Comparison of six ECMO selection criteria and analysis of factors influencing their accuracy

This study compared six extracorporeal membrane oxygenation (ECMO) selection criteria in 42 neonates and analyzed factors influencing the accuracy of outcome predictions. The sensitivity of the criteria in identifying fatal cases varied from 0.44 to 0.94 and the specificity of predictions of survival ranged from 0.42 to 0.69. The criterion having the highest sensitivity had the lowest specificity and conversely the criterion with the lowest sensitivity had the highest specificity. Overall accuracy of the criteria, as measured by the total number of correct outcome predictions, differed little among the criteria (23/42 to 27/42 correct predictions). Three factors influenced predictive accuracy: 1) a primary diagnosis of congenital diaphragmatic hernia (CDH) was associated with a greater mortality (P < 0.001) and a significantly higher positive predictive value (PPV) for all criteria (P = 0.0009–0.012) than that seen in patients with other primary diagnoses; 2) calculating the alveolar‐arterial oxygen gradient using an assumed, rather than measured barometric pressure, or estimating oxygenation index using a calculated, rather than a measured, mean airway pressure, increased false positive mortality predictions in non‐CDH patients; and 3) requiring a peak inspiratory pressure (PIP) of at least 50 cm H2O in the definition of maximal medical management, rather than a PIP of 20–49 cm H2O, significantly increased the PPV for three of four criteria examined (P = 0.02–0.04). Awareness of these factors may facilitate the identification of neonates who need ECMO to survive.