Construction And Evaluation Of Forewarning Model Of Severe Hand,Foot And Mouth Disease

Hand,foot and mouse disease(HFMD)is an enterovirus-caused acute pediatric infectious disease which prevalent in summer and autumn.Most HFMD patients manifest mild symptoms,such as fever and eruption,and can be cure in one week.The minor can progress to severe cases who exhibit nervous system or cardiopulmonary system involvement,and the severe may be capable of death.The conditions of severe HFMD process rapidly and there are short of accurate forewaining biomarkers and models.Therefore,it is of great value in accurately discrimination and prevention of severe HFMD to construct a model of judging the disease progress based on molecular biomarkers.And it is also meaningful in reducing the mortality of HFMD and relieve finical burden.Objective1.To investigate risk factors and build judgement model of severe HFMD.2.To study the dynamic changes of laboratory results and serum cytokines with the number of days after disease onset.3.Construct a model judging the procgress of disease based on laboratory results and serum cytokines,and realize forewarn of severe HFMD progress.Methods1.Questionnaire survey was used to collect the demographic information,clinical symptoms and laboratory results of HFMD patients who were in hospital form May to September in 2016 in the First affiliated Hospital of Xinxiang Medical College,and the serum samples were collected,too.Univariate logistic regression and multivariate logistic regression were applied to build the judgement model of severe HFMD and figure out the risk factors,and the ROC curve and cross validation were employed to assess the efficiency of model.2.The levels of cytokines including IL-6,IL-10,TNF-?and INF-?were assessed by using ELISA kit.Mild and severe patients were divided to diverse subgroups according to the number of days after disease onset based on the nest series case-control study to investigate the potential change regulations of laboratory index and cytokine levels.Furthermore,correlation analysis was used to study the relationship between laboratory index and cytokines.3.Data of laboratory results and cytokine levels were used to construct progress-judging models with classification and regression tree(CATR)and support vector machine(SVM)algorithms,and the efficiency of models were appraised by confusion matrix and cross validation.Results1.The results of logistic regression demonstrated that vomit,anorexia,leucocytosis(>12×10~9/L),neutrophilic leukocytosis(>7.6×10~9/L)and decreased total protein(>12×10~9/L)were the major influence factors of severe HFMD.Vomit(OR=5.13,95%CI:1.78~16.80),anorexia(OR=14.39,95%CI:6.28~35.78)and leucocytosis(OR=3.84,95%CI:1.39~11.04)were risk factors of severe HFMD,and decreased total protein(OR=0.19,95%CI:0.04~0.74)was a protect factor of severe HFMD.There were a sensitivity of 82.61%and a specificity of 76.19%of the model,and the area under ROC curve was 0.877(95%CI:0.828~0.925).Cross validation showed an accuracy of 76.46%,and the Kappa value was 0.53.2.There was no significant difference among subgroups divided by days after disease onset during the laboratory indices including white blood cell counts,neutrophil counts,total protein and glucose,and cytokines including IL-6,IL-10,TNF-?and IFN-?in mild cases.For severe HFMD patients,white blood cell counts,neutrophil counts,total protein and glucose were at high level in the first three days after disease onset,and then descended.IL-6,IL-10 and IFN-?showed trend of an inverse“V”,and the peak of IL-6 was ahead of those of IL-10 and IFN-?.The level of TNF-?showed a decreasing trend with the number of days after disease onset.3.The accuracies of CART model and SVM model in judging the number of days after disease onset were 84.66%and 88.64%.The results of cross validation showed that accuracies of CART and SVM models were 72.64%and 78.49%with the Kappa values of 0.59 and 0.66,respectively.Conclusions1.HFMD patients with symptoms of vomit,anorexia,leucocytosis were more easily to be severe cases,and the judgement model showed a good performance.2.Diverse change regulations were found in laboratory indices including white blood cell counts,neutrophil counts,total protein and glucose,and cytokines including IL-6,IL-10,TNF-?and IFN-?in HFMD patients with disease progress,and this provide a new sight to investigate the process of HFMD.3.CART model and SVM model can be used in forewarning the process of HFMD.