,and the AUC of the verification set was0.844(95%CI:0.747-0.941).Conclusion:A differential diagnosis model for COVID-19 and influenza A based on laboratory indicators is successfully established,and it will help clinical and timely diagnosis of both diseases.