Research on the Transmission of Disease in Airports and on Aircraft (2010)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

5 SESSIoN 1 Understanding How Disease Is Transmitted via Air Travel Jeanne Yu, Boeing Commercial Airplanes (Presenter) Ben S. Cooper, United Kingdom Health Protection Agency (Presenter) The AircrAfT cAbin environmenT Jeanne Yu (Presenter) Travel is all about people moving! The overall travel experience includes many elements as a person moves from one location to another; we think about the travel experience in the context of a “door-to-door experi- ence.” Travelers can experience many environments, moving from ground transport to an airport to an air- plane to another airport and to more ground transport before arriving at their final destination. To further our understanding of disease transmission at airports and on aircraft, it is important to recognize that the airplane flight is just one phase of the overall travel experience and that disease transmission can occur during all phases of the door-to-door experience. This white paper describes the aircraft cabin environ- ment part of the travel experience and how airplane sys- tems work to provide the air you breathe in the aircraft cabin environment. This paper also addresses items that should be considered for aircraft cleaning and disinfec- tion if a significant disease transmission event occurs. Airplanes typically fly at 36,000 ft. To put this num- ber in context, Mt. Everest is about 29,000 ft high. The environment is extreme at 36,000 ft: • Very cold: 245°f (243°C) to 285°f (265°C); • Very dry: less than 1% humidity; • Very low pressure; and • Naturally occurring ozone. To sustain human life, advanced environmental con- trol systems (ECSs) are needed. They control multiple important functions: cabin pressure, ventilation, tem- perature, anti-icing, and fire and smoke protection. Aircraft ECS designs must meet fAA regulatory requirements for safety and health, such as cabin pres- sure (8,000 ft maximum) and ventilation (0.55 lb/min/ person) and should not exceed threshold maximums for carbon monoxide, carbon dioxide, and ozone. The aircraft cabin environment also strives to meet objec- tives for comfort based on industry standards: Tempera- ture (T) [65°f to 85°f, DT < 5°f within a temperature control zone, SAE Aerospace Recommended Practices (ARP) 85]= • Rates of pressurization (climb. 500 ft/min; descent. 300 ft/min, SAE ARP 1270); • Cabin air velocities (<60 ft/min, optimal 20 to 40 ft/min, SAE ARP 85); • Aisle flow considerations for odor, temperature, ventilation mitigation; and • Cabin air treatment (SAE ARP 85). How is air provided to the aircraft cabin? In today’s aircraft design, outside air at 36,000 ft continuously enters the engine. At this altitude, the air is very clean, dry, low in oxygen, and practically particle-free. The air is compressed in the engine compressors and then extracted upstream of the combustion process; it travels in high-pressure ducts along the wing to the wing box of the aircraft. Here the air can pass through a cata-

6 RESEARCH oN THE TRANSMISSIoN of DISEASE IN AIRPoRTS AND oN AIRCRAfT lytic ozone converter to remove the naturally occurring ozone at altitude. The air then travels to the air con- ditioning pack, which houses many components, such as its own compressor, turbine, and heat exchanger. once the air is conditioned to the appropriate pressure and temperature, it goes to the mix manifold where it is mixed with highly filtered recirculated air in about a 50/50 ratio. Boeing aircraft use high-efficiency particu- late air (HEPA) filters with an efficiency of 99.97% at a particle size of 0.3 micrometer (µm) in diameter. In figure 1, the vertical axis shows filter efficiency, and the horizontal axis shows particle size. HEPA filters are ≥99% efficient over a particle size ranging from 0.003 to 10 µm, which encompasses a single virus and bacteria. Air from the mix manifold is supplied to the cabin through the air distribution system via riser ducts to the overhead cabin region and then through downer ducts into air supply nozzles that introduce the air into the aircraft cabin. The ECS is fully automated and air distri- bution is set by aircraft design. The ECS design goal for air supplied to the cabin is to generate a two-dimensional profile in a seat row to mini- mize drafts, temperature gradients, and odor migration. However, some three-dimensional aisle flow is inherent in the design and can be affected by movements such as galleys and occupants moving in the aisle. Air flows continuously into the cabin through the air distribution system and leaves the cabin through return air grilles that run the length of the cabin on both sides where the side wall meets the floor. The Harvard 1997 transportation study and other studies from 1987 to 1998 have measured the microbial level in different indoor environments. The measured levels of contaminants in aircraft cabin air are low compared with other indoor environments. Air also flows continuously out of the airplane through the outflow valve. The outflow valve regu- lates outflow of air and thus cabin pressure. The cabin pressure system controls the cabin pressure so that as the airplane climbs to its maximum certification alti- tude (40,000 to 45,000 ft depending on airplane type), the cabin pressure climbs to about 8,000 ft. Airplanes do not usually fly at their maximum altitude; typi- cally, they fly at an altitude of about 36,000 ft. The resulting aircraft cabin pressure is around 6,000 ft, which is similar to being in a tall building in Denver, Colorado. More detail and an animation showing how the air is provided to the cabin can be found at www.boeing.com/ commercial/cabinair/. ECSs are fully automated so that air flow rates to the cabin and to the flight deck are set by aircraft design. flight decks on some aircraft receive a 50/50 ratio of outside-to-recirculated air and some receive all outside air depending on the requirements and challenges of the flight deck air distribution design: electronic cool- ing, high solar loading from windshields, and higher pressure required in the event of smoke or fire. Pressurized cargo compartments can carry live ani- mals. Depending on the model, systems to heat ven- tilate and air-condition cargo holds are standard or optional. Boeing defers to appropriate authorities for disin- fection of aircraft: the Centers for Disease Control and Prevention (CDC), the U.S. Environmental Protection Agency, and the United Nations World Health organi- zation (WHo) • CDC recommendations for airlines: air travel industry; Particle size in micrometers Efficiency (percent) .01 .02 .03 .04 .05 .1.08 .2 .3 .4 .5 .6 .8 1 2 3 4 5 6 8 10 Good .003 99.97% efficiency airplanes and critical wards of hospitals ** 94% efficiency airplanes ** 80 – 85% efficiency trains * 90 – 95% efficiency hospitals * 60 – 65% efficiency office buildings * 25 – 30% efficiency office buildings * * ASHRAE 52–76 ** (IEST) Filter type “B” VERV17 Common type filters not tested at smaller particle size Single virus Tobacco smoke Bacteria 10 20 30 40 50 60 70 80 90 100 FIGURE 1 Comparative analysis of HEPA filters used in Boeing aircraft versus other applications.

7uNDERStANDiNg How DiSEASE iS tRANSmittED ViA AiR tRAVEl • wHo: website and document, “guide to Hygiene and Sanitation in Aviation;” and • international Air transport Association: website for “Health & Safety for Passengers and Crew.” Boeing also supports the following: • Research and working with the u.S. Department of Agriculture Animal and Plant Health Inspection Service to develop consistent guidelines with all original equip- ment manufacturers on inspecting, cleaning, and disin- fecting contaminated aircraft; and • Airline event response with aircraft cleaning and disinfection guidelines, including an approved material- compatible cleaners list. Aircraft cleaning and disinfection require substances that will not degrade aircraft materials. Boeing tests for material compatibility but does not test for substance efficacy against disease agents. Disinfection materials manufacturers and government agencies are responsible for efficacy testing. Boeing outlines requirements in the following: • Aircraft maintenance manuals that include safety instructions; • Boeing document, “Cleaning interiors of Commer- cial Aircraft;” and • Boeing document, “Evaluation of maintenance Materials.” Boeing research and collaboration are ongoing with academia and industry to further our understanding. We continue to work with the American Society of Heating, Refrigerating and Air Conditioning Engineers (ASHRAE) and industry collaboration to understand potential leverage points in ASHRAE’s strategic research agenda being developed to address the role of heating, ventilation, and air conditioning systems in the spread of infectious disease. We also are working toward maturing computational modeling capabilities. With Purdue University, we are developing model characterization of exhaled airflow from various modes of human respiration, including breathing, talking, and coughing. With the fAA Airliner Cabin Environment Research partners, we are studying additional modeling capabilities of moving bodies in the aircraft cabin. In summary, travel is a phenomenon of people mov- ing; the aircraft flight is one part of a traveler’s door- to-door experience. Aircraft ECSs are fully automated and designed to meet unique requirements for passenger safety and comfort. Aircraft disinfection must take mate- rial compatibility issues into consideration. further inte- grated collaborative research is needed. humAn movemenT PATTerns And The sPreAd of infecTious diseAses Ben S. Cooper (Presenter) Patterns of human movement are fundamental to the persistence, spatial distribution, and dynamics of human infectious diseases. Research aimed at teasing apart the complex relationship between human movement pat- terns and infectious disease dynamics has intensified in recent years, particularly since the 2002–2003 epidemic of coronavirus association with severe acute respiratory syndrome (SARS) and with concerns about a possible influenza H5N1 pandemic. However, the roots of this research go back much further. one way to appreciate the role of travel in the spread of infectious disease is to consider what would happen if people did not move among communities. Research based on mathematical models in the 1950s and 1960s shows that without such movements immunizing infec- tions such as measles would not be able to persist below a critical population size: in the troughs between epidemic peaks the numbers infected would fall to zero, and no further cases would occur without reintroduction from outside the community (1, 2). for measles, this critical population size was found to be about 300,000. The the- ory predicts that island populations below this size would be too small to sustain measles epidemics, and extended periods with no measles cases (until reintroduction of the virus) would be likely. Above this size, such stochastic fadeouts are unlikely and populations are large enough to maintain a continual presence of the pathogen. Later analysis of measles data from island populations has largely confirmed these predictions from mathematical models (3). Such considerations apply not only to actual islands but also to inland islands: the cities, towns, and villages where we live. over the last 20 years theoretical epide- miologists have extensively studied the spread of disease not just in a single population, but in metapopulations, or populations of populations coupled by travel links (4). In these cities and towns, population size plays a role simi- lar to that observed on islands, although coupling (due to human movement) between population centers tends to be stronger. Large populations have a sufficient influx of people susceptible to infection (either through birth, as in the case of measles, or through loss of immunity) to maintain the pathogen throughout the year, typically resulting in a regular seasonal epidemic pattern (5). The smaller the population the more likely stochastic fadeout (epidemic extinction) is to occur. This situation is due to the relative size of the stochastic fluctuations being larger for smaller populations, and the chance of the number infected reaching zero and the epidemic ending is cor- respondingly greater. If these small populations are not

8 RESEARCH oN THE TRANSMISSIoN of DISEASE IN AIRPoRTS AND oN AIRCRAfT linked by travel to other population centers, transmis- sion in these settings will end. Conversely, as coupling via transport networks strengthens, epidemics become more synchronized in the different population centers. Recent studies have shown how epidemic synchrony between different population centers can be explained by human movement patterns (6). At a more fundamen- tal level, many human pathogens (including measles and influenza) are believed to have made the transition from their original animal hosts with the advent of agricul- ture, when humans began to change from living in small relatively isolated groupings of hunter gatherers to larger communities (7). Air travel has an effect similar to that of any other means of human movement: by connecting geographi- cally isolated populations, it allows disease to spread between them and enables pathogens to persist by reduc- ing the chance of local stochastic fadeout. What makes air travel unique is its speed, which allows links between populations separated by large distances to be main- tained for pathogens with short generation times. Using influenza (which has a generation time of about 3 days) as an example, before the advent of the steamship, a pas- senger traveling from Europe to America infected imme- diately before embarkation would have had virtually no chance of transporting the virus between continents. Had Columbus been latently infected with influenza when set- ting out in 1492 for his 70-day Atlantic crossing, about 23 generations of influenza transmission on his carrack would have been required for the epidemic to spread to the Americas. With a crew of 70 men, this feat would have been almost impossible. In contrast, smallpox, with a generation time of 15 days, would have required only four or five generations of transmission on the ship to cross continents, making intercontinental spread quite feasible. With the advent of the steamer, Atlantic crossing times decreased to just a few days (a troop ship cross- ing the Atlantic in 1918 took about 7 days) and only about two generations of transmission were required to transmit influenza between continents, ensuring effi- cient global dissemination of the 20th century’s first pandemic. Air travel now represents by far the most important means for the rapid global dissemination of human pathogens—partly because it is the predominant means of transporting people over large distances but also because the short transit times make it an extremely efficient means of ensuring that even pathogens with very short generation times can be transported over very large distances. These concerns led to work carried out at the United Kingdom’s Health Protection Agency to determine whether practical measures could be taken to reduce this international spread in the event of a major pandemic with a virulent pathogen, particularly pan- demic influenza. first, we examined the potential role of airport entry screening. Entry screening of passengers with thermal imaging technology was used by a number of countries during the SARS epidemic and also by some during the 2009 H1N1 pandemic. A very simple analysis was able to show that, even if the sensitivity and specificity of the imaging technology used to detect symptomatic SARS or influenza infection were perfect (which is very far from being the case), the practice would have almost no value in protecting populations from influenza or SARS (8). This conclusion resulted from an elementary con- sideration of flight times and incubation periods for the two pathogens. only 1% to 6% of passengers incubat- ing SARS when boarding a plane would be expected to develop symptoms by the time they arrived in the United Kingdom (the higher percentage corresponding to the longer flight times), so almost all cases arriving in the United Kingdom would be missed, even with perfect screening. for influenza, which has a shorter incubation period, the corresponding range was 4% to 17%. the large number of passengers infected with influenza while traveling would mean that even if 17% could be detected and isolated, there would be no detectable impact on the epidemic in the destination country. Given that entry screening had been shown not to be an effective strategy, we considered whether canceling flights from affected cities could significantly alter the pattern of global spread in an influenza pandemic (9). Although we did not expect flight cancelation to be able to stop the global spread of influenza (the virus spread around the world quite efficiently in 1918 without the help of air travel), an important question was whether global dissemination could be delayed sufficiently to allow time for the development and production of a vac- cine that would protect against the pandemic virus (a process expected to take about 6 months). To address this question, we built on work started by Rvachev and colleagues working in the former Soviet Union in the 1960s (10). Rvachev had developed meta-population models to study the spatial dissemination of influenza. originally, this work considered population centers linked by rail networks, but it was then extended by Rvachev and Longini to account for the global spread of influenza through the international aviation network (11). our own work further extended these early efforts by recasting the deterministic global metapopulation models into a more realistic stochastic framework (which is important because at the beginning of the epidemic in each city, the numbers infected are small, stochastic effects are dominant, and the times of seeding new epi- demics in each city are expected to show considerable chance variation). In contrast to earlier work, we paid particular attention to a careful parameterization of the model by comparing air travel and influenza data from the 1968–1969 pandemic. This comparison was impor-

9uNDERStANDiNg How DiSEASE iS tRANSmittED ViA AiR tRAVEl tant for arriving at plausible values for the reproduction of pandemic influenza [before undertaking this work, no reliable estimates had been published, but estimates published concurrently with our analysis yielded results similar to those obtained with our model (12)]. This process also informed the modeling of seasonal varia- tion in the transmission potential and differences in sea- sonal variation between tropical and temperate regions (all factors that could have important effects on model predictions). This work was the first to evaluate explic- itly interventions that involved altering the international aviation network with the aim of slowing the global spread of pandemic influenza (figure 2). We considered two possible control policies: first, we evaluated a pol- icy that canceled a proportion, p, of all air travel from countries once they had experienced a certain number, q, of influenza cases (where both p and q were varied); second, we considered policies that did not involve can- celing flights but that reduced local transmission rates in affected countries. Such interventions could include social distancing measures (such as closing schools and promoting hand hygiene) and antiviral treatment and prophylaxis (13, 14). Comparison with the local epidemic peaks from the 1968–1969 pandemic showed that the model, though relatively simple, was able to capture the timing of the global spread of that pandemic with a high degree of accuracy, although some cities, such as Tokyo (where the epidemic peaked more than a month later than predicted by the model), did show departures from the model that were not consistent with chance effects. This analysis also showed that, with contemporary air travel volumes (2002 data), the timing of the epidemic peaks in 1969 would have been expected to occur somewhat earlier, in some cases (for southern hemisphere cities) shifting to an earlier influenza epidemic season. Results of the intervention analysis showed that restric- tions on air travel from affected cities were likely to have little value in delaying epidemics unless almost all travel ceased almost as soon as epidemics were detected in each city (Figure 3). For example, if 90% of air travel from affected cities were canceled after the first 100 influenza cases, the arrival time of influenza in other cities typically would be delayed by only 2 or 3 weeks. Though these delays showed some sensitivity to the city where the pan- demic first emerged and the timing of this event, in no case was the delay achieved close to the 6 months needed to develop and produce a vaccine. Even if 99% of jour- neys from affected cities could have been stopped, we found the delays in the timing of the epidemic peaks were FIGURE 2 Global dissemination of a simulated influenza pandemic originating in Hong Kong at the begin- ning of June to 105 cities, under the assumption that 99.9% of air travel from affected cities is canceled after the first 100 cases in each affected city (and after 1,000 cases in Hong Kong). City shading indicates the probability that each city has experienced a significant epidemic (based on 100 stochastic simulations). Flights connecting cities are shown as blue lines when there is at least a 5% chance that they have not been suspended due to travel restrictions. [Figure adapted from Cooper et al. (9).]

10 RESEARCH oN THE TRANSMISSIoN of DISEASE IN AIRPoRTS AND oN AIRCRAfT only 40 to 50 days, too short to have a significant practi- cal benefit. only if almost all travel from affected cities could be stopped almost as soon as influenza arrived was the intervention able to achieve delays likely to have a significant practical benefit in managing the pandemic. These results are somewhat counterintuitive but can be seen to be a function of the very short generation time of influenza, which results in a rapid initial rate of epidemic growth. If, at the beginning of the epidemic each case infected two other cases after 3 days, we would expect about 10 cases within 10 days of the first case and 100 within 20 days. Thus, even if travel from the city were reduced by a factor of 100 from Day 1, within about 3 weeks there would be the same number of people infected with influenza flying out as there would have been on Day 1 in the absence of any intervention. In contrast, it was found that interventions to reduce local transmission were likely to be more effective at reducing the rate of global spread and less vulnerable to implementation delays. Nevertheless, under the most plausible scenarios, achievable delays were found to be small compared with the time needed to accumulate sub- stantial vaccine stocks. other researchers, working with slightly different sets of assumptions, have reached similar conclusions about the limited role of air travel restrictions in con- trolling influenza pandemics (if the natural history parameters are similar to those for influenza strains we have seen before), and these results have directly informed both national and WHo recommendations for pandemic responses (15–17). While these conclu- sions have been challenged by a correlation found between a reduction in international travel to and from the United States after the terrorist attacks in Sep- tember 2001 and the timing of the seasonal influenza peak in the United States the following winter (18), the modeling work shows that a direct causal relation- ship between the relatively modest reductions in air travel that year and the influenza epidemic timing is extremely unlikely (19). Notably, the timing of influ- enza peaks routinely shows considerable year-to-year variation that cannot be explained by changes in the number of international air travelers. An obvious limitation of modeling studies evaluat- ing the role of the aviation network in the international spread of human pathogens is the failure to account for other modes of travel. However, excluding such travel from global dissemination models will bias model find- ings in favor of interventions that restrict air travel; by ignoring land and sea travel, the models will overesti- mate the impact of air travel restrictions on epidemic spread. Thus, the finding that air travel restriction Percent reduction in air travel from affected cities FIGURE 3 Impact of air travel restrictions on timing of epidemic peaks in the 105 cities shown in Figure 2 during a simulated influenza pandemic. Dots show timing of epidemic peaks in individual cities in the northern temperate zone (red), the tropics (black), and the southern temperate zone (green), where the area of each dot is proportional to the population size. Results from three stochastic simulation runs are shown for reductions in air travel between 0% (far left) and 99.9% (far right).

11uNDERStANDiNg How DiSEASE iS tRANSmittED ViA AiR tRAVEl will have limited value in controlling influenza pan- demic spread should be informative to this simplifying assumption. Recently, the metapopulation modeling framework has been extended again to account for “multiscale mobility networks,” accounting for both long-distance air travel links and shorter-distance com- muting flows, which are an order of magnitude larger (20). Results of this analysis have shown that including such commuting flows has little effect on the pattern and rate of global spread of infectious diseases com- pared with those predicted by air traffic flows alone. The main difference found when including commuting flows in models is increased synchrony of epidemic tim- ing in nearby subpopulations. References Bartlett, M. S. Measles Periodicity and Community Size. 1. Journal of the Royal Statistical Society, Series A, Vol. 120, 1957, pp. 48–70. Bartlett, M. S. The Critical Community Size for Measles in 2. the United States. Journal of the Royal Statistical Society, Series A, Vol. 123, 1960, pp. 37–44. Black, f. L. Measles Endemicity in Insular Populations: 3. Critical Community Size and Its Evolutionary Implica- tion. Journal of Theoretical Biology, Vol. 11, 1966, pp. 207–211. Grenfell, B., and J. Harwood. (Meta) Population Dynam-4. ics of Infectious Diseases. Trends in Ecology & Evolution, Vol. 12, 1997, pp. 395–399. Keeling, M. J. Modelling the Persistence of Measles. 5. Trends in Microbiology, Vol. 5, 1997, pp. 513–518. Viboud, C., o. N. Bjørnstad, D. l. Smith, l. Simonsen, m. 6. A. Miller, and B. T. Grenfell. Synchrony, Waves, and Spa- tial Hierarchies in the Spread of Influenza. Science, Vol. 312, 2006, pp. 447–451. Wolfe, N. D., C. P. Dunavan, and J. Diamond. origins 7. of Major Human Infectious Diseases. Nature, Vol. 447, 2007, pp. 279–283. Pitman, R. J., B. S. Cooper, C. L. Trotter, N. J. Gay, and 8. W. J. Edmunds. Entry Screening for Severe Acute Respi- ratory Syndrome (SARS) or Influenza: Policy Evaluation. British Medical Journal, Vol. 331, 2005, pp. 1242–1243. Cooper, B. S., R. J. Pitman, W. J. Edmunds, and N. J. 9. Gay. Delaying the International Spread of Pandemic Influ- enza. Public Library of Science Medicine, Vol. 3, 2006, p. e212. Baroyan, o. V., l A. Rvachev, u. V. Basilevsky, V. V. 10. Ermakov, K. D. Frank, m. A. Rvachev, and V. A. Shash- kov. Computer Modelling of Influenza Epidemics for the Whole Country (USSR). Advances in Applied Probability, Vol. 3, 1971, pp. 224–226. Rvachev, L. A., and I. M. Longini. A Mathematical Model 11. for the Global Spread of Influenza. Mathematical Biosci- ences, Vol. 75, 1985, pp. 3–22. Mills, C. E., J. M. Robins, and M. Lipsitch. Transmissibil-12. ity of 1918 Pandemic Influenza. Nature, Vol. 432, 2004, pp. 904–906. Bell, D. M., and World Health organization Writing 13. Group. Non-pharmaceutical Interventions for Pandemic Influenza, National and Community Measures. Emerging Infectious Diseases, Vol. 12, 2006, pp. 88–94. Webby, R. J., and R. G. Webster. Are We Ready for 14. Pandemic Influenza? Science, Vol. 302, 2003, pp. 1519– 1522. Hollingsworth, T. D., N. M. ferguson, and R. M. Ander-15. son. Will Travel Restrictions Control the International Spread of Pandemic Influenza? Nature Medicine, Vol. 12, No. 5, 2006, pp. 497–499. Pandemic Influenza Preparedness and Response. WHo, 16. Geneva, Switzerland, 2009. 17. Pandemic Flu: A National Framework for Responding to an Influenza Pandemic. United Kingdom Department of Health, London, 2007. 18. Brownstein, J. S., C. J. Wolfe, and K. D. Mandl. Empirical Evidence for the Effect of Airline Travel on Inter-regional Influenza Spread in the United States. Public Library of Science Medicine, Vol. 3, 2006, p. e401. 19. Viboud, C., m. A. miller, B. t. grenfell, o. N. Bjørnstad, and L. Simonsen. Air Travel and the Spread of Influenza: Important Caveats. Public Library of Science Medicine, Vol. 3, 2006, p. e503. 20. Balcan, D., V. Colizza, B. gonçalves, H. Hu, J. J. Ramasco, and A. Vespignani. multiscale mobility Networks and the Spatial Spreading of Infectious Diseases. Proceedings of the National Academy of Sciences USA, Vol. 106, No. 51, 2009, pp. 21484–21489.