Abstract Introduction Acute hypoxemic respiratory failure (AHRF) carries hospital mortality rates of 30-65%. High-flow nasal cannula (HFNC) therapy is central to its management but shows variable failure rates, highlighting the need for early, accurate predictors of treatment success. The ROX index, though widely used, has delayed predictive capacity due to the need for serial measurements. Lung ultrasound (LUS), a real-time, radiation-free bedside tool that gained prominence during the COVID-19 pandemic, offers potential for earlier prediction of HFNC outcomes but remains underexplored. Methods This prospective observational cohort included adults (≥18 years) with AHRF initiated on HFNC therapy. Lung ultrasound was performed within 24 hours using the Clarius C3 HD3 convex scanner (2-6 MHz), and the Lung Ultrasound Score (LUS) was computed across 12 thoracic zones (range 0-36). ROX indices were recorded at 2, 6, and 12 hours. HFNC failure was defined as escalation to noninvasive or invasive ventilation, death within 28 days, or HFNC reapplication within 24 hours of weaning. Diagnostic accuracy for LUS and ROX was assessed by ROC analysis, identifying optimal cutoffs, sensitivity, specificity, and 95% confidence intervals. Results The ROX index showed modest accuracy at 2 hours (AUC 0.837), with 79.2% sensitivity and 74.5% specificity, performing better for ruling out rather than confirming HFNC failure (NPV 93.8%). Its accuracy improved significantly by 6 hours (AUC 0.963), achieving perfect sensitivity (100%) and 82.4% specificity with no missed failures, and remained stable by 12 hours (AUC 0.962; sensitivity 95.8%, specificity 94.3%). The Lung Ultrasound Score (LUS) demonstrated excellent performance (AUC 0.932) with 95.8% sensitivity, 94.1% specificity, and a LUS cutoff of 14.5. When combined, LUS and ROX achieved near-perfect discrimination (AUC 0.999), with 100% sensitivity and 98% specificity at 6 hours—representing the best balance for ruling in and ruling out HFNC failure. Conclusion Lung ultrasound provides an early, accurate, and noninvasive method for predicting HFNC failure in AHRF. Compared with the ROX index, LUS showed superior diagnostic performance within the first 24 hours, supporting earlier intervention. Combining LUS and ROX further enhanced predictive accuracy, suggesting that an integrated approach may optimize monitoring and improve outcomes. This abstract is funded by: None
Tanedo et al. (Fri,) studied this question.