This study targets adult patients treated with high flow nasal cannula (HFNC) at emergency department (ED) of Severance hospital, Yonsei university. Patients with acute hypoxic respiratory failure presenting to the ED receive conventional oxygen therapy as initial treatment unless immediate endotracheal intubation is required. Partial rebreathing oxygen masks are mainly applied at first. If the patient's condition does not improve despite such treatment, the patient receives HFNC or endotracheal intubation. However, possible treatment range have not been studied, especially in ED. Decisions are made based on the personal experience of the medical staff in charge. Applying HFNC to patients who eventually fail can lead to delayed intubation and increased mortality. Failure prediction models such as ROX index and HACOR score have been developed due to such reasons. However, such models are mostly based on intensive care unit studies and after application of HFNC. Therefore, failure prediction model at the time before application of HFNC and efficacy of existing models in ED are necessary.

This study is a prospective observational study and follows the standard treatment guidelines applied to the patient and the judgment of the attending physician during the patient's treatment process. Immediately before applying HFNC, the patient's respiratory rate, pulse rate, blood pressure, SpO₂, PaO₂, PaCO₂, GCS score are determined, and FiO₂ is measured above upper lips using oxygen analyzer(MaxO2+AE, Maxtec, USA). From these data, ROX index (SF ratio/respiratory rate), ROX-HR (ROX index/pulse rate), POX index (PF ratio/respiratory rate), POX-HR (POX index/pulse rate), and HACOR score (Heart Rate, Acidosis, Consciousness, Oxygenation, Respiratory rate) are calculated. The settings (flow rate, FiO₂, temperature) at the time of HFNC application are also measured. The same indices and HFNC settings are checked 30 minutes, 1 hour, 2 hours, 4 hours, 6 hours, and 12 hours after applying HFNC. Modified Borg score and comfort scale using 5-point Likert scale are additionally determined at 30 minutes for patient's comfort. Primary outcome is HFNC failure at 28 days, defined by endotracheal intubation. Other outcomes include intubation in ED and mortality at 28 and 90 days collected through phone interview.

The receiver operating curve for ROX index, HACOR score, ROX-HR, and POX-HR at baseline, 30 minutes, 1 hour, 2 hours, 4 hours, 6 hours, and 12 hours are drawn for the outcomes. The area under the curve of the above indices are compared and cutoff values are chosen with maximum value of index J by the Youden's Index. A binary variable is created based on the cutoff values and multivariable logistic regression analyses are performed. Cutoff values for maximum specificity are also invested suggesting the lower limit of the indicator to which HFNC can be applied.

Inclusion Criteria:

• Patients with dyspnea and applied to HFNC in ED at Severance Hospital.

Exclusion Criteria:

• age<18
• GCS ≤ 12
• Do not resuscitate patients, whom wishes not to included in the study

1. Respiratory rate more than 20 breaths per minute
2. PF ratio<300mmHg despite conventional oxygen therapy
3. HFNC is needed clinically, such as the use of auxiliary respiratory muscles