Objective: To examine the Extracorporeal Life Support Organization registry data base for all infants and children with respiratory syncytial virus-associated respiratory failure managed with extracorporeal life support, to delineate predictors of outcome.
Design: Retrospective cohort study.
Setting: Extracorporeal Life Support Organization data registry.
Patients: All pediatric patients treated in the United States with extracorporeal life support for severe pediatric respiratory syncytial virus-associated respiratory failure reported to the registry, from 1982 through June 1992.
Interventions: Venoarterial or venovenous extracorporeal life support.
Measurements and main results: As of June 1992, fifty-three pediatric patients meeting study entry criteria were reported to the Pediatric Respiratory Failure Registry (n = 412) as having received extracorporeal membrane oxygenation (ECMO) for severe respiratory syncytial virus infection with pulmonary failure. Forty-nine percent (26/53) were successfully managed and survived to hospital discharge. The mean patient age was 5.0 +/- 8.6 months. Duration of mechanical ventilation before institution of extracorporeal life support was 8.1 +/- 6.2 days. Multivariate logistic regression analysis found four variables to be associated with patient nonsurvival at the p < 0.05 level: male gender, longer duration of mechanical ventilation before ECMO, higher peak inspiratory pressure, and lower ratio of arterial oxygen tension to fraction of inspired oxygen. Era of treatment was not associated with outcome. Receiver operator characteristic curve analysis of this multivariate model resulted in cutoff points of r = 0.5 and 0.1 that resulted in 92% sensitivity and 81% specificity (false-positive ratio 19%) and 96% sensitivity and 73% specificity (false-positive ratio 27%), respectively.
Conclusions: Predictors of outcome of severe respiratory failure caused by respiratory syncytial virus infection managed with ECMO exist, and multivariate predictive models with high sensitivity and low false-positive risk are possible. Similar mathematical models may be helpful in establishing criteria for future trials of ECMO versus conventional respiratory support.