Probability and Estimated Risk of SARS-CoV-2 Transmission in the Air Travel System: A Systemic Review and Meta-Analysis

2.1. Literature Review for Air Travel Transmission Cases

A literature review was performed from May to September 2020 to identify index and secondary COVID-19 cases related to transmission in the air travel system for travel from January to September 2020. Index cases are those who tested positive for COVID-19 upon arrival at their destination. Thus, based on the known incubation period of the disease, these passengers were identified as infected prior to entering the air travel system. Secondary cases are persons who travelled on a flight with an index passenger, tested positive for COVID-19 after the flight, and were identified through contact tracing to have most likely been infected via transmission from the index passenger. Publications were collected using English-language search terms linking COVID-19 and air travel from two primary paths: review of daily news articles and social media posts (referencing published research) or direct search of scientific publication repositories (including arXiv.org, Clarivate, LitCovid, and the CDC’s COVID-19 research database). Additional papers were identified by biomedical experts at International Air Transport Association (IATA) and the Commonwealth Scientific and Industrial Research Organization (CSIRO). As a quality measure, only studies accepted by a journal for publication or by a preprint server were reviewed due to the expected higher quality of information on flights taken by an index passenger and subsequent contact tracing. Two additional studies performed by IATA and the United States Centers for Disease Control and Prevention (CDC) were added to the review due to organizational oversight and data collection capabilities, despite the authors’ limited visibility into the data collection methods and sources. Acknowledging that the variability of recommended behaviors and policy across airlines and governments during the pandemic likely leads to confounding issues in all papers in this study (i.e. mask use, seating), as well as a lack of similar study design, no results were disqualified in the interest of making a conservative, worst case estimate of risk.

The majority of publicly available studies focused on infectious passengers and related transmissions, therefore the study was limited to flights with passenger index cases and did not include transmissions amongst air crew, ground crews, or airport staff. Multiple experts reviewed each study and independently extracted flight information (date, destination and arrival cities, and time of flight), number of index cases, and number of secondary cases as well as available data on passenger behavior. Conservatively, all probable secondary cases identified in the literature are included in risk calculations. Study duplications were recognized by comparing flight information and the study with more information about passenger contact tracing was kept in the review. Discussion was held among reviewers to confirm validity of identified secondary cases being infected in the air travel system, most specifically the airplane cabin, by studying given information about passenger behavior and comparing the index infectiousness period, symptom onset, and positive test dates to COVID-19 incubation period known in September 2020. Group consensus on the number of index and secondary cases was required before proceeding on the study. To verify reasonable interpretations of the review, conversations were held with IATA and the organization’s Medical Advisor, Dr. David Powell (author of [29]), to ensure all publically known flights with transmission were included in the review and to consider the organization’s understanding of air travel risk.

The majority of studies in the literature review focused on a single flight with known secondary cases. Other studies outlined multiple flights, but made qualitative statements on risk based only on the select handful of flights reviewed. News stories released by media and not included in this review indicated index cases aboard many more flights than covered in the published papers, but did not follow up on contact tracing results or identification of secondary cases. A nuanced understanding of air travel risk should therefore consider the full range of air travel experiences during the pandemic from a flight with multiple secondary cases and no mitigation techniques to a flight with no infectious passengers.

2.2. Calculation of Estimated Risk of Transmission

To estimate the risk of air travel transmission, a quantitative approach was taken that considers secondary cases and the overall air travel passenger population. A novel approach to transmission risk was developed which uses two correction factors for asymptomatic transmission and underreporting to address the confounding factors and study heterogeneity identified during the literature review and to more conservatively estimate risk. The number of infected persons in the transmission events identified in Section 2.1 was determined through either mandatory screening or self-reporting, and it is likely that this count of infected persons is much lower than the true number of persons who have been infected while flying. Therefore, a number of caveats must be considered for potential error in the risk calculation. The first is the number of people who are infected, but are asymptomatic. Per the Centers for Disease Control and Prevention (CDC), asymptomatic infections are nominally estimated to be 40% of the total infections and asymptomatic individuals are estimated to be 75% as infectious as a symptomatic individual [12]. This indicates a factor of 1.3x if assuming asymptomatic passengers are not fully counted as index or secondary cases in the data. The second issue is the number of unreported COVID-19 cases, both index and secondary, in relation to air travel which includes:

• Symptomatic passengers who seek medical attention for symptoms, but do not inform medical personnel of their travel;

• Symptomatic passengers who seek medical attention for symptoms, but medical personnel fail to report travel; and

• Symptomatic passengers who do not seek medical attention.

To account for underreported air travel cases, the number of unreported COVID-19 cases in the general global population is assumed to apply. Recent estimates indicate global under-reporting to be a 10-fold [13] [14], 23-fold [15] and 54-fold [16] multiplier of the current reported cases. With these caveats, the equation for estimated risk of transmission is as follows: where n_t is the number of secondary cases due to air travel transmission, n_p is the total passenger population, f_a is the factor to account for number of and infectiousness of asymptomatic cases, and f_c is the factor to account for number of unreported cases.

3.1. COVID-19 Transmission Events in the Air Travel System

During the pandemic it has been estimated that, even with a significant decline in air travel, approximately 1.4 billion passengers traveled by air between January and September 2020 [17]. As of August 2020, there were at least 2866 index passengers documented to have passed through the air travel system (Table 1). Due to the widespread prevalence of the disease and presence of pre-symptomatic, asymptomatic, and mildly symptomatic cases in the population, it is known that there necessarily have been additional COVID-19 infectious passengers entering the travel system, beyond the identified index cases. Nonetheless, review of publicly available data reveals fewer than 50 documented potential secondary cases associated with air travel during the pandemic. Mask use on reviewed flights ranged from use unknown to mandatory N95 use.

Table 1 documents the index and secondary cases for passengers who have flown during the pandemic. International Air Transport Association (IATA) and CDC data are included for thoroughness with the acknowledgement that limited details are known about the data methods and sources and that case counts may be duplicates of other reported cases.

For the purpose of this study, transmission from the index passenger was conservatively assumed to have occurred inside the aircraft cabin in each of these documented cases, although transmission could have occurred elsewhere, such as in the airport. For example, the 15 secondary cases associated with the London-Hanoi flight on March 2, 2020 may not all be the result of in-flight transmission. All of these 15 passengers had either hotel, cruise ship, secondary transportation, travel to a second country, or known contact with other infectious persons not from the flight, or combinations thereof, after the flight from London to Hanoi and prior to testing [11]. Based on the arrival date of the flight (March 2) and the earliest date of the subsequent reverse transcriptase polymerase chain reaction (RT-PCR) tests (March 6), it is possible that the potential secondary cases were infected post-flight since a positive test via PCR can occur on average 2 to 5 days after infection [18]. The testing of the London to Hanoi passengers did not include genomic analysis of the virus and therefore it was not possible to demonstrate that all of the secondary cases map to the same index case by sequence similarity.

For these reasons, the causal connection between the London to Hanoi flight and the eventual positive test results for the 15 cases is in question.

3.2. Epidemiological Study of SARS-CoV-2 in Flight

In comparison to the contact tracing performed for the London-Hanoi flight in Section 3.1, other studies were able to more clearly trace index and secondary case relationships to a flight. An epidemiological study of the January 24, 2020 flight from Singapore to Hangzhou, China with a number of confirmed COVID-19 cases on board (Figure 1) was recently completed [10]. This flight, which lasted approximately five hours, was loaded at 89% capacity. Upon arrival at the destination, all passengers and crew were quarantined and tested as necessary. The study determined that of the 16 passengers who ultimately tested positive for COVID-19, 15 of the passengers were index cases on the flight due to the timing of symptom onset; if they had been infected on the flight, they would not have had symptoms immediately after the flight. The index cases had membership in one of three tour groups with other passengers who also tested positive for COVID-19 and originated in Wuhan. Only one passenger was identified to be a potential secondary case of SARS-CoV-2 in flight due to a different city of origin (Hangzhou), membership in a different tour group with no recorded cases of COVID-19, and reported activity of temporarily moving to an unassigned seat surrounded by infectious and symptomatic passengers during the flight as indicated in Figure 1. All 16 passengers wore masks during air travel, but removed their masks during meal time and to drink water; the mask of the secondary case was not worn tightly and did not cover their nose when sitting in the unassigned seat.

• Download figure
• Open in new tab

Figure 1:

Seat map of flight with 15 index passengers and one possible secondary case.

3.3. Estimated Risk of Transmission

The secondary cases listed in Table 1 are for over 10 flights that occurred March 2020 or earlier, while flight traffic was still high, in areas like Europe and the United States. To take a conservative position on risk, transmission risk was calculated for three time periods of interest: (1) January–June 2020 during which the transmission events listed in Table 1 occurred, (2) the month of March 2020 when the global spread of COVID-19 was occurring, and (3) April– September 2020 to account for the sharp drop in worldwide air travel and increased use of COVID-19 testing. For a slightly more conservative risk calculation, the 3 secondary cases recorded by IATA were included in the multi-month risk calculations (January-June 2020, April-

September 2020) based on the assumption that March was a well-documented month, so it is unlikely that these cases are duplicates of the transmission events in that month. Passenger counts were derived from International Civil Aviation Organization (ICAO) monthly reporting of actual passenger totals [17]. Table 2 lists the estimated risk of SARS-CoV-2 transmission in the air travel system for the global public as calculated with Equation 1.

For 44 secondary cases from January–June 2020 with a 1.3-fold factor for asymptomatic persons (f_a) and a 10-fold factor for underreporting (f_c) against a population of 968 million travelers, the risk of being infected with SARS-CoV-2 in an airplane cabin is estimated to be 5.927×10⁻⁷ or 1:1.7 million. A 54-fold multiplier of reported cases gives a risk of 3.201×10⁻⁶ or a 1:0.3 million (1:312,419) chance of being infected with COVID-19 in the airplane cabin.

Due to developing availability and reliability of COVID-19 testing and localized variability in reporting since the start of the pandemic, statistics for asymptomatic persons, infectiousness, and underreporting carry a degree of uncertainty which affects the overall risk calculation. The CDC indicates that asymptomatic cases could range from 10–70% of the infected population and that asymptomatic infectiousness ranges from 0.25–1.00 [12]. The Chow et al. study indicates a 95% credible interval for an estimated 10-fold underreporting factor ranges from 2.2–50 [14]. For this paper’s treatment of risk, uncertainty is established by first identifying the uncertainty in factors f_a and f_c in Equation 1, next fitting triangular distributions to each statistic with the peak at the baseline values stated above, then running 1 million simulations with a draw from each distribution to calculate risk, and finally selecting the 2.5% and 97.5% risk estimate from the 1 million simulations which set the bounds of the uncertainty interval. Using the above literature estimates of uncertainty in the parameters of f_a and f_c, uncertainty in the transmission risk is estimated to range from 1 case for every 0.7 million (712,000) travelers to 1 case for every 8 million travelers.

At the onset of the pandemic in March, the estimated risk of transmission was 2.353×10⁻⁶ or 1:425,062 (f_c=10). Estimated risk of being infected with COVID-19 could have been as high as 1:78,715 (1.27×10⁻⁵, f_c=54). In comparison, estimated risk during the pandemic (April– September 2020) ranges from 1:1.3 million (f_c=54) to 1:7.1 million (f_c=10) with reduced passenger traffic and few documented secondary cases.