Mechanically Ventilated Patients With Coronavirus Disease 2019 Had a Higher Chance of In-Hospital Death If Treated With High-Flow Nasal Cannula Oxygen Before Intubation

Michael Nurok; Oren Friedman; Matthew Driver; Nancy Sun; Abirami Kumaresan; Peter Chen; Susan Cheng; Daniel S Talmor; Joseph Ebinger

doi:10.1213/ANE.0000000000006211

. 2022 Nov 4;136(4):692–698. doi: 10.1213/ANE.0000000000006211

Mechanically Ventilated Patients With Coronavirus Disease 2019 Had a Higher Chance of In-Hospital Death If Treated With High-Flow Nasal Cannula Oxygen Before Intubation

Michael Nurok ^*,^✉, Oren Friedman ^†, Matthew Driver ^‡, Nancy Sun ^‡, Abirami Kumaresan ^*, Peter Chen ^§, Susan Cheng ^‡, Daniel S Talmor ^∥, Joseph Ebinger ^‡

PMCID: PMC9990488 NIHMSID: NIHMS1826161 PMID: 36730796

BACKGROUND:

The impact of high-flow nasal cannula (HFNC) on outcomes of patients with respiratory failure from coronavirus disease 2019 (COVID-19) is unknown. We sought to assess whether exposure to HFNC before intubation was associated with successful extubation and in-hospital mortality compared to patients receiving intubation only.

METHODS:

This single-center retrospective study examined patients with COVID-19-related respiratory failure from March 2020 to March 2021 who required HFNC, intubation, or both. Data were abstracted from the electronic health record. Use and duration of HFNC and intubation were examined‚ as well as demographics and clinical characteristics. We assessed the association between HFNC before intubation (versus without) and chance of successful extubation and in-hospital death using Cox proportional hazards models adjusting for age, sex, race/ethnicity, obesity, hypertension, diabetes, prior chronic obstructive pulmonary disease or asthma, HCO₃, CO₂, oxygen-saturation-to-inspired-oxygen (S:F) ratio, pulse, respiratory rate, temperature, and length of stay before intervention.

RESULTS:

A total of n = 440 patients were identified, of whom 311 (70.7%) received HFNC before intubation, and 129 (29.3%) were intubated without prior use of HFNC. Patients who received HFNC before intubation had a higher chance of in-hospital death (hazard ratio [HR], 2.08; 95% confidence interval [CI], 1.06–4.05). No difference was found in the chance of successful extubation between the 2 groups (0.70, 0.41–1.20).

CONCLUSIONS:

Among patients with respiratory failure from COVID-19 requiring mechanical ventilation, patients receiving HFNC before intubation had a higher chance of in-hospital death. Decisions on initial respiratory support modality should weigh the risks of intubation with potential increased mortality associated with HFNC.

KEY POINTS.

Question: Is exposure to high-flow nasal cannula oxygen therapy for patients with coronavirus disease 2019 (COVID-19) requiring intubation associated with in-hospital mortality compared to patients who are intubated without prior exposure?
Findings: Patients who received high-flow nasal cannula oxygen therapy before intubation had higher in-hospital mortality than those who were intubated without prior exposure to high-flow oxygenation.
Meaning: Among patients at high likelihood for requiring intubation, the use of high flow-nasal cannula (HFNC) is associated with higher chances of mortality.

See Article, page 689

With the onset and rapid expansion of the coronavirus disease 2019 (COVID-19) pandemic, demand for ventilatory support at times exceeded supply of mechanical ventilators.¹ Although there was some early debate whether COVID-19 caused a distinct form of acute respiratory distress syndrome (ARDS) that should be treated differently,² accumulated research showed that pathology findings^3,4 and physiology^5,6 paralleled classic descriptions of respiratory failure and ARDS. Many advocated for the use of similar non-COVID ARDS strategies for treating respiratory failure, including noninvasive therapies such as high-flow nasal cannula (HFNC).

HFNC has purported benefits in ARDS by correcting gas exchange, obviating the need for sedatives associated with intubation and mechanical ventilation, avoiding muscle atrophy and diaphragmatic dysfunction, and potentially avoiding ventilator-induced lung injury (VILI). Decades of research has established the risk of harm from mechanical ventilation, and there is ongoing interest in maximizing noninvasive therapies and avoiding mechanical ventilation when possible. A multicenter trial in 2015 compared a strategy of HFNC, noninvasive ventilation, and a nonrebreather facemask in ARDS. The investigators found a trend toward decreased intubation rates with HFNC and a statistically significant increase in ventilator free days at 28 days and lower 90-day mortality with HFNC.⁷ A separate meta-analysis found lower intubation rates when HFNC was used for ARDS as opposed to noninvasive ventilation.⁸

Early pandemic practice was characterized by a concern for rapid respiratory deterioration such that patients requiring high levels of support on HFNC were intubated early before frank deterioration. As resources and availability of mechanical ventilators and ICU beds waned and clinicians recognized that respiratory decline was often gradual enough that patients did not need immediate intubation, there was anecdotal migration in practice to using HFNC to support patients for longer periods of time with COVID-19 respiratory failure. Several elements of the pandemic enhanced this practice. First, there was concern for spread of virus with many forms of noninvasive ventilation. HFNC was purported to minimize this risk, making it an attractive alternative to intubation. Second, many patients did not show signs of respiratory distress despite being hypoxemic and HFNC became a modality to address hypoxemia without exposing these patients to mechanical ventilation. Finally, there was also concern for high mortality in intubated patients, and clinicians were interested in exploring different strategies of advanced respiratory support.^9,10. The use of HFNC was ultimately endorsed by national guidelines.^11,12

While the ability of HFNC to oxygenate patients and avoid invasive mechanical ventilation makes it a compelling option, concern has existed that it may lead to patient self-induced lung injury (P-SILI) through the generation of high transpulmonary pressures from persistent vigorous inspiratory effort.^13,14 Mechanical ventilation has been proposed as a mitigation strategy to prevent P-SILI by providing lung-protective ventilation and controlling airway pressure change, cycling frequency, and optimizing transpulmonary pressure.¹⁵

Many experts believe that exposure to HFNC adversely affects outcomes in patients with COVID-19-associated respiratory failure who ultimately require intubation, though evidence is limited and empirical findings have been mixed.^7,16 At least 1 retrospective study suggested that patients placed on HFNC therapy who ultimately required intubation had a higher mortality rate when intubated after 48 hours.¹⁶ Another study of patients supported with HFNC or noninvasive ventilation did not demonstrate increased mortality in the HFNC group; however, median times to intubation were less than 24 hours.⁷ Neither of these last 2 studies were done in patients with COVID-19.

Therefore, we performed a retrospective study of patients with COVID-19-associated respiratory failure to assess whether exposure to HFNC before intubation was associated with successful extubation and in-hospital mortality compared to patients receiving intubation only.

METHODS

Study Population

We performed a retrospective cohort study of all patients admitted to the hospital with a diagnosis of COVID-19 and requiring intubation in our large multisite health system based in Los Angeles, CA (Cedars-Sinai Medical Center [CSMC]) between March 8, 2020 and March 5, 2021). CSMC serves a large and diverse catchment area of >1.8 million people, with over a quarter million inpatient hospital days for admitted patients, greater than 90,000 emergency department visits, and nearly 800,000 appointments each year. COVID-19 diagnosis was based on the presence of a positive severe acute respiratory syndrome-coronavirus 2 (SARS-CoV2) polymerase chain reaction (PCR) or documentation, indicating confirmed diagnosis at another institution, placed by a provider into a discrete data field. Patients intubated in the emergency department before hospital admission were excluded given that they did not have sufficient opportunity to be exposed to HFNC. All patients in the cohort were intubated at some point during their hospitalization; as such, “do not intubate” orders were not present for any of the patients at the time of inclusion in the study. The study was approved by the CSMC Institutional Review Board (IRB) STUDY00000603, and the requirement for written informed consent was waived by the IRB.

Data Collection

We obtained demographic, clinical, and outcomes data from the CSMC electronic health record (EHR). We defined race/ethnicity membership as follows: Asian, Hispanic/Latinx ethnicity (all races), non-Hispanic Black, non-Hispanic White, and other (including individuals with multiple races listed). Clinical characteristics at presentation including smoking status (current versus not current), vital signs (pulse rate, respiratory rate, and temperature), laboratory values (HCO₃, CO₂, Po₂, Pco₂, and pH), oxygen-saturation-to-inspired-oxygen (S:F) ratio, any use of bilevel positive airway pressure (BiPAP), any use of continuous positive airway pressure (CPAP), and length of stay before initiation of advanced respiratory support (HFNC or ventilation) were extracted for all patients. S:F ratio was calculated as the documented percent oxygen saturation from EHR vital sign data divided by the oxygen flow rate. HCO₃ and CO₂ were extracted as nearest value within 24 hours of initiation of advanced respiratory support. Vitals, S:F ratio, Pco₂, Po₂, and pH were extracted as nearest value within 1 hour of initiation of advanced respiratory support. Additionally, International Classification of Diseases-10 (ICD-10) codes were used to identify and extract comorbidities including obesity, hypertension, diabetes, chronic obstructive pulmonary disease (COPD), and asthma.

Exposures and Outcomes

Our primary exposure was the receipt of HFNC before initial intubation. Use of HFNC and intubation were determined based on the oxygen delivery device listed in the EHR. Patients were considered to have received HFNC before intubation if the timestamp for the HFNC delivery preceded the timestamp for the first instance of mechanical ventilation. Duration of HFNC was calculated as the difference between the initiation of HFNC and the initiation of the next non-HFNC device. For patients with multiple HFNC sessions before ventilation, we calculated duration as the cumulative duration across all sessions before mechanical ventilation.

Our primary outcomes were successful extubation following initial mechanical ventilation and in-hospital mortality. Successful extubation was defined as the initiation of a nonventilator device following the initiation of the first instance of mechanical ventilation in the EHR. Time to extubation was defined as the difference between the initiation of the first instance of mechanical ventilation and the initiation of the next nonventilator device. All patients who were not successfully extubated died while intubated, as confirmed by a documented time of death occurring before the initiation of a nonventilator device in the EHR.

In-hospital mortality was determined by vital status documented in the EHR at time of discharge. Time to in-hospital mortality was defined as the difference between initiation of the first instance of mechanical ventilation in the flowsheet and documented time of death. All patients were followed until discharge or in-hospital mortality. For successful extubation, the follow-up time for patients not successfully extubated was set as 1 hour longer than the maximum time to event among patients who were extubated or their own time to death, whichever was longer. For in-hospital mortality, the follow-up time for patients who died before discharge was set as 1 hour longer than the maximum time to event among patients who survived to discharge or their own time to discharge, whichever was longer.^17,18

Statistical Analysis

Comparisons were made between those who were intubated after use of HFNC and those who were intubated without prior use of HFNC. Demographic and clinical variables were summarized using means and standard deviations (SDs) for continuous variables (age, S:F ratio, HCO₃, CO₂, Pco₂, pH, pulse rate, respiratory rate, and temperature) and as counts with percentages for categorical variables (sex, race/ethnicity, smoking status, obesity, hypertension, diabetes mellitus, COPD/asthma, BiPAP, and CPAP). We compared characteristics by respiratory intervention group using t tests for continuous measures and χ² tests for categorical variables to examine differences in baseline characteristics of the groups. Time variables, specifically length of stay before the primary intervention, were summarized by medians and interquartile ranges (IQRs) and compared using the Wilcoxon rank-sum tests by intervention group. We used multivariable Cox proportional hazards models to examine differences in successful extubation and in-hospital mortality among patients receiving HFNC before initial intubation compared to those not receiving HFNC before initial intubation. All analyses were adjusted for age, sex, race/ethnicity, HCO3, CO2, S:F ratio, vitals at time of initiation of advanced respiratory therapy, length of stay before initiation of advanced respiratory therapy, obesity, hypertension, diabetes, and COPD or asthma. Use of BiPAP or CPAP and Po₂, Pco₂, or pH was not included in final models due to issues in model convergence, as <2% of patients required BiPAP or CPAP and <15% had nonmissing Po₂, Pco₂, or pH values.

To evaluate the degree to which unmeasured confounding may impact our findings, we calculated the E-value for all hazard ratio (HR) estimates as well as confidence limits.¹⁹ Briefly, the E-value of an estimate represents the degree to which (ie, the risk ratio) an unmeasured confounder would have to be associated with both the exposure and outcome, even after adjusting for all other covariates, to change an observed estimate to the null value had that unmeasured confounder been included. Similarly, the E-value of a confidence limit represents the degree of association for an unmeasured confounder required to increase the confidence interval (CI) of an estimate to include the null value. P values <0.05 were considered significant. All analyses were conducted using R V4.0.2.

Cohort development was based on all available patients meeting inclusion criteria, and therefore, an a priori power analysis was not performed. A retrospective power calculation estimates the need for 232 in-hospital deaths to detect an HR of 1.5 with 80% power, assuming an alpha level of 0.05. A total of 230 in-hospital deaths occurred in our study, indicative of appropriate power.

RESULTS

A total of N = 440 patients hospitalized with COVID-19 requiring intubation were identified with a mean age of 65.3 ± 14.6 years, 65.5% of whom were male. We identified 311 (70.7%) patients receiving HFNC before intubation and 129 (29.3%) who were intubated without prior use of HFNC. Patients receiving HFNC before intubation were slightly older (66.5 ± 12.7 vs 62.5 ± 18.1; P = .01) than patients with no prior HFNC and had higher rates of obesity, hypertension, and diabetes. At the time of primary intervention, patients receiving HFNC had lower Pco₂ (34.5 ± 8.3 vs 48.8 ± 16.5; P < .001), lower Po₂ (79.1 ± 32.7 vs 165.0 ± 147.3; P < .001), and higher pH (7.4 ± 0.1 vs 7.2 ± 0.2, P < .001) compared to patients with no prior HFNC. The primary intervention (HFNC or primary intubation) was reached earlier in the hospital course for those receiving HFNC compared to those undergoing primary intubation (0.9 hours [0–53.7] vs 10.8 hours [0.7–133.5]; P = .008) (Table). Mean HFNC cumulative duration prior to intubation, in 311 HFNC patients, was 79.44 ± 110.15 hours, or 3.31 ± 4.59 days (median [IQR] 39.03 [15.12, 99.02] hours, 1.63 [0.63, 4.13] days).

Table.

Demographic and Clinical Characteristics of Patients Admitted With COVID-19 Who Required Intubation

Characteristic	Overall (N = 440)	High-flow before intubation (n = 311)	Primary intubation (n = 129)	P value^a
Demographic characteristics
Age, y, mean (SD)	65.31 (14.58)	66.46 (12.73)	62.54 (18.06)	.010
Male sex, n (%)	288 (65.5)	202 (65.0)	86 (66.7)	.815
Race/ethnicity, n (%)
Asian	21 (4.8)	18 (5.8)	3 (2.3)
Hispanic/Latinx	160 (36.4)	131 (42.1)	29 (22.5)
Non-Hispanic Black	75 (17.0)	48 (15.4)	27 (20.9)
Non-Hispanic White	132 (30.0)	87 (28.0)	45 (34.9)
Other^b	30 (6.8)	23 (7.4)	7 (5.4)
Unknown	22 (5.0)	4 (1.3)	18 (14.0)
Smoking status, n (%)	16 (3.6)	11 (3.5)	5 (3.9)	1.000
Comorbidities, n (%)
Obesity	127 (28.9)	102 (32.8)	25 (19.4)	.007
Hypertension	250 (56.8)	189 (60.8)	61 (47.3)	.013
Diabetes mellitus	209 (47.5)	162 (52.1)	47 (36.4)	.004
COPD or asthma	97 (22.0)	70 (22.5)	27 (20.9)	.813
Clinical characteristics at time of primary intervention
Bi-PAP before primary intervention, n (%)	0 (0)	0 (0)	0 (0)	-
CPAP before primary intervention, n (%)	7 (1.6)	5 (1.6)	2 (1.6)	1.000
S:F ratio, mean (SD)	118.34 (43.17)	115.85 (38.54)	125.14 (53.43)	.055
HCO₃, mmol/L, mean (SD)	22.55 (6.02)	22.81 (5.37)	21.92 (7.36)	.224
CO₂, mm Hg, mean (SD)	22.38 (4.95)	22.60 (4.56)	21.75 (5.88)	.139
Pco₂, mm Hg, mean (SD)	39.23 (13.35)	34.53 (8.30)	48.83 (16.45)	<.001
Po₂, mm Hg, mean (SD)	108.15 (97.50)	79.14 (32.69)	165.00 (147.28)	<.001
pH, mean (SD)	7.36 (0.15)	7.43 (0.07)	7.24 (0.19)	<.001
Pulse rate, mean (SD)	92.87 (20.55)	91.03 (20.08)	96.96 (21.07)	.006
Respiratory rate, mean (SD)	24.75 (7.74)	26.13 (6.94)	21.67 (8.54)	<.001
Temperature, mean (SD)	98.42 (1.95)	98.89 (1.42)	97.74 (2.36)	<.001
Length of stay before primary intervention, h, median (IQR)	9.05 (0.25, 62.54)	8.57 (0, 53.70)	10.83 (0.67, 133.45)	.008

Open in a new tab

Abbreviations: BiPAP, bilevel positive airway pressure; COPD, chronic obstructive pulmonary disease; COVID-19, coronavirus disease 2019; CPAP, continuous positive airway pressure; IQR, interquartile range; S:F, oxygen-saturation-to-inspired-oxygen; SD, standard deviation.

P values were generated from t tests for continuous variables, χ² tests for categorical variables, and Wilcoxon rank-sum tests for length of stay before primary intervention.

Other race includes American Indian/Alaska Native Native Hawaiian or other Pacific Islander, and other.

Figure. — Risk of extubation and in-hospital mortality among intubated COVID-19 patients. CI indicates confidence interval; COPD, chronic obstructive pulmonary disease; HR, hazard ratio; S:F, oxygen-saturation-to-inspired-oxygen ratio.

Among the entire cohort, 328 (74.5%) patients were successfully extubated, and 230 (52.3%) patients experienced in-hospital mortality. Among patients receiving HFNC, 224 (72.0%) were successfully extubated, and 177 (56.9%) experienced in-hospital mortality, whereas 104 (80.6%) of patients not receiving HFNC were successfully extubated and 53 (41.1%) experienced in-hospital mortality. In multivariable adjusted analyses, patients receiving HFNC before intubation were more likely to experience in-hospital mortality (HR, 2.08; 95% CI, 1.06–4.05) when compared to patients receiving primary intubation (Figure). No statistically significant differences were observed for successful extubation when comparing patients receiving HFNC to those receiving primary intubation (0.70; 0.41–1.20). To assess for overfitting, analyses were repeated following removal of the vital sign and laboratory data; when doing so, HFNC remained significantly associated with mortality, without a significant association with extubation. In sensitivity analyses examining the E-value for the reported associations, the E-value for the HR for in-hospital mortality was 2.70 with an E-value for the confidence limit of 1.25. The E-value for the HR for successful extubation was 1.88 with an E-value for the confidence limit of 1.

DISCUSSION

This retrospective study of over 400 patients with COVID-19-associated respiratory failure demonstrates that patients who were initially treated with HFNC experienced higher chances of in-hospital mortality compared to those who were intubated as the primary intervention. This association was found to be relatively robust, with a sensitivity analysis indicating that an unmeasured covariate would have to be associated with both the exposure and outcome with a risk ratio of 2.7 to overcome the observed association between HFNC and in-hospital mortality.

While noninvasive ventilation and HFNC oxygenation are widely used as strategies to avoid invasive mechanical ventilation and VILI, the present data raise concerns for introducing harm. P-SILI, a relatively new concept in comparison, may be one of the mechanisms that influences the observed increase in mortality of ARDS patients treated with HFNC. Specifically, the physiological drive to maintain oxygen homeostasis may induce increased effort and thus increased transpulmonary forces, leading to increases in pleural pressures beyond normal physiologic parameters, and, thus, causing inflammation and lung injury. To that end, mechanical ventilation has been proposed as a tool to mitigate P-SILI, by optimizing and controlling lung volumes, forces, and pressures.¹³ While there is compelling biologic plausibility for P-SILI, evidence is limited to animal models and small trials showing unacceptably large tidal volumes in patients supported with noninvasive ventilation.^20,21 Notably, some experts have cautioned that rapid swings in pleural pressure, characteristic of spontaneously breathing patients with increased respiratory effort due to hypoxemia, may herald further lung injury and cautioned against the use of noninvasive options for a COVID-19 ARDS.²

This study adds important evidence to the growing body of literature surrounding the potential benefits and harms of HFNC. At least 1 study has demonstrated increased mortality rates with HFNC when used in disease conditions other than cardiogenic pulmonary edema and COPD.²² Others have demonstrated that HFNC may allow for avoidance of intubation in a portion of patients, including those with COVID-19.^23,24 Future work should focus on identifying patients likely to benefit from HFNC, those who may experience harm, and the thresholds at which HFNC should be abandoned in favor of mechanical ventilation.

On average, those who received HFNC received this earlier on in their hospitalization than those who had primary intubation. While the time to primary intervention and degree of hypoxia (as measured by the S:F ratio) were controlled for in the model, the need for enhanced respiratory support earlier on in the hospital course may indicate a more rapid clinical deterioration in the HFNC group. The relatively rapid transition from HFNC to intubation, a median time of only 1.6 days, indicates that many patients placed on HFNC may have been at a “point of no return” from the need for mechanical ventilation and that any delay adversely contributed to their clinical outcome.

Limitations

Several limitations of this study merit consideration. The retrospective nature of the study precludes the determination of causative relationships. However, the study offers important insight into selection of initial respiratory support among patients suffering from the novel SARS-CoV-2 infection. Data were drawn from a single center, with clinical practice patterns that may not reflect other institutions. Information on the cause of death was not available in a discrete data field within the EHR or in coded data limiting the ability to describe and analyze causes of intrahospital deaths beyond COVID-19. Fortunately, the large size of the institution and the number of clinicians increase the representativeness of clinical decision-making. Further, while our analysis accounted for multiple clinically relevant covariates, unmeasured confounders may be present. Importantly, our sensitivity analysis indicates that such a confounder would require a risk ratio of almost 3 to overcome the association of HFNC and mortality. Also importantly, all covariates were selected a priori based on their clinical relevance. While this robust multivariable adjustment is based on clinical knowledge, it does risk overfitting the model; however, our findings persisted following removal of multiple covariates. Finally, knowledge and experience in the management of COVID-19 evolved rapidly during the first year of the pandemic. Similarly, it is now recognized that viral variants may predispose to varying degrees of respiratory illness.

CONCLUSIONS

In a single-center study of patients with respiratory failure associated with COVID-19, patients exposed to HFNC before intubation had higher chances of death compared to patients intubated without a trial of HFNC. Our data suggest that clinicians should consider early intubation for patients admitted with COVID-19 who do not improve clinically.

DISCLOSURES

Name: Michael Nurok, MBChB, PhD.

Contribution: This author helped with study conception and design; writing the first draft of the manuscript; and material preparation, data collection, and analysis. They also commented on previous versions of the manuscript, and read and approved the final manuscript.

Conflicts of Interest: M. Nurok reported receiving stock options for his role as an adviser to Avant-Garde Health.

Name: Oren Friedman, MD.

Contribution: This author helped contribute to the study conception and design, and perform material preparation, data collection, and analysis. They commented on previous versions of the manuscript, and read and approved the final manuscript.

Conflicts of Interest: O. Friedman participates in a Bristol Myers Squibb speaker bureau.

Name: Matthew Driver, MPH.