Development and Validation of Extracorporeal Membrane Oxygenation Mortality-Risk Models for Congenital Diaphragmatic Hernia

ASAIO J. 2018 Nov/Dec;64(6):785-794. doi: 10.1097/MAT.0000000000000716.

Abstract

The purpose of our study was to develop and validate extracorporeal membrane oxygenation (ECMO)-specific mortality risk models for congenital diaphragmatic hernia (CDH). We utilized the data from the Extracorporeal Life Support Organization Registry (2000-2015). Prediction models were developed using multivariable logistic regression. We identified 4,374 neonates with CDH with an overall mortality of 52%. Predictive discrimination (C statistic) for pre-ECMO mortality model was C = 0.65 (95% confidence interval, 0.62-0.68). Within the highest risk group, based on the pre-ECMO risk score, mortality was 87% and 75% in the training and validation data sets, respectively. The pre-ECMO risk score included pre-ECMO ventilator settings, pH, prior diaphragmatic hernia repair, critical congenital heart disease, perinatal infection, and demographics. For the on-ECMO model, mortality prediction improved substantially: C = 0.73 (95% confidence interval, 0.71-0.76) with the addition of on-ECMO-associated complications. Within the highest risk group, defined by the on-ECMO risk score, mortality was 90% and 86% in the training and validation data sets, respectively. Mortality among neonates with CDH needing ECMO can be reliably predicted with validated clinical variables identified in this study. ECMO-specific mortality prediction tools can allow risk stratification to be used in research and quality improvement efforts, as well as with caution for individual case management.

Publication types

  • Research Support, N.I.H., Extramural
  • Research Support, Non-U.S. Gov't
  • Validation Study

MeSH terms

  • Extracorporeal Membrane Oxygenation / mortality*
  • Female
  • Hernias, Diaphragmatic, Congenital / mortality*
  • Hernias, Diaphragmatic, Congenital / therapy*
  • Humans
  • Infant, Newborn
  • Logistic Models
  • Male
  • Registries
  • Retrospective Studies