Aircraft cabins are served by high‐performance ventilation systems. They also typically hold many persons in close proximity for long durations. Research suggests that seating proximity on aircraft is somewhat associated with increased risk for infection with SARS‐CoV‐2, the virus that causes COVID‐19, but the importance of seating in the context of other factors is not clear. As reported by Murphy et al., ¹ 13 infections among those onboard were later confirmed by nasopharyngeal swab PCR. This 7.5‐h flight had only 17% occupancy. No index passenger was identified, and nine of the 13 infected persons reported wearing masks. A source proximity effect was identified by Khanh et al. ² in finding that most of the 15 confirmed cases and the known index passenger were seated in the business class section during a 10‐h flight. An association of COVID‐19 infection, seating proximity, and mask behavior was reported by Toyokawa et al. ³ for a 2‐h flight within Japan. The authors calculated adjusted odds ratios of 7.29 for no mask, 3.0 for partial wearing of a mask, and 7.46 for being seated within two rows of the index passenger.

A well‐known study by Olsen et al. ⁴ during the Severe Airborne Respiratory Syndrome (SARS) epidemic reported seating locations of probable secondary infections and the index passenger, in one of three flights that were tracked due to one or more passengers having tested positive for SARS. In 2013, after the H1N1 epidemic and during the Middle East Respiratory Syndrome (MERS) outbreaks and ongoing efforts to prepare for a possible influenza pandemic, a National Institute for Occupational Safety and Health (NIOSH) study fit closely the number of infections reported in Olsen to a linear regression, with distance as cabin rows from the index passenger the independent variable. This finding along with regression models of tracer studies conducted in cabin mock‐ups at the University of Illinois and Kansas State University demonstrated that air contaminants, including droplets, travel several seat rows beyond the source row. ⁵ Following on in 2018, a conference paper ⁶ showed that bacteriophage (MS2) droplets emitted from a spray bottle into Boeing 767 and 737 mock‐ups dispersed like other particle tracers. The fluid aliquots collected in a bioaerosol liquid impinger were evaluated for viral particle presence by plaque assay. The number of plaque‐forming units (PFUs), considered proportional to the airborne concentration of viable virus, were fit to an exponential decrease curve using radial distance between source and measurement location.

Aircraft cabin environmental control systems are designed and operated to deliver amounts of clean air per occupant that conform to various standards, including requirements set by the Federal Aviation Administration and Recommendations of the American Society of Heating, Refrigeration, and Air‐Conditioning Engineers. ⁷ , ⁸ Under these conditions, most virus particles are removed within several seat rows from a source on an aircraft, and the recirculated portion of the air supplied to each passenger has passed through high efficiency particle air (HEPA) filters. ⁹ Physical distancing is difficult on crowded flights; and, sitting within 2 m (6 ft) of others, sometimes for hours, might increase risk for SARS‐CoV‐2 exposure. To reduce this risk, CDC issued an order in January 2021 that required the wearing of masks in travel settings (on conveyances and at US transportation hubs) to prevent spread of COVID‐19, including all passengers and crew on aircraft traveling into, within, or out of the United States, ¹⁰ , ¹¹ and recommends all people be up to date with COVID‐19 vaccines before travel. ¹² On April 18, 2022, CDC stopped enforcing the mask order but continued to recommend masking in public transportation settings. ¹² Prevalence of COVID‐19 among the flying public can be estimated from international travelers arriving by air in Canada. From February 21, 2021 to July 4, 2021, 0.90% of all arriving passengers tested positive; from July 5 to September 16, 0.90% of partially or unvaccinated passengers tested positive; and, from September 10 to November 6, 0.64% tested positive. For fully vaccinated passengers, the rate dropped to 0.21% for July 5 to September 16 and to 0.15% from September 10 to November 6 but then increased to 5.0% for November 28, 2021 to January 22, 2022. ¹³

As the SARS‐CoV‐2 pandemic was taking hold in early 2020, many air carriers created socially‐distance seating environments, notably Delta Airlines, which blocked the selection of middle seats and capped seating in every cabin class. ¹⁴ This carrier maintained the practice through April 30, 2021, longer than any other U.S.‐based airline. ¹⁵ Earlier, NIOSH had re‐analyzed the MS2 bacteriophage tracer experiments to estimate the effect of leaving middle seats vacant compared to full occupancy and issued a short report in the Morbidity and Mortality Weekly Report (MMWR), finding that relative exposure to the bacteriophage droplets in vacant middle seat (VMS) scenarios was reduced by 23% to 57%. ¹⁶ A knowledge gap—not accounting for mask wearing—was identified in press coverage of the report. ¹⁷ While the MMWR report did not evaluate the effect of mask wearing (masking) or the interaction of VMS and masking, it reported VMS only as a relative exposure reduction that would act independently from masking if lack of interaction, for example, changing the droplet size distribution, could be reasonably assumed. This assumption is explored in the current study. A New York Times piece stated that the infection risk was low due to ventilation and filtration, ¹⁷ and we reference the popular press article here because it captures a view held widely by researchers and the air travel industry. A contrasting view is that higher density ¹⁸ and longer duration ¹⁹ of occupancy compared to many indoor environments complicate this characterization.

The MMWR report was not alone in predicting lower exposure without adjacent seating. Horstman and Rahai ²⁰ assayed this scenario using CFD in a single‐aisle, three‐on‐each side seating arrangement and report a 40% infection risk decrease due to increased supply air per occupant and specific flow patterns. Pavlik et al. ²¹ compare various seating distance and masking implementations while illustrating the dependence of pattern optimality on occupancy level and spatial risk model. VMS may not be a particularly efficient rule. Barnett and Fleming ²² calculated, from extensive case searches and probabilistic modeling, COVID‐19 infection risk estimates of 1/1000 and 1/6000 for a 2‐h flight, when full during highest and when half‐full during lowest pandemic infection periods, respectively.

In October of 2020, the TRANSCOM Report that detailed aerosol tracer experiments onboard functioning (including flying) Boeing 767 and 777 aircraft was released publicly, ²³ and a corresponding peer‐reviewed journal article was published in December of 2021. ²⁴ This large work included hundreds of trials with and without the source manikin wearing a three‐ply surgical mask. The data were used by Wang et al. ²⁵ as a seat‐specific exposure map to avoid the much criticized completely mixed assumption in their application of the Wells–Riley equation. Authors from Boeing released a pre‐print and published a journal article documenting CFD simulations of a Boeing 737 cabin that tracked a realistic droplet distribution from several source seat locations under varying ventilation and thermal conditions. ²⁶

In all aspects other than release and measurement of a live virus, the TRANSCOM paper by Kinahan ²⁴ provides a much greater wealth of cabin data than the Lynch paper. ⁶ The Zee paper ²⁶ is unique in reporting CFD results at all seat locations in the computational domain. The strength of these data sources and the manner the VMS MMWR report was received motivated substantially extending the VMS work to include the Kinahan and Zee datasets. The current article describes this effort, including characterization of masking, and examines relative exposure reduction in the context of a range of probable infections, using a simplified form of the Wells–Riley equation.

The reduction of 38% when a surgical mask was worn by the particle source manikin in the TRANSCOM study is smaller than the 66% reported for surgical mask material in Hao which, in turn, is smaller than the 72% reported in Lindsley et al. ³⁷ for surgical masks worn by both source and receiving manikins in a conference room setting. Clearly, mask leakage reduces effectiveness, and universal masking is more effective than source‐only application. Liu et al., ³⁸ in a paper on mixing and displacement ventilation and masking found that the aircraft cabin average infection risk (as defined in their paper) was 0.20 for no masking, 0.07 for infectious passenger masking, and 0.02 for universal masking, that is, 65% or 90% reductions when the index or all passengers were masked. Note that using Liu's 65% for both R _S and R _R in Equation (8) results in an 88% equivalent reduction for universal masking, that is, approximately 90%. A second Lindsley paper ³⁹ shows decreasing exposures ordinal to: no masking, receiver masking, source masking, and universal masking, although receiver masking delays the exposure and reduces it early in the emission event compared to source masking. Wang et al. ²⁵ report 73% or 32% decreases in infection probability (rather than exposure level) when all passengers continuously wear high (65.6%) or low (31.0%) efficiency masks. Case studies of COVID‐19 infection on flights with mandated mask wearing ⁴⁰ , ⁴¹ suggests that some virus aerosol is emitted from infectious and inhaled by other masked passengers, and that distance remains an important infection determinant in a masked environment. Results of these studies show that masking alone seems to not eliminate all airborne exposures to infectious droplets and aerosols. This fact points toward multicomponent prevention strategies as good practices.

The three studies analyzed here showed a general decrease in aerosol concentration with distance from a source, although individual measurements showed both high variability close to the emission sources and some countertrend increases in aerosol concentration with distance. This trend broadly agrees with investigations of SARS‐CoV‐2 infection on international flights and high‐speed trains that found seating proximity strongly associated with infection risk: in one instance, 75% of infected passengers were seated within two rows of the symptomatic passenger who likely originated the outbreak ² and in another case study, the seats immediately adjacent to the index cases had a higher risk than other seats, with a relative risk (RR) of 27.8 (95% CI 14.4–53.7) on airplanes and 15.7 (95% CI 9.7–25.5) on trains. ⁴² However, only one of the three randomly generated infection patterns in Figure 5 adheres to the common understanding that most cabin infections occur within two seat rows of the index pax. Olsen's study of SARS1 observed infections several rows distant from the index. ⁴ One study concluded that the risk of COVID‐19 infection on an aircraft is low, even with infectious people onboard. ⁴³ At least qualitatively, the random infection patterns generated in the present study (Figure 5) appear to be more dispersed than is seen in most of the published infection cases. The infection probabilities and exposure patterns in the present study are based on aerosol measurements made at a distance of at least 0.5 m. Even though the NIOSH bacteriophage experiments had a spray source that included droplets, the lack of measurements at very close range favored measurement after droplets had evaporated into aerosols. Inclusion of large droplets would be expected to result in more infections at shorter distances.

None of the three datasets contain information about the level of absolute exposure risk. In the TRANSCOM and Boeing publications, the cabin particle counts and mass are expressed as a percentage of the total released ²⁴ , ²⁶ that are inhaled at given locations. This normalization makes sense for particle tracking and can be used for relative exposure evaluation but must be interpreted correctly in terms of exposure reduction and infection risk. Whenever particles are released into a volume, the number that would be measured at any specific location is much smaller than the number released. Even a person co‐located with the source would inhale a very small fraction of the aerosols expelled. Thus, reductions to a small fraction of the total generated are expected and would apply in most ventilated environments, but these small numbers do not imply small exposure. In broad strokes, aircraft cabin ventilation rates (per passenger) are typically 100 times greater than human respiration rates, meaning the environmental control system reduces exposure by perhaps 99% as a cabin average. Masking can be viewed as reducing the exposure that remains by about 60% and vacant middle seats by on average 37%. The importance of good ventilation in reducing exposure should not be undervalued. Still, a 99% or any other level of exposure reduction is not an indication of the presence or absence of infection risk. The TRANSCOM data were reported down to 0.0000%. However, 0.0001% is still 180 of the approximately 180 million particles released, that is, one millionth. This reporting may be reasonable, however, given experimental uncertainty.

Considering the uncertainties in the infectious fraction of the flying population and in their infectivities, the results are summarized in Table 2 as the probable number of infections per infectious passenger. Perhaps not surprisingly, the relative reduction is nearly independent of the infectivity. That is, VMS reduces the probable infections by about 38%, masking by about 62%, and both combined by about 75%, with only small variations from these numbers due to the infectivity. Table 2 also shows that the infections generated per infectious person change very little with two infectious passengers in the same cabin except at the very high infectivity. For two infectious passengers, then, the number of secondary infections almost doubles—almost, because two infectors might infect the same passenger. The multiplicative effect would decrease further with more infectious passengers. With three infectious passengers, the multiplier is less than three. These fractions are more meaningful when they translate into numbers of people. In the case of vacant middle seats without masking, the lowest % reduction, Table 2 shows nine prevented infections for the highest infectivity of 25 probable infections for one infectious passenger. For the combination of masking and VMS, at the lowest infectivity of one probable infection, these measures are predicted to prevent approximately one infection—essentially the difference between an in‐flight infection occurring and not.

These numbers were generated for very specific example cases, for example, an infected person in an aisle seat near the middle of the cabin. Hundreds of possible cases could be examined and, in the case of two or more infected passengers, thousands of cases. Nevertheless, the consistency of the relative reductions gives some confidence that the results are robust.

The findings in this report have some limitations. The TRANSCOM and Boeing studies did not involve a live virus, meaning that viability in the cabin environment was not considered, and the NIOSH emission source did not represent the variety of respiratory events that could transmit virus (e.g., talking, coughing, sneezing and each of these occurring with varying face covering practices). The NIOSH data were collected in laboratory cabin environments rather than onboard fully operating actual airplanes. Passenger movement during boarding and deboarding and lower ventilation during ground operations were not considered. Even so, from a dose perspective, the in‐flight portion might be the most significant due to much longer exposure duration. ¹⁹ The masking data used in the models did not consider behavioral aspects such as doffing while eating and drinking and noncompliance. Groups having similar immune histories, for example, families were not considered, where spreading their seating to create distance between them would not have a benefit. Finally, the present study did not measure COVID‐19 infection in aircraft cabins, although relative exposure—the %‐difference between exposures in different scenarios—provides useful information. The epidemiological and probabilistic papers cited in the current study support the likelihood that COVID‐19 infection has occurred during commercial passenger flights and, like the analyzes presented here, the dependence of virus exposure on distance. The infection thresholds and infection probabilities examined herein simply cover a reasonable range and are not tied to virus experiments.

Tracer particle data reported by the U.S. Transportation Command (TRANSCOM) and CFD simulations reported by Boeing were used along with NIOSH data, to estimate airborne virus exposure and infection reductions when middle seats are vacant compared to full occupancy and when passengers wear surgical masks in aircraft. Nonlinear regression models that estimate exposure at given distances from the viral source were applied to evaluate exposure reductions from vacant middle seats. Reductions averaged 54% for the seat row where an infectious passenger is located and 36% for a 24‐row cabin containing one infectious passenger. Direct comparison of the source adjacent seat and one set farther away predicted reductions of 20% and 41%. Analysis of the TRANSCOM data showed that universal masking (surgical masks) reduced exposures by 62%.

The extent to which exposure reduction decreases infection risk is not yet fully understood. However, an adaptation of the Wells–Riley equation allowed estimates of probable infections for a range of infectivity. This assessment provided reasonable estimates of relative reduction in the number of infections due to aerosol exposure when masks are worn and middle seats left vacant, although neither the infectious fraction of passengers nor their infectivity are accurately known. For a notional scenario involving 10 infectious passengers compared with no intervention, masking, distancing, and both would prevent 6.2, 3.8, and 7.6 secondary infections, respectively. Across a wide range of scenarios, masking reduced infections by 57%–63%, vacant middle seats by 37%–38%, and both combined by 73%–76%. A vacant middle seat policy comes with the cost of reduced passenger capacity, and masking puts stress on the relationship between airlines and passengers. This assessment shows that these measures, individually and combined provide options to substantially reduce the relative number of infections from aerosol exposure.

James S. Bennett: Conceptualization (equal); formal analysis (equal); methodology (equal); resources (equal); writing – original draft (equal). Seif Mahmoud: Data curation (lead); formal analysis (equal); methodology (equal); writing – original draft (equal). Watts L. Dietrich: Conceptualization (equal); data curation (supporting); formal analysis (equal); methodology (equal); writing – original draft (equal). Byron W. Jones: Formal analysis (equal); methodology (equal); supervision (equal); writing – review and editing (equal). Mohammad H. Hosni: Formal analysis (equal); methodology (equal); resources (equal); supervision (equal); writing – review and editing (equal).

The authors declare no potential conflict of interest.