Risk Factors for Non-invasive Ventilation Failure in Inuenza Infection with Acute Respiratory Failure in Emergency Department

Non-invasive ventilation (NIV) has been widely used in hypoxemic acute respiratory failure (ARF) due to inuenza pneumonia in emergency department (ED). However, the benet of NIV in decreasing intubation rate remains controversial. Previous studies have reported that prolonged use of NIV was associated with increased mortality. Our study aims to identify risk factors for NIV failure in inuenza infection with acute respiratory failure in ED. We perform a retrospective cohort observational study. Enrolled patients older than 18 years who used NIV due to inuenza infection with ARF between 1 January 2017 to 31 December 2018 in Ramathibodi Emergency Department. Patients characteristics, comorbidity, clinical and laboratory outcome, chest imaging, NIV setting and parameter were recorded. We follow the outcome success or failure of the NIV used. Score: Sequential Organ Failure Assessment Score, PF ratio: Ratio of arterial oxygen partial pressure to fractional inspired oxygen, PEEP: Positive end-expiratory pressure, ICU: intensive care unit, OR: odds ratio aOR; adjusted odds ratio, PFCXR: Ratio of arterial oxygen partial pressure to fractional inspired oxygen and chest x-rays score


Abstract
Background Non-invasive ventilation (NIV) has been widely used in hypoxemic acute respiratory failure (ARF) due to in uenza pneumonia in emergency department (ED). However, the bene t of NIV in decreasing intubation rate remains controversial. Previous studies have reported that prolonged use of NIV was associated with increased mortality. Our study aims to identify risk factors for NIV failure in in uenza infection with acute respiratory failure in ED.

Method
We perform a retrospective cohort observational study. Enrolled patients older than 18 years who used NIV due to in uenza infection with ARF between 1 January 2017 to 31 December 2018 in Ramathibodi Emergency Department. Patients characteristics, comorbidity, clinical and laboratory outcome, chest imaging, NIV setting and parameter were recorded. We follow the outcome success or failure of the NIV used.
Results 162 patients were enrolled, 72 (44%) suffered NIV failure in in uenza infection with ARF. We used univariate and multivariate logistic analyses to assess risk factors for NIV failure. The ability of risk factor to predict NIV failure was analyzed using the area under the receiver operating characteristic (AUROC). Risk factors of NIV failure included sequential organ failure assessment (SOFA) score (P = 0.001), PaO2/FiO2 (PF) ratio (P = 0.001) and quadrant in ltrations in chest x-rays (CXR) (P = 0.001).
SOFA score, PF ratio and number quadrant in ltrations in chest radiography have good ability to predict NIV failure, AUROC 0.894 (0.839 -0.948), 0.828 (0.764 -0.892) and 0.792 (0.721 -0.863), respectively and no signi cant difference in the ability to predict NIV failure between three parameters. Use of PF ratio plus number quadrant in ltrations in chest radiography demonstrated higher predictive ability for NIV failure in in uenza infection with ARF.

Conclusions
SOFA score, PF ratio and quadrant in ltrations in chest radiography were good predictors of NIV failure in in uenza infection with ARF.

Background
In uenza infection is a course of acute respiratory failure (ARF). In patients with acute respiratory failure due to in uenza infection admitted to intensive care unit (ICU), the intubation rate is high (60%) and mortality rate after intubation of up to 50% (1-4).
Use of non-invasive mechanical ventilator (NIV) has proved effective in chronic obstructive pulmonary disease (COPD) and acute decompensated heart failure (5,6). In these patients, NIV has achieved reduction in endotracheal intubation rate, ventilator-association pneumonia and mortality rate. NIV is widely used in several clinical settings, nevertheless the bene t of NIV use in patients with hypoxemic ARF remain controversial (7)(8)(9). The effectiveness of NIV use in ARF due to viral pneumonia is controversial and has a high failure rate compared with other causes of ARF. In previous studies, unsuccessful NIV use was found a risk of increased mortality and a complication in patients with ARF (10,11). Therefore, selecting patients with ARF due to in uenza infection that might bene t from NIV use is challenging.
Risk factors are important for physicians to closely monitor if switching to invasive mechanical ventilation is necessary. The aim of this study was to identify risk factors associated NIV failure in ARF due to in uenza infection in ED.

Design and Setting
This retrospective cohort observational study was conducted at Ramathibodi hospital, a tertiary care and university hospital in Bangkok, Thailand, and was approved by the Ethics Committee of Ramathibodi hospital, Mahidol University. The trial was registered in Thai Clinical Trial Registry identi er TCTR In uenza infection was de ned by laboratory con rmed detection of in uenza A and B from nasopharyngeal swab rapid antigen test or PCR test. Acute respiratory failure (ARF) was de ned by: (1) respiratory rate > 25 breaths per minute with accessory muscle use and pulse oximetry < 90% at room air or (2) partial arterial oxygen < 60 mmHg at room air or PF ratio < 300 mmHg with PEEP > 5 cmH 2 0 or partial arterial carbon dioxide > 50 mmHg at room air. NIV failure was de ned by patients switching from NIV to endotracheal intubation in the rst 48 hours.

Data collection and clinical endpoints
Patients who met the inclusion criteria and excluded by exclusion criteria were included in this study. All patients were evaluated by emergency medicine residents, emergency medical staff and internal medicine residents. Patient demographic information, vital signs (systolic blood pressure, respiratory rate), oxygen saturation at triage, SOFA score, APACHE score, PF ratio, arterial blood gas (pH, PaCO2, HCO3, lactate level), NIV setting (mode, initial inspire pressure, PEEP), NIV parameter (Expire tidal volume) and quadrant in ltration in chest radiography were recorded. Data record forms entered in database software (RAMA-EMR) were used. Research was done on all patients who used NIV due to in uenza infection with ARF for data collection.
Clinical endpoint was success or failure of the NIV used in the rst 24 hours. This study aimed to identify risk factors associated NIV failure in ARF due to in uenza infection in ED.

Sample sizes and Statistical Analyses
To evaluate the outcome, patients were divided into 2 groups: (1) those with NIV failure and (2) those who used NIV successfully.
Sample size calculation was aimed at analyzing predictor risks for NIV failure. From our previous hospital data on in uenza infection with ARF NIV failure. Our calculation revealed that around 60 patients with in uenza infection with ARF who failed NIV and 92 patients with in uenza infection with ARF with successful NIV in the emergency department were required to provide an adequate sample size for this study (80% power, 5% alpha error).
Categorical variables were expressed as absolute values and percentages and continuous variables as medians and interquartile ranges (IQRs). Categorical variables were compared using the chi-squared test or with Fisher exact test when appropriate, while continuous variables were compared using the Mann-Whitney U test. The area under the receiver operating characteristic (AUROC) curve and odds ratio from logistic regression were used to assess predicting risk factor for NIV failure in in uenza infection with ARF in ED. The model calibration was also assessed using the Hosmer-Lemeshow (X2) test with P < 0.05 suggesting imperfective calibration.
Data were recorded using Microsoft Excel 365, while Stata version 14.0 and SPSS version 18.0 were used for statistical analyses.

Results
A total of 1453 patients with con rmed in uenza infection in ED over a 2-year period were initially recruited. After reviewing electronic medical records, 180 patients met inclusion criteria and 18 patients were excluded by exclusion criteria (14 patients and family members who signed do-not-attemptresuscitation orders and 4 patients who were transferred to other hospitals), resulting in nal inclusion of 162 patients.
Among these 162 patients, there were 72 patients (44%) who failed NIV use, 64 patients (88.9%) were intubated in 24 hours, average time to intubation was 12 hours (95%CI 11-18) and 12 patients died after being intubated post NIV failure. NIV setting: all patients used BiPAP ST mode and no signi cant differences in tidal volume and initial PEEP. Signi cant differences in patient characteristics, including SOFA score, PF ratio and new quadrant in ltration in chest radiography were observed between NIV failure group and NIV success group. The patient's demographic data are show in Table 1.   (Table 2).

Discussion
In this study, of the162 patients who visited the ED with in uenza infection with ARF from 1 January 2017 to 31 December 2018, 72 patients (44%) had NIV failure in rst 48 hours. Our study showed that SOFA score, PF ratio level and number in ltrated quadrant on chest radiography had the ability to predict NIV failure in in uenza infection with ARF. SOFA score had higher accuracy than PF ratio level and number in ltrated quadrant on chest radiography. However, there was no statistically signi cant difference in ability to predict NIV failure between three parameters.
Many studies have shown an increase in NIV use, both overall or in a speci c group such as pneumonia.
The bene ts of NIV use in ARF due to viral infection remain controversial. ARF with in uenza infection (H1N1) during the pandemic of 2009 had a high failure rate in NIV, which was associated with high mortality (2)(3)(4). Several studies have tried to identify predictive factors for NIV failure in ARF due to in uenza infection. As a result, many predicting factors have been found.
In this study, the ability of each individual predictor for NIV failure was effective and SOFA score had the greatest ability but many parameters are required to calculate SOFA score. We tried to combine PF ratio level and number in ltrated quadrant on chest radiography into a new score, named PFCXR. We create scales with logistic regression to weight parameters. Finally, we summarized this score by rubric scales (Table 3).
PFCXR score showed AUROC 0.881 (Figure 3), higher than PF ratio level and number in ltrated quadrant on chest radiography alone. The cut-off points of PFCXR score >3 for predicted NIV failure in ARF with in uenza infection was sensitivity 66.7%, speci city 93.2 % and positive likelihood ratio 9.81. Patients with higher score of PFCXR score had higher rate failed NIV and needed to be closely monitored for early intubation in 24 hours ( Figure 2).
This study has some limitations. First, this was a retrospective and single center study. Second, the management of NIV and identi cation of NIV failure were based on attending physicians in the ED. However, emergency medicine residents, emergency medical staff and internal medicine residents had been trained in resuscitation care. This variability in shift to shift in medical decision is a part of real life in the emergency department and NIV setting such as mode, initial PEEP and Tidal volume was not clinically and statistically different in this study. Third, data in some confounding variable were not collected in this study such as number of sputum suction and cough re ex. Finally, loss of larger sample sizes due to identi cation of in uenza infection was not performed in all patients who presented with respiratory infection with ARF. Nevertheless, our study showed a strong risk factor to predict NIV failure in in uenza infection with ARF and adequate sample size to nd primary outcome.

Conclusion
SOFA score, PF ratio and quadrant in ltrations in chest radiography were good predictors for NIV failure in in uenza infection with ARF. The combination of PF ratio level and quadrant in ltrations in chest radiography can be used as a predictor for NIV failure in in uenza infection and awareness for intubation. Every physician choosing with NIV should be aware of these risk factors and need closely monitored for early intubate. In pandemic in uenza infection or viral infection with acute respiratory failure, resource of mechanical ventilator and intensive care unit were limited therefor used predictors for triage or choosing who need that resource. Further study needs to validate the results in time to failed NIV and mortality.