Abstract
Background High-flow oxygen therapy (HFOT) has been widely used as an effective alternative to invasive mechanical ventilation (IMV) in some critically ill patients with COVID-19 pneumonia. This study aimed to compare different tools, including the respiratory rate and oxygenation (ROX) index, to predict HFOT failure in this setting. Methodology This single-center retrospective observational study was conducted from September to December 2020 and assessed COVID-19 patients who required HFOT as the first treatment at admission; HFOT failure was defined as IMV use. Prognostic scoring tools were as follows: the Sequential Organ Failure Assessment (SOFA), Acute Physiology And Chronic Health Evaluation (APACHE) II, and Simplified Acute Physiology Score (SAPS) III scores; C-reactive protein; lung consolidation percentage on chest CT; mean partial pressure of oxygen in arterial blood (PaO(2))/fraction of inspired oxygen (FiO(2)) ratio; and ROX index and modified ROX index, calculated using PaO(2) instead of blood oxygen saturation, within the first 24 hours after admission to the intensive care unit (ICU). These scores were analyzed using a multivariate Cox proportional hazard model; optimal cutoffs were computed using the R system for statistical computing. Results The study enrolled 52 patients, 31 (60%) of whom experienced HFOT failure. The best predictors of HFOT failure measured 24 hours after HFOT initiation were as follows: PaO(2)/FiO(2) (threshold 123.6, sensitivity 87%, specificity 81%, hazard ratio [HR] 7.76, and 95% confidence interval [CI] 2.39-17.1); ROX index (threshold 5.63, sensitivity 68%, specificity 95%, HR 6.18, and 95% CI 2.54-13.4); and modified ROX index (threshold 4.94, sensitivity 81%, specificity 90%, HR 8.16, and 95% CI 3.16-21.5) (P < 0.001 for all). Conclusions Early assessment of the ROX index, modified ROX index, and PaO(2)/FiO(2) ratio can adequately predict, with high accuracy, HFOT failure in COVID-19 patients. Because thresholds remain debated and are still not sufficiently validated, we advocate using them with caution for clinical decision-making in this context.