Transparent and Distributed AI Prediction Modeling: A Case Study on Pediatric Covid

Abstract

COVID-19 pandemic has had heterogeneous impacts on pediatric populations with most of them presenting asymptotic or having mild symptoms, nevertheless some will progress to severe disease. One of the most severe conditions is the Multisystem Inflammatory Syndrome in Children (MIS-C), characterized by inflammation of multiple organs and tissues including the heart, lungs, blood vessels, kidneys, digestive system, brain, skin or eyes. Therefore, it is important for healthcare providers to determine such severe patient groups as early as possible so that appropriate interventions can be undertaken to improve pediatric patient outcomes. Machine learning models for predicting such outcomes early using Electronic Health Records (EHR) could potentially help clinical decision making such as advanced and more intensive medical interventions.