Application of a quantitative entry assessment model to compare the relative risk of incursion of zoonotic bat-borne viruses into European Union Member States.

This paper presents a quantitative assessment model for the risk of entry of zoonotic bat-borne viruses into the European Union (EU). The model considers four routes of introduction: human travel, legal trade of products, live animal imports and illegal import of bushmeat and was applied to five virus outbreak scenarios. Two scenarios were considered for Zaire ebolavirus (wEBOV, cEBOV) and other scenarios for Hendra virus, Marburg virus (MARV) and Middle East Respiratory Syndrome Coronavirus (MERS-CoV). The use of the same framework and generic data sources for all EU Member States (MS) allows for a relative comparison of the probability of virus introduction and of the importance of the routes of introduction among MSs. According to the model wEBOV posed the highest risk of an introduction event within the EU, followed by MARV and MERS-CoV. However, the main route of introduction differed, with wEBOV and MERS-CoV most likely through human travel and MARV through legal trade of foodstuffs. The relative risks to EU MSs as entry points also varied between outbreak scenarios, highlighting the heterogeneity in global trade and travel to the EU MSs. The model has the capability to allow for a continual updating of the risk estimate using new data as, and when, it becomes available. The model provides an horizon scanning tool for use when available data are limited and, therefore, the absolute risk estimates often have high uncertainty. Sensitivity analysis suggested virus prevalence in bats has a large influence on the results; a 90% reduction in prevalence reduced the risk of introduction considerably and resulted in the relative ranking of MARV falling below that for MERS-CoV, due to this parameter disproportionately affecting the risk of introduction from the trade route over human travel. Crown

Introduction

Bats are natural reservoir hosts for many viruses which are recognised as serious potential threats to human and/or animal health (Calisher et al., 2006). The bat-borne viruses emerging in the African, Asian and Australian continents have come to the fore more recently with regards to their threat to human health and pandemic potential. Since 2003 there have been a number of large-scale human outbreaks of bat-borne diseases e.g. Zaire ebolavirus (EBOV) and Severe Acute Respiratory Syndrome (SARS) in Western Africa and Asia respectively, whilst a significant number of human cases of Nipah virus (NiV) are reported in Bangladesh every year (IEDCR, 2014). Pteropid bats are known to be the natural host of Hendra virus (HeV) (Halpin et al., 2000), a member of the same genus (Henipavirus) as Nipah virus. Since 1994 HeV has been responsible for seven human cases in Australia, four of which were fatal (Smith et al., 2016). Bats have also been linked with Marburg virus (MARV) (Towner et al., 2007), and, more tenuously, with the emerging Middle East Respiratory Syndrome Coronavirus (MERS-CoV) (Memish et al., 2013). Within the European Union (EU), zoonotic incidents of bat-borne viruses have been restricted to European bat lyssavirus types 1 and 2 which have been responsible for less than 10 human cases since 1977 (Fooks et al., 2003). To date, there have only been a few isolated reports of introduction of bat-borne viruses (e.g. MARV, SARS, MERS-CoV) from outside the EU mainly through entry of infected humans (Puzelli et al., 2013, Desenclos et al., 2004, WHO 2008, Reuss et al., 2014). However, these incidents illustrate that introduction can occur and support the need for some level of surveillance activity to assess the probability of when and where further incursions may take place.

For most bat-borne zoonotic diseases, primary transmission routes for human infection include direct or indirect contact with bat bodily fluids (Leroy et al., 2009, Luby et al., 2006), or via intermediate animal hosts (Parashar et al., 2000, Hanna et al., 2006). Onward transmission of disease is then possible via human-to-human contact or contact with contaminated fomites or the environment, with nosocomial infections being particularly important in some instances (Baron et al., 1983, Chowell et al., 2014). Disease introduction into the EU could therefore potentially occur from a number of routes, including human travel, illegal and legal importation of food products and transport of live animals. These routes have previously been associated with incursion of other viruses into the EU. For example, human travel and immigration are thought to be the primary reasons why individual Member States (MSs) have a high prevalence of the same Human Immunodeficiency Virus (HIV) subtypes as their historical African colonies (Faria et al., 2012). Classical rabies has been detected in imported domestic pets (Suárez-Rodríguez et al., 2013, McQuiston et al., 2008) and avian influenza (H5N1 type A) has been detected in illegal imports of Crested Hawk Eagles from Thailand into Belgium (Van Borm et al., 2005). In the case of trade, illegal importation of food products has been a suggested route of origin for the foot and mouth disease epidemic in the UK in 2001 (Defra, 2001), whilst legal trade in fresh produce, such as fruit and vegetables, has been associated with norovirus (Hjertqvist et al., 2006) and hepatitis A outbreaks (Dentinger et al., 2001). Virus specific transmission characteristics may influence the relative importance of these potential routes of disease introduction.

In terms of government financial resource allocation, it is important to develop methods to assist in efficient targeting of surveillance activities e.g. to inform which pathogen(s) are most likely to enter the EU, where they are most likely to enter and what scenarios would have the most impact with regards to human or animal health and welfare or trade implications. To address these issues, a number of relative risk ranking tools have previously been developed, such as the EU wide DISCONTOOLS (2016) and the UK specific D2R2 (Gibbens et al., 2016). However, these tools are qualitative and are generally based on chosen criteria rather than a defined quantitative assessment. There is, therefore, benefit in a quantitative model that can utilises freely available numerical data from datasets on trade and human travel such as Eurostat (2014) and FaoStat (2014).

To address this need, a generic quantitative risk assessment framework for the entry of bat-borne zoonotic viruses to the EU was developed (Simons et al., 2016), considering the pathways: human travel, live animal movement, legal trade of food products and illegal trade of bushmeat. Using current knowledge of virus characteristics such as environmental survival and host incubation periods, the framework was parameterised for NiV, to provide an assessment of the relative risks of transmission through the known pathways of introduction into the EU. In this paper the model framework is parameterised for a number of other virus outbreak scenarios (MARV, EBOV, HeV and MERS-CoV) and the relative probabilities of introduction to EU MSs are compared and discussed. The impact of uncertainty in the parameter estimates is also investigated though scenario analyses.

Materials & methods

Overview

The entry assessment model parameterised for NiV (Simons et al., 2016), was re-parameterised for five outbreak scenarios (MARV, EBOV, HeV and MERS-CoV), to compare and assess the relative risk of introduction to the 28 EU MSs for these 5 viruses of concern. For EBOV two different outbreak scenarios were considered: 1) Disease geographically distributed in Western Africa, where the human cases are on a similar scale to that observed in the 2014 West Africa outbreak (wEBOV) (i.e. epidemic situation), 2) Disease geographically distributed in Central Africa, where human outbreaks have previously been relatively limited (cEBOV) (i.e. non-epidemic situation). It is acknowledged that the link between bats and MERS-CoV is more tenuous than initially thought when it first emerged (Memish et al., 2013), but the virus was included here to provide an example of a respiratory coronavirus circulating within the Middle East area. Recent evidence of replication and shedding of MERS-CoV in experimentally infected Jamaican fruit bats (Artibeus jamaicensis) (Munster et al., 2016) and discovery of closely related MERS-like CoV (Anthony et al., 2017) further support the hypothesis that bats are ancestral reservoirs for MERS-CoV. As neither live EBOV nor MERS-CoV virus has been isolated from bats a very low prevalence of infection in bats was assumed. Different prevalence values for all the viruses were considered in the scenario analysis.

The assessment was conducted following the World Organisation for Animal Health (OIE) code for import risk analysis (OIE, 2004). Under traditional OIE guidelines, there are three components of risk assessment: entry, exposure and consequence. This model only considered the entry assessment (i.e. it ceases at the point at which virus is released into the EU) and did not consider subsequent potential exposure of the virus to humans, livestock or wildlife on entry to the EU.

Model framework

The model has been discussed in detail by Simons et al. (2016). Briefly, the main outputs of the model were a relative estimate of the annual probability of at least one introduction event into each EU MS, j, P and an overall estimate of the probability of at least one introduction event for the EU as a whole. This estimate took into account factors such as the probability an individual unit is infected (or contaminated) in the exporting country, the survival of the virus over the duration of the journey, whether or not the animal/human displays clinical signs and the annual volume of products being imported. The model equations for the various routes are described by Simons et al. (2016). The model was coded in the R software package and is deterministic; as such no stochastic variability of specific parameters was considered in the baseline model.

The relative risk estimate was derived by combining the probability of at least one introduction event from each of the routes included in the model (Fig. 1 ) from all the potential exporting countries to produce an overall probability for each MS:where R is the total number of routes considered for the virus (human travel (r = 1), live animal movement (r = 2), legal trade of ‘at risk’ products (r = 3), illegal trade of bushmeat (r = 4)) and P is the probability of at least one introduction event via route r to MS j per year. The average number of years to an introduction event was calculated, Y = 1/P. In addition the 28 EU MSs were ranked according to their probability of disease introduction, Z = {1:28} by comparing the average number of years, Y, to an introduction event for each route and MS and ranking the MSs from 1 to 28 accordingly. This provided an indication of where in the EU an introduction event is more likely. Note that the average number of years to an introduction event is based on the input data provided and so does not account for subsequent changes in future years of model factors such as trade patterns or disease prevalence; e.g. for wEBOV it would be assumed that the same number of cases will occur in Western Africa every year as in 2014.

Fig. 1

Overview of model framework up to the point of entry to the EU.

The model considered the four primary routes of introduction based on extensive literature reviews (Simons et al., 2014). While other potential routes exist such as direct exposure from bats through natural migration or accidental exposure via aeroplane strikes, they were not considered here for the viruses of concern. However, as the model framework is adaptable, and Eq. (1) is multiplicative with respect to the routes, the choice of routes can be amended and these pathways can be considered in the future, as and when appropriate data become available.

Parameterisation for individual viruses

The model was parameterised for HeV, MARV, MERS-CoV, wEBOV and cEBOV. The genus Ebolavirus includes five species, each with a single member virus (Kuhn et al., 2013). Due to the potential differences in parameters for the different viruses only Ebola virus (from species Zaire ebolavirus) was parameterised here.

The model considered the probability of introduction to EU MSs from ‘exporting countries’, that is, those in which virus was strongly expected to be circulating in humans, livestock or wildlife (Fig. 2 ). This was determined from peer-reviewed publications of where the viruses had been reported. For livestock and wildlife, including bats, only positive test results for isolation of live virus or detection of viral RNA were considered (active bat infection). Countries that had reported positive seroprevalence or those which had reported a human case known to have arisen from recent travel to another country were not considered as an ‘exporting country’.

Fig. 2

Maps highlighting the exporting countries used in the model for each virus under consideration: NiV (Bangladesh, India, Malaysia, Singapore, Cambodia, East Timor, Indonesia, Thailand); HeV (Australia); MARV (Uganda, Angola, Democratic Republic of Congo, Gabon, Kenya); MERS (Saudi Arabia, United Arab Emirates, Qatar, Jordan, Oman, Kuwait, Iran, Lebanon); wEBOV (Sierra Leone, Liberia, Guinea); cEBOV (Democratic Republic of Congo, Gabon, Republic of Congo).

The full details of the generic model parameters are presented in Simons et al. (2016). In this section an overview of the data sources used to parameterise each route is presented.

Human travel: Passenger travel data from exporting countries to EU MSs were obtained from the Eurostat dataset aviapaexcc (Eurostat, 2014). For EBOV, MARV and HeV for which outbreaks are sporadic, n was estimated using the average number of cases per outbreak over a 15 year period. This value was assumed by the authors to encompass all relevant historical data. However, as human cases of MERS-CoV have been reported regularly since March 2012, n was calculated by dividing the number of reported cases by the number of reporting years assuming a constant rate per year. To account for differences in prevalence between passenger types e.g. business, visiting family, and tourist etc., the baseline prevalence of infection in the exporting country was weighted by the average passenger duration of stay (days) in the exporting country. The ratio of passenger types was assumed to be the same for each exporting country. The sub-clinically infected population was estimated by multiplying the prevalence of infection in passenger type i, θ(i,k) by the incubation period of the virus. Passenger detail such as healthcare employees potentially exposed to infected patients or eco-tourists with the intention of visiting bat caves was not accounted for here although it is acknowledged that these factors could influence the risk outcome as has been documented (WHO, 2008).

Legal trade: To determine whether a product was considered contaminated, the concentration of virus on the product on arrival to an EU MS was estimated where C is a threshold viral load upon arrival at the EU MS, below which the product was considered not to be contaminated. Note, that this value was set to 1 log10 TCID50 for all viruses as a worst case scenario. The model considered the prevalence of contamination in at risk raw products (see Table 1 for definition), the initial concentration of virus on a raw product in the exporting country and any reduction in viral load between initial contamination of the raw product and arrival in the EU MS, including the effect of processing. The default estimate for the prevalence of contamination in raw products, pGraw(k), was based on the estimated prevalence of active virus shedding in bats, pBInf(k), the contact rate of the bat with the product, pBcontact(k) and seasonality of virus shedding, i.e. the proportion of the year that bats can shed the virus. Data on volume of trade from exporting countries to EU MSs were obtained from FaoStat (2014).

Table 1

Summary of virus specific parameter estimates for NiV, HeV, MARV, MERS-CoV and EBOV viruses (see Appendix A for further information and references).

Parameter	Values
Description	NiV	HeV	MARV	MERS-CoV	EBOV
Exporting countries with evidence of virus in human, livestock or wildlife (k)	Bangladesh, India, Malaysia, Singapore, Cambodia, East Timor, Indonesia, Thailand	Australia	Uganda, Angola, Democratic Republic of Congo, Gabon, Kenya	Saudi Arabia, United Arab Emirates, Qatar, Jordan, Oman, Kuwait, Iran, Lebanon	wEBOV (Sierra Leone, Liberia, Guinea) cEBOV(Democratic Republic of Congo, Gabon, Republic of Congo)
Estimated number of human infections in exporting country k, in one year, nHinf(k) for the scenario	Bangladesh = 27, India = 66, all other countries = 0	Australia = 1 (‘rounded up’)	DRC = 154, Uganda = 8, Angola = 374, all other countries = 0	Saudi Arabia = 340, UAE = 24, Jordan = 6, Qatar = 4, Oman =  2, all other countries = 0	West Africa = 16,125, DRC = 75, Gabon = 65, ROC = 79
Average time to clinical symptoms of the virus (days), TIP(k)	9	12.8	7	5.5	8.82
Legal Trade – of at risk products	FaoStat section 8-fruits and derived productsa; pig and pig products	FaoStat section 8-fruits and derived productsa	FaoStat section 8-fruits and derived productsa	FaoStat section 8-fruits and derived productsa; camel meat and milk	As for NiV
Prevalence of active bat infection in exporting country k, PBInf(k)	0.20%	0.47%	0.29%	0.10%	0.10%
Proportion of the year bats may shed active virus, Pseason(k)	0.33	0.33	0.5	0.33	0.5
Initial viral load on product, C0(x)∼logNormal(a,b)	Mean = 2 log10 TCID50/ml,Variance = 2.25 log10 TCID50/ml	Mean = 4.6 log10 TCID50/mlVariance = 1 log10 TCID50/ml	Mean = 3.12 log10 TCID50/mlVariance = 1 log10 TCID50/ml	Mean = 5 log10 TCID50 eq/mlVariance = 1 log10 TCID50/ml	Mean = 3 log10 TCID50/mlVariance = 1 log10 TCID50/ml
Half-life of virus in environment, pre-harvesting (h), CHLenv(k,l)	6.15	2.9	72	0.77	72
Half-life of virus during transport (4 °C) (h), CHLtrans(j,k,l,m)	308	268	144	72	168
Minimum Viral load to consider product contaminated in EU MS, Cmin	1 log10 TCID50	1 log10 TCID50	1 log10 TCID50	1 log10 TCID50	1 log10 TCID50
Live animals: animal species with evidence of infection including serology s	Non-human primate, pig, dog, cat, ferret	Pig, dog, cat, horse	Non-human primate	Dromedary camel	Non-human primate, pig, dog, duiker, rodent, shrew,
Probability bushmeat is of species s, pBMSp(s)	1.5% Bats, 98.5% other species	1.5% Bats, 98.5% other species	1.5% Bats, 6% nonhuman primates, 92.5% other species	1.5% Bats, 98.5% other species (red meat could = camel from Middle East)	1.5% Bats, 6% nonhuman primates, 75% rodents and duikers, 17.5% other species

FAO fruits and derived products see: http://www.fao.org/es/faodef/fdef08e.htm for definition and classification of commodities.

Bushmeat: In this model bushmeat was assumed to enter an EU MS via aircraft passenger luggage. The volume of contaminated bushmeat entering the EU was estimated by combining the probability of a passenger of type i bringing in bushmeat from exporting country k, p and the probability that a consignment of bushmeat was contaminated. The actual number of bushmeat consignments entering the EU from country k was estimated based on the number of bushmeat consignments seized in the EU MS, Nseized(i,j,k), and an under-reporting factor accounting for the proportion of passengers luggage that were searched. The under-reporting factor was estimated to be 0.5% based on literature (Falk et al., 2013) and assuming targeted testing of passengers occurs (Simons et al., 2016).The model did not account for any virus reduction that may occur from processing bushmeat such as smoking or salting and assumed that the possibility of luggage being searched for bushmeat was the same for each exporting country.

Live animals: This route considered the number of animals of species s arriving from an exporting country k in one year and the prevalence of live animal infection of species s in the exporting country to give the probability that at least one infected animal entered a MS. Numbers of live animal exports from exporting countries to EU MSs were obtained from the trans-European TRAde Control and Expert System (TRACES) database which provides data on the number of animals that are brought into the EU and issued with a Common Veterinary Entry Document (TRACES, 2014).

Virus specific parameter estimates used in the model are given in Table 1. Estimates for NiV are also provided for comparison (Simons et al., 2016). Further information (including references) on these estimates is provided in Appendix A.

The model developed here is deterministic for ease of use in an outbreak situation where rapid parameterisation and data availability for all EU MSs are key requirements. Uncertainty and variability in the model were previously considered for NiV by implementing a series of analyses using alternative parameter values (Simons et al., 2016). It was found that while some scenarios had an impact on the absolute values of probability of introduction of NiV, the relative rankings, of both routes and MSs were more robust. However, the estimate for the prevalence of NiV in bats had considerable impact on the average number of years to an EU introduction of NiV relative to the baseline model and much lower estimates for this prevalence were the only scenarios to have an impact on the relative ranking between the routes. Given this, and the complexity involved in assessing multiple uncertainties between multiple scenarios, the scenarios considered here were a 90% and 99% reduction in the virus prevalence in bats as these reductions both had considerable impact in the previous model for NiV; smaller reductions in virus prevalence had little impact.

Results

At the EU level the probability of viral introduction was ranked highest for wEBOV with an overall average prediction of at least one introduction event occurring in one year (Table 2 ), primarily via human travel and associated illegal importation of bushmeat.

Table 2

The expected number of years to EU entry for different viruses, by individual route and all routes combined for the baseline model. Results for 90% and 99% reduction in virus prevalence in bats are shown in brackets respectively for Legal Trade, Bushmeat and all routes (the model assumes no effect on human travel and live animal routes).

Scenario	Human travel	Legal trade	Bushmeat	Live animals	All routes
NiV	540	12 (115, 1147)	70 (682, 5915)	51,649	10   (83, 344)
HeV	3202	45 (441, 4403)	123	39,299	33
			(1220, 11,546)		(292, 1535)
MARV	18	3 (25, 242)	5 (25, 44)	295,015	2 (8, 12)
MERS-CoV	4	8.00E+11	191 (681, 917)	N/Aa	4 (4, 4)
		(7.1e12, 2.9e13)
wEBOV	1	6 (58, 578)	3 (3, 3)	923	1 (1, 1)
cEBOV	19	240	37 (60, 64)	8259	12 (14, 15)
		(2397, 23,962)

The model returned a N/A results due to the probability of introduction being too low to compute.

In relative terms, and given the uncertainties in the absolute value estimates, MARV and MERS-CoV were of a comparable risk whilst the overall probability of introduction was lowest for HeV.

A 90% or 99% reduction in virus prevalence in the exporting country bat population only affected the risk estimates for the legal trade and bushmeat routes. Consequently, for EBOV and MERS-CoV, which had a relatively high probability of introduction from human travel, the decrease in risk from trade and bushmeat was not sufficient to affect the overall probability of disease introduction. For HeV and NiV, however, where the legal trade and bushmeat routes posed the highest risk in the baseline model the decrease in overall risk was substantial. Human travel replaced legal trade and bushmeat as the route with the highest associated probability for HeV, MARV and NiV introduction when the virus prevalence in bats was reduced by 99% (90% for MARV).

The number of imports of live animals was low for all exporting countries resulting in a relatively low probability of introduction via this route for all viruses (Table 2). Only dromedary camels (Camelus dromedaries) have been shown to be a risk factor for MERS-CoV transmission (Azhar et al., 2014), but as there is no legal trade of live camelids to the EU from countries reporting cases of MERS-CoV, the risk from live animals was considered to be negligible.

Within the EU, individual MSs demonstrated different relative probabilities for the various pathogens when the probabilities for all the routes of introduction were combined for each MS (Fig. 3 ). The probability of introduction for MERS-CoV was quite high across most of the EU MSs, but for other viruses it was mainly focussed in a few MSs, usually in Western Europe with the probability of introduction for MSs from Eastern Europe and Scandinavia generally being much lower (Fig. 3). Overall, the probability of introduction was highest for individual viruses in those MSs with strong historical links to relevant exporting countries, e.g. the United Kingdom (UK) for NiV and France for cEBOV. Such links usually correspond to a relatively large volume of human travel or legal trade movements between the countries. It should be noted that this analysis does not consider movement within the EU after the initial entry.

Fig. 3

Average number of years until an introduction event to EU MSs for different viruses; clockwise from top left; NiV, HeV, MERS-CoV, cEBOV, wEBOV and MARV across all routes. Scale shows increasing number of years until an introduction event from left (dark red) to right (light green). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

The countries are ranked according to the probability of introduction for each virus in Table 3 . Overall there was a relatively wide variation in the relative ranking of many of the MSs between the different viruses. Different distributions of risk scores were observed between routes but considering the relative ranking of the MSs (1–28), the UK, France, Germany and the Netherlands generally have the highest probability of introduction for all viruses considered here.

Table 3

Relative ranking of EU MSs by expected number of years until entry of virus. Minimum, maximum and range in EU MS ranking across all viruses are shown. Columns are highlighted with lower ranking or smaller range in ranking of EU MS having darker grey shades.

Discussion

The entry assessment described here shows the potential for application of a quantitative model framework for any pathogens, using zoonotic bat-borne viruses as an example. Although a scarcity of data for virus specific parameters resulted in a high degree of uncertainty in the absolute risk values presented, the main strengths of this model lie in the estimates of relative risks between routes of entry and those MSs which are at greater risk of virus introduction. The model has the capability to allow for a continual updating of the risk estimate using new research data as, and when, it becomes available. Any increase in the model risk estimate output would allow the stakeholder to consider employing suitable risk reduction strategies or heightened surveillance providing a rapid and cost-effective response.

Of particular value was the model's ability to illustrate the relative importance of the different routes of entry between viruses; legal trade of foodstuffs was more important for HeV, NiV and MARV while human travel was more important for MERS-CoV and both EBOV scenarios. These differences could be partly attributed to the virus specific parameters. For example rapid decay of MERS-CoV influenced the relative risk of the transmission pathways; the half-life of MERS-CoV is very short compared to the other viruses (48 min at pre-harvest temperatures (van Doremalen et al., 2013) so it is unlikely to persist in high numbers on any produce imported via legal trade or in contaminated bushmeat.

The probability of introduction to the EU via the pathways under consideration here varies across the EU at MS level; the UK, France, Germany and the Netherlands often had the highest probability of introduction for all viruses considered. In general those countries which ranked the highest, with regard to probability of introduction (Table 3) corresponded to those with the highest population and the highest ‘disposable income’ (calculated as gross domestic product (GDP) derived from purchasing power parity) (see Appendix B). Other contributory factors could include immigration population densities and trade partner characteristics, both of which frequently have a historical basis. The Netherlands was an exception to this in that it ranked highly in the probability of virus introduction yet only 8th for population density and 7th for GDP (See Appendix B). One explanation for this could be that the Netherlands is serving as a hub for travellers and trade entering the EU and that a reasonable proportion entering the Netherlands are going onto other European countries. It is also possible that more Dutch people, compared to other EU MSs, travel to countries with these viruses. Data from Uganda suggest that in 2012 the Netherlands was the 15th most popular country of origin for tourist arrivals, the only European countries with more arrivals were the United Kingdom and Germany, but Dutch tourists represented a higher proportion of their population (Republic of Uganda, 2013).

Freely available statistical data on trade and human travel for Western Europe were, in general, more complete than for Eastern European MSs. Data for countries such as Estonia, Latvia and Lithuania were lacking for some routes resulting in a low ranking for these countries which may not be a true reflection of the actual risk. It is difficult to determine whether this is a true data gap or if the route genuinely has a low probability of introduction for these countries. This was a particular problem for cEBOV where twelve countries were lacking data for specific routes and, therefore, equally ranked as 22.5.

The generic parameters for which EU wide datasets exist have a relatively high degree of completeness although there is a concern that potentially high risk low volume trade products e.g. camel milk may be under-recorded therefore underestimating the risk via these trade products. Virus specific parameters depend more upon focussed research studies and peer reviewed literature and rely upon detection of pathogens in reporting countries. Uncertainty in these virus parameters, in particular, the prevalence of infection in bats in the exporting country, and viral persistence during processing and storage may limit the application of the model by introducing considerable uncertainty. The sensitivity analysis of virus prevalence in bats demonstrated that the results for the relative importance of the routes for EBOV and MERS were quite robust with human travel remaining the route with the highest probability of introduction. With regards to HeV, NiV and MARV, however the sensitivity to the variation in prevalence indicates that further data for this particular parameter would strengthen the model results; this is particularly true of MARV where a 90% reduction in virus prevalence changed the risk ranking order of the routes of introduction. Note that this analysis is to consider the uncertainty about the true prevalence in bats, as such it has no impact on the parameterisation of the human prevalence, which is based on human outbreak data.

It should be noted that the parameterisation of this model uses the best available data at the current time. Some parameters are subject to high uncertainty and the probability of introduction of different viruses will be dynamic, changing over time if a virus spreads amongst different animal species populations or if new human outbreaks occur. Simons et al. previously demonstrated that changes in the exporting country (e.g. if China were to get NiV in the future) or ‘at risk’ product types can have a large effect on the model outputs (Simons et al., 2016). We have demonstrated here that changes in the virus prevalence in bats in the exporting country can have an impact on the average number of years to EU entry for the different viruses and on the relative ranking of the individual routes of entry. We have also highlighted the differences in probabilities for two Ebola scenarios; a relatively small non-epidemic human outbreak and a large epidemic scale outbreak. It is acknowledged that the viruses considered here could have differing sensitivity to stochastic variability of specific parameters given the complex dynamics between the routes and viruses. Alternative scenarios could, therefore, be considered in the future.

All risk pathways were given equal weight within the model as the model predicts probability of introduction not risk of human/animal exposure and consequence as stated in the OIE risk assessment (OIE, 2004). For example, the model results suggest that the legal trade (fruit) route has a high probability of introduction for HeV, although human infection from consumption of contaminated fruit is not a proven transmission route for this virus. This route was considered in the model based on the knowledge that fruit bats are known to consume raw fruit in orchards (Eby and Lunney, 2002) and date palm sap is a known route of transmission for NiV (another henipavirus) (Luby et al., 2006, Khan et al., 2012, Nahar et al., 2014). Similarly, there is currently no evidence of human-to-human transmission of HeV but as this has occurred for the Bangladesh strain of NiV (Gurley et al., 2007) it is plausible that this could occur with or without mutation and adaptation of currently identified strains. Real-time application of the model would allow for removal or addition of pathways if future scientific work provides suitable input data or if trade patterns between third countries and the EU alter. Thus, all pathways were assessed for completeness according to the dogma ‘absence of evidence is not evidence of absence’.

Whilst EU wide trade controls are implicitly accounted for within the model parameters, risk mitigation procedures put in place by individual MSs such as targeted sampling are not taken into account. It is possible that there have already been introduction events of the diseases under consideration here within the EU, but these have remained undetected due to lack of subsequent human/animal infection and/or onward transmission within the individual MS. For example, although the importation of MERS-CoV cases to the EU remains possible, an ECDC risk assessment determined that the risk of sustained human-to-human transmission is low (ECDC, 2015a). However, the outbreak of MERS-CoV in South Korea demonstrated that the potential exists for a serious risk of onward human spread, with >185 cases arising from the importation of 1 human index case (Su et al., 2015).

Validation of such a model presented here is difficult as there are few independent resources for which to compare the results. However, it is of relevance that the five MSs suggested to have the highest probability of introduction of MERS-CoV by the model (Germany, UK, Italy, France and the Netherlands) have already had imported human cases of this pathogen (ECDC, 2014). The model results are also consistent with other reports which predict more imported cases of MERS-CoV to arrive into the EU (ECDC 2014, WHO 2014a, Bialek et al., 2014, Poletto et al., 2014). All cases reported outside of the Middle East have had a recent travel history to the Middle East or contact with a case that had travelled from this region (Su et al., 2015). This is in line with the highest probability of introduction for MERS-CoV predicted by this model to be via human travel (Table 2).

Overall, the approach developed here provides a high-level horizon scanning tool for the probability of introduction of bat-borne zoonotic viruses into the EU. The virus scenario with the highest probability was the wEBOV scenario with an overall average prediction of just under one introduction event per year, primarily via human travel. Due to the wide scope of the model, which necessitated using global datasets sometimes with incomplete data, there was a high degree of uncertainty in the absolute risk values presented. A general lack of data on virus specific parameters also contributed to this uncertainty. Thus, the main strengths of this model lie in the comparison of the relative risks between viruses and routes of entry. Whilst there have been several risk assessments carried out for the introduction of individual pathogens into the EU (Rolin et al., 2013, Durand et al., 2013, Mur et al., 2014, Snary et al., 2012) this model was able to assess a range of viruses and could be adapted for other pathogens, as it has the advantage of easy access to a number of relevant databases. The model also allows for a continual updating of the risk estimate enabling the stakeholder to respond in a rapid and risk appropriate manner, for example, by implementing risk-based surveillance and control strategies.

Funding

This work was funded by the under ANTIGONE (Global Onset of Novel Epidemics) with project number 278976. Additional funding was provided by the UK Department for Environment, Food and Rural Affairs () under project SE4213.

Table A1

Historical review of EBOV outbreaks to the present day (as of November 2016).

Date	Country	Number of cases	Number dead	Strain	Likely source	Exposure to:	Human-to human transmission?
1972	DRC (Zaire)	2	1	Zaire?		Retrospective identification from 1977 case
1976	DRC (Zaire)	318	280	Zaire	Index case had bought antelope and monkey bushmeat prior to infection	Infected needles, nosocomial infections	Not with ease
1976	Sudan	284	151	Sudan	Original cases in factory - not related to exposure to wild living animals	High number of nosocomial infections	Not with ease
1977	DRC (Zaire)	1	1	Zaire	No overt link to 1976 outbreak	1 fatal case 3 unrelated and unconfirmed cases
1979	Sudan	34	22	Sudan	Index case in same factory as 1976		Not with ease
1980	Kenya	1	0	Zaire	Near Mount Elgon	13 yr old girl unknown source - no virus isolated but elevated Ab titre	No secondary transmission
1989/90	USA	0	0	Reston	Cynomolgus monkeys imported from Philippines	4 animal handlers infected but no symptoms
1992	Cote D'Ivoire	0	0	Cote D'Ivoire	Chimpanzee deaths in wild
1992	Italy	0	0	Reston	Cynomolgus monkeys imported from Philippines
1994/95	Cote D'Ivoire	1	0	Cote D'Ivoire	Chimpanzee deaths in wild and infection in human performing autopsy
1994	Gabon	52	31	Zaire	Exposure to dead Chimpanzee?	Deaths in various gold mining camps in rain forest
1995	DRC	315	254	Zaire	Charcoal worker/farmer 1st case		Nosocomial infection and relatives
1996	Gabon	31	21	Zaire	Dead chimpanzee in forest was eaten by hunters	19 human cases directly infected	Family members
1996	Gabon	60	45	Zaire	Dead chimp found to also be infected	Hunter at logging camp	Yes
1996	South Africa	2	1	Zaire	Healthcare worker travelled from Gabon to S. Africa		Transmission to a nurse who died
1996	USA	0	0	Reston	Cynomolgus monkeys imported from Philippines
2000	Uganda	425	224	Sudan	Index cases had attended burials prior to infection		Nosocomial infection high numbers
2001/2002	Gabon	65	53	Zaire	Unusually high number of animals found dead in rainforest mainly NHP	Same outbreak over the border. Epidemiological evidence of 6 different introductions of Ebola virus each related to a hunting episode	At least 2 duikers,2 chimps and 2 gorilla carcasses were suspected of involvement in infection of 6 human index patients.
2001/2002	Republic of Congo	59	44	Zaire	Unusually high number of animals found dead in rainforest mainly NHP	Index cases reported contact with NHP, duikers and porcupines. Ebov was detected in gorilla carcass butchered by index case	At least 2 duikers,2 chimps and 2 gorilla carcasses were suspected of involvement in infection of 6 human index patients.
2002/2003	Republic of Congo	143	128		Gorillas and duikers suspected of infecting 3 human index patients.
2003	Republic of Congo	35	29		Poaching though source of infection not clearly identified
2004	Sudan	17	7	Sudan		Simultaneous outbreaks of measles
2005	DRC	12	10	Zaire
2007	DRC	264	187	Zaire	Preceded by massive fruit bat migration which was hunted by villagers	Putative index case bought freshly killed bats from hunters
2007/2008	Uganda	149	37	Bundibugyo
2008/2009	DRC	32	14	Zaire	Index case believed to be girl who died from post-abortion haemorrhage
2011	Uganda	1	1	Sudan
2012	Uganda	24	17	Sudan		similar to 2000
2012	DRC	57	29	Bundibugyo		Similar to 2007
2012/2013	Uganda	7	4	Sudan		Similar to 2011
2014	West Africa	28,616	11,310	Zaire	Hunting/child bitten by bat		Yes - high percentage of nosocomial transmission
2014	DRC	71	43	Zaire	Preparation of bushmeat	Most closely related to 1995 strain

Table A2

Adapted table from Van Kerkhove et al. (2015) showing estimated time to clinical signs during EBOV outbreaks.

Year	Virus	Estimate	Range	Study number	Ref
1976	Zaire	6.3		318	(Commission, 1978)
1976	Zaire	5.99	5.8–6.18	262	(Camacho et al., 2014)
1995	Zaire	7	1–15	27	(Dowell et al., 1999)
1995	Zaire	6.2	5–8	5	(Bwaka et al., 1999)
1995	Zaire	5.3	–	315	(Chowell et al., 2004)
1995	Zaire	10	–	291	(Lekone and Finkenstadt, 2006)
1995	Zaire	12.7	–	23	(Eichner et al., 2011)
1995	Zaire	7.8	2–19	23	(Ndambi et al., 1999)
2000	Sudan	12	2–21	425	(Okware et al., 2002)
2000	Sudan	3.35	–	425	(Chowell et al., 2004)
2000–01	Sudan	12	1–12	425	(Francesconi et al., 2003)
2007	Bundibugyo	6.3	–	56	(MacNeil et al., 2010)
2007	Bundibugyo	7	2–20	192	(Wamala et al., 2010)
2014–15	Zaire	9.31	2–21	20	(Althaus et al., 2015)
2014–15	Zaire	9.4	–	500	(Team, 2014)
2014–15	Zaire	11.4	–	155	(Team, 2014)
2014–15	Zaire	9	–	1798	(Team, 2015)
2014–15	Zaire	9.9	9–11	193	(Faye et al., 2015)
2014–15	Zaire	12	–	–	(Rivers et al., 2014)
2014–15	Zaire	10	–	–	(Rivers et al., 2014)
Total	All	8.6
Total	Zaire	8.82

Table A3

Detection of EBOV in bats.

Positive bat species	Country	Sample taken	Test	Number tested	Number positive	Number shedding	Prevalence	Ref.
Unknown	DRC (Zaire)	Spleen, liver, kidney, heart	Virus isolation	8	0	0	0	(Germain, 1976)
Numerous	DRC	Liver, kidney and spleen/serum	Virus isolation and IFA	463	0	0	0	(Breman et al., 1999)
Unknown	DRC	Liver, spleen/Serum	Virus isolation/ELISA	539	0	0	0	(Leirs et al., 1999)
Epomops franqueti, Myonycteris torquata, Epomophorus gambianus, Micropteropus pusillus	CAR	Spleen, liver, kidney	RT-PCR (virus isolation only carried out on RT-PCR positive)	23	0	0	0	(Morvan et al., 1999)
Epomops franqueti, Hypsignathus monstrosus, Myonycteris torquata,	Gabon/ROC	Serum/liver & spleen	ELISA/RT-PCR	679	16/13	Not attempted	–	(Leroy et al., 2005)
Epomops franqueti, Hypsignathus monstrosus, Myonycteris torquata,	Gabon/Congo	Serum	IgG ELISA	1390	40	Not attempted	–	(Pourrut et al., 2007)
Epomops franqueti, Hypsignathus monstrosus, Myonycteris torquata, Micropteropus pusillus, Mops condylurus, Rousettus aegyptiacus	Gabon/Congo	Serum/liver & spleen	ELISA/ RT-PCR	1468	95/0	Not attempted	–	(Pourrut et al., 2009)
Eidolon helvum	Ghana	Serum	Indirect fluorescent + western blotting (insufficient material for RT-PCR)	262	1 (Zaire)	Not attempted	–	(Hayman et al., 2010)
Epomops franqueti, Hypsignathus monstrosus, Epomophorus gambianus	Ghana	Serum	ELISA WB	88	5 (Zaire)	Not attempted	–	(Hayman et al., 2012)
Various	China	Serum/pharyngeal & faecal swabs	ELISA/RT-PCR	843/143	10/0 (Zaire)	Not attempted	–	(Yuan et al., 2012)
Rousettus leschenaultii, Cynopterus sp, Megaderma lyra, Macroglossus sobrinus	Bangladesh	Serum/throat, urine/faecal swab	ELISA WB/ RT-PCR	273	5 (R. leschenaultii)by ELISA none by PCR	Not attempted	–	(Olival et al., 2013)

Table A4

Historical review of HeV human cases.

Date	Country	Number of cases	Number dead	Likely source	Exposure to	Human-to human transmission?
1994	Queensland	1	1	Infected horse	Farmer assisted in autopsy of horse. Died 13 months post infection	No
1994	Queensland	2	1	Infected horse	Death of horse trainer and severe illness in stable-hand both with close contact with sick horses	No
2004	Queensland	1	0	Infected horse	Veterinarian tested positive for Hendra virus after performing a post mortem	No
2008	Queensland	2	1	Infected horse	Veterinarian and veterinary nurse were infected after close contact with sick horse. The vet died.	No
2009	Queensland	1	1	Infected horse	Veterinarian died after exposure to Hendra infected horse	No

Table A5

Estimates of average times to clinical symptoms for human HeV cases (days).

Patient	Average time to clinical symptoms (days)	Ref
Patient 1 1994	No accurate data
Patient 1 1995	7	(Selvey et al., 1995)
Patient 2 1995	8	(Selvey et al., 1995)
Patient 1 2004	7	(Hanna et al., 2006)
Patient 1 2008	9 or 16	(Playford et al., 2010)
Patient 2 2008	11	(Playford et al., 2010)
Patient 1 2009	21 (19)a	(Ausvet, 2009)

Patient received antiviral treatment which may delay symptoms 1–2 days.

Table A6

Detection of HeV in bats.

Positive bat species	Country	Sample taken	Test	Number tested	Number positive	Number shedding	Prevalence	Ref.
P. alecto, P. poliocephalus, P. conspicillatus, P. scapulatus	Australia	Uterine fluids	Virus isolation	4	4	4	targeted surveillance no mention of how many	(Halpin et al., 1996)
Pteropus poliocephalus and Pteropus alecto	Australia	Tissue samples	Virus isolation	465	2	2	sampling of recently captured sick or injured wild bats	(Halpin et al., 2000)
P. alecto, P. poliocephalus, P. conspicillatus, P. scapulatus	Australia	Pooled urine	RT-PCR	1672	45a	–		(Field et al., 2011)
P. alecto, P. poliocephalus, P. conspicillatus	Australia	Pooled Urine	Virus isolation	45a samples +ve by q-PCR; 30 of these picked for isolation	4	4		(Smith et al., 2011)
Chalinobus, Miniopterus australis, Nyctophilus, P. alecto, P. poliocephalus, P. conspicillatus, P. scapulatus, Saccolaimus flaviventris, Scotorepens, Synconycterus	Australia	Tissues & serum	RT-qPCR	310	20	_		(Goldspink et al., 2015)

samples are the same reported in different articles.

Table A7

Historical review of MARV virus outbreaks to the present day (as of November 2014) (the 2 Koltsovo laboratory infections which occurred in the former Soviet Union have not been included here).

Date	Country	Number of cases	Number dead	Likely source	Exposure to	Human-to human transmission?	Ref.
1967	Europe	31	7	Imported African green monkeys (Chlorocebus aethiops) from Uganda	Blood, organs, cell cultures	Yes	(Kissling et al., 1970)
1975	South Africa	3	1	Unknown - possibly from Zimbabwe	Visited Sinoia caves 8–9 days prior to onset of symptoms	Yes	(Gear et al., 1975)
1980	Kenya	2	1	Kitum Cave (<70 miles from Lake Kyoga where 1967 monkeys originated)	Possible bat excretions	Yes	(Smith et al., 1982)
1987	Kenya	1	1	Kitum Cave (<70 miles from Lake Kyoga where 1967 monkeys from)	Possible bat excretions	No	(Johnson et al., 1996)
1998–2000	DRC	154	128	Mine workers in Goroumbwa cave	Possible bat excretions	Yes	(Bausch et al., 2006)
2004–2005	Angola	374	329	Unknown	Mostly index cases were children possibly from administration of vaccine using contaminated equipment	Yes	(Fisher-Hoch, 2005)
2007	Uganda	4	2	Mine workers in Kitaka cave	Possible bat excretions	Possibly	(Towner et al., 2007)
2008	USA/Netherlands	2	1	Visit to Python cave in Maramagambo Forest	Possible bat excretions	No	(WHO, 2008)
2012	Uganda	20	9	Same strain as 2007 outbreak	99.3% similar to sequence from bat	Yes	(Amman et al., 2014)
2014	Uganda	1	1	No consumption of bushmeat or contact with bats	Healthcare worker	No	(WHO, 2014b)

Table A8

Detection of MARV in bats.

Positive bat species	Country	Sample taken	Number tested	Number shedding	Prevalence	Concentration	Ref.
Rousettus aegyptiacus	Uganda	liver/spleen tissue	1622 (40 RT-PCR pos)	7	0.40%	∼(>2000 TCID50/ml)	(Amman et al., 2012)
Rousettus aegyptiacus	Uganda	liver/spleen tissue	611 (31 RT-PCR pos)	5	0.80%	1 × 105 pfu/ml	(Towner et al., 2009)
Rhinolophus eloquens, Rousettus aegyptiacus, Miniopterus inflatus	Democratic Republic of the Congo	pooled tissue	381 (12 RT-PCR pos)	0	0		(Swanepoel et al., 2007),
Rousettus aegyptiacus	Gabon & Republic of Congo	liver/spleen tissue	1138 (4 RT-PCR pos)	0	0		(Towner et al., 2007)
Rousettus aegyptiacus	Gabon	liver/spleen tissue	1257 (9 RT-PCR pos)	No virus isolation attempted due to low viral load	–		(Maganga et al., 2011)
Rousettus aegyptiacus	Kenya	faecal & oral swabs/liver, spleen & lung	272 (1 RT-PCR pos)	No virus isolation attempted due to low viral load	–		(Kuzmin et al., 2010)
Rousettus aegyptiacus, Hypsignathus monstrosus, Epomops franqueti, Micropteropus pusillus	Gabon/ROC	liver/spleen tissue	1438 (0 RT-PCR pos)	No virus isolation attempted due to low viral load	–		(Pourrut et al., 2009)
Rousettus aegyptiacus	Uganda	liver/spleen tissue	400 (53 RT-PCR pos)	9	2.25%		(Amman et al., 2014)

Table A9

Global incidence of laboratory confirmed MERS-CoV cases as of 18th November 2016.

Date of onset/most recent case	Country	Number of cases	Number dead
18/11/2016	Saudi Arabia	1484	617
16/06/2016	UAE	84	12
13/06/2016	Qatar	16	5
23/09/2016	Jordan	35	14
31/05/2015	Oman	6	3
19/09/2015	Kuwait	4	2
22/04/2014	Egypt	1	0
17/03/2014	Yemen	1	1
22/04/2014	Lebanon	1	0
18/03/2015	Iran	6	2
25/09/2014	Turkey	1	1
12/09/2016	Austria	2	0
06/02/2013	UK	4	3
07/03/2015	Germany	3	2
08/05/2013	France	2	1
27/05/2013	Italy	1	0
08/04/2014	Greece	1	1
05/05/2014	The Netherlands	2	0
16/05/2013	Tunisia	3	1
23/05/2014	Algeria	2	1
09/04/2014	Malaysia	1	1
01/02/2015	Philippines	3	0
01/05/2014	United States of America	2	0
02/07/2015	South Korea	185	36
30/05/2015	China	1	0
30/07/2016	Thailand	3	0

Table A10

Detection of MERS-CoV in bats.

Positive bat species	Country	Sample taken	Test	Number tested	Number positive	Number shedding	Ref.
Rhinopoma hardwickii, Rhinopoma microphyllum, Taphozous perforatus, Pipistrellus kuhlii, Eptesicus bottae, Eidolon helvum, and Rosettus aegyptiacus	Saudi Arabia	Throat swab, faeces, urine, serum	PCR	110 individual bats and 732 roost faeces samples	1	0	(Memish et al., 2013)

Product	Description	Amount
Milk and Cream	UHT, Sterilisation, HTST	1600 Kg
Powder form, containing added sugar	UHT, Sterilisation, HTST	30 Kg
Added sugar	UHT, Sterilisation, HTST	200 Kg

Table B1

Total population and gross domestic product (GDP) derived from purchasing power parity (PPP).

MS (ISO3)	Population	Population (rank)	GDP(PPP)	GDP(PPP) (rank)
DEU	80,889,505	1	3,704,911	1
FRA	66,206,930	2	2,571,970	2
GBR	64,510,376	3	2,565,070	3
ITA	61,336,387	4	2,128,762	4
ESP	46,404,602	5	1,541,156	5
POL	37,995,529	6	940,179	6
ROU	19,910,995	7	803,313	11
NLD	16,854,183	8	477,949	7
BEL	11,225,207	9	437,803	8
GRC	10,957,740	10	394,485	14
CZE	10,510,566	11	386,300	12
PRT	10,397,393	12	319,599	13
HUN	9,861,673	13	295,209	16
SWE	9,689,555	14	283,555	9
AUT	8,534,492	15	253,309	10
BGR	7,226,291	16	243,786	20
DNK	5,639,565	17	224,893	15
FIN	5,463,596	18	218,441	18
SVK	5,418,506	19	150,155	19
IRL	4,612,719	20	120,040	17
HRV	4,236,400	21	89,897	21
LTU	2,929,323	22	78,336	22
SVN	2,062,218	23	61,790	23
LVA	1,990,351	24	54,307	25
EST	1,313,645	25	45,525	26
CYP	1,153,658	26	35,397	27
LUX	556,074	27	26,368	24
MLT	427,404	28	12,332	28

a Worldbank data 2014 http://data.worldbank.org/data-catalog/Population-ranking-table.

b Worldbank data 2014 in millions of international dollars http://databank.worldbank.org/data/download/GDP_PPP.pdf.