Validating a Neonatal Risk Index to Predict Necrotizing Enterocolitis

Abstract

Necrotizing enterocolitis (NEC) is a costly and deadly disease in neonates. Composite risk for NEC is poorly understood and consensus has not been established on the relevance of risk factors. This two-phase study attempted to validate and test a neonatal NEC risk index, GutCheck(NEC). Phase I used an E-Delphi methodology in which experts (n=35) rated the relevance of 64 potential NEC risk factors. Items were retained if they achieved predefined levels of expert consensus or stability. After three rounds, 43 items were retained (CVI=.77). Qualitative analysis revealed two broad themes: individual characteristics of vulnerability and the impact of contextual variation within the NICU on NEC risk. In Phase II, the predictive validity of GutCheck(NEC) was evaluated using a sample from the Pediatrix BabySteps Clinical Data Warehouse (CDW). The sample included infants born<1500 grams, before 36 weeks, and without congenital anomalies or spontaneous intestinal perforation (N=58,818, of which n=35,005 for empiric derivation and n=23,813 for empiric validation). Backward stepwise likelihood-ratio method regression was used to reduce the number of predictive factors in GutCheck(NEC) to 11 and derive empiric weights. Items in the final GutCheck(NEC) were gestational age, history of a transfusion, NICU-specific NEC risk, late onset sepsis, multiple infections, hypotension treated with Inotropic medications, Black or Hispanic race, outborn status, metabolic acidosis, human milk feeding on both day 7 and day 14 (reduces risk) and probiotics (reduces risk).Discrimination was fair in the case-control sample (AUC=.67, 95% CI .61-.73) but better in the validation set (AUC=.76, 95% CI .75-.78) and best for surgical NEC (AUC=.84, 95% CI .82-.84) and infants who died from NEC (AUC=.83, 95% CI .81-.85). A GutCheck(NEC) score of 33 (range 0-58) yielded a sensitivity of .78 and a specificity of .74 in the validation set. Intra-individual reliability was acceptable (ICC (19) =.97, p<.001). Future research is needed to repeat this procedure in infants between 1500 and 2500 grams, complete psychometric testing, and explore unit variation in NEC rates using a comprehensive approach.