Machine learning to identify febrile children at risk of Kawasaki disease

Authors´ conclusions: the study suggests that the results of objective laboratory tests have the potential to predict Kawasaki disease. Machine learning with XGBoost can help clinicians differentiate Kawasaki disease patients from other febrile patients in pediatric emergency departments with excellent sensitivity, specificity, and accuracy.

Reviewers´ commentary: although the model presented has power to identify patients at risk of Kawasaki disease, it must be externally validated in populations more similar to ours before its use can be recommended.