Extracorporeal Membrane Oxygenation (ECMO) Risk Stratification in Newborns with Congenital Diaphragmatic Hernia (CDH).

BACKGROUND: A means for early postnatal stratification of ECMO risk in CDH newborns could be used to comparatively assess the utilization and outcomes of ECMO use between centers. While multiple CDH mortality risk calculators are available, no validated tool exists specifically for prediction of ECMO use. The purpose of this study was to derive and validate an ECMO risk stratification model.

METHODS: The study population was obtained from CDH Study Group registry for the period between 2007 and 2016. Only centers offering ECMO were included. The cohort was restricted to ECMO candidates and then divided into derivation and validation sets. Using all relevant perinatal predictors in the registry, univariate analysis was performed for the composite outcome of ECMO use or death without ECMO use. The model was derived using the derivation cohort with multivariable logistic regression and automatic stepwise forward selection (P < 0.05 for qualifying variables), and a c-statistic was obtained. The model was then tested on the validation cohort. Sample reuse validation and bootstrap validation were performed. The validated model was then tested for accuracy on CDH subgroups.

RESULTS: There were 1992 patients in the derivation cohort. Four significant variables were identified in the final ECMO risk model: 1-min and 5-min Apgar scores and highest and lowest post-ductal partial pressure of CO2 during the first 24 h of life. The model c-statistic was 0.824 which was confirmed with cross-validation and bootstrap optimism correction. The validation cohort c-statistic was 0.823 (N = 993). The model had good discrimination for left and right CDH, inborn and outborn patients, patients born before and after 2011, and high and low volume centers. The model performed significantly better for postnatally diagnosed patients.

CONCLUSIONS: This study represents proof-of-concept that a risk model can accurately estimate the probability of ECMO use in CDH newborns. This stratification could assist centers as a metric for assessment of ECMO usage and outcomes. Refinement and prospective validation of this model should be carried out prior to clinical application.

LEVEL OF EVIDENCE: 3.