(1154) High-Flow Versus Noninvasive Ventilation in Hypoxic Respiratory Failure: A Propensity-Matched Study

Introduction: Acute hypoxemic respiratory failure (AHRF) is a common cause of ED visits in the US. While both high flow nasal cannula (HFNC) and non-invasive ventilation (NIV) are used in treatment, the optimal non-invasive respiratory support (NIRS) modality is unclear. In addition, determining clinically relevant outcomes for evaluating NIRS remains a challenge. We compared the effectiveness of HFNC vs NIV in ED patients with AHRF using a novel Win Ratio.

Methods: This is a retrospective propensity-score matched (PSM) cohort study of patients with AHRF in the University of Michigan ED (1/2018-12/2022). Patients treated acutely with HFNC or NIV within 24 hours of ED arrival were included. We matched patients 1:1 using a propensity score for odds of receiving NIV derived by a logistic regression model with pre-specified baseline characteristics. The primary outcome was composite major adverse pulmonary events (MAPE): 28-day mortality, ventilator-free days (VFD), and NIRS hours within 72 hours. We compared the effectiveness of HFNC and NIV on both individual outcomes and the composite MAPE variable using a Win Ratio, a hierarchical approach to composite outcomes that allows ranking of outcomes by clinical importance. To calculate the Win Ratio, all possible pairs of patients on HFNC vs NIV were generated; each possible pair was compared first on 28-day mortality, then VFD, then NIRS hours.

Results: Of 1,292 patients who met study inclusion, 795 (61.5%) received HFNC and 470 (36.4%) received NIV. We were able to match 736/1,292 (57.0%) patients, 368/795 (45.9%) on HFNC and 368/470 (78.3%) on NIV. After PSM, baseline characteristics were well-balanced, with all standard mean differences (SMDs) < 0.1 except BMI (SMD 0.109). There was no difference between NIV and HFNC in 28-day mortality (17.7% vs 23.1%, p=0.08) or VFD (median [IQR]: 28 [25,28] vs 28 [13, 28], p=0.50), but patients on NIV required fewer NIRS hours (median 7 vs 13 hours, p< 0.001). However, in Win Ratio analysis, NIV was more effective on MAPE than HFNC: overall wins NIV 54.6% vs HFNC 43.5%; Win Ratio 1.26, 95%CI 1.06-1.49, p< 0.001.

Conclusions: In this observational study of ED patients with AHRF, NIV was associated with lower composite MAPE compared to HFNC. Use of the Win Ratio allowed detection of clinically important outcomes.