Interpretable web-based machine learning model for predicting intravenous immunoglobulin resistance in Kawasaki disease.

Abstract

Background: Kawasaki disease (KD) is a leading cause of acquired heart disease in children that is treated with intravenous immunoglobulin (IVIG). However, 10-20% of cases exhibit IVIG resistance, which increases the risk of coronary complications. Existing predictive models do not integrate multiple machine learning (ML) algorithms or facilitate real-time clinical use. This study presents a region-specific, interpretable ML model for early IVIG resistance prediction in KD.

Methods: A retrospective cohort of 463 children diagnosed with KD at Fuzhou University Affiliated Provincial Hospital (2012-2024) was analyzed. Thirteen ML algorithms were evaluated via cross-validation, with performance assessed by AUC and other metrics. Feature importance was determined using SHapley Additive exPlanations (SHAP), and risk of bias was evaluated using the Prediction Model Risk of Bias Assessment Tool.

Results: The random forest (RF) model demonstrated the highest predictive performance (AUC = 0.78). After feature selection based on SHAP values, a final interpretable RF model incorporating 10 key features was developed, and a web-based tool integrating the Youden index (16.9%) was deployed for real-time risk estimation.

Conclusion: This region-specific, interpretable ML model ( https://milailai.shinyapps.io/data1/ ) is a practical tool for early risk stratification and personalized treatment of IVIG resistance in KD.