Background ROX index after high-flow nasal cannula (HFNC) initiation can be helpful to identify low-risk patients of HFNC failure in patients with acute hypoxemic respiratory failure. However, it remains unknown which variables are useful for predicting HFNC failure in coronavirus disease 2019 (COVID-19) receiving HFNC. This study aims to determine the discrimination ability of the ROX index and SpO2/FiO2 ratio in COVID-19 patients after HFNC initiation. Methods This multicenter study was conducted at seven university-affiliated hospitals in Korea. We retrospectively reviewed the hospitalized patients diagnosed with COVID-19 between February 10, 2020 and February 28, 2021. The ROX index and SpO2/ FiO2 ratio were calculated at 1 hour and 4 hours after HFNC initiation. The primary outcome was HFNC failure, defined as the subsequent use of mechanical ventilation despite using HFNC. Discrimination of prediction for HFNC failure was evaluated by receiver operating characteristic (ROC) curve analysis. Results During the study period, 1,565 patients with COVID-19 were hospitalized and eligible 133 patients who received HFNC were analyzed. Among them, 63 (47.4%) were successfully weaned from HFNC, and 70 (52.6%) were intubated. SpO2/FiO2 ratio at 1 hour after HFNC initiation was a more accurate predictor of HFNC failure compared to ROX index (AUC 0.762; 95% CI: 0.679- 0.846 vs. 0.733; 95% CI: 0.640-0.826) (Figure 1). In the multivariable analysis, patients older than 70 years old had 3.4 times of HFNC failure compared to younger patients (AUC 3.367 [95% CI: 1.358-8.349], p=0.009), and SpO2/FiO2 at 1 hour (AUC 0.983 [95% CI: 0.972-0.994], p=0.003) were associated with HFNC failure. Conclusions SpO2/FiO2 ratio at 1 hour after HFNC initiation demonstrated a useful predictive ability for HFNC failure. SpO2/FiO2 could be a bedside tool to identify the need for mechanical ventilation in COVID-19 patients receiving HFNC.