A team of faculty and a student from the Oregon Health & Science University (OHSU) Department of Medical Informatics & Clinical Epidemiology (DMICE) was awarded an Honorable Mention in the Pediatric COVID-19 Data Challenge, sponsored by the Biomedical Advanced Research and Development Authority (BARDA) in partnership with the National Institute of Health (NIH) National Center for Advancing Translational Sciences (NCATS), the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), and Health Resources & Services Administration (HRSA) Maternal & Child Health Bureau.
The DMICE team – Lorne Walker MD, PhD, Ben Orwoll MD, MS, and Meenakshi Mishra MSc, MPH, and PhD candidate – competed with over 200 participants on 88 teams, and were among three teams recognized with Honorable Mentions in addition to the two winning groups.
This nationwide data science competition utilized the COVID-19 data enclave maintained by the National COVID Cohort Collaborative (N3C), a data repository containing electronic health record (EHR) data from over 4.5 million SARS-CoV-2 positive patients from health systems across the US, including OHSU. Participants in the Pediatric COVID-19 Data Challenge were asked to use EHR data to predict two key outcomes in children with COVID-19: 1) hospitalization after testing positive for COVID-19 and 2) the need for advanced respiratory or cardiovascular support after hospitalization. Severe COVID-19 is rare in children but can have significant health impacts on the most vulnerable. This presents a challenge to pediatric clinicians who must counsel patients and families and apportion drugs and other therapies to the patients who will benefit most. Accurate predictions of disease severity may better inform these key decisions amidst a global pandemic.
The OHSU DMICE team was specifically recognized for feature interpretability and design. The team used a common set of predictors, including demographics, laboratory values, and associated diagnosis codes, to inform an ensemble classifier that combined individual predictions from logistic regression, random forest, gradient-boosted trees, and artificial neural network models. They used Shapley Additive Explanations to provide individual-level and population-level explanations for model predictions. This high-performing approach provides clinicians with an outcome prediction and an individualized explanation with predictors for intervention. The DMICE team is interested in exploring the ways this pediatric COVID-19 model could be used at the bedside to help providers better care for the children of Oregon and southwest Washington.
