Noninvasive Ventilation in Critically Ill Patients With Severe Acute Respiratory Infections

Abstract

BACKGROUND:

The objective of this study was to evaluate the association between noninvasive ventilation (NIV) compared with invasive ventilation and mortality in subjects with severe acute respiratory infection.

METHODS:

This was a retrospective multi-center study of subjects with severe acute respiratory infection treated with ventilatory support between September 2012 and June 2018. We compared the 90-d mortality of subjects managed initially with NIV (NIV group) with those managed with invasive ventilation only (invasive ventilation group), adjusting by propensity score.

RESULTS:

Of 383 subjects, 189 (49%) were in the NIV group and 194 (51%) were in the invasive ventilation group. Of the subjects initially treated with NIV, 117 (62%) were eventually intubated. Crude 90-d mortality was lower in the NIV group versus the invasive ventilation group (42 [22.2%] vs 77 [39.7%]; P < .001). After propensity score adjustment, NIV was associated with lower 90-d mortality than invasive ventilation (odds ratio 0.54, 95% CI 0.38-0.76; P < .001). The association of NIV with mortality compared with invasive ventilation was not different across the studied subgroups.

CONCLUSIONS:

In subjects with severe acute respiratory infection and acute respiratory failure, NIV was commonly used. NIV was associated with a lower 90-d mortality. The observed high failure rate suggests the need for further research to optimize patient selection and facilitate early recognition of NIV failure.

Introduction

Severe acute respiratory infection is defined by the World Health Organization as an acute respiratory illness with a history of fever (≥38°C) and cough, with onset within the past 10 d, and requiring hospitalization. ¹ Severe acute respiratory infection is considered one of the leading causes of morbidity and mortality around the globe, resulting in almost 4 million deaths per year. ^{2 - 4} Of these, approximately 90% are believed to occur in developing countries. ^2-4 Severe acute respiratory infection can be caused by various viral and bacterial pathogens, including several emerging respiratory infectious diseases, such as SARS, pandemic influenza H1N1, and the Middle East Respiratory Syndrome. ^{5 - 8} Patients with severe acute respiratory infection have high ICU admission and case fatality rates primarily due to acute respiratory failure. ⁹

Noninvasive ventilation (NIV) has been increasingly used in managing acute respiratory failure with variable results, depending on the underlying pathology. ¹⁰ Randomized clinical trials have shown NIV to be associated with a reduced need for invasive ventilation, especially in patients with hypercapnic acute respiratory failure related to COPD. ¹¹ NIV has the advantage of maintaining normal swallowing, speech, and airway protection mechanisms. ⁹ However, the role of NIV in patients with acute hypoxemic respiratory failure secondary to pneumonia is still unclear because NIV failure was observed in 25–66% of these patients, with concerns about delayed intubation and increased inspiratory effort that leads to further injury to the lung. ^{12 - 16} This study aimed to evaluate the association between NIV compared with invasive ventilation, and mortality in subjects with severe acute respiratory infection.

Quick Look.

Current Knowledge

Noninvasive ventilation (NIV) has been increasingly used in managing acute hypoxemic respiratory failure with variable results, depending on the underlying pathology. The evidence for the use of NIV in patients who are critically ill with severe acute respiratory infection and acute hypoxemic respiratory failure remains uncertain, leading to considerable practice variation.

What This Paper Contributes to Our Knowledge

Analysis of these data demonstrates that NIV was commonly used in subjects who were critically ill with severe acute respiratory infections and acute respiratory failure. Although NIV was associated with a high failure rate, it was associated with a lower 90-d mortality. These findings suggest the need for further research to optimize patient selection and facilitate early recognition of NIV failure.

Methods

This was a retrospective cohort study. All patients admitted with severe acute respiratory infection to the ICUs of 4 referral hospitals in 4 cities in Saudi Arabia between September 2012 and June 2018 were included in this study. The study was approved by the Ministry of National Guard Health Affairs Institutional Review Board (IRB) and by the IRBs of all participating sites. The IRB waived informed consent due to the retrospective nature of the study. All adult patients (defined according to local policies as ≥14 years old) who met the diagnostic criteria for severe acute respiratory infection and were treated with ventilatory support in the form of NIV (via nasal or face mask) or invasive ventilation were included in the study. Severe acute respiratory infection was defined as an acute respiratory tract infection, including a history of fever and cough within the past 10 days, with clinical or radiologic findings of pulmonary parenchymal disease not already explained by a noninfectious etiology. ¹ Data were extracted from subjects' medical records by using the standardized International Severe Acute Respiratory and Emerging Infection Consortium case report forms. ¹⁷ In this study, we included subjects' demographic features, underlying comorbidities, physiologic and laboratory parameters, radiographic findings, and severity of illness on ICU admission assessed by using the Sequential Organ Failure Assessment (SOFA). We also described ICU management and rescue therapies, including neuromuscular blockade, high-frequency oscillation ventilation, extracorporeal membrane oxygenation, nitric oxide, and prone positioning. The primary outcome was 90-d mortality. Other outcomes included ICU and hospital mortality, and ICU and hospital lengths of stay.

We compared subjects managed initially with NIV (NIV group) with those managed by invasive ventilation only (invasive ventilation group). Categorical variables are presented as percentages and compared by using a chi-square or Fisher exact test. Continuous variables are presented as medians and interquartile ranges (IQR), and were compared by using the Mann-Whitney U test. We constructed Kaplan-Meir survival curves and reported the log-rank test result. A P value of < .05 was used to indicate statistical significance. We developed a propensity score for comparison of baseline characteristics to account for the imbalances between the 2 groups by using the following covariates: age, SOFA score on admission to the ICU, chronic pulmonary disease, chronic cardiac disease, chronic neurologic disease, diabetes with chronic complications, ${\text{P}}_{{\text{aO}}_{\text{2}}}$ / ${\text{F}}_{{\text{IO}}_{\text{2}}}$ , ${\text{P}}_{{\text{aCO}}_{\text{2}}}$ and the Glasgow coma scale score. Overlap of the distribution of the propensity scores between the 2 groups was checked visually by using a butterfly plot (Fig. 1). We imputed for missing variables included in the propensity score by using predictive mean matching.

Fig. 1.

A butterfly plot, showing propensity scores in subjects with severe acute respiratory infection who received noninvasive ventilation (NIV) and those who received invasive ventilation.

To assess the independent association of NIV with 90-d mortality, we performed a PROC GENMODE regression analysis by adjusting for the propensity score. In addition, we also evaluated the association of NIV with 90-d mortality in subgroups of subjects with ${\text{P}}_{{\text{aO}}_{\text{2}}}$ / ${\text{F}}_{{\text{IO}}_{\text{2}}}$ ≤ 100 mm Hg, 101–200 mm Hg, and >200 mm Hg, HCO₃ ≤ and >24 mEq/L, ${\text{P}}_{{\text{aCO}}_{\text{2}}}$ < and ≥ 45 mm Hg, the presence of chronic cardiac disease, chronic respiratory disease, chronic neurologic disease, and the period before and after July 1, 2015. Because of the emerging evidence about the effect of corticosteroids in patients with community-acquired pneumonia and COVID-19, we conducted a subgroup analysis on whether concomitant corticosteroid therapy was or was not used. ^{18 - 20} In these subgroup analyses, we tested for interaction by adding an interaction term to the PROC GENMODE regression analysis, adjusting for the propensity score. We carried out secondary analyses by using regression analysis by adjusting for the same individual variables used in generating the propensity score (age, SOFA score on admission to ICU, chronic pulmonary disease, chronic cardiac disease, chronic neurologic disease, diabetes with chronic complications, ${\text{P}}_{{\text{aO}}_{\text{2}}}$ / ${\text{F}}_{{\text{IO}}_{\text{2}}}$ ${\text{P}}_{{\text{aCO}}_{\text{2}}}$ and Glasgow coma scale) as confirmatory to the primary analysis. We also conducted a secondary analysis by comparing subjects in whom NIV failed with those who had been successfully treated with NIV. SAS software version 9.4 (SAS Institute, Cary, North Carolina) was used.

Results

This study included 383 subjects with severe acute respiratory infection admitted to the ICU with acute respiratory failure requiring ventilatory support. Of these subjects, 189 (49%) were in the NIV group, and 194 (51%) were in the invasive ventilation group. Demographic and baseline characteristics of the NIV group compared with the invasive ventilation group are presented in Table 1 and eTable 1 (see the supplementary materials at http://www.rcjournal.com). On ICU day 1, the NIV group had similar physiologic parameters compared with the invasive ventilation group for ${\text{P}}_{{\text{aO}}_{\text{2}}}$ / ${\text{F}}_{{\text{IO}}_{\text{2}}}$ (median [IQR] 146 [94–213] mm Hg vs 138 (94 –230) mm Hg, respectively; P = .67). The median (IQR) values of ${\text{P}}_{{\text{aCO}}_{\text{2}}}$ , ${\text{P}}_{{\text{aO}}_{\text{2}}}$ , and ${\text{F}}_{{\text{IO}}_{\text{2}}}$ were also comparable between the 2 groups (44 [37–60] mm Hg vs 43 [39–54] mm Hg, 68 [53–85] mm Hg vs 76 [58–99] mm Hg, and 0.45 [0.40 – 0.60] mm Hg vs 0.53 [0.40–0.80] mm Hg, respectively). However, bicarbonate was significantly higher in the NIV group versus the invasive ventilation group (24 [21–29] mEq/L vs 23 [19–27] mEq/L; P = .006). The subjects in the NIV group with severe acute respiratory infection had significantly higher median (IQR) body mass index versus the invasive ventilation group (29.3 [25.5-35.3] kg/m² vs 26.6 [22.3-30.5] kg/m²; P < .001). The subjects in the NIV group were less likely to have chronic neurologic diseases and had higher Glasgow coma scale scores and mean arterial pressure than the subjects in the invasive ventilation group. When adjusted to propensity scores, all baseline characteristics were largely balanced (Table 1). A butterfly plot demonstrated the overlap of the propensity scores of the 2 study groups (Fig. 1).

Table 1.

Baseline Characteristics of the Subjects With Severe Acute Respiratory Infections Treated Initially With NIV Compared With Those Treated Only With Invasive Ventilation; Crude P Values and Propensity Score–Adjusted P Values Are Reported^*

Subjects in the NIV group were less likely to receive vasopressor therapy or renal replacement therapy compared with the invasive ventilation group (90 [48%] vs 146 [75%], P < .001; and 32 [17%] vs 50 [26%], P = .035, respectively) (Table 2). Crude 90-d mortality was significantly lower in the NIV group versus the invasive ventilation group (42 [22.2%] vs 77 [39.7%]; P < .001) (Table 3 and Fig. 2). This association between NIV and 90-d mortality remained after propensity score adjustment (odds ratio 0.54, 95% CI 0.38–0.76; P < .001) (Table 4). Although there was no significant difference in the duration of invasive ventilation between the 2 groups, invasive ventilation–free days and total NIV and invasive ventilation–free days by day 90 were significantly higher among the subjects in the NIV group versus those in the invasive ventilation group (median [IQR] 86 days [67–90] d vs 72 [0–83] d, P < .001, and median [IQR] 82 [60–87] d vs 72 [0–83] d, P < .001, respectively). ICU length of stay was also significantly shorter in the NIV group versus the invasive ventilation group (median [IQR] 10 (5–20) d vs 16 (9–30) d; P < .001). Subgroup analyses showed that the association of NIV with mortality compared with invasive ventilation was not different across all the studied subgroups, as reflected by the nonsignificant P values for interaction.

Table 2.

Main Interventions in Subjects With Severe Acute Respiratory Infection Treated Initially With NIV Compared With Those Only Treated With Invasive Ventilation

Table 3.

Outcomes in Subjects With Severe Acute Respiratory Infection Treated Initially With NIV Compared With Those Only Treated With Invasive Ventilation

Fig. 2.

Kaplan-Meir curves for survival of subjects who received noninvasive ventilation (NIV) and those who received invasive ventilation. P value for the log-rank test is reported.

Table 4.

Association Between NIV Compared With Invasive Ventilation and 90-d Mortality Among Subjects With Severe Acute Respiratory Infection After Adjusting for the Propensity Score

The results of the sensitivity analysis by using regression analysis yielded similar results to the primary analysis (eTable 2 [see the supplementary materials at http://www.rcjournal.com]). Of the subjects initially treated with NIV, NIV therapy failed in 117 (62%) and these subjects had to be intubated. The failed NIV group had significantly lower Glasgow coma scale scores, higher lactate, and higher SOFA scores compared with the successful NIV group (Glasgow coma scale score: median [IQR] 9 [3–15] vs 15 [13–15], respectively; lactate: median (IQR) 2.02 [1.22–3.17] mmol/L vs 1.25 [0.97–2.48] mmol/L, respectively; and SOFA score: median [IQR] 8 [5–10] vs 5 [3–6], respectively). Interestingly, the successful NIV group had higher ${\text{P}}_{{\text{aCO}}_{\text{2}}}$ versus the failed NIV group (median [IQR] 51.5 [40–72] mm Hg vs 42 [35–52] mm Hg; P = .002). Crude 90-d mortality was significantly higher in the subjects in whom NIV failed compared with the successful NIV group (36 [30.8%] vs 6 [8.3%]; P < .001), and ICU length of stay was longer (median [IQR] 17 [10–25] d vs 6 [4–9] d; P < .001) (eTables 3–5 [see the supplementary materials at http://www.rcjournal.com]).

Discussion

In this study, we report that NIV was commonly used in subjects with severe acute respiratory infection and acute respiratory failure. Initial NIV use was associated with lower 90-d mortality compared with the subjects treated only with invasive ventilation. However, NIV was associated with a high failure rate because more than half of the subjects eventually required invasive ventilation. NIV has been demonstrated to be effective in reducing intubation and mortality in patients with COPD and patients with acute pulmonary edema caused by decompensated heart failure. ¹⁰ The evidence for the use of NIV in acute hypoxemic respiratory failure due to pneumonia and ARDS remains uncertain, which leads to considerable practice variation across ICUs. A post hoc analysis of the LUNG SAFE study ²¹ found an increase in mortality rate in the subjects with moderate-to-severe ARDS who were treated with NIV. In contrast, our study demonstrated lower mortality when NIV was initially chosen for the entire population and in each of the 3 ${\text{P}}_{{\text{aO}}_{\text{2}}}$ / ${\text{F}}_{{\text{IO}}_{\text{2}}}$ strata. Further data are required to identify patients likely to benefit from NIV.

We observed a high NIV failure rate of 62%. In a study on subjects with Middle East Respiratory Syndrome, NIV was found to have a very high failure rate (92.4%). ¹³ Unsurprisingly, those subjects in whom NIV failed were more likely to require rescue therapy and to die than those who were successfully treated with NIV. This was similar to what was described in subjects with Middle East Respiratory Syndrome. ¹³ Analysis of these data suggests the need to be selective in choosing NIV in managing patients with severe acute respiratory infection. Factors that have been reported to be associated with a high failure rate include low baseline ${\text{P}}_{{\text{aO}}_{\text{2}}}$ / ${\text{F}}_{{\text{IO}}_{\text{2}}}$ , shock, older age, pneumonia, or failure of improvement after 1 h of treatment. ^{12 , 22 , 23}

Strength and Limitations

The strengths of this study are the inclusion of 4 different centers, the use of standardized case report forms for data extraction, and the use of the propensity score to adjust for known covariates. However, residual unmeasured confounding cannot be excluded despite adjustment. We did not collect detailed data on NIV or invasive ventilation settings, the rationale for choosing NIV or to invasive ventilation as the initial support modality, patient tolerance, or the reasons for NIV failure. To address the temporal bias, we conducted a subgroup analysis based on the time of the study (before vs on or after July 1, 2015). We found that the association of NIV with mortality was not different between the 2 periods. Our study was not designed to address the risk of infection transmission to health-care workers and other patients with NIV versus invasive ventilation.

Our study did not compare NIV with high-flow nasal cannula because high-flow nasal cannula was not commonly used in our institutions before the COVID-19 era, and no subjects received high-flow nasal cannula. Our study did not include subjects treated with helmet NIV, another important area that requires further investigation given the controversy about helmet NIV effectiveness. ^{24 , 25}

Conclusions

In subjects with severe acute respiratory infection and acute respiratory failure, NIV was commonly used. NIV was associated with a lower overall 90-d mortality. The observed high NIV failure rate suggests the need for further research to optimize patient selection and facilitate early recognition of NIV failure.