Dataset from AICORE-kids: Artificial Intelligence COVID-19 Risk AssEssment for kids

Background: This study focused on characterizing pediatric COVID-19 and identifying which children were likely to develop more severe disease. Stratifying patients by projected severity supported earlier and more appropriate treatment decisions and guided care planning, and also informed the risks of outpatient management for families, caregivers, schools, and workplaces. As communities adopted different reopening strategies, these predictions could help shape policy on how to manage pediatric cases in the community.

Materials/Methods: Preliminary analysis showed that combining quantitative serology, inflammatory markers, immune profiles, transcriptomics, epigenomics, longitudinal physiological data, imaging, radiomics, and clinical information including social determinants of health provided enough early information to stratify disease and predict severity. Artificial intelligence and machine learning methods integrated these data to characterize disease patterns and identify biosignatures linked to worsening illness. A Translational Development function guided design control and prepared the methods for regulatory review, with timelines aligned to Emergency Use Authorization (EUA) pathways for SARS-CoV-2 diagnostics.

Outcome/Impact: The intended outcome of this work was to stratify pediatric COVID-19 patients by disease severity using early-stage data, thereby improving treatment planning and informing both individual and community-level decisions. The incorporation of AI/ML approaches and translational oversight was designed to enable regulatory readiness and future deployment under EUA pathways.