Evaluation of droplet dispersion during non-invasive ventilation, oxygen therapy, nebuliser treatment and chest physiotherapy in clinical practice: implications for management of pandemic influenza and other airborne infections

Notes

Permissions

Chapter 1 Introduction

Respiratory viral infections, such as influenza, are spread by droplets, an aerosol of infected material or by direct or indirect contact with contaminated surfaces. The mode of transmission and the factors influencing this are important, as they have key implications for infection control in patients and staff, and therefore pandemic planning. Droplets in the respirable range (around 5 µm) may play a significant part in transmission,1 but the role of aerosols has been questioned2 and there are few studies quantifying viral load in droplets or aerosols. An observational study3 of influenza A and influenza B in exhaled breath showed viral RNA in one-third of infected patients, and 99% of particles had a diameter of < 5 µm when sampled during tidal breathing.

While some individuals recover from seasonal or H1N1 influenza, having experienced minimal symptoms, a subgroup of high-risk patients may develop complications, including respiratory failure,4,5 and, in new more pathogenic strains, such as H5N1, respiratory insufficiency may occur in more than 50% of those affected. These patients are managed with antiviral therapy and antibiotics for secondary bacterial pneumonia, but the mainstay of management is supportive respiratory care, which includes high-flow oxygen therapy (O2) for hypoxaemic patients, and ventilatory support for those with ventilatory insufficiency. 6,7 Adjunctive therapy can include nebulised bronchodilator for patients with underlying asthma or chronic obstructive pulmonary disease (COPD), and physiotherapy is used to facilitate secretion clearance for those in whom influenza has precipitated an infective exacerbation of chronic lung disease, such as COPD, bronchiectasis or cystic fibrosis.

Coughing and sneezing patients can shed relatively large particles (> 10 µm) that travel short distances and may contaminate the bedside environment. Smaller droplets or aerosols will remain airborne for longer periods and disseminate over greater distances. 1 The definition of an aerosol varies but most authorities characterise this as consisting of droplets of < 5 µm. Some medical procedures have been termed ‘aerosol generating’, as the common feature is that they are associated with high or augmented inspiratory and expiratory tidal flows, which may increase viral dissemination but this classification is based on assumptions rather than systematic evidence. The list of aerosol-generating procedures (AGPs) differs a little from country to country but in Department of Health (DH) guidance7,8 these include bronchoscopy, intubation of the airway and invasive ventilation manoeuvres, such as open suctioning, cardiopulmonary resuscitation, non-invasive ventilation (NIV) and continuous positive airway pressure (CPAP) therapy, high-frequency oscillation ventilation, and induction of sputum. Certain other procedures, such as delivery of nebulised medication therapy and high-flow O2 are considered to be possible aerosol generators, but a lesser infective risk. 8 There is an association between some of these AGPs and an increased incidence of severe acute respiratory syndrome (SARS) in health-care workers9–11 and the risk of superspreading events on wards. 12 This has implications for the safe care of patients and risk management for nurses, doctors, physiotherapists and other health-care workers, and has provoked an ethical debate on the duty of care of health-care staff in pandemics. 13,14

Much of the evidence for the link between AGPs and increased transmission of respiratory viral infection was generated during the SARS epidemic. In Toronto and Singapore, health-care workers constituted approximately 20% of critically ill cases. Infection rates were higher in doctors and nurses carrying out endotracheal intubation [relative risk (RR) 13.29, 95% confidence interval (CI) 2.99 to 59.04, p = 0.03], while nurses caring for SARS patients receiving NIV may have been at increased risk (RR 2.23), but this finding did not reach significance (95% CI 0.25 to 21.76, p = 0.5). 9 In a case–control study of dissemination of SARS from an index case to other patients on the same ward, Yu et al. 12 showed an increased risk associated with the index patient requiring O2 or bilevel NIV. Case reports15,16 have also linked transmission of infection to nebuliser use in the index patient. However, patient variables are also likely to be important, as sicker patients who may have a higher viral load are more likely to require O2 and ventilatory support, and those with underlying asthma who require nebuliser therapy may cough more due to airway hyper-reactivity. For these reasons specific infection control precautions have been introduced for unavoidable AGPs and these include use of high-efficiency FFP3 (or N95) masks, eye protection, gowns, aprons and gloves. 8 Guidelines also suggest that AGPs should only be used if necessary, and controversy has arisen over the role of NIV. 17,18 Its use is recommended with appropriate precautions in some national guidelines,7 but not in other guidelines, and NIV use is cautioned against by some authorities. 19–21

Non-invasive ventilation (NIV) and CPAP are unlikely to have a role in acute lung injury caused by influenza or in secondary bacterial pneumonia, or in patients with multisystem failure. 17 However, NIV was used successfully in some SARS cases,22,23 and as indicated in DH guidelines,24 there is potential for NIV to reduce the need for intubation in influenza pneumonia in those with chronic respiratory disease,25 to facilitate extubation, and to widen the provision of ventilatory support outside the intensive care unit. It may also be used as a ceiling of ventilatory care in patients with COPD, congestive cardiac failure and other serious comorbidities, and to palliate symptoms in those with end-stage disease in whom ICU admission is not indicated. These indications should be set against the risks of droplet dissemination during the delivery of NIV – yet at present those risks have not been quantified.

It is also important to note that there are problems in interpreting the evidence of transmission of infection during SARS. This is because transmissibility could have been increased by an inadequate use of protective personal equipment (PPE) in early cases;11,26 NIV equipment has evolved since 2003–4, and there have been subsequent experimental studies that have investigated air flows around oxygen masks and during NIV. 27–30 These studies used human simulator models or normal subjects mimicking respiratory distress. Hui et al. 28,31 have carried out a series of experimental studies analysing particle spread from NIV and oxygen masks,32 using smoke particles as a proxy of droplets in expired air. However, human simulators may not closely reflect the behaviour of sick patients, and smoke particles are considerably smaller (< 1 µm) than droplets generated by coughing and sneezing (range 5 to > 10 µm). Therefore, the behaviour of smoke particles may not accurately represent droplet dispersion. Other workers have used a Schlieren optical visualisation technique33 to demonstrate exhaled air flows in normal subjects when coughing with and without masks. These provide useful information on expiratory flow profiles but none of the investigations has been carried out using the range of common clinical interventions defined as AGPs, analysed droplet size or studied patients with respiratory infections.

This background therefore provided the rationale of this study, the aim of which was to investigate droplet dispersion during O2, NIV and nebuliser treatment in patients with coryzal symptoms, patients with an infective exacerbation of chronic lung disease and a control group of normal subjects, to inform safe use. We reasoned that patients with a chronic exacerbation of lung disease or a coryzal infection would generate droplets regardless of the aetiology of the infection, therefore we did not specify that the infection had to be due to H1N1 or any other subtype of influenza A or influenza B. We sought to:

1. determine droplet size and concentration

2. determine geographical distribution of droplets

3. compare and contrast droplets generated during different interventions

4. examine whether modifications of treatment delivery affect droplet dissemination

5. estimate droplet decay after the intervention had ceased.

Although not classified as an AGP, we added an analysis of droplet counts and dispersion during a standardised session of chest physiotherapy in the chronic respiratory patients. This was because there was a high level of concern by physiotherapists that droplet dissemination would be considerable, thus putting these health-care workers at risk. In addition, in 40 patients admitted to the respiratory wards at the Royal Brompton Hospital NHS Foundation Trust with suspected swine flu in the first and second wave of H1N1 in 2009, all had underlying respiratory disease (predominantly cystic fibrosis and asthma) or neuromuscular disease, and required chest physiotherapy as part of their clinical management.

Chapter 2 Methods

Interventions

Normal controls and coryzal patients

On arrival in the side room, subjects were seated in a semirecumbent position on the bed. The Aerotrak counters were aligned to the subject as described, and baseline readings of droplet counts at the two sampling positions D1 and D2 were obtained over 40 minutes, sampling at 5-minute intervals. Subjects were then asked to do a series of spontaneous coughs both without and with a surgical mask. They then received O2 via a 60% Ventimask for 20 minutes, then NIV delivered through the non-vented hospital full-face mask (ResMed) using the modified filtered circuit for 20 minutes, then NIV via standard circuit with a vented mask for 20 minutes, and, finally, 4 ml of nebulised normal saline via the mouthpiece. Between interventions there were periods of 40 minutes to allow background droplet counts to fall to baseline levels (Figure 1).

FIGURE 1.

Example of coryzal subject experimental run. bc, baseline count; mNIV, modified NIV; neb, nebuliser; ox, oxygen therapy.

Patients

Measurements were carried out as above but with the following differences. All patients had arterial oxygen saturation (SaO₂), heart rate and transcutaneous PCO₂ monitored throughout the experimental interventions. After baseline measurements patients received 24% oxygen via Venturi mask for 20 minutes. Those who were using NIV received 20 minutes of ventilatory support at their current clinically indicated IPAP and EPAP settings, with oxygen entrained to maintain saturation to > 90%, first with the modified circuit and non-vented mask (2) and secondly with the vented face mask (1). The patients also underwent a standardised session of chest physiotherapy over 10 minutes. This consisted of cycles of deep breathing with percussion or shaking to loosen any secretions, followed by an assisted cough initiated manually, augmented by the physiotherapist performing inward and upwards pressure on the lower thorax to aid expectoration, after which the patient rested and cycles were repeated for 10 minutes. Throughout the study only two physiotherapists performed the physiotherapy in order to standardise the techniques as far as possible. Intervals of 40 minutes between interventions were added as in normal subjects and coryzal patients to re-establish baseline droplet levels (Figure 2).

FIGURE 2.

Example of patient experimental run. bc, baseline count; neb, nebuliser; ox, oxygen therapy.

Chapter 3 Results

In total, 44 subjects and patients were studied: 12 normal controls, 11 with coryzal symptoms and 21 patients. Subject and patient characteristics are given in Tables 1–3. The patients had a range of chronic lung conditions and all had been admitted because of an acute infective exacerbation. None of the patients or coryzal individuals had an H1N1 infection. All normal subjects and 10 of the coryzal subjects completed the 60% O2, NIV, mod NIV and nebuliser therapy. One coryzal patient completed all interventions except NIV modes, as these provoked claustrophobia. All patients received physiotherapy, but normal subjects or coryzal patients did not; all patients received 24% O2 via Ventimask and nebuliser therapy. Eight patients received NIV and mod NIV, as this was indicated to manage hypercapnic respiratory failure. A total of 19 coryzal subjects and patients therefore underwent the NIV and mod NIV interventions.

Physiotherapy

In the patients there was an increase in > 10-µm droplets at D1, but this has fallen at 1 m (D2) (p < 0.003) (Figure 3). There was no increase in the other droplet ranges.

FIGURE 3.

Physiotherapy results: droplet size > 10 µm at D1 and D2 in patients.

Modified NIV

Using the circuit modification the mean difference was not significantly different from baseline values on NIV in any group at D1 or D2, indicating that droplet count was significantly reduced compared with standard NIV with vented mask (Figure 4).

FIGURE 4.

Non-invasive ventilation circuit (NIV) (top row) and modified NIV results (bottom row) in patients at D1 in ranges 5–10 µm and > 10 µm.

Oxygen therapy

In normal controls, the coryzal group and in patients no significant increase in droplets in aerosol or large droplet range was seen either at D1 or D2 (Figure 5).

FIGURE 5.

Oxygen results in patients at D1 and D2 in droplet ranges 3–5, 5–10 and > 10 µm.

Nebuliser therapy

In all groups there was a significant increase across all droplet size ranges on therapy at D1 and D2 in normal subjects. In coryzal subjects and patients there were increases in the 0.3–0.5, 0.5–1, 1–3 and 3.5-µm aerosol ranges both at D1 and D2, but no significant mean difference in the larger droplet ranges of 5–10 and > 10 µm (Figure 6).

FIGURE 6.

Nebuliser results in patients at D1 and D2 in droplet ranges 0.3–0.5 to 1–3 µm.

Mean differences and p-values for all interventions in each group are given in Table 4.

TABLE 4 - Effect of intervention in droplet ranges D – difference between mean value pre and during intervention

D1, distance 1; D2, distance 2; p, p-value.

p-values < 0.05 are highlighted in bold text.

Microns	D1	0.3–0.5	0.5–1	1–3	3–5	5–10	> 10	D2	0.3–0.5	0.5–1	1–3	3–5	5–10	> 10
NIV
Normal	D	–0.097	–0.103	0.096	–0.073	–0.108	0.148		–0.062	–0.065	–0.057	–0.013	0.028	0.100
	p	0.0991	0.992	0.863	0.653	0.636	0.379		0.968	0.938	0.873	0.544	0.449	0.407
Patient	D	–0.003	0.004	0.021	0.143	–0.228	0.666		–0.002	–0.002	0.027	0.067	0.165	0.184
	p	0.565	0.412	0.400	0.231	0.806	0.042		0.470	0.298	0.298	0.308	0.243	0.339
Coryzal	D	–0.038	–0.068	–0.043	0.175	–0.060	0.807		–0.045	–0.055	0.036	0.233	0.422	0.574
	p	0.894	0.912	0.639	0.143	0.552	0.044		0.908	0.909	0.238	0.047	0.018	0.052
Mod NIV
Normal	D	–0.071	–0.073	–0.054	–0.078	–0.084	–0.086		–0.073	–0.081	–0.089	–0.114	–0.125	0.057
	p	0.971	0.963	0.703	0.703	0.690	0.588		0.975	0.970	0.963	0.836	0.734	0.431
Patient	D	–0.020	0.005	0.027	0.013	–0.039	0.087		–0.018	–0.017	–0.003	–0.014	–0.005	0.236
	p	0.829	0.437	0.393	0.463	0.558	0.402		0.776	0.687	0.521	0.545	0.513	0.244
Coryzal	D	–0.024	–0.026	–0.011	–0.147	–0.208	0.151		–0.016	–0.037	–0.092	–0.125	–0.153	0.011
	p	0.776	0.774	0.539	0.861	0.770	0.240		0.671	0.815	0.857	0.813	0.745	0.486
Oxygen
Normal	D	–0.063	–0.083	–0.041	–0.003	–0.234	–0.013		–0.050	–0.057	–0.062	–0.049	–0.077	0.081
	p	0.961	0.942	0.572	0.502	0.576	0.519		0.947	0.902	0.847	0.709	0.713	0.400
Patient	D	–0.003	–0.009	–0.023	–0.114	–0.068	–0.007		–0.011	–0.027	–0.051	–0.090	–0.102	–0.069
	p	0.578	0.636	0.577	0.748	0.565	0.507		0.728	0.789	0.736	0.764	0.685	0.554
Coryzal	D	–0.047	–0.067	–0.075	–0.108	–0.171	–0.012		–0.037	–0.045	–0.032	–0.001	0.004	0.058
	p	0.955	0.913	0.773	0.750	0.732	0.511		0.905	0.846	0.699	0.503	0.489	0.429
Physio
Patient	D	–0.005	0.057	0.123	0.128	–0.010	1.393		–0.011	0.024	0.070	0.169	0.175	0.424
	p	0.610	0.118	0.164	0.260	0.511	0.003		0.702	0.206	0.151	0.134	0.228	0.158
Nebuliser
Normal	D	15.660	109.480	71.681	27.054	404.932	2.270		25.878	87.932	46.887	1.549	0.232	0.207
	p	< 0.0001	< 0.0001	< 0.0001	< 0.0001	< 0.0001	< 0.0001		< 0.0001	< 0.0001	< 0.0001	< 0.0001	0.120	0.270
Patient	D	15.516	91.193	56.320	3.967	0.426	0.253		23.080	90.576	45.920	1.642	0.149	0.309
	p	< 0.0001	< 0.0001	< 0.0001	< 0.0001	0.111	0.261		< 0.0001	< 0.0001	< 0.0001	< 0.0001	0.241	0.281
Coryzal	D	11.204	64.822	38.341	1.871	0.197	0.349		17.994	49.458	30.454	1.144	0.234	0.384
	p	< 0.0001	< 0.0001	< 0.0001	< 0.0001	0.229	0.192		< 0.0001	< 0.0001	< 0.0001	< 0.0001	0.097	0.133

Chapter 4 Discussion

The results suggest that NIV using a vented mask in patients with an acute exacerbation of chronic lung disease disseminates large droplets locally. However, at a distance of 1 m the count has fallen significantly. There was no evidence of the generation of small droplets, i.e. an aerosol. Coryzal subjects also produced large droplets that spread for at least 1 m, which indicates that those with rhinorrhoea/upper airway inflammation also generate droplets. This group might be more representative of patients with an early progressive viral infection who are unlikely to produce large volumes of infected sputum when compared with those with cystic fibrosis or bronchiectasis. However, we did not see a difference in counts in those with markedly productive coughs compared with those with minimal sputum production (asthma, obesity hypoventilation) on the day of study. The large droplet count proximal to the mask was significantly reduced in both the patients and coryzal group in the NIV circuit with exhalation port filter, indicating that this modification minimises large droplet dissemination. These filters do not appear to increase the work of breathing if changed regularly. The finding that bilevel NIV with a vented mask disseminates large droplets is in keeping with the superspreading episodes seen in the SARS outbreak, where NIV use was found to be a risk factor on multiple logistic regression analysis. 12

Physiotherapy has not previously been included in the list of interventions in which PPE and FFP3 masks are indicated. Indeed the results show that it is not an AGP but, perhaps not surprisingly, given that the point of chest physiotherapy is to clear secretions, there was a significant increase in large droplets. As expected, these levels have dropped by D2 but the findings indicate that use of full PPE may be prudent for physiotherapists and respiratory therapists carrying out these procedures in patients with chronic respiratory disease in whom the H1N1 virus has generated an infective exacerbation or secondary bacterial pneumonia.

Oxygen therapy was not associated with an increase in droplets in any group, in any aerosol or droplet range. While it may be possible that 24% O2 might be an insufficient flow rate to shear droplets from the upper airway or disseminate those generated by spontaneous coughing or sneezing, we did not see an increase in droplet count in coryzal subjects who used 60% O2 – a flow rate more typical of that required by patients with acute lung injury due to viral pneumonia. These results should be contrasted with those from Yu et al. ,12 who showed O2 was a significant risk factor for superspreading events. However, their results were based on correlation rather than direct measurement of droplet densities and may be affected by the fact that sicker patients with higher viral loads are more likely to require O2 than those with milder disease who do not.

The association of spread of SARS with nebuliser use is controversial. Although there are case reports,15,16 in this study it is not possible to separate out droplets generated by the nebuliser itself from those generated by the patient. In addition, in clinical practice, patients being treated with nebulised bronchodilator are likely to have air flow obstruction due to asthma or COPD and are therefore more likely to be coughing and wheezing spontaneously. It is plausible that the flow from the nebuliser (either powered by a compressor or oxygen) would disseminate spontaneously generated droplets further. It is notable that the nebuliser was the only intervention that produced in droplets in the aerosol range (< 5 µm). This is entirely in line with the droplet range designed to be generated by this device, and means that this intervention also acts as a quality control confirming that the Aerotrak counters were fully able to detect particles in this range in clinical circumstances. However, in both the coryzal group and the patients we did not detect droplets in the 5- to 10-µm and > 10-µm ranges as occurred during NIV and physiotherapy. This indicates that the vast majority of droplets are likely to be nebulised saline as opposed to patient droplet secretions.

Chapter 5 Conclusions

Despite the limitations, this study indicates that NIV, O2 and physiotherapy are not AGPs. Physiotherapy and NIV generate large droplets adjacent to the patient, but these fall significantly at 1 m from the patient. A mod NIV circuit using a non-vented mask and filtered exhalate reduces the number of large droplets produced. Nebulised saline delivered by a mouthpiece produces an aerosol of droplets, but most are in the expected droplet range for the device and large droplets were not seen in patients and coryzal subjects. O2 at 60% and 24% did not appear to be an aerosol or droplet-generating procedure.

What are the implications for clinical practice and infection control? These results imply that during NIV and physiotherapy use of full PPE should be considered for health-care team members working within 1 m of the patient, as droplet count is increased. As the droplets are large and many drop out within 1 m over bedside surfaces, the crucial importance of handwashing and decontamination of near surfaces is evident, as transmission in these circumstances is likely to be direct via droplet spread or from fomites and direct contact with the patient’s local environment. As small and aerosolised particles were not demonstrated, the role other protective measures, such as negative pressure rooms, which have been advocated in some pandemic flu guidelines, may be less important.

Recommendations for further research:

1. Droplet sampling should be carried out in patients with pandemic influenza to confirm that droplets generated in this situation are comparable to those produced by patients in this study.

2. Droplet sampling sizing could be carried out in human simulator models with laser droplet imaging to corroborate results.

3. Viral carriage in different size droplets should be assessed to test whether using droplets as a proxy of infectivity risk is a realistic clinical substitute.

Acknowledgements

With thanks also to Winston Banya, Senior Statistician, Research Department, and Professor Paul Cullinan, Occupational Medicine Department, Royal Brompton Hospital, who provided statistical advice. We are grateful to the NIHR HTA panel for suggesting the addition of the coryzal group to study design.

The project was supported by the NIHR Respiratory Disease Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College London.

References

1. Brankston G, Gitterman L, Hiriji Z, Lemieux C, Gardam M. Transmissin of influenza A in human beings. Lancet Infect Dis 2007;7:257-65.
2. Tellier R. Review of aerosol transmission of influenza A virus. Emerg Infect Dis 2006;12:1657-62.
3. Fabian P, McDevitt JJ, DeHaan WH, Fung ROP, Cowling BJ, Chan KH, et al. Influenza virus in human exhaled breath: an observational study. PLoS ONE 2008;3.
4. The ANZIC. Influenza Investigators. Critical care services and 2009 H1N1 influenza in Australia and New Zealand. N Engl J Med 2009;361:1925-34.
5. Jain S, Kamimoto L, Bramley AM, Schmitz AM, Benoit SR, Louie J, et al. Hospitalised patients with 2009 H1N1 influenza in the United States, April–June 2009. N Engl J Med 2009;361:1935-44.
6. Department of Health . Pandemic Flu: A National Framework for Responding to an Influenza Pandemic 2009.
7. Department of Health . Pandemic Influenza: Managing Demand and Capacity in Healthcare Organisations (surge) 2009.
8. Department of Health and Health Protection Agency . Pandemic (H1N1) 2009 Influenza. A Summary of Guidance for Infection Control in Healthcare Settings 2009.
9. Fowler RA, Guest CB, Lapinsky SE, Sibbald WJ, Louie M, Tang P, et al. Transmission of severe acute respiratory syndrome during intubation and mechanical ventilation. Am J Respir Crit Care Med 2004;169:1198-202.
10. Scales DC, Green K, Chan AK, Poutanen SM, Foster D, Nowak K, et al. Illness in intensive care staff after brief exposure to severe acute respiratory syndrome. Emerg Infect Dis 2003;9:1205-10.
11. Lau JTF, Fung KS, Wong TW, Kim JH, Wong E, Chung S, et al. SARS transmission among hospital workers in Hong Kong. Emerg Infect Dis 2004;10:280-6.
12. Yu IT, Xie Z, Tsoi K. Why did outbreaks of severe acute respiratory syndrome occur in some hospital wards and not others?. Clin Infect Dis 2007;44:1017-25.
13. Sokol DK. Virulent epidemics and scope of healthcare workers’ duty of care. Emerg Infect Dis 2006;12:1238-41.
14. Simonds AK, Sokol DK. Lives on the line? Ethics and practicalities of duty of care in pandemics and disasters. Eur Respir J 2009:1-7.
15. Lee N, Hui DS, Wu A, Chan P, Cameron P, Joynt GM, et al. A major outbreak of severe acute respiratory syndrome in Hong Kong. N Engl J Med 2003;348:1986-94.
16. Wong RS, Hui DS. Index patient and SARS outbreak in Hong Kong. Emerg Infect Dis 2004;10:339-41.
17. Simonds AK, Simonds AK. Non-invasive respiratory support: a practical handbook. London: Hodder Arnold; 2007.
18. McCracken J. Should noninvasive ventilation be considered a high-risk procedure during an epidemic?. CMAJ 2009;181:663-4.
19. World Health Organization (WHO) . Clinical Management of Human Infection With Influenza A (H5N1) Virus 2007.
20. Gomersall CD, Tai DYHT, Derrick JL, Goh MS, Buckley TA, Chua C, et al. Expanding ICU facilities in an epidemic: recommendations based on experience from SARS epidemic in Hong Kong and Singapore. Intensive Care Med 2006;32:1004-13.
21. World Health Organization (WHO) Interim Guidelines . Infection Prevention and Control of Epidemic- and Pandemic-Prone Acute Respiratory Diseases in Health Care 2007.
22. Cheung TMT, Yam LYC, So LKY, Lau ACW, Poon E, Kong BMH, et al. Effectiveness of noninvasive positive pressure ventilation in the treatment of acute respiratory failure in severe acute respiratory syndrome. Chest 2004;126:845-50.
23. Yam LYC, Chan AYF, Cheung TMT, Tsui ELH, Chan JCK, Wong VCW. Non-invasive versus invasive mechanical ventilation for respiratory failure in severe acute respiratory syndrome. Chin Med J 2005;118:1413-21.
24. Department of Health . Pandemic Influenza: Surge Capacity and Prioritisation in Health Services 2008.
25. Confalonieri M, Potena A, Carbone G, Della Porta R, Tolley E, Meduri GU. Acute respiratory failure in patients with severe community-acquired pneumonia. A prospective randomised evaluation of noninvasive ventilation. Am J Respir Crit Care Med 1999;160:1585-91.
26. Seto WH, Tsang D, Yung RW, Ching TY, Ng TK, Ho M, et al. Effectiveness of precautions against droplets and contact in prevention of nosocomial transmission of severe acute respiratory syndrome (SARS). Lancet 2003;361:1519-20.
27. Somogyi R, Vesely AE, Azami T, Preiss D, Fisher J, Fowler RA. Dispersal of respiratory droplets with open vs closed oxygen delivery masks: implications for transmission of severe acute respiratory syndrome. Chest 2004;125.
28. Hui DS, Hall SD, Chan MTV, Chow BK, Tsou JY, Joynt GM, et al. Noninvasive positive-pressure ventilation: an experimental model to assess air and particle dispersion. Chest 2006;130:730-40.
29. Ip M, Tang JW, Hui DS, Wong ALN, Chan MTV, Joynt GM, et al. Airflow and droplet spreading around oxygen masks: a simulation model for infection control research. Am J Infect Control 2007;35:684-9.
30. Mardimae A, Slessarev M, Han J, Sasano H, Sasano N, Azami T, et al. Modified N95 mask delivers high inspired oxygen concentrations while effectively filtering aerosolized microparticles. Ann Emerg Med 2006;48:391-9.
31. Hui DS, Chow BK, Ng SS, Chu LCY, Hall SD, Gin T, et al. Exhaled air dispersion distances during non-invasive ventilation via different Respironics facemasks. Chest 2009;136:998-1005.
32. Hui DS, Hall SD, Chan MTV, Chow BK, Ng SS, Gin T, et al. Exhaled air dispersion during oxygen delivery via a simple oxygen mask. Chest 2007;132.
33. Tang JW, Liebner TJ, Craven BA, Settles GS. A Schlieren optical study of the human cough with and without wearing masks for aerosol infection control. J R Soc 2009:S727-36.
34. Rello J, Rodriguez A, Ibanez P, Socias L, Cebrian J, Marques A, et al. Intensive care of adult patients with severe respiratory failure caused by influenza A (H1N1) in Spain. Critical Care 2009;13.
35. Kumar A, Zarychanski R, Pinto R, Cook DJ, Marshall J, Lacroix J. Critically ill patients with 2009 influenza A(H1N1) infection in Canada. JAMA 2010;302:1872-9.
36. Hui DS, Chow BK, Chu LCY, Ng SS, Hall SD, Gin T, et al. Exhaled air and aerosolized droplet dispersion during application of a jet nebuliser. Chest 2009;135:648-54.
37. Perneger TV. What’s wrong with Bonferroni adjustments. BMJ 1998;316:1236-8.

Appendix 1 Protocol

List of abbreviations

AGP

aerosol-generating procedure

CI

confidence interval

COPD

chronic obstructive pulmonary disease

CPAP

continuous positive airway pressure

DH

Department of Health

EPAP

expiratory positive airway pressure

IPAP

inspiratory positive airway pressure

mod NIV

modified NIV circuit with exhalation filter

NIV

non-invasive ventilation

O2

oxygen therapy

PaCO2

partial pressure of arterial carbon dioxide

PaO2

partial pressure of arterial oxygen

PPE

protective personal equipment

RR

relative risk

SaO2

arterial oxygen saturation

SARS

severe acute respiratory syndrome

SD

standard deviation

TcCO2

transcutaneous partial pressure of carbon dioxide

WHO

World Health Organization