A machine learning model combining fungal biomarkers and routine laboratory indicators shows promise for improving early diagnosis of invasive fungal infections in critically ill children.

We retrospectively analyzed 3,062 PICU patients (247 IFIs, 8.1%) and used clinical features, comorbidities, and laboratory indicators to build binary classification models with multiple machine learning algorithms. In the Test Group List, Diagnosis of fungi1 represents the diagnostic grouping used for data analysis, and missing data indicate that the relevant clinical tests were not performed.External Verification Form refers to the dataset used for external validation.