Prospective validation of HACOR score for HFNO failure in emergency department patients with acute respiratory...

High-flow nasal oxygen (HFNO) failure occurs in 30-42% of emergency department (ED) patients with acute respiratory failure (ARF). The HACOR score was derived in intensive care patients on mask-based noninvasive ventilation; its performance during HFNO in unselected ED patients has not been prospectively validated.

Adults started on HFNO for ARF with arterial blood gas data at HFNO initiation (T0) and at 60 min (T60) were prospectively enrolled at a single tertiary academic ED. The primary outcome was HFNO failure (intubation or death from any cause within 48 h). We assessed discrimination (AUC, DeLong method), calibration (cross-validated slope) and clinical utility (decision curve analysis); the Youden-optimal T60 threshold was derived, bootstrap-corrected for optimism, and compared with the original cutoff (HACOR >5).

Of 332 patients (median age 71 years; 47.0% male), 117 (35.2%) failed HFNO, of whom 115 (98.3%) were intubated. T60 HACOR outperformed T0 (AUC 0.890 [95% CI 0.853-0.927] vs. 0.701 [0.645-0.757]; ΔAUC +0.189; p < 0.001), showed no evidence of miscalibration (slope 0.98; calibration-in-the-large -0.001), and showed positive net benefit at threshold probabilities of 10-60%. HACOR ≥8 at T60 gave sensitivity 86.3% and specificity 76.7% (bootstrap-corrected 85.5% and 76.3%; selected in 75.5% of resamples); HACOR >5 gave sensitivity 94.9%, specificity 42.3% and NPV 93.8%. Subgroup AUCs ranged 0.856-0.912.

The 60-min HACOR score discriminates HFNO failure well, with no evidence of miscalibration, in an unselected ED cohort, substantially outperforming the admission score. The >5 threshold remains the reference for ruling out failure; the ≥8 threshold derived here is exploratory and requires prospective validation.