Kawasaki Disease Coronary Aneurysm Prediction Model from Chinese Patient Data

A retrospective study of 2,777 Kawasaki disease patients from two centers in China developed a machine learning model to predict medium-to-giant coronary artery aneurysms. The support vector machine model uses seven key clinical features and was validated internally and externally, achieving an AUC of 0.732 and 0.689, respectively. The model is deployed as an R Shiny-based online tool.

Use Cases

• Validate the published support vector machine model's performance using the seven key features: time to diagnosis, monocyte percentage, rash, eosinophil percentage, C-reactive protein, triglycerides, and neutrophil percentage.
• Analyze the global feature importance and individualized explanations provided by the SHAP analysis to understand model predictions for clinical risk assessment.
• Reproduce the intercept-only recalibration method described to correct model miscalibration using the provided clinical and laboratory data.
• Use the decision curve analysis framework from the study to evaluate the net clinical benefit of the prediction model across different threshold probabilities.

Strengths

• Model development and validation is based on data from 2,777 patient records.
• The study employed 11 different machine learning algorithms and performed both internal and external validation.
• Model interpretability is addressed using SHAP analysis for both global and individual predictions.

Limitations

• The underlying clinical dataset is not directly provided; the primary file is a research paper describing the model.
• Model performance, while validated, shows a moderate external validation AUC of 0.689, indicating room for improvement in generalizability.
• The 3.2 MB DOC file contains the study manuscript, not a structured, analysis-ready dataset of patient records.

Provenance

Source

Retrospective electronic medical records from two medical centers in China.

Collection Method

Clinical and laboratory data extracted from patient records for model development and validation.

Freshness

Last updated March 2026.

Geography

China