| Objective:To explore the risk factors of IgA vasculitis with nephritis(IgAVN)in children,and to establish a IgAVN risk prediction model.And then,to provide a scientific quantitative tool for individual risk prediction and benefit assessment of patients.Methods:The clinical data of 1007 children with IgA vasculitis(IgAV)in Hunan Children’s Hospital in 2018 and 287 cases from 2018 to 2019 in other hospitals(Hunan Provincial people’s Hospital,Changsha Central Hospital and Changsha first people’s Hospital)were analyzed retrospectively.The data of 715children with IgAV in Hunan Children’s Hospital were selected by sample splitting method to construct the clinical prediction model of IgAVN.The remaining data of 292 IgAV cases in Hunan Children’s Hospital were used for internal verification,and the data of 287 IgAV cases in other hospitals were used for external verification.The data of 1007 cases in Hunan Children’s Hospital were analyzed by univariate analysis.Multivariate logistic regression analyzed meaningful indicators of single-factor analysis in modeling data sets,which aim was to screened out the risk factors of IgAVN and established the IgAVN prediction model according to its regression coefficient.The clinical efficacy including discrimination,accuracy and clinical practicability was evaluated by the area under ROC curve(AUC),calibration curve and clinical decision curve(DCA).Internal verification dataset and external verification dataset further verified the clinical efficacy of the IgAVN model to evaluate the repeatability and portability of that model.At last,the IgAVN model was rendered as a nomogram.Results:(1)Multivariate analysis showed that age(OR:1.184,95%CI:1.112～1.259),persistent skin purpura(OR:2.467,95%CI:1.625～3.745),red blood cell distribution width(OR:1.314,95%CI:1.145～1.507),complement C3(OR:0.226,95%CI:0.092～0.554),immunoglobulin G(OR:0.888,95%CI:0.837～0.943),triglycerides(OR:1.902,95%CI:1.424～2.541)and urea nitrogen(OR:1.190,95%CI:1.037～1.365)were statistically significant.(2)The prediction model of IgAVN based on multivariate logistic regression analysis was logit P=-4.169+0.169×age+0.903×persistent purpura+0.273×red blood cell distribution width-1.485×complement C3-0.118×Ig G+0.643×TG+0.174×urea nitrogen.(3)The AUC of the IgAVN prediction model was 0.777(95%CI:0.743～0.811).The calibration curve of the IgAVN model revolved around the ideal curve as well as the Hosmer-Lemeshow test wasX~2=10.47,P=0.40.And The DCA of the model was higher than two extreme lines when the prediction probability was about 15%～82%.(4)The AUC of the internal verification dataset was 0.734(95%CI:0.678～0.791),compared with the modeling dataset AUC by Z test,P=0.244.The calibration curve of the internal verification dataset fluctuated around the ideal curve and the Hosmer-Lemeshow test wasX~2=8.72,P=0.559.Meanwhile,the DCA of the internal verification dataset was higher than extreme lines when the prediction probability is between 18%and80%.(5)The AUC of the external verification dataset was 0.751(95%CI:0.688～0.814),compared with the modeling dataset AUC by Z test,P=0.474.The calibration curve drawn by the external validation dataset fluctuateed around the ideal curve and the Hosmer-Lemeshow test wasX~2=11.2,P=0.342.The clinical decision curve was higher than two extreme lines.When the prediction probability was between 14%and 85%,the DCA of the external verification dataset was higher than extreme lines.Conclusion:(1)Age,persistent skin purpura,red blood cell distribution width,complement C3,immunoglobulin G,triglycerides and urea nitrogen were independent influencing factors of IgAVN.(2)The clinical prediction model of IgAVN established by independent influencing factors has good discrimination ability,accuracy and clinical practicability.(3)The application of IgAVN model to internal verification data and external verification data also has better clinical efficacy,indicating that the model was worth popularizing.(4)The IgAVN model is presented with a nomogram,which makes the results more readable and was helpful to make clinical decisions for patients. |
