key: cord-0674815-9cr4sxik authors: Barreda, Santiago; Asadi, Sima; Cappa, Christopher D.; Wexler, Anthony S.; Bouvier, Nicole M.; Ristenpart, William D. title: The Impact of Vocalization Loudness on COVID-19 Transmission in Indoor Spaces date: 2020-09-09 journal: nan DOI: nan sha: afb94359833788e4a8d9c11d1576d730e7b1c7e1 doc_id: 674815 cord_uid: 9cr4sxik There have been several documented outbreaks of COVID-19 associated with vocalization, either by speech or by singing, in indoor confined spaces. Here, we model the risk of in-room airborne disease transmission via expiratory particle emission versus the average loudness of vocalization and for variable room ventilation rates. The model indicates that a 6-decibel reduction in average vocalization intensity yields a reduction in aerosol transmission probability equivalent to doubling the room ventilation rate. The results suggest that public health authorities should consider implementing"quiet zones"in high-risk indoor environments, such as hospital waiting rooms or dining facilities, to mitigate transmission of COVID-19 and other airborne respiratory diseases. There is an emerging consensus that COVID-19 is transmissible via airborne aerosol particles that are emitted when infected individuals breathe, speak, sneeze, or cough [1 -8] . The relative contributions of these expiratory activities to airborne transmission remains unclear, but multiple outbreaks have been documented in which asymptomatic carriers were speaking or singing in confined indoor spaces with susceptible individuals [9, 10] . Vocalization causes micron-scale droplets of respiratory mucosa to form via a "fluid-film-burst" mechanism, either in the lungs during inhalation due to expansion of the alveoli, or in the vocal cords due to rapid opening and closing of the glottis during phonation [11] [12] [13] . Upon exhalation into the ambient air these droplets rapidly evaporate to leave behind dried aerosol particles large enough to carry viable virus that, although too small to see by eye, are lightweight enough to remain suspended for long times; particles smaller than approximately 5 µm will typically be removed from rooms by air exchange rather than gravitational settling [14] [15] [16] . Expiratory particles in this size range from exhaled breath are known to carry infectious influenza virus [17] ; likewise, viable SARS-CoV-2, the virus responsible for COVID-19, has been observed in micron-scale aerosol particles sampled from hospital air several meters away from infected patients [18] . We recently demonstrated that the emission rate of micron-scale respiratory aerosol particles strongly correlates with the loudness of speech [19, 20] . An increase in vocalization intensity of about 35 decibels, roughly the difference between whispering and shouting, yields a factor of 50 increase in the particle emission rate. We also reported that the size distribution of the dried particles is independent of vocalization loudness, and that certain individuals, for unclear reasons, act as superemitters during vocalization, releasing an order of magnitude more particles than average. We hypothesized that airborne disease transmission might occur more readily in noisy environments where infected individuals must speak loudly, thus causing enhanced emission of infectious expiratory particles into the air [19] . Epidemiologists have speculated that recent COVID-19 outbreaks in churches [9] , bars [21] , or meat processing facilities [22, 23] might be due in part to the loudness of these environments. In response, various public health authorities have provided official recommendations that discourage [24] [25] [26] [27] or even explicitly prohibit [28] singing and other loud vocalizations, or prohibit conditions like playing loud music that necessitate raising of voices [29] . Much remains unknown, however, about the possible link between vocalization loudness and airborne disease transmission. If virus-laden particles are emitted via vocalization, and if louder vocalization yields more particles, then a key question is: how does the loudness of vocalization affect the transmission probability? As a starting point to addressing this question, we use the simplest quantitative theoretical model for airborne disease transmission, named the Wells-Riley model after the early investigators who performed this pioneering work [30, 31] . Detailed derivations and assessments of the accuracy of the Wells-Riley model are provided elsewhere [32, 33] ; here we simply use the model framework, which is that the transmission probability follows the complement of a Poisson distribution, where is the expected number of infectious pathogens that a susceptible individual inhales. This probability distribution assumes that only one pathogen is necessary to initiate infection, but more complicated expressions are available to account for larger minimum infectious doses [14] . In the classic Wells-Riley formulation, is calculated with the assumption that pathogens are emitted at a rate q pathogens per second from one or more infected individuals in a room with instantaneously well-mixed air, so that the relative positions of the infected and susceptible individuals are irrelevant. As such, the model does not account for potential enhanced transmission by direct inhalation of the respiratory plume emitted by an infected individual, but the assumption of wellmixed air serves in part as the basis for minimum ventilation standards promulgated by CDC [34] and ASHRAE [35] because it yields a lower bound for transmission risk to all room occupants regardless of position. The Wells-Riley model further assumes that the room has a ventilation rate of Q liters per minute delivering fresh (pathogen-free) air, and that susceptible individuals are moving B liters of air in and out of their lungs per minute of breathing (i.e., the minute ventilation). In the case where there is just one infected individual, the expected value is where t is the total exposure time. The parameter here represents an infection efficiency (0 < < 1) that includes physical effects, like the deposition efficiency within the respiratory tract of the susceptible individual, and immunological effects, like the ability of the immune system to repress the infection. For a minimum infectious dose of 1 pathogen, the quantity is equivalent to the "quanta" of infectivity initially used by Wells and Riley in their models [30, 31] . It is already well known from equations (1) and (2) that increasing the exposure time or decreasing the room ventilation rate will increase the expected number of inhaled pathogens and the corresponding transmission probability [36] . What is new here is consideration of the impact of vocalization intensity on the virus aerosolization rate q. The particle emission rates that we previously reported were measured in a laboratory environment while using a microphone and decibel meter placed near the mouth [19, 20] . Importantly, the particle emission rate varied linearly with the root-mean-square amplitude as measured by the microphone; the amplitude varies nonlinearly with the corresponding sound pressure level in decibels (Fig. S1 ). Using these measurements, we can relate expected particle emission rates to different sound pressure levels, measured in C-weighted decibels (dBC). Full details are presented in the Supplementary material; the final result is that the average particle emission rate is estimated to depend on the vocalization intensity 1 , measured in dBC at 1 m from a non-masked speaker, as where � and � are scaled expiratory particle emission rates for breathing and vocalization, respectively, that depend on the expiratory flowrates. The parameter represents the fraction of time the infected individual is vocalizing during the exposure time; is close to zero for individuals who vocalize rarely such that breathing-related emission dominates, and approaches one for those who vocalize continuously, such as in singing or chanting. The average virus aerosolization rate then is where is the viral concentration in the respiratory fluid of the infected individual, and is the pre-evaporation volume of droplets emitted during vocalization. The parameter = / ≈ 0.5 is the volume ratio of droplets emitted via breathing versus vocalization; several researchers have found that vocalization yields significantly larger droplets than breathing [11, 13, 19] . Combination of equations (2) -(4) into (1), and noting that the ventilation rate in a room with volume is related to the air changes per hour as = , yields the desired probability, Here = is an effective rate constant composed of parameters that, for a given room and specific virus, are not readily alterable by human interventions. The striking feature of equation (5) is the large power-law dependence on the vocalization intensity. A contour plot of the transmission probability versus vocalization intensity and duration illustrates this pronounced impact for a 1-hour exposure time in a room with three ACH ( Figure 1 ). The transmission probability is lowest in the bottom left corner, corresponding to infectors who vocalize rarely and quietly, as might be observed in a library or quiet office space. In contrast, the transmission probability increases gradually with duration and rapidly with intensity. It reaches maximal values in the top right corner, corresponding to infectors who vocalize loudly and close to continuously, as might be observed in a noisy bar environment or at a choir practice. The model also gives insight on the cost-benefit analysis of increasing the room ventilation rate. Fig. 2A shows the transmission probability versus vocalization intensity for different ACH values. As expected, doubling the ventilation rate of fresh (pathogen-free) air decreases the transmission probability. A notable feature, however, is that a similar reduction in transmission probability can be gained, without changing the ventilation rate, simply by decreasing the vocalization intensity by approximately 6 dBC. This reduction can be quantified via a risk reduction factor, where is the probability at some initial condition and is the adjusted probability via an intervention either with an increased ventilation rate or decreased vocalization. For simplicity, we can focus on small values of such that asymptotically ≈ , in which case the risk reduction factor for doubling the room ventilation rate is = 1 2 . If the infected individual simply vocalizes half as often (i.e., is halved), then to good approximation ≈ 1 2 as well. Furthermore, keeping the room ventilation rate and the vocalization duration fixed, the risk reduction factor for decreasing the vocalization intensity by decibels is To achieve a 50% reduction in risk for vocalization that ordinarily would occur at 60 dBC would require a decrease of only = 5.4 dBC. More precise calculations of the risk reduction factor (Fig. 2B) show that in general, a 10 dBC decrease in average vocalization intensity is always more effective at reducing risk of aerosol transmission than doubling the ventilation rate. The risk reduction achieved either by increasing room ventilation or by decreasing the loudness of vocalization is insensitive to the pathogen concentration in respiratory emissions or their infection efficiency, though those quantities do affect the actual probability of transmission. In other words, the numerical values of the probabilities shown in Figs 1 and 2A will vary with the viral load of the infector, but the overall shape of the curves will remain the same. Similarly, wearing of masks will reduce the particle emission rate of the infector and decrease the effective deposition efficiency in susceptible individuals and thus decrease the overall probability, but the relative risk reduction as characterized here will remain unchanged. We also emphasize that the Wells-Riley model explicitly assumes the air is well mixed, and that more sophisticated plume or puff models [37, 38] or computational fluid dynamics models [39, 40] are required to account for the directionality and turbulent diffusivity of the airflow and proximity of individuals. Whatever transport model is used, however, the vocalization source terms in equations (3) and (4) suggest that reductions in vocalization intensity will strongly decrease the amount of virus available to be transported, and thus decrease the overall transmission probability. To relate these proposed decibel decreases to real-world situations, we consider typical noise levels in different indoor environments, often measured in A-weighted decibels (dBA), which are thought to better reflect subjective perceptions of loudness. Ambient noise in restaurants is typically between 65-80 dBA, with an average of 73 dBA [41] , and background noise levels of 75 dBA have been observed at day-care facilities [42] . Music plus crowd noise in bars and nightclubs can average as high as 90-100 dBA [43] . The relationship between ambient noise levels and the speech loudness necessary for comprehension is complex, but in general speech must be nearly the level of the background noise to be understood, and speakers adjust their vocalization intensity to maintain a positive signal-to noise ratio when possible [44] [45] [46] . As a result, all other things being equal, a reduction in background noise on the order of 5-10 decibels will facilitate, if not directly result in, a corresponding reduction in average speaking levels. Further, the relatively high amount of background noise in many public spaces suggests that there is considerable room to reduce noise levels behaviorally (e.g., turning music down, encouraging silence), since noise is not inherent to the operation of many of these spaces (as opposed to industrial facilities). When wearing facemasks, the reduction in the background noise necessary to achieve a similar magnitude reduction in transmission risk may be larger owing to the need to speak more loudly through the mask [47] . A more detailed analysis of mask filtration efficacy and vocalization through masks is necessary to characterize the impact of this effect on transmission probability. There are tremendous installation, maintenance, and energy costs associated with increased ventilation rates, especially in air conditioned or heated indoor spaces [48] . In practice many ventilation systems recycle a substantial fraction of the room air, so increasing the flow rate of Table S1 . draws in 5 liters/min of air from the funnel, of which 80% comprised a sheath flow and 20% a sample flow ( ) measured in the detector. A microphone and a decibel meter placed near the funnel entrance simultaneously measured the root-mean-square amplitude, , of the vocalization and the corresponding sound pressure level (SPL) in C-weighted decibels. The key finding, shown in Fig. S1b , is that the rate of particles moving through the detector in the sample flow, in particles per second (p/s), varied linearly with the vocalization amplitude, . (S1) The where is the total particle emission rate from vocalization (p/s). A similar statement pertains to the (non-vocalization) particle emission rate during breathing, . Over sufficiently long time periods, the average total particle emission rate will reflect the relative amounts of time spent breathing versus vocalizing, viz., where 0 ≤ < 1 is the fraction of time the individual spends vocalizing. Inserting the relationships defined in (1) and (2) Next, we note that the microphone amplitude is related to the sound pressure level in decibels via a power-law relationship of the form as shown in Fig. S1C . Nonlinear regression yields best fit values of = 0.094 and = 105 dBC. The decibel readings were recorded 6.5 cm from the mouth, but it is standard to report sound pressure levels at a distance of 1 m from the noise source. Accordingly, we adjust the sound pressure level as where ∆= 25 dBC for 1 = 1 m. Combination of (1), (2), (5) and (6) For convenience we define � = , and � = , and substitution of the empirical coefficients , , and ∆ yields equation (3) in the main text. The independent variables of interest in equation S7 for modeling the transmission probability are and 1 . All other parameters are known from the empirical measurements reported by Asadi et al., except for the expiratory flowrates and . As noted by several authors, the relationship between measured sound pressure level and the expiratory flow rate is quite complicated, and depends on the pitch (fundamental frequency), the "open quotient" of the vocal cords, the lung pressure and vocalization pressure threshold, and the glottal and epiglottal resistances (Schneider and Baken 1984 , Titze 1992 , Jiang et al. 2016 ). As first summarized succinctly by Rubin et al., there is a "lack of any consistent relationship between sound pressure levels and air flow" (Rubin et al. 1967) . Accordingly, as a first approximation here we simply treat the average flow rate during vocalization as a fixed constant independent of the sound pressure level, which in general will yield a conservative underestimate of the total particle emission rate as sound pressure level increases. Model parameters and sources are listed in Table S1 . Rapid Expert Consultation on the Possibility of Bioaerosol Spread of SARS-CoV-2 for the COVID-19 Pandemic The coronavirus pandemic and aerosols: Does COVID-19 transmit via expiratory particles? It is Time to Address Airborne Transmission of COVID-19 Airborne Transmission of COVID-19 Re-thinking the Potential for Airborne Transmission of SARS-CoV-2 Reducing Transmission of SARS-CoV-2 Consideration of the Aerosol Transmission for COVID-19 and Public Health Relative contributions of transmission routes for COVID-19 among healthcare personnel providing patient care High SARS-CoV-2 Attack Rate Following Exposure at a Choir Practice When will it be safe to sing together again? Size distribution and sites of origin of droplets expelled from the human respiratory tract during expiratory activities The Mechanism of Breath Aerosol Formation Modality of human expired aerosol size distributions The airborne lifetime of small speech droplets and their potential importance in SARS-CoV-2 transmission Infectious virus in exhaled breath of symptomatic seasonal influenza cases from a college community Viable SARS-CoV-2 in the air of a hospital room with COVID-19 patients Aerosol emission and superemission during human speech increase with voice loudness Effect of voicing and articulation manner on aerosol particle emission during human speech A Virus Walks into a Bar…, The New York Times COVID-19 Among Workers in Meat and Poultry Processing Facilities -19 States What explains the high rate of SARS-CoV-2 transmission in meat and poultry facilities? Centre for Evidence Based Medicine Corona Virus Disease 2019 -Considerations for Events and Gatherings COVID-19 Planning Considerations: Guidance for School Re-entry Keeping workers and clients safe during COVID-19 in close contact services Singing as a risk for transmission of SARS-CoV-2 virus COVID-19 Industry Guidance: Places of Worship and Providers of Religious Services and Cultural Ceremonies Amendment No. 2 to Miami-Dade County Emergency Order No. 26-20 On air-borne infection -Study II Droplets and droplet nuclei Airborne spread of measles in a suburban elementary school How far droplets can move in indoor environments -revisiting the Wells evaporation-falling curve Review and comparison between the Wells-Riley and doseresponse approaches to risk assessment of infectious respiratory diseases Guidelines for Environmental Infection Control in Health-Care Facilities American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE), Ventilation for Acceptable Indoor Air Quality Role of ventilation in airborne transmission of infectious agents in the built environment -a multidisciplinary systematic review A Comprehensive Breath Plume Model for Disease Transmission via Expiratory Aerosols Airborne spread of infectious agents in the indoor environment A study of the dispersion of expiratory aerosols in unidirectional downward and ceiling-return type airflows using a multiphase approach Dispersion of exhalation pollutants in a twobed hospital ward with a downward ventilation system Restaurant noise: Levels and temporal characteristics Loud speech in realistic environmental noise: phonetogram data, perceptual voice quality, subjective ratings, and gender differences in healthy speakers Pilot study of patron sound level exposure in loud restaurants, bars, and clubs in New York city Average speech levels and spectra in various speaking/listening conditions Speech recognition in multitalker babble using digits, words, and sentences A comparison of two word-recognition tasks in multitalker babble: Speech Recognition in Noise Test (SPRINT) and Words-in-Noise Test (WIN) Speech intelligibility assessment of protective facemasks and air-purifying respirators HVAC filtration for controlling infectious airborne disease transmission in indoor environments: Predicting risk reductions and operational costs Aerosol emission and superemission during human speech increase with voice loudness Effect of voicing and articulation manner on aerosol particle emission during human speech Viral kinetics and exhaled droplet size affect indoor transmission dynamics of influenza infection Characterizing exhaled airflow from breathing and talking The measurement of airflow using Singing helmet that allows free movement of the jaw Volumes and breathing patterns during speech in healthy and asthmatic subjects Deposition efficiency of inhaled particles (15-5000 nm) related to breathing pattern and lung function: an experimental study in healthy children and adults Vocal intensity, subglottic pressure and air flow relationships in singers Influence of lung volume on the airflow-intensity relationship Vocal Intensity in Speakers and Singers Temporal profiles of viral load in posterior oropharyngeal saliva samples and serum antibody responses during infection by SARS-CoV-2: an observational cohort study We thank the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (NIAID/NIH), grant R01 AI110703, for supporting this research.