Construction and subgroup validation of a predictive model for IVIG-resistance in kawasaki disease.
Kawasaki disease is an acute immune vasculitis, and intravenous immunoglobulin (IVIG) therapy was an important treatment method, although a small proportion of children exhibited IVIG resistance. Our study aimed to construct a risk model for Kawasaki disease with IVIG resistance and to validate it across different clinical characteristic subgroups, with the goal of optimizing personalized and precise management to improve patient outcomes.
We first compared various factors between the groups with and without IVIG resistance. Then, we used LASSO analysis to screen for factors that significantly predicted outcomes. The selected factors were used to develop the risk model. We evaluated the model using ROC curves, calibration curves, and decision curve analysis, and conducted internal validation through 5-fold cross-validation. Finally, we performed subgroup analyses based on age, sex, and other factors.
Through univariate analysis, LASSO analysis, and correlation analysis, we identified WBC, PLT, CRP, fever days, PLR, and ALT as key factors in constructing the risk model. The model achieved an area under the curve of 0.816 (95% CI: 0.775-0.857). Additionally, calibration curves, DCA, and 5-fold cross-validation demonstrated that the model had good predictive performance. The model's effectiveness was also satisfactory across various subgroups.
Our study successfully constructed a risk model for Kawasaki disease with IVIG resistance in the Chinese population. The model demonstrated strong predictive ability and was validated across multiple subgroups, showing potential clinical application value.
We first compared various factors between the groups with and without IVIG resistance. Then, we used LASSO analysis to screen for factors that significantly predicted outcomes. The selected factors were used to develop the risk model. We evaluated the model using ROC curves, calibration curves, and decision curve analysis, and conducted internal validation through 5-fold cross-validation. Finally, we performed subgroup analyses based on age, sex, and other factors.
Through univariate analysis, LASSO analysis, and correlation analysis, we identified WBC, PLT, CRP, fever days, PLR, and ALT as key factors in constructing the risk model. The model achieved an area under the curve of 0.816 (95% CI: 0.775-0.857). Additionally, calibration curves, DCA, and 5-fold cross-validation demonstrated that the model had good predictive performance. The model's effectiveness was also satisfactory across various subgroups.
Our study successfully constructed a risk model for Kawasaki disease with IVIG resistance in the Chinese population. The model demonstrated strong predictive ability and was validated across multiple subgroups, showing potential clinical application value.
Authors
Li Li, Gong Gong, Zhang Zhang, Huang Huang, Bai Bai, Song Song, Chen Chen, Yang Yang, Liu Liu
View on Pubmed