Current issues and trends in Respiratory therapy

High-Flow Nasal Cannula Therapy in COVID-19: Using the ROX Index to Predict Success

Abhimanyu Chandel, Saloni Patolia, A Whitney Brown, A Claire Collins, Dhwani Sahjwani, Vikramjit Khangoora, Paula C Cameron, Mehul Desai, Aditya Kasarabada, Jack K Kilcullen,

Steven D Nathan, and Christopher S King

BACKGROUND: Optimal timing of mechanical ventilation in COVID-19 is uncertain. We sought to

evaluate outcomes of delayed intubation and examine the ROX index (ie, [SpO2=FIO2]/breathing fre- quency) to predict weaning from high-flow nasal cannula (HFNC) in patients with COVID-19.

METHODS: We performed a multicenter, retrospective, observational cohort study of subjects with

respiratory failure due to COVID-19 and managed with HFNC. The ROX index was applied to pre-

dict HFNC success. Subjects that failed HFNC were divided into early HFNC failure (^ 48 h of

HFNC therapy prior to mechanical ventilation) and late failure (> 48 h). Standard statistical compari- sons and regression analyses were used to compare overall hospital mortality and secondary end

points, including time-specific mortality, need for extracorporeal membrane oxygenation, and ICU

length of stay between early and late failure groups. RESULTS: 272 subjects with COVID-19 were

managed with HFNC. One hundred sixty-four (60.3%) were successfully weaned from HFNC, and

111 (67.7%) of those weaned were managed solely in non-ICU settings. ROX index >3.0 at 2, 6, and 12 hours after initiation of HFNC was 85.3% sensitive for identifying subsequent HFNC success. One

hundred eight subjects were intubated for failure of HFNC (61 early failures and 47 late failures).

Mortality after HFNC failure was high (45.4%). There was no statistical difference in hospital mor-

tality (39.3% vs 53.2%, P 5 .18) or any of the secondary end points between early and late HFNC failure groups. This remained true even when adjusted for covariates. CONCLUSIONS: In this ret-

rospective review, HFNC was a viable strategy and mechanical ventilation was unecessary in the

majority of subjects. In the minority that progressed to mechanical ventilation, duration of HFNC

did not differentiate subjects with worse clinical outcomes. The ROX index was sensitive for the

identification of subjects successfully weaned from HFNC. Prospective studies in COVID-19 are

warranted to confirm these findings and to optimize patient selection for use of HFNC in this

disease. Key words: COVID-19; SARS-CoV-2; high-flow nasal cannula; hypoxemic respiratory failure; viral pneumonia; respiratory insufficiency. [Respir Care 2021;66(6):909–919. © 2021 Daedalus Enterprises]

Introduction

Patients with coronavirus disease 2019 (COVID-19) face

substantial morbidity and mortality related to viral pneumo-

nitis that can progress to ARDS.1 The optimal management

strategy for respiratory failure related to the novel severe

acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is

still evolving. Patients with COVID-19 who require mechan-

ical ventilation are at high risk for poor outcomes and have a

likelihood of mortality estimated at approximately 40%.2

Dr Chandel is affiliated with the Department of Pulmonary and Critical

Care, Walter Reed National Military Medical Center, Bethesda,

Maryland. Dr Patolia is affiliated with the Virginia Commonwealth

University School of Medicine, Richmond, Virginia. Drs Brown,

Khangoora, Nathan, and King are affiliated with the Department of

Advanced Lung Disease and Transplant, Inova Fairfax Hospital, Falls

Church, Virginia. Dr Collins is affiliated with Advanced Lung Disease

Research, Inova Fairfax Hospital, Falls Church, Virginia. Dr Sahjwani is

affiliated with the Department of Pediatrics, Inova Fairfax Hospital,

Falls Church, Virginia. Ms Cameron is affiliated with Respiratory

Therapy, Inova Fairfax Hospital, Falls Church, Virginia. Drs Desai,

RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6 909

Though overall mortality of the disease, including the mor-

tality of patients in the ICU, has decreased over the course of

the pandemic, COVID-19 remains a significant burden on

the worldwide health care infrastructure.3 Mortality may be

related to the progressive course of the viral infection, but it

could be perpetuated by the inherent complications of me-

chanical ventilation itself.

High-flow nasal cannula (HFNC) devices can deliver

warmed, humidified oxygen at flows up to 60 L/min and

FIO2 up to 1.0. This modality of oxygen delivery can

reduce the need for intubation and mechanical ventila-

tion for patients with acute hypoxemic respiratory fail-

ure.4,5 Data also suggest that early use of this therapy

may decrease the need for invasive mechanical ventila-

tion in COVID-19.6 Success of HFNC can be predicted

by the ROX index (ie, [SpO2=FIO2]/breathing frequency), which is a score that has been validated in the treatment

of pneumonia and ARDS. This clinical score was ini-

tially applied based on clinical data at 2 h, 6 h, and 12 h

after application of HFNC.7 The score has been subse-

quently applied to the use of HFNC in the treatment of

COVID-19, and investigators have proposed values that

correlate with subsequent failure of HFNC and need for

endotracheal intubation.8-11 Most prior research related

to HFNC use in patients with COVID-19 has focused

efforts on utilizing the ROX index to identify patients at

risk of subsequent endotracheal intubation, and data

regarding the use of the index to select patients who may

ultimately be weaned from HFNC are lacking.

Substantial controversy exists as to the optimal timing

of initiation of invasive mechanical ventilation in the

management of COVID-19 respiratory failure. Some have

argued for more aggressive, early intubation to avoid pos-

sible patient self-induced lung injury.12-14 Others have

advocated for longer trials of noninvasive supplemental

oxygen modalities as a means to avoid endotracheal intu-

bation and associated complications.15,16 Thus, despite

possible hazards associated with delayed intubation, many

clinicians have utilized extended trials of HFNC in

patients with COVID-19 respiratory failure.16,17 The aim

of this study was to evaluate predictors of successful

weaning and overall outcomes in subjects managed with

HFNC for the support of respiratory failure related to

COVID-19.

Methods

Study Population

We performed a multicenter, retrospective, observational

study of subjects treated for acute respiratory failure second-

ary to COVID-19 and managed with HFNC within the Inova

Health System. The Inova Health System consists of 5 hospi-

tals, including a large tertiary care center and 4 community

hospitals. Subjects were included if they were $ 18 y old,

had a laboratory-confirmed diagnosis of COVID-19 by poly-

merase chain reaction testing, and were treated with HFNC

for $ 2 h. Patients were excluded if endotracheal intubation

was performed prior to initiation of HFNC (eg, following

extubation to reduce the risk of re-intubation) or performed

on an elective basis (eg, for elective surgical care). To mini-

mize heterogeneity of the studied population, patients who

were switched to noninvasive ventilation prior to endotra-

cheal intubation were also excluded. Given the objective to

compare outcomes associated with early versus late endotra-

cheal intubation, patients for whom endotracheal intubation

was not within their goals of care were also excluded.

QUICK LOOK

Current knowledge

High-flow nasal cannula (HFNC) is routinely used as

part of the care of patients with respiratory failure

related to COVID-19. Significant debate exists as to

the optimal timing of progression to invasive mechani-

cal ventilation in the event of clinical worsening or fail-

ure to wean from HFNC.

What this paper contributes to our knowledge

In this multicenter, observational, cohort study, HFNC

was frequently successful in avoiding the need for

invasive mechanical ventilation. The ROX index (ie,

[SpO2=FIO2]/breathing frequency) was sensitive for the identification of subjects who could be managed with

HFNC without the subsequent need for endotracheal

intubation. Clinical outcomes did not differ between

subjects based on the duration of HFNC therapy prior

to the initiation of mechanical ventilation. Extended

use of HFNC may be reasonable in the care of patients

with COVID-19 as a measure to avoid invasive me-

chanical ventilation.

Kasarabada, and Kilcullen are affiliated with Medical Critical Care

Service, Inova Fairfax Hospital, Falls Church, Virginia.

The authors have disclosed no conflicts of interest.

Correspondence: Abhimanyu Chandel MD, Walter Reed National Military

Medical Center, Department of Pulmonary and Critical Care, 8901 Rockville

Pike, Bethesda, MD, 20814. E-mail: [email protected].

DOI: 10.4187/respcare.08631

SEE THE RELATED EDITORIAL ON PAGE 1044

HFNC FOR COVID-19 RESPIRATORY FAILURE

910 RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6

Data were collected for subjects admitted to the Inova

Health System between March 1, 2020, and June 9, 2020.

The study was approved by the institutional review board

(U20-06-4134) at Inova Fairfax Hospital. All data were col-

lected from the electronic medical record.

Inova Health System’s COVID-19 Management

Protocol

The strategy for the management of acute respiratory fail-

ure was fairly homogenous across our health care system.

Efforts were made to avoid intubation where feasible with the

use of HFNC (Optiflow, Fisher & Paykel, Auckland, New

Zealand). Noninvasive ventilation was largely avoided early

on due to concerns regarding aerosolizing the SARS-CoV-2

virus but was increasingly utilized over time. Inhaled nitric

oxide was delivered in a blend with oxygen via HFNC, and

self-proning was incorporated where deemed clinically

appropriate. Failure of HFNC was defined as the need for me-

chanical ventilation despite HFNC application. The need for

endotracheal intubation after HFNC was at the discretion of

the treating clinician, but it was generally based on the pres-

ence of hypoxemia with a failure to maintain SpO2 > 88% de- spite receiving the maximum FIO2 allowed by the HFNC,

breathing frequency > 35 breaths/min with associated respi- ratory distress, severe metabolic acidosis, cardiopulmonary

arrest, or altered mental status requiring intubation for avoid-

ance of aspiration. In the event of the need for mechanical

ventilation, subjects were typically managed initially with

moderate PEEP (10–12 cm H2O) and a lung-protective venti-

lator strategy. Neuromuscular blockade and prone positioning

were frequently utilized in subjects with severe ARDS. The

choice of sedation and analgesia was at the discretion of the

attending intensivist and was targeted to a Richmond

Agitation Sedation Scale of 0 to –2.18 Subjects were consid-

ered for extracorporeal membrane oxygenation (ECMO) if

they were < 60 y old, were on invasive mechanical ventila- tion for < 10 d, had SpO2=FIO2 < 100, and failed lung-protec- tive ventilation despite neuromuscular blockade and prone

positioning.

Adjunct therapeutics targeting COVID-19 disease were

administered at the discretion of the attending physician and

commonly included systemic glucocorticoids and remdesi-

vir. The use of convalescent plasma was infrequent during

the study period. Given high patient volumes related to the

COVID-19 pandemic across the Inova Health system,

changes in the usual protocol for treatment and monitoring

of patients with respiratory failure at our facilities were nec-

essary. All patients managed with invasive mechanical venti-

lation were treated in an intensive care environment.

However, expansion of the level of acuity managed outside

of an intensive care setting was required, and many subjects

were managed with HFNC in augmented step-down units up

to the point of requiring endotracheal intubation.

Data Collection

Data were abstracted in a structured format by 3 of the

authors (AC, SP, and DS), including demographics, comor-

bid diseases (as documented in the admitting history and

physical), and clinical data (eg, vital signs within 1 h prior

to HFNC application and for 12 h thereafter, common labo-

ratory results, and illness severity as estimated with the

Sequential Organ Failure Assessment [SOFA]). The ROX

index was calculated and recorded at 2 h, 6 h, and 12 h after

HFNC application. Laboratory data were collected when

available within 6 h of initiation of HFNC. Adjunctive

measures provided while subjects were receiving HFNC,

such as the use of prone positioning or the administration

of inhaled nitric oxide, remdesivir, or systemic steroids

(ie, the equivalent of prednisone $ 20 mg/d) were also recorded. The primary outcome examined was overall hos-

pital mortality. Secondary outcomes included the need for

ECMO, mortality at 14 d and at 28 d after HFNC and endo-

tracheal intubation, and ICU length of stay. Data were also

collected and compared for ICU-related complications,

including the development of ventilator-associated pneu-

monia (ie, a combination of new or progressive radio-

graphic infiltrate with a positive respiratory specimen and a

clinically documented diagnosis), pneumothorax, second-

ary infection (ie, a positive culture or related microbiologic

data thought to be pathologic by the treating clinician),

acute kidney injury (ie, a rise in serum creatinine of $ 0.3 mg/dL over 48 h), need for renal replacement therapy, and

imaging-confirmed venous thromboembolism (ie, based on

the finalized radiographic report or documented point-of-

care ultrasound findings in subjects with acute decompen-

sation and suspected pulmonary embolism).

Subjects were first divided into those managed with

HFNC who were successfully weaned from this modality

and those who were ultimately intubated. Those who under-

went endotracheal intubation after HFNC failure were then

divided into 2 groups; early failure (defined as # 48 h of HFNC therapy prior to endotracheal intubation) and late

failure (intubation after > 48 h of HFNC therapy).

Statistical Analysis

Distribution of all continuous data were examined for

normality using visual inspection and the Wilk-Shapiro

test. Characteristics of the groups are presented using the

mean 6 SD for normally distributed data and compared between groups using the 2-sample t test. Data that were not normally distributed are presented as median (interquar-

tile range) and compared using the Wilcoxon rank-sum

test. Categorical data are presented as counts with prop-

ortions and compared using the Fisher exact test (2-tailed).

The diagnostic accuracy of the ROX index to predict suc-

cess of HFNC (ie, application without subsequent need for

HFNC FOR COVID-19 RESPIRATORY FAILURE

RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6 911

mechanical ventilation) is presented using a receiver oper-

ating characteristic curve together with sensitivity, specific-

ity, and predictive values at the defined cutoffs, and

summarized using the area under the curve together with

the 95% CI. To compare clinical outcomes between early

and late HFNC failure, we performed logistic regression

(overall ICU mortality, 14-d mortality, and 28-d mortality).

ICU length of stay demonstrated a positively skewed distri-

bution. To minimize the effects of outliers and to account

for this distribution, negative binomial regression was uti-

lized to compare this outcome. P values < .05 were consid- ered statistically significant. Univariate and multivariate

logistic regression analysis of factors possibly associated

with mortality were performed. Variables were included in

the model if they were statistically significant based on uni-

variate analysis and subsequently removed by means of the

stepwise backward elimination method with P < .15. Outcome data were available for all subjects at the time of

analysis. Any missing clinical data were handled via com-

plete case analysis (only cases with available data were ana-

lyzed). All statistical analyses were performed using

STATA 14 (StataCorp, College Station, Texas).

Results

During the study period, our search strategy identified 393

subjects with respiratory failure secondary to COVID-19

who required the use of HFNC within the Inova Health

System. Patients who did not receive HFNC therapy prior to

endotracheal intubation (n ¼ 27), were switched to noninva- sive ventilation (n ¼ 21), were intubated for an elective rea- son (n ¼ 1), or were < 18 y old (n ¼ 6) were excluded. Given that the primary study objective was to analyze the

outcomes of subjects who ultimately underwent endotracheal

intubation, 66 patients were excluded as intubation and me-

chanical ventilation did not align with their goals of care. Of

the remaining 272 subjects, 164 (60.3%) recovered without

intubation and were weaned successfully from HFNC,

whereas 108 (39.7%) subjects were intubated after failing

HFNC, with 61 intubated after # 48 h of HFNC and 47 intu- bated after > 48 h of HFNC application (Fig. 1). The characteristics of the 164 subjects managed with

HFNC who were successfully weaned from this modality

are presented in Table 1. Compared to those who underwent

intubation, subjects who were successfully weaned from

HFNC were more likely to be younger and have no comor-

bidities. A history of active cancer, higher initial SOFA

score, higher lactate, higher procalcitonin, and lower neutro-

phil to lymphocyte ratio were all associated with subsequent

failure of HFNC. Subjects weaned successfully from HFNC

received this therapy for longer and had a higher median

ROX index at the defined cutoffs compared to those subjects

who required mechanical ventilation. None of the subjects

successfully weaned from HFNC died prior to hospital dis-

charge. Receiver operator curves based on the ROX index

were estimated at 2 h, 6 h, and 12 h after initiation of HFNC

Patients with confirmed COVID-19 respiratory failure treated with

HFNC 393

Subjects enrolled 272

Improved and weaned from HFNC 164 (60.3%)

Intubated after HFNC failure 108 (39.7%)

Failure of HFNC ≤48 h 61 (56.5%)

Failure of HFNC >48 h 47 (43.5%)

Do-not-intubate: 66 Switched to NIV: 21 No trial of HFNC prior to endotracheal intubation: 27 Electively intubated for surgical procedure: 1 Age <18 y: 6

Excluded 121

Figure 1. Flow chart. HFNC ¼ high-flow nasal cannula, NIV ¼ noninvasive ventilation.

HFNC FOR COVID-19 RESPIRATORY FAILURE

912 RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6

to predict the success of HFNC. Overall diagnostic accuracy

was good, and this improved with a longer duration of

HFNC application (Fig. 2). The diagnostic accuracy of a

ROX index at 12 h was the best (area under the curve 0.78

[95% CI 0.72–0.84]), and an index of > 3.67 had a sensitiv-

ity of 84.1%, specificity of 49.4%, positive predictive value

of 71.5%, and a negative predictive value of 67.1% for pre-

dicting success of HFNC, thus satisfying the closest-to-(0,1)

criterion for threshold selection. For subjects who were not

intubated or weaned from HFNC within the first 12 h after

HFNC initiation, a ROX index > 3.0 at each time point (ie,

2 h, 6 h, and 12 h) had a sensitivity of 85.3%, specificity of

51.1%, positive predictive value of 75.5%, and a negative

predictive value of 66.7% for the subsequent success of

HFNC.

The characteristics of the 108 subjects intubated after

HFNC failure are displayed by group in Table 2. The mean

age was 60 y, and the majority were male (69.4%) and non-

White (87.0%). Most had comorbidities (78.7%), of which

the most common were hypertension (48.1%), diabetes

mellitus (41.7%), and hyperlipidemia (31.5%). Most clini-

cal characteristics were similar between the 2 groups; how-

ever, SOFA score was significantly higher in the early

HFNC failure group compared to the late HFNC failure

Table 1. Baseline Characteristics of Subjects Treated With HFNC

All Subjects

(n ¼ 272) Weaned from HFNC

(n ¼ 164) HFNC Failure

(n ¼ 108) P

Age, y 57 6 13 54 6 14 60 6 13 < .001

Female 92 (33.8) 60 (36.6) 32 (29.6) .24

Race, non-White 248 (91.2) 154 (93.9) 94 (87.0) .08

Body mass index, kg/m2 28.7 (25.2–33.4) 28.6 (25.5–33.2) 28.7 (24.9–33.6) .90

HFNC duration, d 3 (1–6) 4 (2–7) 2 (1–4) < .001

Comorbid diseases

No comorbid disease 83 (3.5) 60 (36.6) 23 (21.3) .01

Hypertension 116 (42.6) 64 (39.0) 52 (48.1) .17

Diabetes mellitus 101 (37.1) 56 (34.1) 45 (41.7) .25

Chronic kidney disease 20 (7.4) 8 (4.9) 12 (11.1) .061

End-stage renal disease 8 (2.9) 4 (2.4) 4 (3.7) .72

Coronary artery disease 9 (3.3) 5 (3.0) 4 (3.7) .74

Hyperlipidemia 74 (27.2) 40 (24.4) 34 (31.5) .21

Asthma 13 (4.8) 9 (5.5) 4 (3.7) .57

COPD 2 (0.7) 1 (0.6) 1 (0.9) > .99

Active cancer 7 (2.6) 1 (0.6) 6 (5.6) .02

HFrEF 4 (1.5) 2 (1.2) 2 (1.9) .65

Systemic anticoagulation 9 (3.3) 8 (4.9) 1 (0.9) .09

Clinical data at HFNC initiation

Heart rate, beats/min 93 (80–104) 89 (80–103) 95 (82–104) .19

Mean arterial pressure, mm Hg 89.7 6 13.0 89.3 6 12.9 9.3 6 13.2 .57

Breathing frequency, breaths/min 29 (24–36) 28 (24–36) 30 (26–37) .059

Oxygen saturation 93 (90–96) 93 (90–96) 93 (89–95) .22

SOFA score 3 (1–5) 2 (1–4) 4 (2–7) < .001

White blood cells, �109 per mL 8.3 (6.0–11.4) 8.0 (6.0–1.9) 8.9 (6.1–11.6) .40 Neutrophil to lymphocyte ratio 6.5 (4.2–11.7) 6.1 (3.9–1.6) 8.1 (4.9–12.0) .02

Lactate, mmol/L 1.7 (1.3–2.3) 1.5 (1.3–2.1) 1.9 (1.4–2.8) < .005

C-reactive protein, mg/L 16.8 (10.0–24.2) 16.7 (9.8–23.6) 17.2 (1.8–26.3) .51

D-dimer, mg/mL 1.3 (0.9–2.5) 1.3 (0.8–2.2) 1.3 (0.9–2.7) .25

Procalcitonin, ng/mL 0.3 (0.1–0.6) 0.2 (0.1–0.5) 0.3 (0.1–1.0) .033

ROX index

2 h after HFNC 4.5 (3.3–6.0) 4.9 (3.7–6.7) 3.6 (2.8–4.8) < .001

6 h after HFNC 4.6 (3.6–6.3) 5.1 (4.1–6.9) 3.9 (3.0–4.8) < .001

12 h after HFNC 4.7 (3.4–6.2) 5.3 (4.3–6.9) 3.8 (2.6–4.5) < .001

Data presented as mean 6 SD, median (interquartile range), or n (%) unless otherwise indicated.

HFNC ¼ high-flow nasal cannula HFrEF ¼ heart failure with reduced ejection fraction SOFA ¼ Sequential Organ Failure Assessment

HFNC FOR COVID-19 RESPIRATORY FAILURE

RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6 913

group. Additionally, subjects who failed HFNC late were

more likely to have received adjuvant therapies such as

self-proning (39.3% vs 72.3%, P < .001), inhaled nitric

oxide (14.8% vs 42.6%, P < .002), remdesivir (19.7% vs

40.4%, P ¼ .031), and systemic steroids (27.9% vs 53.2%, P ¼ .01) prior to intubation compared to those intubated af- ter early HFNC failure.

Clinical outcomes are summarized in Table 3. Overall

hospital mortality for subjects requiring invasive mechanical

ventilation was high (45.4%), which did not differ signifi-

cantly between the early and late failure groups (39.3% vs

53.2%, P ¼ .18). Furthermore, mortality at 14 d after initia- tion of HFNC (24.6% vs 25.5%, P > .99), at 14 d after intu- bation (24.6% vs 34.0%, P ¼ .29), at 28 d after initiation of HFNC (34.4% vs 42.6%, P ¼ .43), and at 28 d after intuba- tion (34.4% vs 51.1%, P ¼ .12) were not significantly differ- ent between the groups. ECMO requirements (13.1% vs

14.9%, P ¼ .79) and median (IQR) ICU length of stay were also similar (14 d [IQR 9–20] vs 15 d [IQR 8–23], P ¼ .95). Table 4 demonstrates the relationship between clinical

factors and overall hospital mortality for subjects intubated

after HFNC failure. In univariate regression analysis,

significant factors were age, male gender, heart rate, mean

arterial pressure, and SOFA score. After adjustment for

multiple variables, no significant difference between the

primary or secondary end points was noted for either group

(Table 5).

Additional ICU complications by early versus late HFNC

failure are displayed in Table 6. Notably, pneumothorax, sec-

ondary infection, and acute kidney injury were common,

occurring in 11.1%, 29.6%, and 55.6% of the study popula-

tion, respectively. There were no significant differences for

any of the complications between the groups.

Discussion

Our study documents the clinical outcomes of 272 sub-

jects with respiratory failure related to COVID-19 that was

treated with HFNC. A significant portion (60.3%) of subjects

with respiratory failure related to COVID-19 were managed

successfully with HFNC and never required initiation of me-

chanical ventilation. Strikingly, 111 (67.7%) of these sub-

jects were managed successfully in non-ICU settings. Of the

108 subjects treated with HFNC who ultimately required

1 0.

75 0.

25 0

0. 50

S en

si tiv

ity

1 0.

75 0.

25 0

0. 50

S en

si tiv

ity

1 0.

75 0.

25 0

0. 50

S en

si tiv

ity

0 0.25

* *

* †

0.50 0.75 1

1 - Specificity 0 0.25 0.50 0.75 1

1 - Specificity

0 0.25 0.50 0.75 1

1 - Specificity

A B

C

Figure 2. Receiver operator characteristic curves for ROX index at 2 h (A), 6 h (B), and 12 h (C) as predictor of high-flow nasal cannula success.

A: Area under the curve (AUC) ¼ 0.70 (CI 0.63–0.76). * ROX index > 3.41, 83.5% sensitivity, 42.6% specificity, positive predictive value (PPV) 68.8%, negative predictive value (NPV) 63.0%. B: AUC ¼ 0.72 (CI 0.65–0.79). * ROX index > 3.46, 89.3.% sensitivity, 41.8% specificity, PPV 69.9%, NPV 71.4% C: AUC ¼ 0.78 (CI 0.72–0.84). C: AUC ¼ 0.78 (CI 0.72–0.84). * ROX index > 3.67, 84.1.% sensitivity, 49.4% specificity, PPV 71.5%, NPV 57.1% † ROX index > 4.57, 72.4% sensitivity, 75.9% specificty, PPV 82.1%, NPV 64.6%.

HFNC FOR COVID-19 RESPIRATORY FAILURE

914 RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6

endotracheal intubation, we noted high overall mortality

(45.4%), significant use of ECMO (13.9%), and a longer me-

dian stay in the ICU of 14 d (IQR 8–21).

HFNC has previously been reported to have several posi-

tive physiologic and clinical advantages in the treatment of

acute respiratory failure. HFNC can enhance patient comfort

through a reduction of important subjective patient-reported

symptoms, including dyspnea and oral dryness, compared to

conventional oxygen delivery.4 Additionally, HFNC may

provide physiologic benefit from a reduction in patient work

of breathing and a decrease in physiologic dead space though

high air flows.19 HFNC has been used successfully in the

management of respiratory distress related to other viral ill-

nesses, and data suggest that the use of HFNC in COVID-19

has the potential to decrease the need for mechanical ventila-

tion.6,20 Avoidance of intubation may allow for a reduction

in complications commonly associated with endotracheal

intubation such as pneumonia, ventilator-associated lung

injury, or secondary infections. Furthermore, avoidance of

mechanical ventilation through the use of HFNC may help

conserve this valuable resource in the event of ventilator

shortages.

However, despite these advantages, there is concern that

poor patient selection or prolonged trials of HFNC may

Table 2. Baseline Characteristics of Subjects Intubated After HFNC Failure

All Subjects

(n ¼ 108) Early HFNC Failure

(n ¼ 61) Late HFNC Failure

(n ¼47) P

Age, y 60 6 13 58 6 13 62 6 11 .07

Female 33 (3.6) 18 (29.5) 15 (31.9) .84

Race, non-White 94 (87.0) 55 (9.2) 39 (83.0) .39

Body mass index, kg/m2 28.7 (24.9–33.6) 3.2 (26.3–35.7) 27.9 (23.5–32.9) .08

HFNC duration, d 2 (1, 4) 1 (0, 1) 4 (3, 8) < .001

Comorbid diseases

No comorbid disease 23 (21.3) 17 (27.9) 6 (12.8) .063

Hypertension 52 (48.1) 25 (41.0) 27 (57.4) .12

Diabetes mellitus 45 (41.7) 23 (37.7) 22 (46.8) .43

Chronic kidney disease 12 (11.1) 7 (11.5) 5 (1.6) > .99

End-stage renal disease 4 (3.7) 3 (4.9) 1 (2.1) .63

Coronary artery disease 4 (3.7) 2 (3.3) 2 (4.3) > .99

Hyperlipidemia 34 (31.5) 16 (26.2) 18 (38.3) .21

Asthma 4 (3.7) 2 (3.3) 2 (4.3) > .99

COPD 1 (.9) 1 (1.6) 0 (0) > .99

Active cancer 6 (5.6) 5 (8.2) 1 (2.1) .23

HFrEF 2 (1.9) 0 (0) 2 (4.3) .19

Systemic anticoagulation 1 (.9) 1 (1.6) 0 (0) > .99

Clinical data at HFNC initiation

Heart rate, beats/min 95 (82–104) 99 (85–104) 93 (80–100) .22

Mean arterial pressure, mm Hg 9.3 6 13.2 90.4 6 13.6 9.1 6 12.9 .91

Breathing frequency, breaths/min 30 (26–37) 30 (25.5–37) 31 (26–37) .65

Oxygen saturation 93 (89–95) 93 (88–94) 93 (90–96) .42

SOFA score 4 (2–7) 5 (2–8) 4 (2–5) .02

White blood cells, �109 per mL 8.9 (6.1–11.6) 9.2 (6.1–11.5) 8.4 (6.1–11.9) .93 Neutrophil to lymphocyte ratio 8.1 (4.9–12.0) 9.0 (4.3–12.9) 7.4 (5.6–11.6) .80

Lactate, mmol/L 1.9 (1.4–2.8) 1.8 (1.3–2.8) 2.0 (1.5–3.0) .41

C-reactive protein, mg/L 17.2 (1.8–26.3) 18.0 (11.1–28.2) 16.7 (9.7–23.3) .47

D-dimer, mg/mL 1.3 (0.9–2.7) 1.5 (0.9–2.5) 1.2 (0.8–2.9) .97

Procalcitonin, ng/mL 0.3 (0.1–1.0) 0.3 (0.1–1.2) 0.3 (0.1–0.6) .13

Adjunctive measures prior to intubation

Self-proning 58 (53.7) 24 (39.3) 34 (72.3) < .001

Inhaled nitric oxide 29 (26.9) 9 (14.8) 20 (42.6) < .002

Remdesivir 31 (28.7) 12 (19.7) 19 (40.4) .031

Systemic steroids 42 (38.9) 17 (27.9) 25 (53.2) .01

Data presented as mean 6 SD, median (interquartile range), or n (%) unless otherwise indicated.

HFNC ¼ high-flow nasal cannula HFrEF ¼ heart failure with reduced ejection fraction SOFA ¼ Sequential Organ Failure Assessment

HFNC FOR COVID-19 RESPIRATORY FAILURE

RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6 915

Table 3. Primary and Secondary Outcomes of Subjects Intubated After HFNC Failure

All Subjects

(n ¼108) Early HFNC Failure

(n ¼ 61) Late HFNC Failure

(n ¼ 47) P

Primary outcome

Overall hospital mortality 49 (45.4) 24 (39.3) 25 (53.2) .18

Secondary outcomes

Progression to ECMO 15 (13.9) 8 (13.1) 7 (14.9) .79

Mortality at 14 days following HFNC 27 (25.0) 15 (24.6) 12 (25.5) > .99

Mortality at 14 days following intubation 31 (28.7) 15 (24.6) 16 (34.0) .29

Mortality at 28 days following HFNC 41 (38.0) 21 (34.4) 20 (42.6) .43

Mortality at 28 days following intubation 45 (41.7) 21 (34.4) 24 (51.1) .12

ICU length of stay, d 14 (8–21) 14 (9–20) 15 (8–23) .95

Data presented as n (%) or median (interquartile range).

HFNC ¼ high-flow nasal cannula ECMO ¼ extracorporeal membrane oxygenation

Table 4. Factors Associated With Overall In-Hospital Mortality in Subjects Intubated After HFNC Failure

Variables Univariate Analysis Multivariate Analysis

Odds Ratio (95% CI) P Odds Ratio (95% CI) P

Age 1.11 (1.05–1.14) < .001 1.10 (1.05–1.16) < .001

Male 2.49 (1.04–5.95) .040 2.40 (0.81–7.10) .11

Race, non-White 1.12 (0.36–3.49) .84

Body mass index, kg/m2 1.00 (0.95–1.05) .98

Comorbidities

Hypertension 1.43 (0.67–3.07) .35

Diabetes mellitus 1.09 (0.51–2.36) .82

Chronic kidney disease 1.23 (0.37–4.10) .73

Coronary artery disease 3.78 (0.38–37.58) .26

Hyperlipidemia 1.10 (0.49–2.49) .81

Obstructive lung disease 0.79 (0.13–4.96) .81

Active cancer 6.59 (0.74–58.45) .09

HFrEF 1.21 (0.07–19.83) .90

Clinical data prior to HFNC initiation

Heart rate, per 10 beats/min 1.34 (1.04–1.63) .02 1.63 (1.10–2.16) < .001

Mean arterial pressure, per 10 mm Hg 1.34 (1.10–1.97) .049

Breathing frequency, breaths/min 1.01 (0.97–1.06) .55

Oxygen saturation 0.99 (0.94–1.04) .69

SOFA score 1.12 (1.00–1.25) .049

White blood cells, 1.04 (0.95–1.13) .42

Neutrophil to lymphocyte ratio 1.01 (0.98–1.05) .56

Lactate 1.33 (0.96–1.83) .08

C-reactive protein 1.00 (0.97–1.03) .98

D-dimer 1.06 (0.96–1.19) .26

Procalcitonin 1.01 (0.99–1.03) .38

Adjunctive measures prior to intubation

Self-proning 0.71 (0.33–1.51) .37

Inhaled nitric oxide 0.80 (0.34–1.90) .61

Remdesivir 1.19 (0.51–2.73) .69

Systemic steroids 1.59 (0.73–3.46) .24

HFNC ¼ high-flow nasal cannula HFrEF ¼ heart failure with reduced ejection fraction SOFA ¼ sequential organ failure assessment

HFNC FOR COVID-19 RESPIRATORY FAILURE

916 RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6

result in worse clinical outcomes. In an observational study

prior to the emergence of SARS-CoV-2, delayed failure of

HFNC was associated with worse overall ICU mortality

and fewer ventilator-free days at day 28.21 However, it is

not clear how these prior data translate to the unique clini-

cal syndrome of COVID-19.

The use of HFNC in the treatment of COVID-19 has

become common, with multiple case series reporting

high proportions of critically ill subjects receiving this

therapy.13,22 Despite this, controversy exists regarding

the timing of progression from HFNC to mechanical

ventilation should patients fail to wean from HFNC or

their clinical condition worsen. Some have argued that

vigorous spontaneous inspiratory efforts can lead to

volutrauma and self-induced lung injury after large

swings in transpulmonary pressure and associated lung

stress. Therefore, in some instances, experts have advo-

cated that intubation should be performed as soon as

possible.12 It has been further suggested that, given the

prolonged duration of COVID-19 illness, the use of

noninvasive ventilation may have an unacceptably high

failure rate and may delay endotracheal intubation.14

However, others have argued that the liberal use of

early mechanical ventilation for the respiratory failure

associated with COVID-19 is not justified. This latter

argument has led some clinicians to consider prolonged

trials of HFNC in an effort to avoid endotracheal intu-

bation and its associated complications.15,16

Our study provides evidence that prolonged trials of

HFNC in patients with respiratory failure related to COVID-

19 may be reasonable and are not clearly associated with

adverse clinical patient outcomes. We failed to demonstrate

any difference in our primary end point (ie, overall hospital

mortality) or any of the secondary end points including the

need for ECMO, mortality at 14 d and 28 d after HFNC and

endotracheal intubation, and ICU length of stay. Despite these

findings, we recognize that the decision regarding the optimal

strategy and timing of intubation is nuanced and patient-spe-

cific. Poor patient selection, lack of appropriate monitoring,

and failure to recognize clinical deterioration in patients on

HFNC are likely to be related to adverse clinical outcomes.

Table 5. Outcomes of Subjects Intubated After HFNC Failure With Adjustment for Confounders

Unadjusted Odds Ratio (95% CI) P* Adjusted Odds Ratio (95% CI) P*

Primary outcome

Overall hospital mortality 1.75 (0.81–3.78) .15 2.13 (0.80–5.62) .13

Secondary outcomes

Progression to ECMO 1.16 (0.39–3.46) .79 1.78 (0.43–7.32) .42

Mortality at 14 d after HFNC 1.05 (0.44–2.53) .91 0.97 (0.35–2.69) .95

Mortality at 14 d after intubation 1.58 (0.68–3.66) .28 1.45 (0.53–3.96) .46

Mortality at 28 d after HFNC 1.41 (0.64–3.09) .39 1.39 (0.51–3.81) .52

Mortality at 28 d after intubation 1.99 (0.91–4.33) .08 2.53 (0.91–7.00) .07

ICU length of stay 0.95 (0.72–1.26)† .73‡ 0.92 (0.71–1.18)† .50‡

Early failure was used as the reference for comparison.

* Statistical comparison of the data were performed using logistic regression analysis. † Relative ratio by negative binomial regression analysis.

‡ Statistical comparison performed using negative binomial regression analysis.

HFNC ¼ high-flow nasal cannula ECMO ¼ extracorporeal membrane oxygenation

Table 6. Complications During ICU Stay of Subjects by Early Versus Late HFNC Failure

All Subjects

(n ¼ 108) Early HFNC Failure

(n ¼ 61) Late HFNC Failure

(n ¼ 47) P

Pneumothorax 12 (11.1) 6 (9.8) 6 (12.8) .76

VAP 19 (17.6) 10 (16.4) 9 (19.1) .80

Secondary infection 32 (29.6) 16 (26.2) 16 (34.0) .40

Acute kidney injury 60 (55.6) 32 (52.5) 28 (59.6) .56

Need for renal replacement therapy 29 (26.9) 20 (32.8) 9 (19.1) .13

Venous thromboembolism 12 (11.1) 8 (13.1) 4 (8.5) .55

Data are presented as n (%).

HFNC ¼ high-flow nasal cannula VAP ¼ ventilator-associated pneumonia

HFNC FOR COVID-19 RESPIRATORY FAILURE

RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6 917

Our results further suggest that, as in other causes of

hypoxemic respiratory failure, the ROX index has a high

sensitivity in identifying patients likely to succeed on

HFNC. It may help select patients who could benefit from

HFNC and those who could safely undergo prolonged trials

of HFNC as a means of avoiding intubation in respiratory fail-

ure related to COVID-19. The ROX index was first described

and validated in subjects with respiratory failure prior to the

outbreak of COVID-19. This index has also been applied to

predict the need for endotracheal intubation after HFNC

application in subjects with COVID-19. However, unlike our

current analysis, most other research has emphasized the iden-

tification of patients likely to ultimately fail HFNC.8-10 In a

previous cohort of subjects with COVID-19,11 in which the

ROX index was applied to predict successful weaning from

HFNC, the authors described similar model accuracy. In the

analysis of this cohort, a ROX index cutoff was identified at a

single time interval (ie, 4 h) following the application of

HFNC. Our results build on this earlier work by applying the

ROX index at multiple time intervals after the application of

HFNC and demonstrating that monitoring the ROX index

over time may aid in the identification of patients who can

ultimately be weaned from HFNC. Given the similarity in

clinical outcomes between early and late failure subjects in

our cohort, prediction of HFNC success may be of clinical

utility. Our results indicate that the index appears to perform

similarly in respiratory failure related to COVID-19 compared

to the non-COVID-19 cohort in which it was initially eval-

uated. We identified ROX index cutoffs that may be useful in

selecting patients who could be successfully weaned from

HFNC without the need for endotracheal intubation.

Subjects in the early failure group were more likely to

have a higher overall illness severity, but the magnitude of

this difference was small and may not be clinically impor-

tant. Though most of the baseline clinical characteristics

were otherwise similar between the groups, subjects with

late HFNC failure were more likely to have received adjunc-

tive therapies such as self-proning, inhaled vasodilators,

remdesivir, and steroids prior to endotracheal intubation. We

postulate this may reflect additional therapies trialed by clini-

cians in an effort to stave off mechanical ventilation, and we

acknowledge that this treatment difference may confound

outcome differences between the groups. In univariate analy-

sis, these therapies were not significantly associated with

mortality in this small sample size. Of note, no health care

workers in the Inova Health System were suspected of iatro-

genic infection with COVID-19 during the study period,

which is likely a testament to the safety of HFNC in this set-

ting, staff diligence, and the efficacy of appropriate personal

protective equipment utilized per clinical practice guidelines.

This analysis has several limitations. First, this was a

retrospective observational study. Though attempts were

made to correct for covariates, all confounders may not

have been accounted for and likely cannot be in the absence

of a randomized clinical trial. Second, this trial was per-

formed within a single hospital system. Heterogeneity in

practice patterns is likely, and similar data from multiple

hospital systems would be informative. Additionally, given

the small size of our study population, it is feasible our study

lacked the statistical power to detect differences in clinical

outcomes between the groups. Finally, although the ROX

index is well suited for application in clinical care, given the

ability to rapidly calculate it at the bedside on the basis of

universally available clinical data, other prediction models

may be more accurate or even easier to apply in clinical

practice. Efforts to validate additional predictors or combina-

tions of predictors to identify patients with COVID-19 likely

to be weaned from noninvasive ventilation is a continued

area of interest and deserves future research. Prospective tri-

als with larger sample sizes are required to further explore

these important clinical questions.

Conclusions

Respiratory failure related to COVID-19 is a unique

condition for which strategies regarding noninvasive and

invasive ventilation management are still being optimized.

In this retrospective review, we noted that HFNC was uti-

lized frequently, and many subjects with hypoxemic respi-

ratory failure related to COVID-19 did not require

intubation after management with this therapy. Prolonged

use of HFNC was not associated with worse clinical out-

comes compared with shorter trials in those who ulti-

mately required mechanical ventilation. The ROX index

was sensitive for the identification of subjects who were

successfully managed with HFNC without the subsequent

need for endotracheal intubation. A ROX index > 3.67 at 12 h after the application of HFNC was an accurate pre-

dictor of successful weaning in our cohort. Prospective

study of HFNC in COVID-19 is warranted to confirm

these findings and to optimize patient selection for use of

this device in this evolving care setting.

REFERENCES

1. Gattinoni L, Coppola S, Cressoni M, Busana M, Rossi S, Chiumello

D. COVID-19 does not lead to a “typical” acute respiratory distress

syndrome. Am J Respir Crit Care Med 2020;201(10):1299-1300.

2. Chang R, Elhusseiny KM, Yeh Y-C, Sun W-Z. COVID-19 ICU and

mechanical ventilation patient characteristics and outcomes: a system-

atic review and meta-analysis. medRxiv 2020 [Epub ahead of print]

doi: 10.1101/2020.08.16.20035691.

3. Dennis JM, McGovern AP, Vollmer SJ, Mateen BA. Improving sur-

vival of critical care patients with coronavirus disease 2019 in

England: a national cohort study. Crit Care Med Mar to Jun 2020;2020

4. Roca O, Riera J, Torres F, Masclans JR. High-flow oxygen therapy in

acute respiratory failure. Respir Care 2010;55(4):408-413.

5. Rochwerg B, Granton D, Wang DX, Helviz Y, Einav S, Frat JP, et al.

High flow nasal cannula compared with conventional oxygen therapy

for acute hypoxemic respiratory failure: a systematic review and meta-

analysis. Intensive Care Med 2019;45(5):563-572.

HFNC FOR COVID-19 RESPIRATORY FAILURE

918 RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6

6. Deng L, Lei S, Jiang F, Lubarsky DA, Zhang L, Liu D, et al. The out-

come impact of early vs late HFNC oxygen therapy in elderly patients

with COVID-19 and ARDS. medRxiv 2020. 2020.2005.2023.20111450.

7. Roca O, Caralt B, Messika J, Samper M, Sztrymf B, Hernandez G,

et al. An index combining respiratory rate and oxygenation to predict

outcome of nasal high-flow therapy. Am J Respir Crit Care Med

2019;199(11):1368-1376.

8. Mueller AA, Tamura T, Crowley CP, DeGrado JR, Haider H, Jezmir

JL, et al. Inflammatory biomarker trends predict respiratory decline in

COVID-19 patients. Cell Rep Med 2020;1(8):100144.

9. Panadero C, Abad-Fernandez A, Rio-Ramirez MT, Acosta Gutierrez

CM, Calderon-Alcala M, Lopez-Riolobos C, et al. High-flow nasal

cannula for acute respiratory distress syndrome (ARDS) due to

COVID-19. Multidiscip Respir Med 2020;15(1):693.

10. Calligaro GL, Lalla U, Audley G, Gina P, Miller MG, Mendelson M,

et al. The utility of high-flow nasal oxygen for severe COVID-19

pneumonia in a resource-constrained setting: a multi-centre prospec-

tive observational study. EClinicalMedicine 2020;28:100570.

11. Zucman N, Mullaert J, Roux D, Roca O, Ricard JD, Contributors.

Prediction of outcome of nasal high flow use during COVID-19-

related acute hypoxemic respiratory failure. Intensive Care Med

2020;46(10):1924-1926.

12. Gattinoni L, Chiumello D, Caironi P, Busana M, Romitti F, Brazzi L,

et al. COVID-19 pneumonia: different respiratory treatments for dif-

ferent phenotypes? Intensive Care Med 2020;46(6):1099-1102.

13. Meng L, Qiu H, Wan L, Ai Y, Xue Z, Guo Q, et al. Intubation and

ventilation amid the COVID-19 outbreak: Wuhan’s experience.

Anesthesiology 2020;132(6):1317-1332.

14. de Simone G, Mancusi C. COVID-19: timing is important. Eur J

Intern Med 2020;77:134-135.

15. Tobin MJ, Laghi F, Jubran A. Caution about early intubation and me-

chanical ventilation in COVID-19. Ann Intensive Care 2020;10(1):78.

16. Rola P, Farkas J, Spiegel R, Kyle-Sidell C, Weingart S, Duggan L,

et al. Rethinking the early intubation paradigm of COVID-19: time to

change gears? Clin Exp Emerg Med 2020;7(2):78-80.

17. Villarreal-Fernandez E, Patel R, Golamari R, Khalid M, DeWaters

A, Haouzi P. A plea for avoiding systematic intubation in severely

hypoxemic patients with COVID-19-associated respiratory failure.

Crit Care 2020;24(1):337.

18. Ely EW, Truman B, Shintani A, Thomason JWW, Wheeler AP,

Gordon S, et al. Monitoring sedation status over time in ICU patients:

reliability and validity of the Richmond Agitation-Sedation Scale

(RASS). JAMA 2003;289(22):2983-2991.

19. Dysart K, Miller TL, Wolfson MR, Shaffer TH. Research in high flow

therapy: mechanisms of action. Respir Med 2009;103(10):1400-1405.

20. Rello J, Perez M, Roca O, Poulakou G, Souto J, Laborda C, et al.

High-flow nasal therapy in adults with severe acute respiratory infec-

tion: a cohort study in patients with 2009 influenza A/H1N1v. J Crit

Care 2012;27(5):434-439.

21. Kang BJ, Koh Y, Lim C-M, Huh JW, Baek S, Han M, et al. Failure of

high-flow nasal cannula therapy may delay intubation and increase

mortality. Intensive Care Med 2015;41(4):623-632.

22. Bhatraju PK, Ghassemieh BJ, Nichols M, Kim R, Jerome KR, Nalla

AK, et al. Covid-19 in critically ill patients in the Seattle region - case

series. N Engl J Med 2020;382(21):2012-2022.

Thisarticle is approved forContinuingRespiratoryCareEducation credit. For information and to obtain your CRCE

(free to AARC members) visit www.rcjournal.com

HFNC FOR COVID-19 RESPIRATORY FAILURE

RESPIRATORY CARE � JUNE 2021 VOL 66 NO 6 919