A quantitative evaluation of aerosol generation during cardiopulmonary resuscitation

Summary It is unclear if cardiopulmonary resuscitation is an aerosol‐generating procedure and whether this poses a risk of airborne disease transmission to healthcare workers and bystanders. Use of airborne transmission precautions during cardiopulmonary resuscitation may confer rescuer protection but risks patient harm due to delays in commencing treatment. To quantify the risk of respiratory aerosol generation during cardiopulmonary resuscitation in humans, we conducted an aerosol monitoring study during out‐of‐hospital cardiac arrests. Exhaled aerosol was recorded using an optical particle sizer spectrometer connected to the breathing system. Aerosol produced during resuscitation was compared with that produced by control participants under general anaesthesia ventilated with an equivalent respiratory pattern to cardiopulmonary resuscitation. A porcine cardiac arrest model was used to determine the independent contributions of ventilatory breaths, chest compressions and external cardiac defibrillation to aerosol generation. Time‐series analysis of participants with cardiac arrest (n = 18) demonstrated a repeating waveform of respiratory aerosol that mapped to specific components of resuscitation. Very high peak aerosol concentrations were generated during ventilation of participants with cardiac arrest with median (IQR [range]) 17,926 (5546–59,209 [1523–242,648]) particles.l‐1, which were 24‐fold greater than in control participants under general anaesthesia (744 (309–2106 [23–9099]) particles.l‐1, p < 0.001, n = 16). A substantial rise in aerosol also occurred with cardiac defibrillation and chest compressions. In a complimentary porcine model of cardiac arrest, aerosol recordings showed a strikingly similar profile to the human data. Time‐averaged aerosol concentrations during ventilation were approximately 270‐fold higher than before cardiac arrest (19,410 (2307–41,017 [104–136,025]) vs. 72 (41–136 [23–268]) particles.l‐1, p = 0.008). The porcine model also confirmed that both defibrillation and chest compressions generate high concentrations of aerosol independent of, but synergistic with, ventilation. In conclusion, multiple components of cardiopulmonary resuscitation generate high concentrations of respiratory aerosol. We recommend that airborne transmission precautions are warranted in the setting of high‐risk pathogens, until the airway is secured with an airway device and breathing system with a filter.


Introduction
Whether cardiopulmonary resuscitation (CPR) generates high concentrations of respiratory aerosol is uncertain and controversial [1][2][3].Most resuscitation guidelines have adopted a precautionary approach, prioritising rescuer safety and designating CPR as an aerosol-generating procedure [2,[4][5][6].These precautions, designed to reduce the risk of airborne disease transmission, may delay initiation of resuscitation with consequent impact on survival whilst rescuers don personal protective equipment [4].This approach may have contributed to worse outcomes following cardiac arrest during the COVID-19 pandemic [7,8].It remains unknown whether the precautionary approach confers appropriate protection to providers or causes unnecessary and inappropriate delays to commencement of CPR.
Case reports and retrospective epidemiological studies suggest an association between performing CPR and the risk of airborne disease transmission [9][10][11][12].However, to our knowledge, the amount of respiratory aerosol generated during CPR has never been measured in humans.This reflects the inherent challenges of using highly sensitive aerosol detection devices in a range of uncontrolled environments given the unpredictable nature of cardiac arrest.
The risk of making assumptions about aerosol generation from medical procedures is well illustrated by recent clinical quantitative aerosol studies.These studies show that several procedures previously considered to be high risk for aerosol generation, such as tracheal intubation and facemask ventilation, do not generate high concentrations of aerosol [13][14][15][16][17][18][19].This work has resulted in their removal from the list of aerosol-generating procedures in England [20,21].It is important to ensure appropriate precautions are being taken by rescuers performing CPR whilst minimising the risk of patient harm.As such, there is a pressing need to quantify respiratory aerosol generation during CPR in humans to help inform infection prevention and control guidelines.
We conducted a prospective study to quantify aerosol generation in patients with out-of-hospital cardiac arrest to test the hypothesis that CPR generates high concentrations of respiratory aerosol.As we were unable to measure respiratory aerosol concentrations in participants before cardiac arrest, or to separate individual components of CPR during cardiac arrest management, we conducted complementary investigations in human participants under elective general anaesthesia without cardiac arrest, and in a porcine cardiac arrest model, to measure the extent and nature of aerosol generation during each CPR component.

Methods
Ethics approval for the human studies was granted by the Greater Manchester Research Ethics Committee.Human participants with cardiac arrest were recruited prospectively via deferred nominated consultee consent following an initial consent waiver.The patients who participated in the general anaesthesia part of the study gave written consent before enrolment.The pigs were made available to the study team from two separate research projects (project licences PC545FB99 and PP4585512, UK Home Office).
Approval to study aerosol generation in these animals was granted by the University of Bristol Animal Welfare and Ethical Review Body in accordance with the 3Rs animal welfare principles (UIN-22-087) [22].Aerosol recording started after arrival of a designated pre-hospital critical care team with an additional researcher.
The researcher timestamped all events including: ventilatory breaths; chest compressions; defibrillation; rhythm checks; breathing system disconnections; and return of spontaneous circulation.Recordings were terminated if the i-gel was exchanged for a tracheal tube, or the airway was contaminated by gastric contents.
Aerosol samples were collected from a closed breathing system (Fig. 1).Patients' lungs were manually ventilated using a Mapleson C breathing system connected to the i-gel, with an oxygen flow of 5 l.min - Aerosol was sampled from within the breathing system via the tracheal tube using the same sampling technique as for the human studies (Fig. 1).This approach enabled respiratory aerosol identification by excluding the influence Particle size distribution analysis was undertaken using normalised aerosol concentrations (dN/dlogD p ) to account for the sampling bin widths and lognormal particle size distribution; dN is the total concentration of particles and dlogD p is the difference in the log of the bin width [27].
Aerosol concentrations recorded per cm 3

Results
Complete recordings were obtained during resuscitation of Time-averaged aerosol concentrations were greater for breath 2 compared with breath one (p < 0.001, Table 1),  and the breath two aerosol wave consistently outlasted the decay in the ETCO 2 trace sampled at the same time (Fig. 2a).
Specific periods were identified during resuscitation when chest compressions were relatively isolated from ventilatory breaths, most commonly after a rhythm check (identified in 11/18 participants).When chest compressions resumed following the rhythm check (and in the absence of immediately preceding ventilatory breaths), there was a clear increase in the concentration of aerosol emitted, which was maintained above the baseline (11/11 recordings) (Fig. 3a).Four participants with cardiac arrest had a single defibrillation shock (120-200 J) delivered during aerosol sampling.In all cases there was a transient increase in the aerosol concentration detected following the shock (Fig. 3b).

Time
Particle concentration analysis for the 10 s following the shock showed both time-average (32, It was also not possible to determine the effect of external cardiac defibrillation as an isolated event during resuscitation attempts because all shocks were preceded and immediately followed by chest compressions in accordance with resuscitation guidelines (Fig. 3b).
Comparator aerosol data generated using an equivalent concentrations during ventilation of patients under general anaesthesia (breaths 1 and breaths 2) were more than an order of magnitude lower than those recorded in participants with cardiac arrest (all p < 0.001, Table 1 and Fig. 4).
Aerosol concentrations for breath one and breath two were not different in patients under general anaesthesia (Table 1), and the decay of breath two was more rapid and showed an almost identical profile to the decay of the ETCO 2 waveform (Fig. 2b).
After Size distribution analysis of the particles generated during ventilation of the pigs showed an increase in all particle sizes following cardiac arrest; however, there was a marked increase in particles < 1lm (Fig. 5e and f).The profile of the particle distributions was similar between the pre-arrest and post-arrest recordings, providing further evidence they were respiratory in origin.

Discussion
We found that CPR in humans generates high concentrations of respiratory aerosol and these concentrations were consistently higher than those seen in previous studies of awake and anaesthetised humans (by up to 100-fold), even when coughing or undertaking forced expiratory activities [13][14][15][16].This finding is important because respiratory aerosols generated during CPR may pose a high risk of airborne infectious disease transmission.It should be noted that high aerosol concentrations do not necessarily mean high risk of airborne disease transmission and more work is required to determine the risk of airborne disease transmission from this aerosol and its environmental dispersion.However, as very high concentrations of respiratory aerosol are generated during CPR, we recommend airborne transmission precautions for providers in the setting of high-risk pathogens, until the airway is secured and sealed with an advanced airway device and breathing system with a filter.
We extended our observational findings in human patients to a porcine cardiac arrest model.This model facilitated paired respiratory aerosol measurements before and after cardiac arrest (recordings not possible for participants recruited following cardiac arrest).Here, we showed that ventilatory breaths delivered after cardiac arrest generate 270-fold higher aerosol concentrations than those before cardiac arrest, strongly suggesting it is the physiological changes following cardiac arrest that cause the high aerosol concentrations measured.As these recordings were undertaken within the same animal they account for inter-subject variability, increasing the confidence in the findings.The porcine model also enabled each component of CPR to be studied both in isolation and as part of the standard CPR algorithm directly analogous to that delivered to human participants; this is not possible during CPR for patients with cardiac arrest because it would be a deviation from standard care that could compromise outcomes.These porcine recordings show isolated chest compressions and external cardiac defibrillation generate aerosol but at lower concentrations than during the ventilatory components of CPR.There was, however, a synergistic interaction when ventilatory breaths, chest compressions and external defibrillation are combined (as also reported in a recent porcine CPR model [28]).
Delivering manual chest compressions requires significant physical activity; rescuers are close to the patient and likely to be taking large volume breaths increasing the risk of inhaling aerosol generated during CPR.It is likely that chest compressions, with or without external cardiac defibrillation, will release aerosol near the rescuer at the start of resuscitation before the airway is secured.Furthermore, the delivery of `rescue breathsẃ ithout a filter, whether by mouth-to-mouth or mask ventilation, are likely to be associated with a particularly high-risk of aerosol inhalation and potential for acquisition of respiratory pathogens.
Our previous studies of respiratory aerosol emissions were undertaken in ultraclean operating theatres, an environment with very low (practically zero) background aerosol concentrations, enabling resolution of aerosol generated by natural respiratory events such as breathing, speaking and coughing [13-16, 23, 24, 29].These studies show the presence of detectable respiratory aerosol at distances of 20 cm associated with tidal breathing and 50 cm with coughs in healthy subjects (who generate far lower aerosol concentrations than those detected during cardiac arrest) [13][14][15][16]24].The inference is that during CPR, bioaerosol will be detectable in much larger quantities and so poses an appreciably higher risk of aerosolised disease transmission; however, once an airway device with a filter is inserted, it is likely that the aerosol will be contained.
Nonetheless, caution when breaking the breathing circuit (e.g. when disconnecting the oxygen for defibrillation) is likely to be prudent, unless the filter is left attached to the airway device.
We we were able to eliminate the influence of background aerosol by sampling directly from the respiratory circuit [23].
The inline bacterial and viral filter removes particles > 0.3 lm (the lower limit of detection of the aerosol sampling device) from the inspired gas, giving confidence that all aerosol detected was participant-generated.
It should be noted that our study has only measured respiratory aerosol and not assayed the presence of transmissible pathogens such that we can only make inferences about transmission risk.However, patients with an acute viral illness generate more respiratory aerosol, and respiratory pathogens such as SARS-CoV-2 are concentrated in the smaller aerosol particles, generated predominantly in the distal respiratory tract by the bronchiolar fluid burst mechanism [35,36].All aerosol detected in this study was in the respirable size range (< 10 lm diameter) and likely to pose a substantial risk to first responders from those patients with cardiac arrest who have a transmissible respiratory pathogen.
Our sampling methodology meant that assessments could not be made during the early bystander or `first responder´phase of CPR.However, sampling air from the vicinity of the participant in the natural environment would present major challenges to ascertain whether detected aerosol originated from their respiratory tract.Advances in aerosol sampling technology (such as biofluorescence signatures) may enable future assays to distinguish respiratory aerosol from other particles, but this is not yet possible.It is likely that the degree of aerosol emission will be affected by airway patency when CPR is first started (with a similar effect on gas exchange) but this should not be considered a sufficient mitigation for risk, as it is likely to be unpredictable from case to case.It should be noted that CPR aerosol dispersion models and simulation studies require prior knowledge of the amount of aerosol emitted at the source; our study has, to our knowledge, provided these data for the first time.
Our sample sizes are relatively small, reflecting the challenges in collecting these unique data but are in line with typical sample sizes in other clinical aerosol studies.• Cardiopulmonary resuscitation generates very high concentrations of respiratory aerosol.• These high aerosol concentrations potentially increase the risk of airborne disease transmission.
• We recommend airborne transmission precautions during CPR in the setting of high-risk pathogens.
• Risk is reduced once the airway is secured and connected to a breathing system with a filter.
The primary outcome measure was to determine the respiratory aerosol concentrations generated by ventilation during CPR (compared with that generated in control participants under general anaesthesia).Secondary outcome measures included determination of aerosol generation by chest compressions and defibrillation.This is the first study of its kind and so no relevant data were available to inform a power calculation.The planned number of human subjects (n = 18) was estimated from our previous aerosol sampling studies, which have been adequately powered to detect aerosol generated from natural respiratory events or aerosol-generating procedures [14-16, 23, 24].A power calculation was conducted for the subsequent porcine study, based on a two-tailed, non-parametric paired comparison of aerosol generation before and following cardiac arrest.The human data comparing ventilation of participants under general anaesthesia vs. ventilation of participants with cardiac arrest resulted in an estimated effect size of 1.67.Therefore, eight pigs were required to achieve a power of 95% (a 0.05) [25].For patients in cardiopulmonary arrest, aerosol recordings were non-invasive and inclusion in the study did not alter standard clinical care.The first-to-scene paramedics provided standardised advanced life support to adults aged ≥ 18 y in cardiac arrest, comprising insertion of a supraglottic airway device (i-gelâ, Intersurgical, Wokingham, UK) and CPR according to Resuscitation Council UK guidelines (a ratio of 30:2 compressions to breaths) [5, 26].Patients were supine and received either manual or mechanical chest compressions (Lund University Cardiac Assist System (LUCAS), Stryker, Sweden).

Figure 1
Figure1Breathing system with aerosol sampling used for all studies.HME, heat and moisture exchange filter.
18 adults in out-of-hospital cardiac arrest (11 male, median (IQR [range]) age 59 (56-75 [40-85]) y).All patients had bystander CPR before arrival of the paramedics.The initial cardiac rhythm was asystole in 13 patients, pulseless electrical activity in three patients, and ventricular fibrillation in two patients.Four patients received external cardiac defibrillation during aerosol sampling.Median time between the initial call to emergency services and start of aerosol sampling was 32 (24-43 [10-55]) min.One patient was successfully resuscitated and transferred to hospital, the other 17 were declared dead at the scene.There was no deviation from standard resuscitation care and no adverse events were associated with aerosol sampling for any patient.Median (IQR [range]) duration of aerosol sampling was 7.3 (5.3-10.7 [2.0-16.1])min.This showed a consistent, repeating pattern of aerosol generation that matched the timestamped sequence of breaths and chest compressions periods (Fig. 2a).This was corroborated when the ETCO 2 waveforms were superimposed, showing a very similar pattern confirming the association of aerosol concentration peaks with the expiratory phase of ventilation.During 30:2 compression to ventilation CPR, the aerosol profile of the first breath of each pair (breath one: comprising a single breath) showed a much sharper peak compared with the prolonged, slowly decaying plateau that lasted > 10 s seen with the second breath (breath two: composed of a breath plus chest compressions).

Figure 2
Figure2Comparison of timelines demonstrating aerosol concentrations and their association with exhaled carbon dioxide recorded during sampling.Blue, aerosol concentration; red, ETCO 2 ; red arrows, ventilatory breaths.Note same y-axis for comparison.(a) Representative timeline for a participant with cardiac arrest during CPR (30:2 compressions to breaths).This shows the repeating pattern of aerosol peaks associated with ventilatory breaths (seen also on corresponding ETCO 2 trace).Note the prolonged aerosol plateau associated with the second breath and chest compressions; (b) aerosol concentrations generated by an equivalent pattern of ventilation in a participant under general anaesthesia and not in cardiac arrest, receiving two ventilatory breaths and no chest compressions.CC, chest compressions; RC, rhythm check.

Figure 3
Figure 3 Timelines demonstrating aerosol concentrations and exhaled carbon dioxide values associated with chest compressions and defibrillation in human participants.Blue, aerosol concentration; red, ETCO 2 ; red arrows, breaths.(a) Aerosol concentrations during a period of chest compressions following a 5 s rhythm check dissociating it from ventilatory breaths; (b) aerosol generation associated with external cardiac defibrillation during CPR (black arrow, external cardiac defibrillation shock).(Data shown in (a) and (b) are from two different participants in cardiac arrest).

Figure 5
Figure 5 Representative respiratory aerosol concentration and ETCO 2 during components of resuscitation in the porcine cardiac arrest model.Blue, aerosol concentration; red, ETCO 2 ; red arrows, breaths; CC, chest compressions; shock, external cardiac defibrillation shock.(a) Aerosol recorded during ventilation of an anaesthetised pig before termination and starting CPR; (b) aerosol generation during: compression only CPR; compressions followed by external defibrillation; and isolated defibrillation shocks; (c) aerosol generated by sequences of two ventilatory breaths followed by a 20-s pause (x 3) without any chest compressions.This is followed by 30 x chest compressions and a 20-s pause; followed by an isolated defibrillation shock; (d) aerosol generated during cycles of two ventilatory breaths followed by 30 x chest compressions.Note the similar aerosol and ETCO 2 profile to the human data in Fig. 2a; (e) comparison of the particle size distribution, during 60 s of ventilation of the anaesthetised animal vs. 60 s of ventilation following cardiac arrest.Aerosol concentrations normalised to dN/dlogD p to account for the lognormal size distribution, n = 8.Blue line, pre-arrest median; black line, post-arrest median; (f) data in (e) normalised to the maximum value per animal.Blue line, normalised pre-arrest median; black line, normalised post-arrest median; shaded area, IQR.Note changes of y-axis aerosol scale between (a-d) to best represent the data graphically.

Furthermore, the large 1
differences detected in aerosol generation with specific components of CPR increases confidence of our findings.The observational human cardiac arrest data were supported by the porcine model to enable separation of the various CPR components.The animals included in the study were of a similar weight to the participants under general anaesthesia; respiratory aerosol concentrations generated by ventilation of the pigs before cardiac arrest were similar to those seen during ventilation of human participants under general anaesthesia; respiratory aerosol concentrations generated during CPR manoeuvres in the porcine cardiac arrest model were similar to those measured during resuscitation of participants with cardiac arrest; and the profile of aerosol Box Findings and recommendations.

Table 1
Aerosol concentrations (particles.l - ) for participants with cardiac arrest, participants under general anaesthesia, and pigs in a cardiac arrest model.Values are median (IQR [range]).
Breath 1 is a single ventilatory breath and breath 2 also includes a period of chest compressions for participants with cardiac arrest and the pigs in a model of cardiac arrest.*Aerosol concentrations recorded following cardiac arrest.