ROX Index for Prediction of 48‐h Reintubation After Extubation in Medical ICU Patients Receiving High‐Flow Nasal Cannula

Background and Objective The respiratory rate–oxygenation (ROX) index predicts HFNC failure in pneumonia with hypoxemic respiratory failure, but its utility in the post‐extubation HFNC setting remains unclear. Methods We retrospectively analysed data from a prospectively enrolled observational ICU cohort at Samsung Medical Center (Seoul, Korea). Among 1881 mechanically ventilated adults, we analysed 726 extubation episodes managed with post‐extubation HFNC. The primary outcome was reintubation within 48 h. ROX was calculated hourly for 24 h after extubation. Diagnostic discrimination was assessed using area under the receiver operating characteristic curve (AUROC) and clustered receiver operating characteristic methods, stratified by PaO 2 /FiO 2 on the day of extubation. Results Reintubation occurred in 112/726 extubation episodes (15.4%). Overall, ROX demonstrated limited discrimination (AUROC 0.681 at 2 h, 0.681 at 6 h, 0.677 at 12 h, 0.730 at 18 h, and 0.614 at 24 h). Clustered AUROC varied according to hypoxaemia severity: 0.545 (PaO 2 /FiO 2 ≥ 300; n = 118), 0.684 (PaO 2 /FiO 2 200–< 300; n = 460), and 0.762 (PaO 2 /FiO 2 < 200; n = 148). In PaO 2 /FiO 2 < 200, even‐hour–derived ROX thresholds performed similarly in odd hours: rule‐in 4.21 and 6.60, balanced 9.70 (Youden), and rule‐out 12.50 and 16.96. Median time from ROX < 9.7 to reintubation was 14.1 h. Conclusion ROX demonstrated limited discriminative performance for predicting reintubation after extubation in patients receiving post‐extubation HFNC. However, discrimination varied by hypoxaemia severity, and ROX may serve as an adjunct for post‐extubation risk stratification in patients with PaO 2 /FiO 2 < 200.