Prognostic Performance of PaO2/FIO2 versus SpO2/FIO2 Ratios in Acute Hypoxemic Respiratory Failure

Concurrently measured PaO2/FIO2 (PF) and SpO2/FIO2 (SF) exhibit moderate-to-high correlation. However, their performance for risk stratification when using PF if available, or SF if not, as operationalized in recent acute respiratory distress (ARDS) and sequential organ failure assessment score (SOFA-2) guidance, is unknown.

Measure the prognostic performance of PF- and SF-defined hypoxemic respiratory failure severity.

We conducted a retrospective observational study of patients receiving invasive or non-invasive mechanical ventilation, or high-flow nasal oxygen across 58 Common Longitudinal ICU data Format (CLIF) consortium hospitals. We used the lowest PF, if available, or SF if not, from the first 24 hours of respiratory support to define ARDS and SOFA-2 severity categories. We characterized risk of in-hospital death or hospice discharge by day 28 by severity category, and evaluated prognostic performance using logistic regression models with area under the receiver operating characteristic curves (AUROC). We then evaluated whether PF- or SF-defined patients' risk of death or hospice differed after adjustment for other patient risk factors.

Oxygenation severity was defined with PF in n = 40,820 patients, and SF in n = 16,906. In-hospital death or hospice discharge was higher for PF (33.4%) versus SF (23.4%) patients, overall, and within each severity category. The SF-AUROC was numerically higher (0.60 [95%-CI:0.58-0.61]) than PF-AUROC (0.57 [95%-CI:0.57-0.58]). After adjustment, there was no statistical difference in outcome between PF and SF-defined patients.

PF and SF thresholds for severity categories recommended as equivalent in ARDS and SOFA-2 guidance define populations with different mortality, but have similar risk discrimination.