Association of extreme hyperoxemic events and mortality in pediatric critical care: an observational cohort study

Objective: Our aim was to confirm whether extreme hyperoxemic events had been associated with excess mortality in our diverse critical care population. Methods: Retrospective analysis of 9 years of data collected in the pediatric and cardiothoracic ICUs in Children's Hospital Los Angeles was performed. The analysis was limited to those mechanically ventilated for at least 24 h, with at least 1 arterial blood gas measurement. An extreme hyperoxemic event was defined as a PaO2 of >= 300 torr. Multivariable logistic regression was used to assess the association of extreme hyperoxemia events and mortality, adjusting for confounding variables. Selected a-priori, these were Pediatric Risk of Mortality III predicted mortality, general or cardiothoracic ICU, number of blood gas measurements, as well as an abnormal blood gas measurements (pH < 7.25, pH > 7.45, and PaO2 < 50 torr). Results: There were 4,003 admissions included with a predicted mortality of 7.1% and an actual mortality of 9.7%. Their care was associated with 75,129 blood gas measurements, in which abnormal measurements were common. With adjustments for these covariates, any hyperoxemic event was associated with excess mortality (p < 0.001). Excess mortality increased with multiple hyperoxemic events (p < 0.046). Additionally, treatment resulting in SpO2 > 98% markedly increased the risk of a hyperoxemic event. Conclusion: Retrospective analysis of critical care admissions showed that extreme hyperoxemic events were associated with higher mortality. Supplemental oxygen levels resulting in SpO2 > 98% should be avoided.