Establishment And Analysis Of LASSO Regression Based Rosette Model Of Kawasaki Disease Complicated With Coronary Artery Disease

Background:Kawasaki disease（KD）is a systemic vasculitis syndrome.Coronary artery lesion（CAL）is the most serious complication of KD,which affects the long-term prognosis of children.In recent years,the incidence of KD has been increasing year by year,so it is helpful to improve the prognosis of children by exploring the risk factors of CAL and taking effective treatment in time.Objective:To explore the risk factors of KD concurrent CAL,establish a line chart model to predict CAL risk.Methods :The clinical data and auxiliary examination results of children with KD treated in Suizhou Central Hospital affiliated to Hubei University of Medicine from June 2017 to May 2021 were retrospectively analyzed.According to the results of coronary artery detection by echocardiography,the children were divided into coronary artery disease group（CAL group）（70 cases）and non-coronary artery disease group（NCAL group）（95 cases）.LASSO regression was used to analyze the risk factors of KD combined with CAL,and multivariate Logistic regression was used to establish the prediction model and construct the line chart.Then the model is verified by ROC curve,calibration curve and decision curve analysis to evaluate the advantages and disadvantages of the model.Result :Five predictors（fever time ≥10d,combined mycoplasma infection,WBC > 20×109 /L,PLT count,CRP/ALB ratio）were screened by LASSO regression,which were incorporated into Logistic regression and constructed a column chart.The c-index of the line graph was 0.846.ROC curve analysis was performed on the prediction probability of the model,with AUC of 0.841（95%CI: 0.782 0.900）,sensitivity of 85.7% and specificity of 71.6%.The calibration curve of the model fits well with the ideal curve,and can reflect the risk factors of CAL well.The decision curve indicates the applicability of the model in clinic.Conclusions :.In this study,it was found that fever time ≥10d,combined mycoplasma infection,WBC > 20×109 /L,PLT count and CRP/ALB ratio were independent risk factors for CAL,and the KD concurrent CAL histogram model established based on the above independent risk factors had good differentiation and accuracy.It is helpful for clinical workers to screen out children at high risk for CAL.