Status: Funded - Open
Emily Bucholz, MD, PhD, MPH
BACKGROUND: Infants born with single ventricle (SV) heart disease have the highest morbidity and mortality of all patients with congenital heart disease. GAP: Multiple studies have investigated risk factors for mortality and neurodevelopmental outcomes in SV heart disease; however, they have invariably faced the same methodological limitations and typically explain only 20-30% of the variance in outcomes. HYPOTHESIS: We hypothesize that the application of machine learning techniques will improve model prediction and lead to the identification of risk factors for SV outcomes that were not previously recognized. METHODS: Using data from the Pediatric Heart Network Single Ventricle Reconstruction trial and extension studies, which represent the largest cohort of children with SV heart disease followed for 15 years, we will leverage machine learning techniques to develop risk prediction algorithms for SV-related outcomes. Our primary outcome will be 1-, 5-, and 10-year transplantation-, reintervention-, and major morbidity-free survival. Secondary outcomes will include neurodevelopment and exercise performance after the Fontan operation. These analyses will build upon prior models by expanding the list of candidate predictors to include variables that have not been traditionally evaluated, by incorporating polynomial terms, interaction terms, and genetic features, and by using novel variable selection algorithms. We will then use these models to create online decision support tools and risk calculators that can be applied to individual SV patients. RESULTS: Pending. IMPACT: Better prediction of adverse outcomes in SV heart disease can help caregivers make more informed healthcare decisions, develop strategies to improve outcomes, and refine discussions with families to improve shared decision-making.