European Journal of Wildlife Research

, Volume 57, Issue 5, pp 1065–1075 | Cite as

Wildtool, a flexible, first-line risk assessment system for wildlife-borne pathogens

  • Paul Tavernier
  • Jeroen Dewulf
  • Sophie Roelandt
  • Stefan Roels
Original Paper

Abstract

We describe the prototype of an electronic tool for risk assessment with dynamic ranking of wildlife-borne pathogens in function of their need for surveillance. Data about pathogens, their hosts and occurrences are obtained from literature and are classified as qualitative scores under six main criteria with their sub-criteria, corresponding to the elements of a standard risk assessment. Pathogen-specific data are reviewed by experts. The information is processed per pathogen through an algorithm and through summing up of the values obtained by converting four-tiered qualitative sub-criteria scores to weighted five-tiered numerical values. For a consistent comparison between pathogens, the “unknown” sub-criteria scores are assigned a median value of 3, allowing preservation of the sub-criteria concerned and their weights for the risk assessment, but minimizing the effect of this score on the outcome. Irregular data availability is further accommodated by a different data processing for comprehensiveness and refinement requirements, which is realised by a respective first- and second-level ranking of pathogens, the latter using additional quantitative and qualitative data for the release assessment. Continuous data updates are necessary to reflect the current situation in the field. Output flexibility is implemented by the possibility to run queries based on the choice of a region, a specific target group susceptible to the pathogens and a set of weights for the sub-criteria.

Keywords

Wildlife pathogens Risk assessment Prioritization Surveillance Wildtool 

Introduction

Disease outbreaks in man and animals are associated worldwide with pathogens emanating from wildlife (Kuiken et al. 2005). In the last decades, socio-economic and environmental changes have contributed substantially to the increased probability of such events (Jones et al. 2008; Williams et al. 2002). Consequently, the need for a continuous risk assessment has increased as well, for example in order to delineate surveillance goals or to direct the organisation of surveillance networks for wildlife-borne pathogens.

At national level, an integrated organization of sampling, diagnosis and reporting of wildlife diseases has become necessary, not only for reasons of public and animal health, but also to allow international communication, including the annual reporting of notifiable diseases to the World Organisation for Animal Health (OIE). In many European countries, there is a clear interest in surveillance in wildlife but national programs are generally lacking (Ryser-Degiorgis et al. 2009). Insufficient awareness, budgetary limitations and the complexity of administrative procedures are main obstructions towards the realisation of national surveillance schemes in wildlife. This is particularly the case in federally structured countries, where the sharing of competences between regional and federal levels adds to this complexity (Doherty 2000). In this context, active surveillance has to be restricted to a number of prioritized pathogens and a first-line methodology for risk assessment is required to determine at each moment where available resources should be allocated to initiate more detailed assessments (McKenzie et al. 2007).

Among other characteristics, a “decision support system” is often dealing with unstructured or poorly defined problems and should offer flexibility to accommodate changing circumstances (Thrusfield 2007a). These two characteristics are particularly relevant for a first-line risk assessment of wildlife-borne pathogens. The system should also meet a number of requirements. Full transparency of methodology and data recording is needed (Doherty 2000). Furthermore, the system should be performant enough to allow comparison of pathogens based on available information, even if the latter is incomplete, which is the case for many wildlife-borne pathogens. Next, a first-line tool should be sufficiently comprehensive to include the broadest range of infectious risks from wildlife for which information is available. Finally, the scoring process should be refined to a maximum in order to obtain a risk assessment as accurate as possible and corresponding to the situation in the field.

Due to the scarcity of data about many wildlife-borne pathogens, a multi-purpose risk assessment system, using the available data for different objectives, is desirable. A satisfying answer to these two main challenges, being the irregular data availability and the intended multi-purpose nature of the system, is difficult to find in existing risk assessment methods for wildlife-borne pathogens since they mostly focus on specific pathogens or wild animal species. A comparative risk assessment between pathogens is seldom addressed and, if it is, important information such as detailed data about the presence of particular wild animal species, determining the release assessment of each pathogen, are not taken into account. Furthermore, criteria for which no information is available for evaluation in certain pathogens are often eliminated or overestimated, causing an inconsistent comparison between pathogens.

Within the context of the Belgian federal government project “Wildsurv” a dynamic risk assessment system for pathogens in wildlife was developed. The aim of this paper is to present a prototype of this system that was named Wildtool.

Methods

Definitions

According to Thrusfield (2007b), a host is defined as a wild animal capable of being infected and give sustenance to an infectious agent. Replication or development of the agent usually occurs in the host. We define target species as humans or domestic animal species that are reported to be affected by the pathogens of concern and hereby experience an adverse impact to a greater or lesser extent. Wildlife species can be susceptible to the pathogens they carry and can be considered as targets as well. The target species are classified in four main target groups, including man, production animals, companion animals and wildlife. End users are defined as the responsible decision makers at various official departments but include also interest groups such as conservationists or hunters organizations.

Risk assessment

The methodology of Wildtool is based on the risk analysis concept for import of animal diseases as described in the Terrestrial Animal Health Code of the OIE (2008a). It was developed by the Wildsurv cooperators and supervisory committee with input from internal and external reviewers. Only the first two parts of a standard risk analysis, being the hazard identification and the risk assessment are carried out in this first-line system (Cf. “Discussion”).

A broad hazard identification (HI) selects the pathogens that are taken into account for the prioritization procedure. It includes those pathogens causing the diseases that are listed in the 2008 OIE notification form “Wildlife” (OIE 2008b). However, any other pathogen can be included as well, from the moment that information is found in the literature about a possible release from wildlife and a possible exposure of target species in the territory of concern. Consequently, the HI consists in a non-restrictive list of pathogens. The current version of Wildtool is limited to pathogens originating from wild mammal and bird species, those from lower vertebrates are not included as yet.

The risk evaluation includes a release, exposure and consequence assessment. The release assessment evaluates the presence of the respective pathogens in wildlife or the probability of their introduction on the Belgian territory through wildlife. For the release assessment, data about the presence of wildlife hosts for the respective pathogens on the territory of concern and data about occurrences of the pathogens in wildlife are used. The exposure and the consequence assessment are carried out separately for the various target groups. The exposure assessment evaluates the probability for different target groups to be exposed to pathogens from wildlife. The consequence assessment evaluates the impact that the pathogens can have on different target groups. The final risk estimation integrates for each pathogen the results of the foregoing steps (OIE 2008a) and allows ranking the pathogens in function of their risk and their consequential need for surveillance in wildlife.

Data collection

Firstly, the Wildsurv cooperators collected data from the literature about the pathogens included in the hazard identification and about their hosts, with the preliminary objective to find out which kind of information was currently available for evaluation of the elements of an OIE-type risk assessment. The kind of information that was most frequently mentioned for many pathogens determined the choice of the criteria, the sub-criteria and the detailed items composing the sub-criteria (Appendix 1). After discussion and revision of the preliminary sub-criteria list with external reviewers, the final list was agreed on. The way of scoring and processing of the scores was approved accordingly.

Data collected by the database administrators were classified as qualitative scores, according to six main criteria with their sub-criteria, which correspond to the elements of a standard OIE risk assessment (Table 1). The main criteria determine the pathway of the data processing as represented in Fig. 1.
Table 1

Elements of risk evaluation, main criteria, data sources and types of scores

Corresponding part of risk evaluation

Main criteria (data sources)

Types of scores

Release assessment

Host presence (literature)

Y/N

Subcriteria scores (1–100)

Release assessment

Vector presence (expert consultation)

Y/N

Subcriteria scores (1–100)

Release assessment

Occurrence in wildlife (literature and national and international reports)

Y/N

Subcriteria scores (1–100)

Release + exposure assessment

Transmission (literature + expert consultation)

Subcriteria scores (H/M/L/NE/NA)

Exposure assessment

Occurrence in targets (literature and national and international reports)

Y/N

Consequence assessment

Impact (literature + expert consultation)

Subcriteria scores (H/M/L/NE/NA)

Fig. 1

Algorithm, first- and second-level ranking of each pathogen. (Y yes, N no, PA pathogen, “Region” region chosen by end user)

For the main criterion “Host presence”, all the wild living mammal and bird species known to occur on the Belgian territory were included in the database. Based on biological literature (Burfield and Van Bommel 2004; Devos et al. 2005; Lamotte 2006; Libois 2006; Peeters 2007; Vlaamse Avifauna Comissie 1989; Verkem et al. 2003; Vermeersch et al. 2004; Weiserbs and Jacob 2007), the presence of each species in the regions of concern is scored simply with “Yes” or “No”. The five sub-criteria for “Host presence” contain detailed qualitative and quantitative data about population numbers, geographical and temporal distributions, trends and migration patterns (Table 2). These data, obtained by standard methods, are mainly known for bird species and far less for mammal species for which only presence, distribution data and rough population estimates are available. Vectors are not considered as “hosts” for the various pathogens, therefore information about vectors is listed separately under the main criterion “Vector presence”. Data for this main criterion were obtained directly from experts involved in the inventory of arthropod vectors for infectious diseases.
Table 2

Subcriteria host presence, vector presence and occurrence in wildlife for second-level ranking

Host presence

Number of months present in “region” (1–12)

(1/12 to 12/12) × 100

Migrating (Y or N)

100 or 1

% UTM squares occupied in “Region” (host distribution)

1–100

Fraction of total European population in “Region”

n × 100/total European population

Increasing trend in “region” (Y or N)

100 or 1

Vector presence (only if vector-borne pathogen)

Vector presence in Belgium (Y or N)

100 or 0

Vector present in “region” (Y or N)

100 or 0

Occurrence in wildlife

Number of European countries in which reported in wildlife

n × 100/total European countries

Number of neigbouring countries (including Belgium) in which reported in wildlife

n × 100/total neighbouring countries

Most recent year of reporting in wildlife anywhere in Europe: subtract from current year: 0–1/2–5/6–10/11–20/20–100 years ago

100/80/60/40/20

Most recent year of reporting in neighbouring countries (including Belgium) in wildlife: subtract from current year: 0–1/2–5/6–10/11–20/20–100 years ago

100/80/60/40/20

Y yes, N no, UTM universal transverse mercator, “Region” region chosen by end user

“Occurrence in wildlife” and “Occurrence in targets” data are obtained from literature (case reports, incidences, direct and indirect prevalences from Belgium and Europe), national (Federal Agency for the Safety of the Food Chain 2010; Ducoffre 2010) and international reports (WAHID 2010; Handistatus II 2010). Besides simple Yes/No scores, refined information about “Occurrence in wildlife” is recorded under four sub-criteria including the year of occurrence and the number of European or neighbouring countries in which occurrences in wildlife were reported (Table 2). The latter sub-criteria were chosen as a means to obtain a standard measure for the frequency of occurrence of a given agent. One single occurrence is often reported in more than one publication and on the other hand it is not always clear whether multiple-linked occurrences in one area should be considered as such, or rather as a measure for the potential of spread of the agent and thus as one occurrence. By adding up the number of countries in which at least one occurrence is reported, double or multiple reporting is eliminated through which a more objective measure for the frequency of reporting is obtained, sufficient for the demands of a “first-line” system.

Eleven “Impact” and nine “Transmission” sub-criteria were retained (Table 3), some of which were further subdivided per target group or otherwise (e.g. notifiable disease, resistance) in the scoring form to be used by the data collectors and experts (Appendix 2). In the “Guidelines for scoring” (Appendix 1), detailed items are presented to score each sub-criterion.
Table 3

Subcriteria for impact and transmission: translation of qualitative to numerical scores

Impact

Case fatality (man, production animals and companion animals)

NE

L

M

H

 

U

Morbidity (all target groups)

NE

L

M

H

 

U

Mortality (all target groups)

NE

L

M

H

 

U

Impact on life comfort (all target groups)

NE

L

M

H

 

U

Risk of population decrease (game, threatened spp. and pest spp.)

NE

L

M

H

 

U

Economic impact (all target groups)

NE

L

M

H

 

U

Notifiable disease (national, OIE)

N

   

Y

 

Probability of eradication achievement (all target groups)

 

H

M

L

NE

U

Treatment possibilities (all target groups)

 

H

M

L

NE

U

Vaccination efficiency (all target groups)

 

H

M

L

NE

U

Risk concerning use as a weapon in bioterrorism

NE

L

M

H

 

U

Transmission characteristics

Contagiousness and/or efficiency of transmission by vectors

 

L

M

H

 

U

Genetic stability

 

H

M

L

 

U

Importance of occupational/circumstantial exposure

 

H

M

L

 

U

Probability of introduction

NE

L

M

H

End

U

Probability of transmission from wildlife to “target group”

 

L

M

H

 

U

Transmission efficiency between different wildlife species

NE

L

M

H

 

U

Probability of secundary transmission (intra- or extra-target group)

NE

L

M

H

 

U

Transmission influenced by extrinsic factors (environment and anthropogenic)

NE

L

M

H

 

U

Resistance: in environment to desinfectants

 

L

M

H

 

U

Numerical score

1

2

3

4

5

3

Y yes, N no, H high, M medium, L low, NE non-existent, End endemic, U unknown

The information on which the scores are based is obtained from peer reviewed literature. The “impact” and “transmission” sub-criteria are scored with one of four qualitative evaluations (high, medium, low, and non-existent) or with an “Unknown” or “Not applicable” score. “Not applicable” is used when the target group to which the sub-criterion refers is not adversely affected by the pathogen (this option, included in the scoring form, does not appear in Table 3 because the sub-criteria concerned are represented in a general way and not per target group). For a target group that is not affected by a given pathogen, all the sub-criteria referring to that target group are scored “NA” simultaneously. Subsequently, these scores were reviewed by experts who were identified based on their familiarity with the pathogen(s) of concern. In case of disagreement they were asked to provide supporting arguments for their opinions, allowing a further discussion of the scores where necessary. The identification of experts for review of the literature based scores was validated by an inquiry of the heads of the infectious diseases departments of the Veterinary and Agrochemical Research Centre (VAR, Brussels), the Institute of Tropical Medicine (ITM, Antwerp), and the Veterinary Faculty of Ghent University.

Choices of parameters

Host presence and occurrence of pathogens may differ between regions, in contrast to impact and transmission characteristics which are inherent to the pathogen itself. However, impact and transmission characteristics are different for different target groups. Consequently, corresponding to his particular competences, the end user should choose for which region and for which target group he wants to carry out a prioritization of pathogens from wildlife.

Moreover, the relative importance attached to the sub-criteria of concern can be different to some degree, depending on the objectives of the end user. Stockbreeding managers are mainly interested in the economic impact, morbidity and mortality caused by infectious hazards and in prevention (e.g. vaccination). Animal welfare and treatment possibilities are criteria that will be highlighted by companion animal owners. For conservationists the case fatality and treatment possibilities may be less important than mortality (responsible for population decrease), risk of introduction in certain habitats and transmission efficiency between wildlife species. Corresponding to the relative importance attached to the various impact and transmission sub-criteria, weights can be given to the sub-criteria.

Data processing

The actual processing of the data is based on MySQL software and starts with an algorithm (Fig. 1). Through Yes/No choices, those pathogens are selected from the hazard identification list, for which at least one wildlife host is reported in the chosen region and for which at least one occurrence is reported in wildlife or in at least one target group in Europe. If it concerns a mainly vector-borne pathogen, the “Vector presence” in Europe is taken into account as well.

As a direct consequence of the limited data availability for many wildlife-borne pathogens, a differentiated approach of the data processing and the output of the results is necessary. Detailed data for the release assessment are often lacking which involves that a choice will have to be made between two different objectives, being the comprehensiveness or the refinement of the risk assessment. For example, Yes/No choices, necessary for selecting the agents for ranking, can create an image that is very different from reality. A confirmed host presence for a given pathogen can in reality be very limited (e.g. one only host species is present in a very limited area or in a very short period in the year, or with low distribution density). The same is true for a limited occurrence of a given pathogen in wildlife (e.g. reported once in a small geographical area in Europe, or in a very distinct ecological context). Refinement of the release assessment (second-level ranking) necessitates the use of detailed data about “host presence” and “occurrence in wildlife”. On the other hand these data are often not available, for example the host presence of mammal species, or undetected occurrences of pathogens in wildlife. A refined risk assessment is not possible for the pathogens where these kinds of data are missing. This does not mean that they could not pose a possible threat and if comprehensiveness is aimed at, a first level of ranking, using the only available “rough” information (Yes/No choices), has to be maintained.

The “Impact” and “Transmission” sub-criteria-scores further determine the ranking of the pathogens as follows. The qualitative scores are translated to numerical scores ranging from 1 to 5, the specific value depending on the meaning for the risk evaluation of each sub-criterion (Table 3). For example, if the “risk of population decrease in threatened species” is scored “High”, this qualitative score is translated to the numerical score “4”, but if “probability of eradication in production animals” is scored “High” this is translated to the numerical score “2”. “Non-existent” scores obtain the numerical value 1 or 5 depending on the sub-criterion. “Non-applicable” scores are given the value “0” and the sub-criteria concerned are thus eliminated from the risk assessment.

Sum of the scores

After multiplying each numerical score with the weight assigned to the specific sub-criterion, the products are added up to obtain the total score per pathogen. For normalization reasons, the latter is expressed as a percentage of the maximum total score that is possible for that particular pathogen in the chosen target group (the maximum possible score being calculated as: ∑ (maximum numerical score × maximum weight/sub-criterion)). This expression is used for consistency reasons, to allow compilation (within the same range of values) of the first-level scores with the refined scores in the second-level ranking.

Through the use of “NA” scores, the number of sub-criteria used to evaluate the pathogens that affect a particular target group is always the same, which is necessary for consistency (calculation of the “maximal total score” within that target group)

Subsequently the pathogens can be ranked in decreasing order of total scores.

Processing of the “unknowns”

For some sub-criteria in some pathogens it is impossible to assign a score because this score is unknown (no information in the literature and unknown by experts). Yet, a score has to be included in order to allow a consistent comparison with the other pathogens. In our system the median numerical score “3” is assigned to all the “unknown” sub-criteria for each pathogen. This score is also each time multiplied with the weight given to the sub-criterion concerned.

Uncertainty estimation

The uncertainty estimation is based on the relative number of “unknown” impact and transmission sub-criteria for each pathogen. The degree of uncertainty can be expressed as: number of unknown sub-criteria × 100/total number of sub-criteria (expressed as a %). This degree of uncertainty is indicated next to the total score for each pathogen.

Two-level ranking

Only pathogens with reported occurrences in wildlife in Europe are considered for the second-level ranking. Detailed qualitative and quantitative information concerning “Host presence” and “Occurrence in wildlife” is used. Occurrence in targets is not used because it is not necessarily wildlife borne. For “Host presence”, if more than one host exists for a pathogen the host with the highest scores is taken into account.

The second-level sub-criteria are not weighted because there is no apparent reason to assume that they can have different relative importance as they concern only the release assessment, which is the same for all the target groups or end users objectives. The scores for these sub-criteria are directly expressed within a range of 1–100 (Table 2) and they are summarised by calculating the mean value of the “known” sub-criteria (unlike the first level). Due to the fact that the second-level scores are a refinement based on information of irregular availability, the choice was made to include only the “known” sub-criteria for the second-level ranking. Thus, if no information is found in the database for one or more of the second-level sub-criteria, these “unknown” sub-criteria are not given the “median” value as in the first-level ranking but they are simply not taken into account for calculating the mean refined score. By averaging the “known” scores, the total second-level score for each pathogen retained in the second-level ranking can be compared consistently with the others.

The final second-level score for each pathogen is calculated by averaging the earlier determined “Impact” and “Transmission” scores with the second-level refined score concerning “Host presence” and “Occurrence in wildlife”. The percentage of “unknown” second-level sub-criteria scores is included in the total uncertainty estimation (compiled as a mean value with the uncertainty estimation of the first level).

Output

At the moment of writing, 222 pathogens are included in the hazard identification. A user interface has been developed allowing the querying of Wildtool by different end users in function of their responsibilities. The end users can choose their own parameters. These choices include the weights they want to give to the various sub-criteria “Impact” and “Transmission”, the region and the target group for which they want to carry out the risk assessment. For Belgium the choices for the region include Flanders, Wallonia, the Brussels Region, or the country as a whole. Possible choices of target groups include man, production animals, companion animals or wildlife itself. Within the wildlife target group a choice has to be made between game species, unprotected and pest species, or protected and threatened species.

The result of each query is presented as a ranked list of pathogens. An example for the target group “Production animals” using the data contents of the database at the time of writing this paper and a determined set of weights (Appendix 3) is presented in Fig. 2 (Cf. “Discussion”). The number of highest ranked pathogens to be shown in the end result can be left to the user. In the current Wildtool version, the first 15 prioritized pathogens are shown.
Fig. 2

Example of the results of first- and second-level ranking (target groups, production animals; region, Flanders; determined set of weights)

Discussion

Other risk assessment systems including prioritization of pathogens have been described with different objectives. They use different scoring methods for a number of predetermined criteria. Some of them use weighting to differentiate the importance of the criteria, whereas others do not. Human infectious disease prioritization systems have been proposed by Doherty (2000), Ciotti (2003) and Krause (2008). A full quantitative approach for prioritising zoonoses is applied in the Dutch Emerging Zoonoses project (Van der Giessen et al. 2010). In the semi-quantitative prioritization method worked out by the scientific committee of the Belgian Federal Agency for the Safety of the Food Chain (Cardoen et al. 2009), zoonoses transmitted through food were ranked. A pure qualitative approach, without using scores, was used in the ranking of non-foodborne zoonotic pathogens in France by the Institut de Veille Sanitaire (Valenciano 2001). Prioritization of animal infectious diseases is carried out in a British project of the Department for Environment, Food and Rural Affairs (DEFRA 2009): starting from “disease profiles”, animal diseases are ranked in five separate lists, of which four are focusing on impact issues representing the four “reasons for intervention” as stated by the British Animal Health and Welfare Strategy, and a fifth list on “risk and mitigation”.

Risk assessments for wildlife-borne pathogens usually focus on specific issues, for example the introduction of West Nile virus (Patel et al. 2009) or the incursion of diseases through wild boar (Hartley 2009). Decision support tools for risks related to wildlife translocations have been developed by Hartley and Gill (2010) and Sainsbury et al. (2010). To our knowledge, the only first-line risk assessment method using prioritization and focusing entirely on wildlife-borne pathogens is the “rapid risk analysis” of McKenzie et al. (2007) that is based on the same principles as our Wildtool system because it also refers to the OIE standards and it allows a risk analysis for different target groups. In the McKenzie method, three criteria were used, being the risk of introduction, the likelihood of spread and the consequences of the pathogens on different target groups. For the processing, unlike the other methods, the three criteria scores were multiplied with each other to get to the final score per pathogen within a target group.

In the Wildtool system, we propose a first-line and multi-purpose risk assessment method for wildlife-borne pathogens. An important objective of such a system is that it identifies the pathogens to consider for in-depth risk analysis and targeted surveillance. Therefore, the risk communication and risk management are left to the end user, according to his responsibilities.

We precluded a pure quantitative methodology because of the broad range of pathogens that is considered in this first-line approach and the scarcity of numerical data about many wildlife-borne pathogens. Though numerical data are used as well, the majority of data where we start from are of qualitative nature. However, in the processing procedure, the qualitative expressions are converted to numbers and the final result is based on the ranking of the end scores. Therefore our method should be classified as a semi-quantitative risk evaluation.

While other prioritization methods are nearly exclusively based on extensive expert agreements concerning the methodology and the scoring, our method starts from a broad literature search, providing the sub-criteria and their scores. This approach is imposed by the irregular data availability in the field of pathogens from wildlife. Subsequently these scores are reviewed by a large number of experts, each of them being familiar with one or more of the pathogens. Their task is to refine the scores as much as possible and to reduce the number of “unknown” scores. Experts were selected based on recommendations provided by the infectious diseases departments of the institutes mentioned above. Though our risk assessment method is substantially literature based, for the expert review part more specialised expert validation tools should be considered to further improve the data quality. One or more experts per pathogen were consulted but for some pathogens no experts were found in which case we had to rely on the literature solely.

We deliberately used a large number of sub-criteria keeping in mind the broad range of pathogens that are covered and for which very different features may be of importance for the different objectives of the end users. Yet the sub-criteria list has to be uniform in order to be comparable for the different pathogens. In the Wildtool system, in order to offer a flexible answer to different objectives of the end users, a different set of sub-criteria is used for each target group whereby specific sub-criteria applicable to that particular target group are taken into account, together with the general sub-criteria applicable to all the target groups. Furthermore, the relative importance that the end user may attach to the different sub-criteria used within a given target group is taken into account as “weights” applied to the sub-criteria. While in other methods the weights are determined by experts or policy makers, most often on a once-only base, the choice of the weights in Wildtool is left to the end user for reasons of flexibility. A second objective of using weights is that a more distinct differentiation is obtained between the end scores of the different pathogens in the ranking list. A range of weights from 1 to 100 was used, which proved to result in a sufficient degree of separation between the end scores for identifying the most important pathogens for our purposes, though more pronounced in the second- than in the first-level ranking. The full range of values should be used when assigning weights to the sub-criteria. In the context of a first-line risk assessment it is probably more important to know that a given pathogen belongs to those which are to be considered for in-depth risk assessment than to know which specific position each pathogen occupies within the prioritized group of pathogens. In order to highlight the individual ranking position of a pathogen within the prioritized pathogens group, a higher degree of separation can be obtained by replacing the 1/2/3/4/5 numerical score range for the sub-criteria scores by a 20/40/60/80/100 score range, analogous to the one used for the second-level sub-criteria scores. Weighting is not applied by Doherty (2000), Ciotti (2003) and McKenzie et al. (2007).

The final ranking list uses the median score (value 3) for all the “unknown” sub-criteria, which allows comparing the pathogens starting from an equal number of sub-criteria. This is an important difference with the methodology of Doherty (2000) who doesn’t take into account the “unknowns” and that of McKenzie et al. (2007) who assigns the highest value to a criterion if the score is unknown. Krause (2008) uses the score “0” not only as a “medium” score (average importance), but also for the “unknown” scores. However, a zero score means that the unknown criterion is eliminated in the total score. In contrast, using the median value “3” for all the “unknown” sub-criteria of a given pathogen allows the preservation of these sub-criteria and their weights in the risk assessment, while the given score itself will not influence the prioritization result.

As a consequence of the data availability, the uncertainty estimation in Wildtool is not expressed as a “probability distribution” nor as a “confidence interval”, as is often the case in (semi-) quantitative risk evaluations, but indicates the relative amount of “unknown” information.

Limited data availability in the field of wildlife-borne pathogens also involves a differentiated approach of the data processing. A more accurate risk assessment for pathogens prioritized in the first-level ranking is only possible if detailed information is available. The first-level ranking focuses on comprehensiveness and can be considered as a filter with broad meshes, retaining all the pathogens from wildlife that can constitute a possible risk. Due to the small surface of the Belgian territory and the current scarcity of data about occurrences and vector presence, these two main criteria are scored Yes or No at the European scale (Fig. 1). Host presence however is well known and is checked at the regional/national scale. One occurrence in targets or in wildlife in Europe and the presence of one wildlife host in the given region are sufficient to include the pathogen in the first-level ranking (for mainly vector-borne pathogens the presence of a vector in Europe is necessary as well). In the second-level ranking, the host presence information at the regional/national scale is refined. The occurrence is now checked only in wildlife, as well at the regional/national scale (if information is available) as at the European scale. In case of mainly vector-borne pathogens, the vector presence is checked at the regional/national scale (Table 2).

Interpretation of the results is done by comparing the first and the second-level ranking order of a pathogen. As illustrated in the example shown in Fig. 2, some pathogens ranking high in the first-level result such as Eastern Equine Encephalitis virus, West Nile virus and Bluetongue virus, disappear in the second-level “top-15” due to the refined assessment of their release from wildlife. In this comparative result, although they obtained a high first-level ranking and should be considered for surveillance, their release in the given circumstances is less probable than for the higher ranked pathogens in the second level. For others, it is remarkable that a high ranking order, comparable to their first-level ranking, is conserved even after refinement with detailed release data, for example Chlamydophila psittaci (consequences for turkey breeding), Salmonella enterica subsp. enterica (different target species) and Avian influenza HP strains (consequences for poultry industry). The confirmation of their high first-level rank by focusing on the release assessment means that regarding the parameters chosen by the end user, these are most probably the pathogens to be considered in the first place for active surveillance.

Dependent on the programmation, some possibly important items may not be distinguished by a computerised system for the processing of qualitative data. For example, a query to assess the risk of wildlife-borne pathogens for production animals will also prioritize pathogens that have only economic importance in one particular production sector, even if the animal species concerned is of minor economic importance for the country (for example, myxomatosis in rabbits and duck virus hepatitis in Belgium, Cf. Fig. 2). Therefore, it is important that the end user has access to the information the scores are based on. For this reason a sufficiently detailed data input is important. Simultaneously, the way of data processing should be understood well by the end user. Alternatively, in this example, the sub-criterion “economic impact” could be split up in “economic impact for the sector” and “economic impact for the country/region”. This discussion illustrates the determinative role of the programming and the difficulty to find a workable balance between the comprehensiveness of the system and the unambiguousness of the results.

An important question concerns the balance between the different elements of the risk assessment. In the first-level ranking, the number of sub-criteria used for scoring respectively the impact, release or exposure assessment is already slightly different, but the second-level ranking highlights the release assessment to a large extent. We could find no information in the literature about how the proportion between the elements of a risk assessment should be conceived, but this information could have implications for the programming. In our opinion, the release assessment is a most important element in the particular field of wildlife-borne pathogens, and emphasising the release assessment is a choice that can be made, as far as pathogens are compared consistently i.e. based on the same kind of information. On the other hand, the consequence assessment being proportionally underestimated in the second-level ranking, some pathogens with a lower or questionable impact will appear in the ranking list as illustrated in Fig. 2, for example Mycobacterium microti or LP avian influenza strains.

A flexible and multi-purpose risk assessment system as proposed here requires continuous updating. Linking Wildtool to existing databases, providing the information that is the most susceptible to regular updates (i.e. “occurrences”), could offer practical advantages. Regarding the extent of the system, the practical implementation will rely by preference on the support of national or international organisms.

The scoring results offer baseline information about which pathogens to focus on for in-depth risk analysis by risk managers with different interests. At the national level prioritization results can be used in the targeted implementation of surveillance networks in wildlife with the aim of coordinating the needs of different departments with a minimum of sampling efforts. In Wildtool, the existing Belgian networks having access to the various wildlife hosts have been identified and are automatically linked to the prioritized pathogens. Data about wild animal species and the prioritized pathogens can be consulted by the end user.

Conclusions

In Wildtool, we have pursued a first-line approach and a maximum flexibility as compared with other prioritization systems. The choices offered to the different end users and the continuity of data updates correspond to realistic needs considering the sharing of competences between various official departments and the ever changing situation “in the field” concerning host presence, pathogen occurrences and knowledge about pathogens.

As mentioned above, the system can only be workable if it is updated continuously by a team of administrators. The current version of Wildtool should be considered as a prototype. Other versions of Wildtool could be developed for supranational prioritization of wildlife-borne pathogens.

Notes

Acknowledgements

This work was funded by the Belgian Federal Public Service of Health, Food Chain Safety and Environment (contract RT 07/5 WILDSURV). We acknowledge the contributions of the members of the WILDSURV steering committee and the many experts who kindly responded to our call for advice concerning pathogens and vectors. A special word of thanks is given to Marc Artois, Joke van der Giessen and Andrew Frost for their highly appreciated comments and support.

Supplementary material

10344_2011_520_MOESM1_ESM.pdf (28 kb)
Appendix 1Guidelines for scoring (PDF 27 kb)
10344_2011_520_MOESM2_ESM.pdf (35 kb)
Appendix 2Scoring form (PDF 35 kb)
10344_2011_520_MOESM3_ESM.pdf (32 kb)
Appendix 3Examples of weights assigned to the sub-criteria (PDF 32 kb)

References

  1. Burfield I, Van Bommel F (2004) Birds in Europe. Population estimates, trends and conservation status. BirdLife Conservation series no12. BirdLife Int, CambridgeGoogle Scholar
  2. Cardoen S, Van Huffel X, Berkvens D, Quoilin S, Ducoffre G, Saegerman C, Speybroeck N, Imberechts H, Herman L, Ducatelle R, Dierick K (2009) Evidence-based semiquantitative methodology for prioritization of foodborne zoonoses. Foodborne Pathog Dis 6(9):1083–1096PubMedCrossRefGoogle Scholar
  3. Ciotti M (2003) Establishing a list of priority diseases for the countries of the Dubrovik pledge. The Dubrovnik pledge on surveillance and prioritization of infectious diseases. Doc EUR/03/5042861. Report on a WHO meeting, Bucharest, Romania, 21–23 Nov 2002. pp 2–5Google Scholar
  4. DEFRA (2009) Veterinary surveillance: prioritisation project. Available at: http://www.defra.gov.uk/foodfarm/farmanimal/diseases/vetsurveillance/strategy/programme/prioritisation.htm
  5. Devos K, Kuijken E, Verscheure C, Meire P, Benoy L, De Smet W, Gabriëls J (2005) Overwinterende wilde ganzen in Vlaanderen, 1990/91-2003/04. Natuur. Oriolus 71(Bijl):4–20Google Scholar
  6. Doherty J-A (2000) Establishing priorities for national communicable disease surveillance. Can J Infect Dis 11(1):21–24PubMedGoogle Scholar
  7. Ducoffre G (2010) Surveillance van infectieuze aandoeningen door een netwerk van laboratoria voor microbiologie 2008 + epidemiologische trends 1983–2007. IPH/EPIREPORTS, nr 2010–16 Mei 2010Google Scholar
  8. Federal Agency for the Safety of the Food Chain (2010). Zoosanitaire toestand in België. Available at: http://www.favv.be/sp/pa-sa/zoosani_nl.asp
  9. Handistatus II (2010). Available at: http://www.oie.int/hs2/report.asp?lang=en
  10. Hartley M (2009) Qualitative risk assessment of the role of the feral wild boar (Sus scrofa) in the likelihood of incursion and the impacts on effective disease control of selected exotic diseases in England. Eur J Wildl Res. Available at: http://www.springerlink.com/content/r501536w45801w2r/?p=7c8532f0dbb146bebed1268740024f38&pi=1
  11. Hartley M, Gill E (2010) Assessment and mitigation processes for disease risks associated with wildlife management and conservation interventions. Vet Rec 166:487–490PubMedCrossRefGoogle Scholar
  12. Jones KE, Patel NG, Levy MA, Storeygard A, Balk D, Gittleman JL, Daszak P (2008) Global trends in emerging infectious diseases. Nature 451:990–993PubMedCrossRefGoogle Scholar
  13. Krause G (2008) Prioritisation of infectious diseases in public health—call for comments. Eurosurveillance 13(40). Available at: http://www.eurosurveillance.org/ViewArticle.aspx?ArticleId=18996
  14. Kuiken T, Leighton FA, Fouchier RAM, LeDuc JW, Peiris JSM, Schudel A, Stöhr K, Osterhaus ADME (2005) Pathogen surveillance in animals. Science 309:1680–1681PubMedCrossRefGoogle Scholar
  15. Lamotte S (2006) l’Erosion de la biodiversité: les mammifères. Partim: Chauves-souris. Dossier scientifique. Direction de la Nature, Division Nature et Forêts, Direction générale des ressources naturelles et de l’environnement, JambesGoogle Scholar
  16. Libois R (2006) L’Erosion de la biodiversité: les mammifères. Partim: Les mammifères non volants. Dossier scientifique. Unité de recherches zoogéographiques, Faculté des Sciences, Université de LiègeGoogle Scholar
  17. McKenzie J, Simpson H, Langstaff I (2007) Development of methodology to prioritise wildlife pathogens for surveillance. Prev Vet Med 81:194–210PubMedCrossRefGoogle Scholar
  18. OIE (2008a) Import Risk Analysis. Chapter 2.1. In: Terrestrial Animal Health Code, 17th edn. Available at: http://www.oie.int/eng/normes/mcode/code2008/a_summry.htm
  19. OIE (2008b) Questionnaire on wildlife diseases. Available at: http://www.oie.int/wildlife/eng/New_Wild_Questionnaire-English_2008.xls
  20. Patel K, Lopez M, Roberts H, Sabirovoc M (2009) West Nile Virus: Potential risk factors and the likelihood for the introduction of the disease into the United Kingdom. Working Document. Food and Farming Group, International Animal Health., DEFRA, London. pp 1–34. Available at:http://www.defra.gov.uk/foodfarm/farmanimal/diseases/monitoring/documents/qra-wnv-090210.pdf
  21. Peeters L (2007) Trekkalender roofvogels over Nederland en België. Bijlage: Themanummer roofvogels. Natuur Oriolus 73(3):98–99Google Scholar
  22. Ryser-Degiorgis M-P, Gavier-Widen D, Gortazar-Schmidt C, Kuiken T (2009) EWDA network for wildlife health surveillance in Europe. Summary of data presented at the inaugural meeting. Brussels 15 Oct 2009. Available at: http://www.ewda.org/EWDA%20Summary%20of%20reports,%20Brussels%20Oct%202009.pdf
  23. Sainsbury AW, Vaughan RJ, Curtiss P (2010) Disease risk analysis for the reintroduction of the white tailed sea eagle (Haliaeetus albicilla) to England. Report to Natural England. Institute of Zoology, Zoological Society of London, LondonGoogle Scholar
  24. Thrusfield M (2007a) Veterinary information systems. In: Veterinary Epidemiology, 3rd edn. Blackwell, Oxford, pp 205–207Google Scholar
  25. Thrusfield M (2007b) The transmission and maintenance of infection. In: Veterinary Epidemiology, 3rd edn. Blackwell, Oxford, pp 205–207Google Scholar
  26. Valenciano M (2001) Définition des priorités dans le domaine des zoonoses non alimentaires, 2000–2001. Rapport de l’Institut de Veille Sanitaire. Available at: http://www.invs.sante.fr/publications/2002/def_priorite_zoonoses/index.html
  27. Van der Giessen JWB., Van de Giessen AW., Braks MAH. (2010) Emerging zoonoses information and priority-setting system. In: Emerging zoonoses: early warning and surveillance in the Netherlands. RIVM-rapport 330214002, Bilthoven. pp 29–31.Google Scholar
  28. Verkem S, De Maeseneer J, Vandendriessche B, Verbeylen G, Yskout S (2003) Zoogdieren in Vlaanderen. Ecologie en verspreiding van 1987 tot 2002. Natuurpunt.Studie en JNM Zoogdierenwerkgroep, Mechelen/GentGoogle Scholar
  29. Vermeersch G, Anselin A, Devos K, Herremans M, Stevens J, Gabriëls J, Van Der Krieken B (2004) Atlas van de Vlaamse broedvogels 2000–2002. Mededelingen van het Instituut voor Natuurbehoud 23, BrusselsGoogle Scholar
  30. Vlaamse Avifauna Comissie (1989) Vogels in Vlaanderen, voorkomen en verspreiding. Vlavico, IMP, BornemGoogle Scholar
  31. Weiserbs A, Jacob J-P (2007) Oiseaux nicheurs de Bruxelles, 2000–2004. Répartition, effectifs, évolution. Aves, Liège.Google Scholar
  32. Williams ES, Yuill T, Artois M, Fisher J, Haigh SA (2002) Emerging infectious diseases in wildlife. Rev Sci Tech 21(1):139–157PubMedGoogle Scholar

Copyright information

© Springer-Verlag 2011

Authors and Affiliations

  • Paul Tavernier
    • 1
  • Jeroen Dewulf
    • 2
  • Sophie Roelandt
    • 1
  • Stefan Roels
    • 1
  1. 1.Wildsurv Project, Operational Direction Interactions and SurveillanceVeterinary and Agrochemical Research CenterBrusselsBelgium
  2. 2.Veterinary Epidemiology Unit, Department of Obstetrics, Reproduction and Herd Health, Faculty of Veterinary MedicineGhent UniversityMerelbekeBelgium

Personalised recommendations