International travel between global urban centres vulnerable to yellow fever transmission

Abstract Objective To examine the potential for international travel to spread yellow fever virus to cities around the world. Methods We obtained data on the international flight itineraries of travellers who departed yellow fever-endemic areas of the world in 2016 for cities either where yellow fever was endemic or which were suitable for viral transmission. Using a global ecological model of dengue virus transmission, we predicted the suitability of cities in non-endemic areas for yellow fever transmission. We obtained information on national entry requirements for yellow fever vaccination at travellers’ destination cities. Findings In 2016, 45.2 million international air travellers departed from yellow fever-endemic areas of the world. Of 11.7 million travellers with destinations in 472 cities where yellow fever was not endemic but which were suitable for virus transmission, 7.7 million (65.7%) were not required to provide proof of vaccination upon arrival. Brazil, China, India, Mexico, Peru and the United States of America had the highest volumes of travellers arriving from yellow fever-endemic areas and the largest populations living in cities suitable for yellow fever transmission. Conclusion Each year millions of travellers depart from yellow fever-endemic areas of the world for cities in non-endemic areas that appear suitable for viral transmission without having to provide proof of vaccination. Rapid global changes in human mobility and urbanization make it vital for countries to re-examine their vaccination policies and practices to prevent urban yellow fever epidemics.


Introduction
In December 2015, Angola reported its first locally acquired case of yellow fever in nearly a decade. The ensuing epidemic was first recognized in Luanda, then spread across Angola's 18 provinces, resulting in 4347 suspected or confirmed cases and 377 deaths. 1 International travellers departing from Angola then imported yellow fever virus into Kenya and the Democratic Republic of the Congo, 2 where another epidemic ensued, causing 2987 suspected or confirmed cases and 121 deaths. 1 Furthermore, 11 foreign workers infected in Angola travelled to urban centres in China, the first time imported cases of yellow fever have been reported in Asia. 3 Four cases were recently imported into Europe over an 8-month period by travellers returning from South America. 4 The time period is in stark contrast to the 27 years during which the previous four cases of travel-associated yellow fever were imported into Europe. 4 In early 2018, nine cases were exported from Brazil and led to three deaths. 5 Increased air travel and globalization is making it easier for humans to transport yellow fever virus across international borders, potentially catalysing deadly urban epidemics. 3 An essential tool in the fight against yellow fever is a live-attenuated vaccine developed in 1937. 6 This vaccine is vital for the prevention and control of yellow fever epidemics since no effective antiviral therapy exists. 7 However, a substantial proportion of the world's yellow fever vaccine stock was recently consumed in response to epidemics in Africa 8 and Brazil. 9 As a stopgap measure, the World Health Organization (WHO) approved fractional dosing to extend the vaccine supply, while recognizing that the duration of immunity may be compromised. 10 With only four WHOqualified yellow fever vaccine manufacturers in the world, rapid replenishment of the global emergency stockpile stretches finite resources, potentially resulting in vaccine shortages for preventive campaigns. 11 In late 2017, stocks of YF-VAX ® (Sanofi Pasteur, Lyon, France) in North America were depleted because of manufacturing difficulties. 5 Should another urban epidemic occur in the near future, vaccine demand could easily exceed the available supply.
Although many countries have vaccination policies to prevent international spread of the yellow fever virus, implementation is inconsistent. 12 Most, but not all countries where yellow fever is endemic require arriving international travellers without medical contraindications to provide official documentation of vaccination as a prerequisite for entry. As the vaccine provides protective immunity to 90% and 99% of individuals 10 and 30 days after vaccination, respectively, 13 most travellers are protected from acquiring and exporting the yellow fever virus. Furthermore, some countries where the disease is not endemic, but where the competent mosquito vector Aedes aegypti is present require travellers arriving from a yellow fever-endemic country to provide proof of vaccination. 14 The confluence of climate change, 15 rapid urbanization 16 and international air travel 17 are accelerating the globalization of mosquito-borne viruses such as dengue, chikungunya and Zika viruses. Here we examined the potential for the yellow fever virus to spread via international air travel into the world's cities, in order to guide global epidemic prevention efforts.
Objective To examine the potential for international travel to spread yellow fever virus to cities around the world. Methods We obtained data on the international flight itineraries of travellers who departed yellow fever-endemic areas of the world in 2016 for cities either where yellow fever was endemic or which were suitable for viral transmission. Using a global ecological model of dengue virus transmission, we predicted the suitability of cities in non-endemic areas for yellow fever transmission. We obtained information on national entry requirements for yellow fever vaccination at travellers' destination cities. Findings In 2016, 45.2 million international air travellers departed from yellow fever-endemic areas of the world. Of 11.7 million travellers with destinations in 472 cities where yellow fever was not endemic but which were suitable for virus transmission, 7.7 million (65.7%) were not required to provide proof of vaccination upon arrival. Brazil, China, India, Mexico, Peru and the United States of America had the highest volumes of travellers arriving from yellow fever-endemic areas and the largest populations living in cities suitable for yellow fever transmission.
Conclusion Each year millions of travellers depart from yellow fever-endemic areas of the world for cities in non-endemic areas that appear suitable for viral transmission without having to provide proof of vaccination. Rapid global changes in human mobility and urbanization make it vital for countries to re-examine their vaccination policies and practices to prevent urban yellow fever epidemics.

Methods
To identify gaps in yellow fever vaccination policies around the world, we assessed the potential for the international spread of yellow fever from areas deemed by WHO to be at risk of transmission to areas where conditions are known, or predicted, to be suitable for transmission. Our goal was to provide a global perspective on urban exposure to imported yellow fever virus, irrespective of past or present epidemics.

Global endemicity
We considered places where WHO recommended yellow fever vaccination in 2016, including recently identified parts of Brazil, to be areas where humans were at risk of local infection. [18][19][20] We refer to these areas as yellow fever-endemic areas, although we recognize that they may not have been experiencing yellow fever transmission. We excluded places where yellow fever vaccination was generally not recommended by WHO. For non-holoendemic countries (i.e. where only part of the country was at risk of yellow fever), 20 we delineated subnational areas of risk using ArcGIS v. 10.4.1 (Esri, Redlands, United States of America). We then used LandScan (Oak Ridge National Laboratory, Oak Ridge, USA) 21 to estimate the total population living within the global range of the yellow fever virus.

International dispersion
To account for the possibility that individuals infected with yellow fever virus within an endemic area might travel by land to a nearby airport in a non-endemic area, we used ArcGIS v. 10.4.1 to identify all commercial airports registered with the International Air Transport Association (IATA): (i) within 200 km of any yellow fever-endemic area worldwide (base scenario); and (ii) within 200 km of any city within a yellow fever-endemic area (urban scenario). In the base scenario, we considered travellers departing from areas of potential sylvatic or urban transmission as possible sources of exported yellow fever virus. In the urban scenario, we focused on travellers departing from airports within 200 km of a city (i.e. an urban centre with more than 300 000 residents, as defined by the United Nations' World Urbanization Prospects) 22 located in a yellow fever-endemic area. We mapped the final destination air-ports and the number of international travellers (determined from unique trips on commercial flights) departing from airports in each scenario by analysing worldwide tickets sales data from IATA between 1 January and 31 December 2016. 23 These data included the travellers' full itineraries: their initial airport of embarkation, their final destination airport and, where applicable, connecting airports. The data did not detail uncompleted trips due, for example, to cancelled or missed flights. Overall, these data accounted for an estimated 90% of all trips on commercial flights worldwide; the remaining 10% were modelled using airline market intelligence. 23 Such data have been used previously to anticipate the global spread of emerging infectious diseases. 24

Potential for urban transmission
To identify cities where yellow fever was not endemic, but which may have been suitable for viral transmission, we used a high-resolution, global, ecological model of dengue virus transmission, which was developed using empirical data on the real-world occurrence of dengue fever and associated environmental and climatic predictors of dengue virus transmission. 25 We assumed that cities predicted to be suitable for dengue virus transmission were also ecologically suitable for yellow fever virus transmission, because both viruses are primarily transmitted by Aedes aegypti, an anthropophilic mosquito highly adapted to urban settings. 25 Adopting a conservative approach, we excluded cities where the predicted probability of dengue-suitability was below 50%. As our analysis focused on urban importation and transmission of yellow fever virus, we did not consider its introduction into rural, sylvatic areas or transmission among non-human primates. We defined a yellow fever-suitable city as a population centre with at least 300 000 residents in an area where the yellow fever virus was not endemic but which was predicted to be suitable for viral transmission. We excluded cities above 2300 m because environmental conditions at these elevations are considered unsuitable for yellow fever virus transmission. 26 We assessed the potential for importation of the yellow fever virus by quantifying the volume of airline passengers travelling from yellow feverendemic areas of the world, according to our base and urban scenarios, to yellow fever-suitable and -endemic cities. We also considered the possibility that individuals infected with the virus might arrive at an airport in a nonendemic area and then travel by land to a neighbouring city within a yellow fever-endemic or -suitable area: in our analysis, we included all commercial airports located within 200 km of these mutually exclusive geographical areas. We then categorized traveller flows according to the official yellow fever travel vaccination policy in each endemic and non-endemic country: (i) no proof of yellow fever vaccination required; (ii) proof of vaccination required if arriving from a yellow fever-endemic country; and (iii) proof of vaccination required if arriving from any country. 27 Finally, we aggregated the resident populations of all yellow fever-suitable and -endemic cities.

Results
We estimated that 923 million people lived in areas of the world where yellow fever was endemic in 2016, spanning 25 holoendemic and 17 non-holoendemic countries or territories (Box 1).
In our base scenario, 45.2 million travellers departed from yellow feverendemic areas for international destinations in 2016. Of these, 7

Research
Global vulnerability to yellow fever Shannon E Brent et al.
within or adjacent to yellow feverendemic cities, 11.7 million (25.8%) had destinations at airports within or adjacent to yellow fever-suitable cities and 25.6 million (57.8%) had other destinations ( Fig. 1). Of the 7.9 million travellers with international destinations at or near other yellow fever-endemic cities, 0.86 million (11.0%) landed in a country where proof of yellow fever vaccination was not required upon arrival: one holoendemic country (i.e. South Sudan) and three non-holoendemic countries (i.e. Argentina, Brazil and Peru). Of the 11.7 million travellers with destinations at or near yellow fever-suitable cities, 7.7 million (65.7%) landed in a country where proof of yellow fever vaccination was not required: four non-holoendemic countries (i.e. Argentina, Brazil, Ecuador and Peru) and 12 non-endemic countries (e.g. the United States). Conversely, 14.9 million travellers departed non-endemic areas of the world for airports within or adjacent to yellow fever-endemic cities; 11.4 million (76.4%) of these travellers landed in countries where proof of yellow fever vaccination was not required on arrival.
In our urban scenario, 32.2 million travellers departed airports within or near yellow fever-endemic cities for international destinations in 2016. Of these, 6.1 million (18.9%) arrived at or near yellow fever-endemic cities (Table 1); there was one fewer destination city than in our base scenario. In addition, 8.4 million (26.1%) arrived at or near yellow fever-suitable cities; there were six fewer destination cities than in our base scenario ( Table 2). As the urban scenario considered only travellers departing from airports within 200 km of a city within a yellow feverendemic area, it represents the potential for dispersion during an urban outbreak rather than dispersion secondary to urban or sylvatic transmission, as in the base scenario.
Among countries with yellow feverendemic cities, Brazil, Colombia and Nigeria had the highest traveller numbers from other yellow fever-endemic areas of the world and the largest populations living in yellow fever-endemic cities (Fig. 2). Colombia and Nigeria required proof of yellow fever vaccination from travellers arriving from other yellow fever-endemic countries but not from non-endemic countries. In contrast, Brazil did not require proof of vaccination from travellers arriving from yellow fever-endemic countries. Among countries with yellow fever-suitable cities, Brazil, China, India, Mexico, Peru and the United States had the highest traveller numbers arriving from yellow fever-endemic areas and the largest populations living in yellow feversuitable cities (Fig. 3). Of these, Brazil, Mexico, Peru and the United States did not require proof of yellow fever vaccination from travellers arriving from yellow fever-endemic areas. Fig. 4 and

Discussion
The 2016 yellow fever epidemic in Angola and the associated exportation of cases into urban areas of China exposed shortcomings in existing yellow fever travel vaccination policies and practices. As a holoendemic country, Angola has a policy that requires all international travellers to provide proof of yellow fever vaccination upon arrival. In addition, China has the same requirement for travellers arriving from yellow fever-endemic countries. Yet both lines of defence failed, leading to the first cases of imported yellow fever in Asia. Recent research has confirmed the role played by air travel between Angola and China in increasing the risk of importing the disease. 28 This event illustrates that urban areas that have never experienced yellow fever transmission, or have not  Notes: A yellow fever-endemic area was a national or subnational area where the World Health Organization recommended yellow fever vaccination. A yellow fever-endemic city was a city located in an area where vaccination was recommended. A yellow fever-suitable city was a city that was suitable for dengue virus transmission (see main text for details). Other destinations were: (i) all destinations where yellow fever was not endemic and which were not suitable for yellow fever transmission; and (ii) areas where yellow fever was endemic or which were suitable for yellow fever transmission but did not contain a settlement with a population greater than 300 000.
experienced it in modern times, are increasingly susceptible to epidemics. We elected to study the travel conduits that could facilitate the international spread of yellow fever virus into the world's cities.
First, our analysis revealed that 89% of travellers departing from yellow fever-endemic areas for yellow feverendemic cities in other countries (both holoendemic and non-holoendemic) in 2016 were required to provide proof of vaccination upon arrival. This high proportion presumably reflects countries' desire to protect themselves against importation of yellow fever virus. To reduce the risk of importation, and of the consequent potential for domestic transmission and of possible exportation of yellow fever virus, these countries NA: not applicable. a All destination countries and territories were yellow fever-endemic areas. b Countries and territories were ranked according to the number of travellers arriving from yellow fever-endemic areas, which was determined by examining all outbound international flights from airports within areas where the World Health Organization (WHO) recommended yellow fever vaccination and all airports within 200 km of such areas. 17-19 c Nationally aggregated population living in cities. d The base scenario considered international travellers arriving from airports within areas where WHO recommended yellow fever vaccination and all airports within 200 km of such areas. e The urban scenario considered international travellers arriving from airports within 200 km of a city (population ≥ 300 000) in an area where WHO recommended yellow fever vaccination. f We did not take into account Brazil's temporary yellow fever vaccination requirements for incoming passengers from Angola and the Democratic Republic of the Congo during the 2016 outbreak. g There were 10 other yellow fever-endemic destination countries with an airport within 200 km of a yellow fever-endemic city with a population of at least 300 000: Argentina, Burundi, Central African Republic, Chad, Ethiopia, Gambia, Guinea-Bissau, Niger, Peru and Sierra Leone. We did not show the 7 countries where there was no city with at least 300 000 residents located in a yellow fever-endemic area: Ecuador, Equatorial Guinea, French Guiana, Guyana, Mauritania, Suriname and Trinidad and Tobago.

Research
Global vulnerability to yellow fever Shannon E Brent et al.
should focus on implementing existing yellow fever travel vaccination policies effectively. However, some travellers may purchase counterfeit international vaccination certificates, 29 which makes this line of defence potentially fallible.
Second, we found that less than 35% of travellers departing yellow feverendemic areas for cities that appeared suitable for yellow fever transmission, were required to provide proof of vaccination upon arrival. Countries that did not require proof of yellow fever vaccination might have assumed that the historical absence of yellow fever was predictive of its future absence. In other instances, nationally implemented vaccination policies may be obfuscated a Destination cities in these countries and territories were ecologically suitable for yellow fever virus transmission but were not in yellow fever-endemic areas. b Countries and territories were ranked according to the number of travellers arriving from yellow fever-endemic areas, which was determined by examining all outbound international flights from airports within areas where the World Health Organization (WHO) recommended yellow fever vaccination and all airports within 200 km of such areas. 17-19 c Nationally aggregated population living in yellow fever-suitable cities. In the urban scenario, there were six fewer yellow fever-suitable destination cities than in the base scenario: Satna, India (population 0.31 million); Ibb, Yemen (population 0.45 million); Al Hudaydah, Yemen (population 0.57 million); Taiz, Yemen (population 0.69 million); Aden, Yemen (population 0.88 million); and Sana'a, Yemen (population 2.7 million). d Our base scenario considered international travellers arriving from airports within areas where WHO recommended yellow fever vaccination and all airports within 200 km of such areas. e Our urban scenario considered international travellers arriving from airports within 200 km of a city (population ≥ 300 000) in an area where WHO recommended yellow fever vaccination. f United States' territory included all continental states and Hawaii. Puerto Rico was not included and is listed separately. Other United States territories, such as Guam, American Samoa and the United States Virgin Islands, do not have cities with at least 300 000 residents and are thus not included. g We did not take into account Brazil's temporary yellow fever vaccination requirements for incoming passengers from Angola and the Democratic Republic of the Congo during the 2016 outbreak. h There were 29 other countries or territories suitable for yellow fever transmission (details available from the corresponding author on request).
because only a small geographical area within a country may be ecologically suitable for yellow fever transmission; for example, the 9.5 million United States' residents who live in five urban areas that appear suitable for yellow fever transmission represent less than 3% of the country's population. Nonetheless, countries should carefully consider whether the risk of yellow fever virus importation and subsequent domestic transmission warrants a change to existing yellow fever travel vaccination policies or practices. Of note, administering yellow fever vaccine at national ports of entry to individuals who do not hold a record of vaccination will increase immunity among susceptible travellers but will not prevent importation of the virus by travellers who are already infected. Third, we found that less than 25% of travellers who departed from areas of the world where yellow fever was not endemic for yellow fever-endemic cities were required to provide proof of vaccination upon arrival. This reveals a policy gap in protecting international travellers against becoming infected and subsequently exporting the virus. This low proportion may reflect the absence of national incentives because countries with entry requirements for yellow fever vaccination are protecting international travellers and the global community without realizing any domestic benefit.
Although broader use of yellow fever vaccine by international travellers could limit dispersion of the virus and reduce the risk of urban epidemics, its use in non-epidemic settings must be carefully weighed against the risk of vaccine-associated neurological and viscerotropic events. Infants younger than 9 months, adults aged 60 years and older and individuals with thymus disorders and weakened immune systems are at an elevated risk of these potentially life-threatening events. 30 Furthermore, if international changes in vaccination policy and practice are implemented and enforced, travellers could face difficulties accessing yellow fever vaccine, given current diminished stocks and constrained manufacturing capacity. Even though an estimated 50 million vaccine doses were produced in 2017, 11 a new yellow fever epidemic in a populated urban centre could readily deplete global emergency vaccine stockpiles.
We made several important assumptions in our analysis. First, we assumed that the risk of yellow fever virus dispersion across all yellow feverendemic areas of the world was uniform, because we were not attempting to model the spread of the virus out of a particular geographical area that was experiencing epizootic or epidemic activity. Rather, our goal was to describe global pathways via which the yellow fever virus could disseminate to trigger epidemics in the world's cities, thereby identifying crucial gaps in existing yellow fever travel vaccination policies and practices. Since the potential for international dispersion of the virus out of rural areas presumably differs from that out of urban areas, our urban scenario focused solely on travellers departing airports in or immediately adjacent to cities in yellow fever-endemic areas. However, the recent case of a traveller who acquired a yellow fever virus infection in rural Suriname and then flew to

Research
Global vulnerability to yellow fever Shannon E Brent et al.
the Netherlands indicates that there is still a risk of yellow fever exportation from rural areas. 4 Our assumptions about the suitability of cities for yellow fever virus transmission were based on a global ecological model of dengue virus transmission. A recently published modelling analysis of suitability for yellow fever transmission globally predicted a similar pattern to the pattern of dengue suitability we assumed, 31 especially in urbanized regions, which were the primary focus of our study. However, we may have overestimated the risk of yellow fever transmission in areas where dengue is known to be active but where Ae. albopictus rather than Ae. aegypti is the dominant vector (e.g. in China, Hong Kong Special Administrative Region). On the other hand, although Ae. aegypti is the primary vector for transmission of yellow fever virus, some studies have indicated that Ae. albopictus might also be a competent vector in nature. 32 As our analysis focused on the importation of yellow fever virus into cities and ignored downstream transmission among nonhuman primates in rural sylvatic cycles, we believe our model of urban dengue suitability closely approximates suitability for yellow fever virus transmission.
Our model of dengue suitability represents an annualized view of potential yellow fever transmission. The model does not account for seasonal variability due to changing climatic conditions. 33 Furthermore, we did not take into account seasonal patterns in local (i.e. urban-rural) or international travel despite the possibility that interactions between the ecological seasonality of yellow fever transmission and the seasonality of human mobility could influence the risk of yellow fever virus importation. In addition, we did not attempt to quantify variations in the intensity of transmission between tropical and subtropical climates or between industrialized and developing areas of the world. For example, because of differences in climate and the built environment, 34 some cities in the southern United States have experienced sporadic transmission of dengue, chikungunya and Zika viruses, whereas cities in Latin America have experienced sustained and intense transmission of the same pathogens. Moreover, we did not attempt to estimate how the underlying level of population immunity influences the potential for epidemics. Although we presumed that populations in yellow fever-suitable cities would have negligible immunity to the yellow fever virus, we made no assumptions about immunity in yellow fever-endemic cities, because high-resolution data on yellow fever vaccination and natural infection were lacking. Lastly, we did not take into account Brazil's temporary yellow fever vaccination requirements for travellers who came from Angola and the Democratic Republic of the Congo during the 2017 yellow fever outbreak and therefore categorized Brazil as not requiring proof of vaccination upon arrival from yellow fever-endemic countries.
With more than 3 billion domestic and international passengers now boarding commercial flights each year,

Research
Global vulnerability to yellow fever Shannon E Brent et al.