Construction and internal validation of a predictive model for risk of severe gastrointestinal involvement in children with IgA vasculitis

Early identification of the severity of gastrointestinal (GI) involvement in children with IgA vasculitis (IgAV) is crucial for prognosis assessment and formulation of treatment plans. This study aimed to develop a risk predictive model for severe GI involvement in pediatric IgAV to assist clinicians in decision-making. In a single-center retrospective cohort study, 293 eligible children were divided into two cohorts based on the time of admission: a training set and a validation set. Candidate risk indicators were identified using binary logistic regression, and a predictive model was developed to predict severe GI involvement in pediatric IgAV based on training set. The receiver operating characteristic (ROC) curve was utilized as a tool to evaluate the performance of the logistic regression model. Furthermore, a nomogram model was developed, and its calibration curve was visualized using a 1000-time bootstrap resampling method. The validity of these results was verified on an independent validation set. Vomiting, lymphocyte count, C-reactive protein and thickening of intestinal wall displayed on abdominal ultrasound were identified as the potentially predictors for severe GI involvement in pediatric IgAV. The predictive model had area under the ROC curve of 0.816 and 0.802 in the training and validation set, with a sensitivity and specificity of 80.0% and 72.4%, 82.8% and 69.8%, respectively. The calibration curve demonstrated favorable predictive accuracy of the predictive model. Decision curve analysis indicated that the predictive model possesses clinical applicability. Our research has identified potentially predictive indicators for severe GI involvement in children with IgAV. The risk prediction model we established can assist pediatricians in early prediction and identification of high-risk children, especially those at high risk of severe GI bleeding.