Arterial oxygen tensions in mechanically ventilated ICU patients and mortality: a retrospective, multicentre, observational cohort study

Background: Supplemental oxygen therapy is commonly required for respiratory failure requiring mechanical ventilation in the ICU. However, hyperoxaemia may be injurious and may increase mortality. We evaluated the relationship amongst the degree of hyperoxaemia and changes in fraction of inspired oxygen (FIO₂) in response to hyperoxaemia, as well as associations with mortality in mechanically ventilated ICU patients. Methods: We retrospectively identified all invasively mechanically ventilated patients admitted to five ICUs, and retrieved all oxygen tension (PaO₂) and FIO₂ data. We assessed the time between arterial blood gas (ABG) samples, proportions of patients with hyperoxaemia, and changes in FIO₂ when hyperoxaemia was present. The primary outcome was the association between PaO₂ (assessed by mechanically ventilated exposure-time-divided area under the curve [AUC]) and mortality (in-ICU and post-ICU discharge) using a multistate illness–death model with transition intensities estimated by Cox proportional hazards models. Results: We assessed 177 769 ABG analyses obtained from 4998 patients between January 2012 and June 2016. The median time between ABGs was 3 h (inter-quartile range: 2–4 h); the median PaO₂ was 11.3 kPa (9.8–13.6 kPa), and FIO₂ was 0.40 (0.35–0.50). Hyperoxaemia (PaO₂ >13.7 kPa) was present in 23.9% of the ABGs, and hyperoxaemia seemed to be disregarded when FIO₂ was <0.40, as >50% of these FIO₂ values were not subsequently reduced. AUC PaO₂ >16.0 kPa was associated with increased ICU mortality (adjusted hazard ratio: 1.75; 95% confidence interval: 1.28–2.40). Conclusions: In mechanically ventilated ICU patients, hyperoxaemia was common. Although oxygen supplementation was often reduced when hyperoxaemia was observed, several patients remained hyperoxaemic. Hyperoxaemia was associated with increased ICU mortality in these patients.