Impact of using non-rebreathing mask in patients with respiratory failure

Liberal oxygen therapy might increase the mortality rate of patients. Non-rebreathing mask (NRM) is a high-ow, non-invasive oxygen device that can provide oxygen concentration up to 95%. This study aimed to determine the impact of using NRM in patients with respiratory failure. This retrospective cohort study was conducted in four medical institutions in Taiwan. Data were extracted from the Chang Gung Research Database between January 2010 and December 2016. The association between mortality and NRM use in patients with respiratory failure in the emergency department was analysed. Before receiving ventilator support, patients were divided into the NRM treatment and no NRM treatment groups. A 1:4 propensity score matching was conducted. Regarding the duration of NRM use, treatments were grouped as 0 hour, 0–1 hour, 1–2 hours, and > 2 hours.


Introduction
Supplemental oxygen is a common therapy in clinical practice. Physicians started to use oxygen to treat patients since the nineteenth century. 1 Considering that supplemental oxygen had no adverse clinical effects, several clinicians used oxygen liberally in patients even without hypoxaemia. [2][3][4] However, several studies have indicated that liberal oxygen therapy is not safe considering the deleterious effects of hyperoxia such as increased levels of in ammatory cytokines and reactive oxygen species, 5,6 reduced cardiac output, and pulmonary vessel constriction. 3,7 A recent systematic review and meta-analysis study reported that liberal oxygen therapy increases the risk of mortality. 8 High-concentration oxygen therapy is commonly used in critically ill patients. 9,10 Non-rebreathing mask (NRM) is a relatively high-ow and non-invasive oxygen device that can provide oxygen concentration up to 95% under a continuous oxygen ow of 10-15 L/minute. 11 NRM is usually used to preoxygenate patients before induction and intubation [12][13][14] and is not used as a long-term oxygen therapy because it might result in carbon dioxide retention and nasal and oral mucosa irritation. 3 In emergency departments (EDs), patients experiencing shortness of breath are often treated with highconcentration oxygen. 15 Emergency physicians (EPs) usually preoxygenate these patients with NRM before induction and intubation to increase the duration of safe apnoea. 16,17 However, excess supplemental oxygen could increase the mortality rate in critically ill patients because of oxygen toxicity. 18,19 The advantages and disadvantages of NRM in patients with respiratory failure are still unclear. Therefore, this study aimed to analyse the association between NRM use before intubation and mortality in patients with respiratory failure.

Study setting
The data were obtained from the largest healthcare system in Taiwan, the Chang Gung Memorial Hospital, which receives approximately 10-12% of the National Health Insurance budget according to government statistics. The Chang Gung Research Database (CGRD) is a multi-institutional and original medical record-based research database. 20 In this study, the data from Keelung, Linkou, Chiayi, and Kaohsiung branches located from northern to southern Taiwan were used.

Study participants
All patients aged older than 17 years who visited the ED for non-traumatic disease and who experienced respiratory failure with ventilator support between January 2010 and December 2016 were included in the study. Patients who experienced out-of-hospital cardiac arrest (OHCA) or had do-not-resuscitate (DNR) order were excluded. Patients were divided into the NRM group and the non-NRM group.

Measurements
Data of patients' demographics and comorbidities were extracted from the CGRD database. Patients were divided into three diagnostic groups according to the causes of respiratory failure including pulmonary disease, cardiovascular disease, and other metabolic diseases. Diagnoses were grouped according to the diagnostic codes from the International Classi cation of Diseases, Ninth and Tenth Revision, Clinical Modi cation. According to the duration of NRM use, patients using NRM were further grouped into the following treatment groups: 0 hour, 0-1 hour, 1-2 hours, and > 2 hours. Death in hospital within 30 days was de ned as in-hospital mortality.

Data analysis
For continuous variables, age was summarised as means ± standard deviations. The distributions of categorical data were presented as numbers and percentages. Student's t-test, one-way analysis of variance, and chi-square test were used.
Propensity score matching Propensity score was calculated using logistic regression. Variables of age, sex, and co-morbidities (including myocardial infarction, heart failure, peripheral arterial disease, cerebrovascular accident, chronic obstructive pulmonary disease, peptic ulcer disease, liver cirrhosis, diabetes mellitus, chronic kidney disease, malignancy, and bedridden status) and main diagnoses were used to estimate the probability of receiving NRM before intubation. Main diagnoses, which were divided into other metabolic diseases, diseases of the circulatory system, and diseases of the respiratory system, were diseases considered to cause respiratory failure in a single episode. Propensity score matching (PSM) was performed using NCSS version 12.0.4 (NCSS statistical software, LLC, Kaysville, Utah, USA). The greedy method was used to create a 1:4 matched study groups with a 0.25 standard deviation width. The caliper half-width was 0.17579, and the average match Mahalanobis distance was 0.00014.
To determine the associations between the variables and NRM use, survival analysis using the Kaplan-Meier estimator and Cox regression was performed. The effects were estimated using adjusted odds ratios (aORs) and the corresponding 95% con dence intervals (CIs). Results were considered statistically signi cant for a two-tailed test at P < 0.05. In addition to PSM, the IBM Statistical Package for the Social Sciences for Windows version 22.0 (released 2013, IBM Corp., Armonk, NY) was used for statistical analyses.

Results
After excluding patients with OHCA and DNR order, 18749 patients who developed respiratory failure in ED were enrolled. Among them, 1074 patients had used NRM (Fig. 1). After a 1:4 PSM, the age, sex, and distribution of co-morbidities between 1028 patients using NRM and 4112 patients not using NRM were similar. Tables 1 and 2 present the demographic characteristics of patients before and after PSM matching.   Figure 2 shows the 72-hour and 30-day mortality rates in patients using and not using NRM. Before PSM matching, patients not using NRM had a higher 72-hour mortality rate than those using NRM, but after PSM matching, there were no signi cant differences in 72-hour mortality rates between patients using NRM and patients not using NRM (Fig. 2a and 2b). The 30-day mortality rate was relatively higher in patients using NRM > 2 hours than in patients belonging to the remaining three treatment groups (i.e. 0 hour, 0-1 hour, 1-2 hours treatment groups) both before and after PSM matching. However, there was no signi cant difference in the 30-day mortality rate between patients using NRM and patients not using NRM ( Fig. 2c and 2d). Figure 3 shows the survival curve of patients not using NRM and patients using NRM for different durations. After performing a strati ed analysis according to the cause of respiratory failure, patients with pulmonary disease and using NRM for more than 2 hours had a higher mortality rate than patients with pulmonary disease but without using NRM (p-value: 0.016) (Fig. 3b). There was no association between mortality and NRM use in patients with respiratory failure due to cardiovascular disease and other metabolic diseases. Figure 4 shows the result of Cox regression analysis. Patients with respiratory failure due to pulmonary disease and who use NRM > 2 hours had a higher mortality rate than patients not using NRM (OR, 1.4; 95% CI, 1.06-1.74) (Fig. 4b).
Discussion A recent systematic review and meta-analysis showed that liberal oxygen therapy increases the mortality rate of patients and does not improve other important patient clinical outcomes. 8 The study used different oxygen delivery devices including nasal prongs, face masks, and invasive mechanical ventilators. Different from a previous study, our study used a high-ow and non-invasive oxygen device, NRM, before providing ventilator support. In EDs, EPs usually preoxygenate patients with NRM before induction and intubation to increase the duration of safe apnoea. 16,17 EPs also use non-invasive but highow oxygen therapy in elderly patients considering the provision of hospice care rather than intensive resuscitation; however, this requires further communication and discussion among healthcare professionals. [21][22][23] In Taiwan, NRM is commonly used. Several patients hesitate to receive intubation considering the possibility of delayed weaning and tracheostomy, subsequently delaying the provision of ventilator therapy. Considering that excess supplemental oxygen causes oxygen toxicity in critically ill patients 18 , we determined whether NRM use in patients with respiratory failure could increase their mortality rate.
According to a previous study, the 30-day mortality rate in patients with respiratory failure ranged from 29.7-41.7%. 24 In this study, the mortality rates were 30.0% and 30.4% in patients using NRM and patients not using NRM, respectively, a result consistent with the results of the previous studies. However, according to clinical practice, patients using NRM as an alternative oxygen treatment are considered less critical than patients using ventilator support immediately. We found that the 72-hour mortality rate in patients using NRM was 4.5% lower than the mortality rate of patients not using NRM. This suggests that the initial conditions of patients using NRM were less critical than the initial conditions of patients not using NRM. Therefore, this study performed PSM and analysed the in uence of NRM according to its duration of use. Patients using NRM for a prolonged period had a higher mortality rate than patients using NRM for a limited time. This is possibly attributed to the high oxygen toxicity as a result of highconcentration oxygen therapy reported in a previous study. 18 Although liberal oxygen therapy has not been recommended considering that it results in an increased mortality rate, it is still unclear which groups of patients are evidently affected with the increase in mortality rate following liberal oxygen therapy. The recent systemic review enrolled patients with sepsis, critical illness, stroke, trauma, myocardial infarction, or cardiac arrest and patients undergoing emergency surgery. 8 While liberal oxygen therapy might increase patients' mortality rate, there was no association between liberal oxygen treatment and mortality in patients with sepsis, stroke, trauma, and myocardial infarction. In our study, after performing a strati ed analysis, the in uence of NRM use on mortality was observed only in patients with respiratory failure due to pulmonary disease. This result is relatively consistent with the results of a systemic review, although there was no clear de nition of critically ill patients in the systemic review article. Clinically, high-concentration oxygen therapy was not recommended in patients with asthma and chronic obstructive pulmonary disease, as it might result in carbon dioxide retention and acidosis. [25][26][27] Several studies have also reported that prolonged breathing with a signi cantly high fraction of inspired oxygen (FIO 2 ) (FIO 2 ≥ 0.9) uniformly causes severe hyperoxic acute lung injury, and when the FIO 2 is not reduced, death may occur. 28 Hence, patients with pulmonary disease who prolonged use NRM have higher mortality rates than patients with pulmonary disease who are not using NRM.

Limitations
This study has some limitations. First, the diagnoses that cause respiratory failure were divided into only three groups, namely, pulmonary disease, cardiovascular disease, and other metabolic diseases. However, these diseases are signi cantly different regardless of being in the same group. For example, respiratory failure caused by pneumonia and chronic pulmonary disease have different mechanisms, but are all grouped as pulmonary disease, possibly in uencing the result of the study. Second, the four study settings belonged to the same medical system; this may limit the implications of the conclusions. Finally, considering the retrospective design of this study, there might be some confounding factors that were not measured in this analysis that could in uence the outcome of the study. For example, ventilator setting is associated with mortality. 29 High-concentration oxygen ventilation and prolonged use of ventilator, which were not well controlled in the study, could also cause lung damage. 30,31 In this study, data regarding ventilator setting were unavailable; hence, the confounding effects of ventilator were not controlled. Thus, further prospective studies are required to determine the association between NRM use and patient mortality. Figure 1 Flowchart of enrolment and status of patients. The 1:4 propensity score matching by gender, age, myocardial infarction, heart failure, peripheral arterial disease, cerebrovascular accident, chronic obstructive pulmonary disease, peptic ulcer disease, liver cirrhosis, diabetes mellitus, chronic kidney disease, malignancy, bedridden status, and main diagnosis. Abbreviation: out-of-hospital cardiac arrest (OHCA), do not resuscitate (DNR), non-rebreathing mask (NRM).

Figure 2
Before 1:4 propensity score matching, a) 72-hour mortality at different time of using the non-rebreathing mask, and b) 30-day mortality at different time of using the non-rebreathing mask; after 1:4 propensity score matching, c) 72-hour mortality at different time of using the non-rebreathing mask, and d) 30-day mortality at different time of using the non-rebreathing mask. propensity score matching. In pulmonary disease, the  other metabolic problem. Patients with pulmonary disease and using NRM for more than 2 hours had a higher mortality rate than patients with pulmonary disease but without using NRM (p-value: 0.016) (Fig.   3b).