Airflow and droplet spreading around oxygen masks: a simulation model for infection control research.

BACKGROUND: Respiratory assist devices, such as oxygen masks, may enhance the potential to spread infectious aerosols from patients with respiratory infections.
METHODS: A technique was developed to visualize exhaled aerosols during simulated patients' use of oxygen masks in a health care setting and tested using the simple, the nonrebreathing, and the Venturi oxygen masks. A smoke tracer was introduced into one of the lungs of the model to enable it to mix with the incoming oxygen and then to be further inhaled/exhaled by the model according to a variety of realistic respiratory settings (14, 24, and 30 breaths per minute, with tidal volumes of 500, 330, 235 mL, respectively) and oxygen supply flow rates (between 6 and 15 liters per minute). Digital recordings of these exhaled airflow patterns allowed approximate distances to be estimated for the extent of the visible exhaled air plumes emitted from each oxygen mask type at these settings.
RESULTS: It was found that the simple, the nonrebreathing, and the Venturi-type oxygen masks produced exhaled smoke plumes over minimum distances of 0.08 to 0.21 m, 0.23 to 0.36 m, and 0.26 to 0.40 m, respectively.
CONCLUSION: Health care workers may therefore consider any area within at least 0.4 m of a patient using such oxygen masks to be a potential nosocomial hazard zone.

Since the 2003 severe acute respiratory syndrome (SARS) outbreaks, there has been an increased interest in aerosol transmission of infectious agents. Several reviews and studies discussing the potential for aerosol transmission of various infectious agents have since been published,1, 2 including those on SARS-associated coronavirus,3, 4, 5, 6, 7, 8, 9, 10, 11, 12 varicella zoster virus (causing chickenpox), tuberculosis, and influenza.14, 15, 16 Others have studied the potential for health care procedures to act as generators or amplifiers of sources of infection, particularly the use of oxygen masks and ventilation methods.17, 18, 19

Visualization of airflows is difficult, and recent techniques have involved the use of tracer particles of nebulized liquid droplets in a cloud or solid particulates in a smoke.18, 19 Furthermore, the quantitative analysis of such suspended liquid or solid particles is difficult because it depends, ultimately, on their visibility on the recorded image. Although the movement of these tracer particles may follow the airflow movements precisely,18, 19, 20, 21 the exhaled liquid (including mucous) droplets from a potentially infected patient will probably not do so, especially because evaporation will be continuously changing its size and mass.2, 22 Such limitations notwithstanding, the visualization of exhaled air plumes from oxygen masks in health care settings is still useful for planning the control of potentially infectious aerosols, not only for a possible influenza pandemic but for daily encounters with tuberculosis (TB), measles, and chickenpox in hospitals and other health care and community environments. Here, we present some estimates of dispersal distances of smoke-visualized airflows produced using a realistic lung model at a variety of physiologic settings with the simple, the nonrebreathing, and the Venturi-type oxygen masks with different air supply rates.

Methods

Masks

Three different types of commonly used oxygen masks were used: a simple oxygen mask, a nonrebreathing mask, and a Venturi-type mask (all from Salter Labs, Arvin, CA).

Lung model and smoke generator

A Laerdal airway management trainer with an artificial lung model (INGMar Medical Adult/Pediatric Demonstration Lung Model; IngMar Medical, Ltd, Pittsburgh, PA) was used to simulate human respiration. A smoke generator (CONCEPT Colt 4 Basic Smoke Generator, 1.1 kW, 220V/50 Hz; Concept Smoke Systems, Maidenhead, Berks, UK) was used to inject smoke into the left lung of the Laerdal trainer, such that the smoke mixed with the inhaled and exhaled oxygen that was supplied at different flow rates. This is a portable smoke gun that produces a persistent, dense smoke, ideal for smoke visualization. The smoke source is canister oil, which is safe to humans. The variable airflow was supplied by an oxygen cylinder and could be set to a flow rate of between 1 and 15 liters per minute.

The Laerdal trainer has realistic anatomic features, and the lung model could simulate a spontaneously breathing patient at a variety of settings, including adjustable respiration rates (RR), tidal volumes (TV), and peak flows. Within the limits of the lung model performance, combinations of RR in breaths per minute and TV in liters were chosen to create 3 “respiratory models,” each representative of a commonly encountered patient situation (see Table 1).

Table 1

Different physiologic settings or “respiratory models” used in this study

Settings	Model 1	Model 2	Model 3
Respiratory rate (breaths/min)	14	24	30
Tidal volume (mL)	500	330	235

Environment

A clean room was constructed for the study with the following dimensions: floor space 2 × 2 m, ceiling height 2.4 m. The ventilation was provided by adjustable speed airflow on the ceiling with ventilation windows on at the bottom of the walls in the corners. An airflow meter monitored the ambient airflow. During the study, the air change rate was maintained at 12 air changes per hour, with the room temperature and humidity remaining at a relatively constant 22.3°C and 62%, respectively. Illumination was provided by normal room strip fluorescent lighting, with the model filmed against a black backdrop to enhance the contrast ( Fig 1A).

Fig 1

An approximation was made of the maximum visible dispersal distance traveled by the exhaled smoke plume (y1 in A, shown by dotted line arrow), shown here for the Venturi mask (at 40% O2, 6 L/min, at respiratory model setting No. 1). This distance had to be estimated, taking into account the geometric distortion due to the different relative positions of the Laerdal trainer, the digital camera, and the direction traveled by the exhaled smoke plume. It was estimated by knowing the real value of the chin-to-chest distance (x2 in B) on the Laerdal trainer, the measured screen distances of x1 and y1 (in A), and the relative error of these measured screen distances, estimated by the ratio (x1/x2)/(y1/y2), where the real lengths of the wooden rods, x2 and y2, were known. Note that these ratios x1/x2 and y1/y2 will not be exactly the same because they are measured in different directions and different planes, so they will be at different distances from the camera. It is this difference between ratios x1/x2 and y1/y2 that gives the value of the geometric corrections shown in Table 2.

Table 2

Estimated maximum visible smoke dispersal distances, with geometrical correction, at peak of exhalation, for each oxygen mask and for each respiratory model number shown in Table 1

Oxygen mask type	Oxygenflow rate (L/min)	Respiratory model number	Estimated maximum visible dispersal distance ± geometrical correction (cm), (ie, the difference in the ratios x1/x2 and y1/y2)
Simple	10	1	9.5 ± 0.6 (ie, about 6%)
	10	2	12.5 ± 0.8
	10	3	8.3 ± 0.5
	15	1	11.2 ± 0.7
	15	2	20.7 ± 1.2
	15	3	16.3 ± 1.0
Nonrebreathing	8	1	34.1 ± 3.1 (ie, about 9%)
	8	2	35.8 ± 3.2
	8	3	23.5 ± 2.1
	10	1	24.6 ± 2.2
	10	2	24.6 ± 2.2
	10	3	26.3 ± 2.4
Venturi-type	6 (35% O2)	1	27.2 ± 1.1 (ie, about 4%)
	6 (35% O2)	2	33.8 ± 1.4
	6 (35% O2)	3	27.4 ± 1.1
	6 (40% O2)	1	39.7 ± 1.6
	6 (40% O2)	2	39.7 ± 1.6
	6 (40% O2)	3	26.3 ± 1.1

NOTE. Each of the value of percentages shown are in the last column (6%, 9%, 4%) applies, approximately, to all measurements made for that particular oxygen mask type.

Image capture and smoke dispersal distance estimates

A digital video (DV) camera (Sony DCR-DVD100E) was used to capture the images of exhaled flows from the respiratory model wearing the different masks at different airflow and respiratory model settings (shown in Table 1). The DV images were examined carefully, and only those that showed the exhaled air plumes most visibly are presented here. The images selected for further analysis were saved as high-resolution bitmaps (approximately 760 × 570 pixels) and enlarged in Microsoft Powerpoint (Microsoft Corp, Redmond, WA) to estimate the maximum smoke dispersal distances.

The maximum dispersal distance was estimated using the chin-to-chest distance in one direction (measured and real distance x1 in Fig 1A) of the Laerdal trainer as a scale for estimating the distance traveled in another direction by the visible smoke (measured and real distance y1 in Fig 1A) in the captured images. To estimate the error in making such distance measurements from the 2-dimensional (2-D) screen images that were enlarged and printed to perform this measurement, the Laerdal trainer was again set up with the 3 types of oxygen masks, with wooden rods. The rods were actually tapered chopsticks, and the pointed ends conveniently allowed their secure insertion into the particularly small exit holes of the simple and nonrebreathing masks. These rods were of known lengths and were positioned to represent approximately the directions of the chin-to-chest (x1) and exhaled smoke plume (y1) measurements for each mask. These smoke plumes were emitted in a roughly conically shaped space, with the cone apex adjacent to the oxygen mask, and the rod in the y1 direction was positioned to lie within this space, accordingly. The Laerdal trainer with the mask and attached rods was carefully positioned to be as similar as possible to that of the corresponding captured digital image for the same mask with its exhaled smoke plumes then photographed at a resolution of 1024 × 768 pixels with a digital camera (Canon Ixus II, Canon Inc., Tokyo, Japan).

This was necessary because of the geometric distortion present because of distances x and y not lying in the same direction or the same plane, which was caused by the relative positions of the Laerdal trainer, the camera, and the direction taken by the exhaled smoke plumes (compare Fig 1A and 1B). The distances x1 and y1 were measured from enlarged, printed still images taken from the digital film footage for each mask, at each setting (as shown in Fig 1A). Knowing the real length of the rods (real lengths x2 and y2 in Fig 1B), the 2 ratios x1/x2 and y1/y2 were compared. If these distances had been measured in the same plane and the same direction on the same screen image, then these 2 ratios should be equal. Any difference between the ratios x1/x2 and y1/y2 would be due to geometric distortion and give an estimate of the error incurred by measuring and scaling up distances measured from the enlarged, printed, 2-D still images of the exhaled smoke plumes. Therefore, scaled up, real value of y1 could be estimated as the following:

real y2 = measured y1/(measured x1/real x2).

Small differences in the position of the Laerdal trainer and the rods between the original captured DV images and the still photographed images would not affect the real distance estimate y1 too much because the exhaled smoke plumes were conical in shape. As long as the rod representing the y1 direction did not move too far in any direction, it would probably have remained within the cone of the exhaled smoke plume. Hence, the positioning of the Laerdal trainer with the rods attached to the masks, using a direct eye comparison with the capture DV image, for the still photograph, would still allow a realistic estimate to be made for y1.

Results

The most visible exhaled airflows were seen from the simple oxygen mask at oxygen flow rates of 10 L/min and 15 L/min, from the nonrebreathing mask at 8 L/min and 10 L/min, and from the Venturi-type mask at 35% O2 at 6 L/min and 40% O2 at 6 L/min. Table 2 shows the estimated distance (with geometric corrections) traveled by the exhaled smoke plumes, based on the scale factor calculated by the following:

real y2 = measured y1/(measured x1/real x2).

For the simple oxygen mask (at 10 L/min and 15 L/min), the exhaled smoke plumes appeared to travel the least distance. The exhaled smoke plumes from the nonrebreathing mask (at 8 L/min and 10 L/min) were more visible and appeared to travel farther than for the simple oxygen mask. Although there is some overlap with the distances traveled by exhaled smoke plumes from the nonrebreathing mask, overall, those from the Venturi-type mask (as settings of 35% O2 at 6 L/min and 40% O2 at 6 L/min) appeared to travel the farthest. Overall, the visible dispersal distances of the exhaled smoke plumes from the simple, the nonrebreathing, and the Venturi-type oxygen masks, at the different oxygen flow rates, with the different respiratory models used in this study (shown in Table 1) ranged from 0.08 to 0.21 m, 0.23 to 0.36 m, and 0.26 to 0.40 m, respectively. From these distance estimates, respiratory model No. 2 (24 breaths/min, TV 0.33 L) appeared to produce the greatest visible dispersal distance of the exhaled smoke plume.

Discussion

In this study, we have captured and characterized digital images of the exhaled airflows produced from a variety of oxygen masks using an artificial lung model at several physiologic settings. From these results, there seem to be relative differences in the visible distances traveled by the exhaled plumes from each type of mask. Although the relatively higher oxygen flow rates of 10 L/min and 15 L/min used for the simple oxygen mask were not used for the nonrebreathing and the Venturi-type masks, from the results shown in Table 2, it is not unreasonable to assume that the maximum visible dispersal distances (and hence the minimum distances traveled) of the exhaled smoke plume for these latter 2 masks would almost certainly exceed those of the simple oxygen mask at these higher oxygen flow rates. These images should be useful in assessing the infection control risk from potentially infectious exhaled plumes from patients with respiratory symptoms using such masks in health care and community settings.

There are relatively few papers that attempt to visualize exhaled airflows as a potential source of infection. Somogyi et al used a very simple setup, using flash photography, to illustrate the potential dangers of exhaled airflows. Inhaled saline mist was exhaled through 3 commonly used types of oxygen masks, demonstrating that exhaled plumes could spread infectious aerosols over a distance of approximately 1 to 2 head diameters, ie, up to approximately 0.3 m, where an average head diameter is approximately 0.13 to 0.16 m.

Hui et al18, 19 used a 2-D green (527-nm wavelength) laser sheet to illuminate oil-based smoke particles (<1-μm diameter) to illustrate the dispersion of potentially infectious exhaled plumes from a variety of respiratory support devices and procedures, including oxygen masks and noninvasive positive pressure ventilation. From these latter 2 studies, the estimated dispersal distances of exhaled, potentially infectious, air from patients using such respiratory assist devices were in the range of 0.40 to 0.50 m. These studies used a more sophisticated lighting system with monochromatic, coherent laser light-sheet. They also tested a more limited range or a totally different respiratory assistance setup, ie, a simple oxygen mask at 1 respiratory setting (RR = 12 breaths/min; TV = 0.5 L; oxygen flow rate = 4 L/min), and noninvasive positive-pressure ventilation. This present study also differs from those of Hui et al18, 19 in that it was performed in a more typical isolation room environment, with ambient lighting and a background ventilation rate of 12 air changes per hour. The exact effect of the 12 air changes per hour on the extent of the smoke plumes is difficult to quantify, although their dispersal patterns as shown here may be more representative of those seen in isolation rooms commonly used for patients suspected to be infected with respiratory pathogens of higher morbidity/mortality.

It is interesting that the extent of visible smoke dispersion recorded is similar in all these setups (up to a maximum of approximately 0.4 m from the patient), which may well be due to the different lighting conditions used, particularly between this present study and that of Hui et al.18, 19 With the more intense laser-sheet lighting used by Hui et al, it is likely that further distances would have been recorded for the limits of visible smoke dispersal, at the higher oxygen flow rates used in this study. Hence, the 0.4-m distance estimated here can only be considered a lower limit for the extent of any “nosocomial hazard zone,” as defined by the visible dispersal limits of the smoke tracer used here.

Although current guidelines for the infection control of aerosol transmission is mainly limited to 3 pathogens (measles, chickenpox, and tuberculosis),2, 24 other infectious agents such as influenza and whooping cough may be just as transmissible in certain situations, as suggested by previous relatively high estimates of their basic reproductive numbers (R0) of approximately 2 to 20 (for influenza) and 15 to 17 (for whooping cough). These guidelines and R0 values apply to natural methods of dissemination, such as normal breathing, coughing, sneezing, and talking. However, when respiratory assist devices are used and infectious exhaled air is mixed and expelled with oxygen/air at high flow rates, the potential for increased transmissibility is evident.

It is important to recognize the limitations of this visualization approach. The visible boundaries of exhaled flows can only be a guide to the real behavior of infectious droplets in exhaled air. Although Hui et al18, 19 suggested upper limits of exhaled-plume distances as possible safety zones, some of the more distantly dispersed particles may not be visible in their image capture systems. Somogyi et al presented their results in a slightly different manner by stating that the visible cloud (or smoke), in fact, indicates the minimum distance traveled by the exhaled air, which may be a more cautious approach that can be used to interpret the results presented in this study. Hence, from this study, those of Hui et al,18, 19 and to a lesser extent that of Somoygi et al, there is a suggestion of a minimum baseline “nosocomial hazard zone” extending to at least 0.4 m away from patients using such respiratory support devices. It is likely to extend farther than this, but this could not be quantified from this study. Such a “nosocomial hazard zone” may well increase in size when the patient coughs or sneezes. This could not be examined using the same respiratory model used in this study (there is no cough function) but is certainly worthy of further investigation using other models, eg, perhaps using human volunteers.

With concerns for a possible influenza pandemic, aerosol transmission infection control is becoming more important. Although direct contact transmission predominates as the main route of nosocomial transmission for most pathogens, the relevance of long-distance transmission has become a concern in the design of new hospitals. This has led to an increased number of single-bed, negative pressure isolation rooms, as well as greater distances between beds in such new facilities. Since the 2003 SARS outbreaks, the modeling of airflows in health care institutions has been performed in a variety of ways and on different scales by both engineers7, 8, 9, 26, 27, 28, 29 and physicians.3, 5, 6, 12, 17, 18, 19 Studies assessing the effectiveness of personal protective equipment, such as masks, have also been performed.

With influenza, the relative risk from airborne of contact transmission is still being hotly debated, and even existing infection control guidelines have been questioned in this regard. The airborne route of influenza has been well documented,31, 32, 33, 34, 35 so why do some guidelines still treat influenza as a short-range rather than a long-range airborne disease? Perhaps the most useful debate has been published by the UK Health Protection Agency's Guidance for Pandemic Influenza that comprehensively summarizes the evidence for the different transmission routes of influenza. It still concludes, however, that influenza is mostly transmitted by large droplets and direct contact. Tellier goes one step further and presents a convincing argument as to why influenza should be considered as a true airborne infection and recommends N95 respirators as the minimum level of personal protective equipment for the purposes of pandemic influenza planning.

These debates notwithstanding, the use of respiratory assist devices, similar to those shown in this study, on the basis of the physics alone, has the potential to increase the distance over which influenza and other aerosol-transmitted infections can be naturally transmitted, thus elevating its potential transmission risk to that of a true airborne disease in such situations.

Most nosocomial and community-acquired respiratory infections are mild and unlikely to cause severe morbidity or mortality during a nonpandemic period. However, when a new respiratory pathogen arises, with the potential to cause high morbidity and mortality, eg, SARS and more recently avian H5N1 or possibly some other future pandemic influenza strain, this baseline data should be useful in reducing nosocomial transmission and enhancing health care workers' awareness of the risks posed by the use of such respiratory assist devices. In addition, it is known that, unlike SARS-associated coronavirus, influenza may be presymptomatically transmissible.36, 37 Hence, there is a convincing argument for all staff working in the immediate vicinity, ie, within a zone of 0.4 m, of such patients, to be wearing N95 masks as their minimum personal protective equipment during an influenza pandemic or when dealing with any other respiratory pathogen with the potential to cause high morbidity and mortality.