key: cord-0900723-8taigwel
authors: Strand-Amundsen, Runar; Tronstad, Christian; Elvebakk, Ole; Martinsen, Tormod; Dybwad, Marius; Lingaas, Egil; Tønnessen, Tor Inge
title: Quantification of aerosol dispersal from suspected aerosol generating procedures
date: 2021-09-24
journal: ERJ Open Res
DOI: 10.1183/23120541.00206-2021
sha: 27b02f0694bd4ac8216ec5996b31f0ea17c6d9a0
doc_id: 900723
cord_uid: 8taigwel

BACKGROUND: Oxygen delivering modalities like humidified high-flow nasal cannula (HFNC) and non-invasive positive-pressure ventilation (NIV) are suspected of generating aerosols that may contribute to transmission of disease such as COVID-19. We sought to assess if these modalities lead to increased aerosol dispersal compared to the use of non-humidified low-flow nasal cannula oxygen treatment (LFNC). METHODS: Aerosol dispersal from 20 healthy volunteers using HFNC, LFNC and NIV oxygen treatment was measured in a controlled chamber. We investigated effects related to coughing and using a surgical facemask in combination with the oxygen delivering modalities. An aerodynamic particle sizer measured aerosol particles (APS3321, 0.3–20 µm) directly in front of the subjects, while a mesh of smaller particle sensors (SPS30, 0.3–10 µm) was distributed in the test chamber. RESULTS: Non-productive coughing led to significant increases in particle dispersal close to the face when using LFNC and HFNC but not when using NIV. HFNC or NIV did not lead to a statistically significant increase in aerosol dispersal compared to LFNC. With non-productive cough in a room without air changes, there was a significant drop in particle levels between 100 cm and 180 cm from the subjects. CONCLUSIONS: Our results indicate that using HFNC and NIV does not lead to increased aerosol dispersal compared to low-flow oxygen treatment, except in rare cases. For a subject with non-productive cough, NIV with double-limb circuit and non-vented mask may be a favourable choice to reduce the risk for aerosol spread.

For COVID-19 patients experiencing reduced lung function with insufficient oxygenation, using humidified High-Flow Nasal Cannula oxygenation (HFNC) and non-invasive ventilation (NIV) have been reported to reduce the frequency of intubation and subsequent invasive mechanical ventilation [1] . HFNC has been associated with a reduction in intensive care unit length of stay [2] . As these treatment modalities are suspected to be aerosol generating procedures [3] , some medical centres were initially reluctant to use them and thereby lowered the threshold for intubation. With increasing numbers of COVID-19 patients overwhelming the capacity of ICU beds with invasive ventilation, many patients have been treated in stepdown units with HFNC or NIV. For the health and safety of healthcare workers, it is important to assess to what extent these procedures generate aerosols. Documentation for airborne transmission being a component for the spread of COVID-19 is surfacing [4, 5] . At the same time, there are reports that the potential viral load of aerosolized particles may be low [6] .

Several recent approaches have investigated aerosol dispersal and spread during oxygen therapies and research is still ongoing to answer these questions [7] [8] [9] [10] [11] [12] [13] . While the present evidence is non-conclusive if HFNC or NIV oxygen treatment has high aerosol dispersal potential, results from recent studies suggests that HFNC and NIV does not lead to significantly increased aerosol dispersal compared to low-flow oxygen modalities [14] [15] [16] [17] .

In our study aerosol levels were measured in a controlled chamber, where 20 healthy volunteers used HFNC, non-humidified low flow nasal cannula (LFNC), and NIV in BIPAP mode. We investigated the parameter "coughing" with all oxygen modalities, and for HFNC we also investigated using a "surgical facemask". An aerodynamic particle sizer was used to measure (optical) aerosol particles (0. 3 -20 µm) directly in front (30 cm) of the subject, while smaller optical particle sensors (0.3 -10 µm) were distributed in the test chamber.

We sought to assess the following questions:

1. Does the use of HFNC or NIV lead to increased aerosol dispersal compared to the use of LFNC? 2. How does non-productive coughing contribute to aerosol dispersal in settings with HFNC/NIV/LFNC? 3. How is the spatial distribution of aerosols in a confined space influenced by HFNC/NIV/LFNC?

We recruited healthy adult volunteers age >18 years. The recruitment process and study were approved by the regional committee for medical and health research ethics in Norway (approval reference: REK 153325) and written consent was obtained from the subjects.

The protocol comprised three oxygen modalities; LFNC, HFNC and NIV ( Figure 1B) , split into eight events.

We included coughing events with all modalities and included a test with a surgical facemask (Medline, EN14683 Type II) within the HFNC modality. The subjects were instructed to provoke coughing (voluntary intensity) every 30 seconds during the coughing tests. To counter potential carryover effects by protocol sequence, the modality order was separated into four differently ordered subsets where the 20 subjects were 

The test chamber ( Figure 1A ) with no external active air supply or internal air currents had an internal volume of 11.36 m 3 (l:w:h: 234.5 X 234.5 X 206.5 cm), with an anteroom to reduce aerosol contamination.

A City M Air Purifier (CAMFIL, USA) with an airflow of 7.2 m 3 /min was used to zero the particle levels before and between each event. Each subject wore a disposable polypropylene non-woven coverall (Worksafe).

Two different classes of aerosol instruments were used to count and measure particle sizes. A TSI model 3321 Aerodynamic Particle Sizer® (TSI Incorporated, MN, US) was used to sample the air in the breathingzone 30 cm in front of the subject. Nine small particle sensors (SPS30, Sensirion AG, ZH, Switzerland) were used (custom-made wireless setup) for simultaneous particle counting at different locations ( Figure   1A ). Both instruments measure the aerosol concentration over a range of particle sizes, from 0.3 µm (optical size) or 0.5 µm (aerodynamic size) to 20 µm for the APS 3321, and from 0.3 µm (optical size) to 10 µm for the SPS30. For readability we use the name "breathing-zone single sensor" for the APS 3321, and "mesh sensors" for the SPS30 sensors.

Prior to performing the experiments, the particle sensors were tested with particle generation from humans and with a NaCl aerosol generator. Temperature and relative humidity in the test chamber were continuously logged during the experiments, with an AM2320 sensor (Guangzhou Aosong Electronics Co., Ltd, China). There was low variation with an average of 26.0 +/-0.7 STD degrees C and 47.1 +/-5.6 STD % relative humidity.

Comparisons were made of the average particle concentration between all pairwise combinations of events relevant to the research questions. As the distributions in concentration values were highly skewed and deviated from a normal distribution (confirmed by Shapiro-Wilk test), comparisons of medians were conducted. To quantify differences between events, median differences were calculated, and their 95% confidence intervals were estimated using the bootstrap method. The Wilcoxon signed-rank test was used to quantify the statistical significance of the differences. These comparisons were done on both datasets (breathing-zone single sensor and mesh sensors).

To obtain statistical estimates of changes in particle concentration over time, the particle concentration was modelled as a function of time using a linear mixed-effects model with random intercept and slope, using the averaged particle concentrations over each minute as model input. A full covariance matrix based on Cholesky parameterization was used as covariance structure in the models, selected based on the Akaike information criterion on models fit on the breathing-zone single sensor dataset. The particle concentration dependency on the four distance-categories of the mesh sensors ( Figure 1A ) was modelled statistically for each event using a linear mixed-effects model with random intercept and fixed slope.

Estimation of Spearman correlations between emitted particle levels and age, gender and weight were done in Graphpad 9.0.0. All other statistical calculations were done in Matlab R2019b.

Descriptive statistics 20 healthy adult subjects were included in the study, seven were female, mean age was 43 years (11.3 SD) and mean weight 80 kg (18.7 SD). Descriptive statistics for the measurements by the breathing-zone single sensor are presented in Table 1 . As visualized in Figure 2 (A, B) , there was a large spread in the particle concentration among subjects, with a consistent dispersion around 100% for most events. The distributions were positively skewed with tails of extreme values. Most of the particles measured by the breathing-zone single sensor were in the range of 0.3 -5 µm ( Figure 2E ). Our measurement-setup fulfilled the criteria for representative aerosol sampling (intake) and high transport efficiency (tubing) for particle sizes ≤5 µm. The particle dispersal from coughing produced spiked measurements, with the largest spikes attributed to only a few individuals ( Figure 3 and Figures E1-E3 ). No correlations were found between aerosol dispersion and age, gender, or weight. 

The results from pairwise comparisons of events are shown in Table 2 (breathing-zone single sensor dataset). There was a median increase in particle concentration during HFNC compared to LFNC, but the confidence intervals were wide, including changes in both directions. For particles > 5µm, we measured a statistically significant difference between HFNC and LFNC, but the total particle count in this size range was small (median difference 0.5 particles/litre). Comparing events with and without coughing, there was a relatively large increase in particle concentration when coughing during LFNC, and to a lower extent during HFNC. The median particle concentration was lower with HFNC+cough compared to LFNC+cough, but this difference was not statistically significant. Compared to LFNC+cough and HFNC+cough, NIV+cough led to lower levels of particle dispersion, especially in the particle range ≤ 1.0 µm. Wearing a surgical facemask reduced the mean particle concentration during HFNC, both with and without coughing, but the differences were not statistically significant. Comparing NIV with LFNC, the particle concentrations were similar. None of the mesh sensors measured a statistically significant difference in particle concentration between HFNC and LFNC. In agreement with the breathing-zone single sensor dataset, the largest differences were attributed to coughing, with the largest effects registered by the sensors near the breathing zone. For details see online supplement (Table E1 ).

Trends over time were close to zero for the events without coughing (Table 3 , Figure 3 ). Although not statistically significant, the estimated trends were largest for LFNC and HFNC with coughing. With HFNC there were typically some early spikes before a slow reduction over time, resulting in slightly negative estimates. For details view Figures E1-E3 (online supplement) . The results were similar for the three mesh sensors that were close to the subject ( Figure 1 , Table E2 ). 

With the mesh sensors there were few distance-dependent differences detected except for events involving coughing (Table 4, Figure 4 ). With coughing, the differences between 30 and 100 cm were small, but the particle concentration was significantly reduced at 180 and 285 cm. Notably for HFNC-M, the particle concentration was lowest at the closest distance but significantly increased 100 cm in front of the subject. 

We investigated the aerosol dispersal from the respiratory tract of 20 healthy adult volunteers in a test chamber with a close to zero particle baseline level, while using LFNC, HFNC or NIV, combined with the parameters "coughing" and "surgical facemask". The key findings were that HFNC or NIV did not lead to large increases in aerosol dispersal compared to LFNC, except in rare cases (3/20) . When investigating changes in particle concentration at different distances, there were no large differences between LFNC, HFNC and NIV.

There was a tendency for small increases in median particles during HFNC compared to LFNC, for the lower particle ranges (< 5 µm). Helgeson et al. reported a similar non-significant increase at 4 cm from the mouth of the subjects when using HFNC or oxymask. For the particle range > 5µm we measured a statistically significant difference at 30 cm, but the overall number of particles measured in this range was low, with a median difference of just 0.5 particles per litre ( Table 2 ). The main contribution to this increase came from three subjects that dispersed more particles (of all sizes measured) when using HFNC compared to LFNC (Table 2, Figure 2 .C/D), indicating that in rare cases healthcare workers could expect patients connected to HFNC to disperse more particles. Comparing NIV with LFNC, the particle concentrations were similar, indicating that NIV did not lead to increased aerosol dispersal (Table 2, Figure 2 .C/D). The mesh sensors dataset showed similar results for pairwise comparisons between events for all sensors (online supplement, Table E1 ). There was a trend with HFNC with spikes during the first minutes ( Figures E1-E3) that could be related to the nasal irritation that some of the subjects (30%) reported during this period. This nasal irritation might explain the slightly higher median with HFNC compared to LFNC.

Our results generally agree with recent studies that indicate that HFNC does not lead to significantly increased aerosol dispersal compared to LFNC in a group of healthy adults [7] [8] [9] 11] . We used a controlled chamber and zeroed particles between each event, to ensure accurate measurement of aerosol dispersal from the subjects without dampening effects from active air circulation.

Most of the recent studies have investigated particle dispersion in a negative-pressure room to reduce the ambient background [11, 16, 18] , but this does not completely exclude ambient particles that potentially can contribute to the measurements. Wilson et al. used zeroing of particle levels between measurements like we did [15] . (60 L/min), but to achieve the spatial resolution they measured eight separate times with a handheld instrument. We measured simultaneously at 10 different locations to the sides and in front of the subject for a distance up to 2.85 meter (Figure 1 ). Our results corroborate the results reported by Bem et al. in that there were no significant differences particle levels between LFNC and HFNC for the particle range below 5 µm [16] .

While we found few differences in particle levels vs different distances for LFNC, HFNC and NIV without coughing, we did detect a difference in particle levels as a function of distance for episodes involving coughing. LFNC+cough and HFNC+cough led to relatively large increases in particle concentration close to the subjects (Table 3 , Figure 3 ), indicating how coughing can lead to a build-up of aerosol particles (10-minute period) in an enclosed space with no air circulation, where aerosol concentration drop-off occurred between 100 cm and 180 cm ( Figure 4 , Table 4 ). In a hospital setting there will typically be an active room ventilation creating airflows moving in the room. The turbulences formed by these air currents will vary with each location. Consequently, making exact predictions about safe distances are difficult. If the clinical setting involves low air-change rates like in a standard patient room,

our results indicate that a clinician treating (LFNC or HFNC) a coughing patient may be exposed to significantly lower particle levels at 180 cm distance compared to a 100 cm distance or closer.

There was a significant reduction in measured particle levels when comparing HFNC+cough or Wearing a surgical mask (HFNC+mask+cough) had a dampening effect on the large cough-spikes of particles in the medium (>1µm & ≤ 5 µm) and large (> 5µm) particle size range (Figures E2 and E3) . As a comparison Wilson et al. reported that using a surgical mask halved the number of particles measured in most instances they tested when using HFNC and NIV on 10 healthy volunteers.

Our primary interest was in the 0.3 -5.0 µm particle size range ( Figure 2E ) which are typically the dehydrated remains from slightly larger primary respiratory droplets. While particles in this size range represents only a fraction of the total volume emitted from the respiratory tract and have lower viral load potential than the larger particles, they are emitted in huge numbers compared to the larger particles [19] , and can remain airborne for longer periods of time (hours) under all temperature and relative humidity conditions [20] . At the same time, particles in this range are known as the "breathable/inhalable fraction" and will typically reach and be deposited in the bronchioles and the alveoli of the lower airways [21] . The smallest particle size (0.3µm) in the range we measured ( Figure 2E) , is known as "the most penetrating particle size" due to the difficulty of capturing this particle size with particle filters [22] . Provoked coughing by healthy adults may be a poor representation of actual aerosol dispersal by patients that have airway related diseases. Also, there was large individual variation in coughing intensity.

Although the measurements of the low-cost mesh sensors tended to agree with the more precise breathingzone single sensor, the results based on these sensors are likely less accurate [24] . While we aimed at creating a similar setting for each subject, there were minor variations in sitting positions and the angles between the breathing-zone of the subjects and the sensors, that might have influenced the accuracy of the particle measurements. The poor ability for the SPS30 sensor to separate particle sizes in bins [24] was a limitation for our ability to discern how different particle sizes propagate spatially.

In the study group of 20 healthy individuals, using HFNC and NIV did not lead to increased aerosol dispersal compared to low-flow oxygen treatment, except in rare cases. For a subject with non-productive cough, NIV with double-limb circuit and non-vented mask may be a favourable choice to reduce the risk for aerosol spread. 

High-Flow Nasal Cannula in Critically III Patients with Severe COVID-19

High-flow nasal oxygen in patients with COVID-19-associated acute respiratory failure

Aerosol generating procedures and risk of transmission of acute respiratory infections to healthcare workers: a systematic review

Transmission of COVID-19 virus by droplets and aerosols: A critical review on the unresolved dichotomy

Identifying airborne transmission as the dominant route for the spread of COVID-19

Aerosol persistence in relation to possible transmission of SARS-CoV-2

Variation in Aerosol Production Across Oxygen Delivery Devices in Spontaneously Breathing Human Subjects

Aerosol Risk with Noninvasive Respiratory Support in Patients with COVID-19

Effects of surgical masks on droplet dispersion under various oxygen delivery modalities

Comparison of high-flow nasal cannula versus oxygen face mask for environmental bacterial contamination in critically ill pneumonia patients: a randomized controlled crossover trial

Aerosol Generation from the Respiratory Tract with Various Modes of Oxygen Delivery

Author's Reply on High-Flow Nasal Cannula for COVID-19 Patients: Low Risk of Bio-Aerosol Dispersion

Airborne spread of SARS-CoV-2 while using high-flow nasal cannula oxygen therapy: myth or reality?

Particulate generation with different oxygen delivery devices

The effect of respiratory activity, non-invasive respiratory support and facemasks on aerosol generation and its relevance to COVID-19

Risk of Aerosol Formation During High-Flow Nasal Cannula Treatment in Critically Ill Subjects

Placing a mask on COVID-19 patients during high-flow nasal cannula therapy reduces aerosol particle dispersion

High-flow nasal cannula for COVID-19 patients: risk of bioaerosol dispersion

Particle sizes of infectious aerosols: implications for infection control

Airborne transmission of viral respiratory infections from an aerosol physics perspective

Quantity and size distribution of cough-generated aerosol particles produced by influenza patients during and after illness

Laboratory evaluation of particle-size selectivity of optical low-cost particulate matter sensors