Can a national dataset generate a nomogram for necrotizing enterocolitis onset?

Abstract

Mother's own milk and donor human milk use is increasing as a means of necrotizing enterocolitis (NEC) prevention. Early onset of enteral feeding has been associated with improvement of many outcomes but has not been shown to reduce the incidence of NEC. Better definition of the window of risk for NEC by gestational strata should improve resource management with respect to donor human milk and enhance our understanding of NEC timing and pathogenesis. Our objective was to establish a NEC dataset of sufficient size and quality, then build a generalizable model of NEC onset from the dataset across gestational strata. We used de-identified data from the Pediatrix national dataset and filtered out all diagnostic confounders that could be identified by either specific diagnoses or logical exclusions (example dual diagnoses), with a specific focus on NEC and spontaneous intestinal perforation (SIP) as the outcomes of interest. The median day of onset was plotted against the gestational age for each of these diagnoses and analyzed for similarities and differences in the day of diagnosis. Onset time of medical NEC was inversely proportional to gestation in a linear relationship across all gestational ages. We found the medical NEC dataset displayed characteristics most consistent with a homogeneous disease entity, whereas there was a skew towards early presentation in the youngest gestation groups of surgical NEC (suggesting probable SIP contamination). Our national dataset demonstrates that NEC onset occurs in an inverse stereotypic, linear relationship with gestational age at birth. Medical NEC is the most reliable sub-cohort for the purpose of determining the temporal window of NEC risk.