Early Career
Status: Funded - Open
Jennifer Schramm, MD
Summary
BACKGROUND: Necrotizing enterocolitis (NEC) is a sometimes deadly GI infection that can complicate the care of children with severe congenital heart disease. Studies thus far have not been able to identify specific and modifiable risk factors and a different approach is needed. GAP: Traditional single variate and multivariate studies of NEC in congenital heart disease (CHD) have failed to identify modifiable risk factors to reduce rates of the disease. Machine learning and deep phenotyping are advanced artificial intelligence methods that are able to study these risk factors; especially when the disease itself is more rare. HYPOTHESIS: We hypothesize that machine learning with deep phenotyping will identify novel, modifiable risk factors for the development of NEC that will lead to a predictive model to reduce rates of the disease. METHODS: We will use a large dataset from the pediatric cardiac critical care consortium (PC4) to do univariate, multivariate and deep phenotyping to understand the risk factors associated with NEC development in children with CHD. Approximately 356 cases of necrotizing enterocolitis in infants with CHD in the PC4 database have been found across the 72 institutions that contribute data. RESULTS: Pending. IMPACT: This approach will allow us to identify multiple modifiable risk factors to work towards prevention of NEC in children with CHD. This will save the lives of our most vulnerable children and increase the use of machine learning and deep phenotyping in pediatric cardiac intensive care.