Model for ranking freshwater fish farms according to their risk of infection and illustration for...

P

Mrs

BCNa

b

c

d

Ue

f

g

a

ARRA

KFDRRVD

f

0

ARTICLE IN PRESSG ModelREVET-3567; No. of Pages 17

Preventive Veterinary Medicine xxx (2014) xxx–xxx

Contents lists available at ScienceDirect

Preventive Veterinary Medicine

j ourna l h om epa ge: www.elsev ier .com/ locate /prevetmed

odel for ranking freshwater fish farms according to theirisk of infection and illustration for viral haemorrhagicepticaemia

irgit C. Oidtmanna,∗, Fiona M. Pearcea,b, Mark A. Thrusha, Edmund J. Peelera,hiara Ceolinc, Katharina D.C. Stärkd, Manuela Dalla Pozzac, Ana Afonsoe,icolas Diserens f, R. Allan Reesea, Angus Camerong

Centre for Environment, Fisheries and Aquaculture Science (Cefas), Barrack Road, Weymouth, Dorset DT4 8UB, United KingdomThe Ministry for Primary Industries NZ, 25 The Terrace, Wellington 6011, New ZealandIstituto Zooprofilattico Sperimentale delle Venezie, Viale dell’Università, 10, 35020 Legnaro, PD, ItalyDepartment of Production and Population Health, Royal Veterinary College, Hawkshead Lane, North Mymms, Hatfield AL9 7TA,nited KingdomEuropean Food Safety Authority, Via Carlo Magno 1A, 43126 Parma, ItalyCentre for Fish and Wildlife Health, Vetsuisse-Faculty, University of Bern, Länggassstrasse 122, 3001 Bern, SwitzerlandAusVet Animal Health Services, 140 Falls Road, Wentworth Falls, 2782 NSW, Australia

r t i c l e i n f o

rticle history:eceived 20 December 2013eceived in revised form 5 April 2014ccepted 8 April 2014

eywords:ishiseaseisk factorisk-based surveillanceiral haemorrhagic septicaemiaisease freedom

a b s t r a c t

We developed a model to calculate a quantitative risk score for individual aquaculturesites. The score indicates the risk of the site being infected with a specific fish pathogen(viral haemorrhagic septicaemia virus (VHSV); infectious haematopoietic necrosis virus,Koi herpes virus), and is intended to be used for risk ranking sites to support surveillancefor demonstration of zone or member state freedom from these pathogens. The inputs tothe model include a range of quantitative and qualitative estimates of risk factors orga-nised into five risk themes (1) Live fish and egg movements; (2) Exposure via water; (3)On-site processing; (4) Short-distance mechanical transmission; (5) Distance-independentmechanical transmission. The calculated risk score for an individual aquaculture site is avalue between zero and one and is intended to indicate the risk of a site relative to the riskof other sites (thereby allowing ranking). The model was applied to evaluate 76 rainbowtrout farms in 3 countries (42 from England, 32 from Italy and 2 from Switzerland) with theaim to establish their risk of being infected with VHSV. Risk scores for farms in England andItaly showed great variation, clearly enabling ranking. Scores ranged from 0.002 to 0.254
(mean score 0.080) in England and 0.011 to 0.778 (mean of 0.130) for Italy, reflecting thediversity of infection status of farms in these countries. Requirements for broader applica-tion of the model are discussed. Cost efficient farm data collection is important to realisethe benefits from a risk-based approach.
Crown

Please cite this article in press as: Oidtmann, B.C.,

according to their risk of infection and illustration fohttp://dx.doi.org/10.1016/j.prevetmed.2014.04.005

∗ Corresponding author. Tel.: +0044 1305 206661/+0044 1305 206601;ax: +0044 1305 206601.

E-mail address: [email protected] (B.C. Oidtmann).

http://dx.doi.org/10.1016/j.prevetmed.2014.04.005167-5877/Crown Copyright © 2014 Published by Elsevier B.V. All rights reserved

Copyright © 2014 Published by Elsevier B.V. All rights reserved.

1. Introduction

et al., Model for ranking freshwater fish farmsr viral haemorrhagic septicaemia. PREVET (2014),

An environment where animals can be grown with alow risk of losses due to the occurrence of animal diseasesis beneficial for agri- and aqua-culture and eventually ben-efits society as a whole. Competent authorities (CAs) take

.

dx.doi.org/10.1016/j.prevetmed.2014.04.005


http://www.sciencedirect.com/science/journal/01675877

http://www.elsevier.com/locate/prevetmed

mailto:[email protected]


IN PRESSG Model

Veterinary Medicine xxx (2014) xxx–xxx

Fig. 1. Viral haemorrhagic septicaemia virus (VHSV) in Europe: Areasshown in red were reported as infected with VHSV by EU member states(EU MS) to the Animal Disease Notification System (ADNS; http://ec.europa.eu/food/animal/diseases/adns/adns en.htm) during time period

ARTICLEPREVET-3567; No. of Pages 17

2 B.C. Oidtmann et al. / Preventive

a role in achieving and maintaining a high animal healthstatus through a range of activities, including surveillance.In times of limited resources, there is an increased need todevelop cost saving surveillance methods.

A recent review of risk-based methods for fish andterrestrial animal disease surveillance (Oidtmann et al.,2013) found that although risk-based surveillance (RBS)approaches are applied in the design or assessment of anumber of terrestrial animal diseases, there are few exam-ples of risk-based approaches applied to aquatic animals(Oidtmann et al., 2009, 2011c; Kleingeld, 2010; Diserenset al., 2013), or scenario tree modelling approaches for theevaluation of surveillance systems (Oidtmann et al., 2008;Lyngstad et al., 2011).

The European Council Directive 2006/88/EC (Anon,2006) on aquatic animal health requires that risk-basedanimal health surveillance is applied to aquaculture pro-duction businesses (APBs) in the EU. The frequency ofinspections should take account of the likelihood thatthe fish farm may contract and spread disease, thus therisk must be assessed for each APB. Five disease cat-egories (not to be confused with risk categories) forcountries, zones or compartments are defined by the Direc-tive: Category I—approved pathogen-free status; CategoryII—not declared disease-free, but subject to a surveil-lance programme to achieve disease-free status; CategoryIII—infection status is unknown; Category IV—subject toan eradication programme, and Category V—where somefarms (but not necessarily all) are known to be infected.Since there are multiple notifiable fish diseases, a singleAPB may be in multiple disease categories (e.g. in CategoryI for viral haemorrhagic septicaemia (VHS), and Category IVfor infectious haematopoietic necrosis (IHN)). The Directiverequires that a risk-based approach is used for both diseasesurveillance (article 10 of the Directive) and complianceinspections (article 7) for all disease categories. This paperpresents a quantitative model to rank fish farms based onthe likelihood of disease introduction. The model can calcu-late the risk of introduction for both freshwater salmonidand cyprinid pathogens. Parameter estimates for 3 fresh-water fish diseases listed by European Council Directive2006/88/EC, VHS, IHN, and Koi herpes virus disease (KHD)were obtained through an expert consultation (Oidtmannet al., in press). We also present the application of the modelfor VHS, in three geographic regions. VHS was chosen forthis case study as it is one of the most important viral dis-eases of freshwater farmed rainbow trout (Oncorhynchusmykiss) in Europe (Smail, 1999), and is responsible for esti-mated annual production losses of 20–30% (Baruchelli et al.,1990). VHS virus (VHSV) is a rhabdovirus, genus Novirhab-dovirus, and four genotypes of VHSV are recognised basedon nucleic acid sequencing (OIE, 2012), which are broadlyassociated with geographic location. In Europe, the dis-ease became of relevance with expanding rainbow troutaquaculture. Historically, VHSV genotype Ia was detectedin most EU member states (MS). VHS is listed in EU legisla-tion (Anon, 2006) and EU MS can, through surveillance and



biosecurity measures, demonstrate freedom and restrictimports of live susceptible fish species to regions with thesame health status. VHS is absent from the UK and a numberof geographic zones throughout the EU. Denmark recently

1.1.2010–19/03/2013. ADNS is a notification system designed to regis-ter and document the evolution of the situation of important infectiousanimal diseases in EU MS.

succeeded with the eradication of the disease from most ofits territory (Anonymous, 2013; Bang Jensen et al., in press).The expected benefits of freedom from the virus include,a more productive and less disrupted (due to absence ofdisease control measures) rainbow trout aquaculture, andaccess to wider markets. VHSV remains a continued threatto trout production in Europe. Recent notifications of VHSVdetection in EU MS are shown in Fig. 1.

We previously presented a model for risk ranking offarms for pathogen introduction and spread for freshwa-ter salmonid fish farms (Oidtmann et al., 2011c). The modelpresented here develops this methodology further by revis-ing its structure, extending its application to a broaderrange of fish diseases, using parameter estimates derivedfrom a wider expert consultation exercise, and having astochastic functionality.

2. Materials and methods

2.1. The model

The purpose of the model is to calculate a risk scorefor individual aquaculture sites. The score indicates therisk of the site being infected with a specific fish pathogen(VHSV, IHN virus (IHNV), Koi herpes virus (KHV)), andis intended to be used for risk ranking sites to sup-port surveillance for demonstration of zone or memberstate freedom from these diseases. The model inputsinclude a range of quantitative and qualitative esti-mates of risk factors organised into five risk themes: (A)Live fish and egg movements; (B) Exposure via water;(C) On-site processing; (D) Short-distance mechanicaltransmission; (E) Distance-independent mechanical trans-mission (Table 1). The estimates are based on location(relative to potential sources), pathogen introduction path-ways, and biosecurity practices of the farm site being


assessed. Farm data are based on site records and siteinspection.

For each theme, a risk score between zero and one iscalculated as described in Appendix A. The final risk score


http://ec.europa.eu/food/animal/diseases/adns/adns_en.htm

http://ec.europa.eu/food/animal/diseases/adns/adns_en.htm

ARTICLE IN PRESSG ModelPREVET-3567; No. of Pages 17

B.C. Oidtmann et al. / Preventive Veterinary Medicine xxx (2014) xxx–xxx 3

Table 1Model for ranking freshwater fish farms. Definitions of the five risk themes.

Theme name Theme definition

Live fish & egg movements Introduction of pathogen through consignments of infected live fish (for ongrowing, stockenhancement or processing) and contaminated eggs from other sites, but not includingpathogen present in transport water or contaminated transport tanks, lorries, nets orpackaging

Exposure via water Direct introduction of pathogen to farm source water by susceptible fish populations andactivities upstream: i.e. other farmed and wild stocks (including fish released by farms forrestocking or conservation purposes); fish processing

On-site processing Introduction of pathogen through dead infected fish sourced from other sites for on-farmprocessing

Short distance mechanical transmission Introduction of pathogen from sources in close proximity to the farm through routesincluding: piscivorous birds or other animals

Distance independent mechanical transmission Introduction of pathogen via contaminated fomites including: fish transporters andassociated equipment; other equipment or personnel shared with other fish farms; fisheryactivities; visitors; vehicles etc.

Live fish and eggmovements on Fraction of 1

On-site processing

Risk theme

Short distance

mechanical

transmission

X weight

x i

x k

x m

x l

=

= B

= A

= C

= E

= D

∑ A-E

Exposure via water

Score based on number of

movements and riskassociated with each

movement

Fraction of 1x j

Type of estimate Number feedinginto calculation

Distance

independent Fraction of 1

Final score for risk ofintroduction

Final

result fortheme

B

A

C

E

D Fraction of 1

Fraction of 1

Score based on number andtype of upstream risks (and

distance)

Score based on presence ofon-site processing and

source of fish

Score based on number and

type of nearby risks (anddistance)

Score based on presence /absence of mechanical

of com

foatappfae

TMee

mechanical

transmissiontransmission pathways

Fig. 2. Model for ranking freshwater fish farms. Overview

or the farm site is then calculated as the weighted averagef the theme risk scores. The calculated farm risk score is

value between zero and one and is intended to indicatehe risk of a farm site relative to all other sites (therebyllowing ranking) (Fig. 2). It is not a direct measure of therobability of being infected. Weights and probabilities for



athways of pathogen introduction could not be obtainedrom existing published data and were obtained through

formal expert elicitation (Oidtmann et al., in press). Thexperts had been consulted to provide weights for the 5

able 2odel for ranking freshwater fish farms, VHSV case study. Risk theme weights ap

t al., in press). Weight for risk theme A was determined by subtracting sum of wxpert consultation).

Theme

1Low prevalence,endemic zone

2Medium prevalence,endemic zone

Live fish movements A 0.745 (0.70) 0.705 (0.68)

Water B 0.1 0.13

Processing C 0.05 0.04

Short distance D 0.025 0.035

Distance-independent E 0.08 0.09

bining scores to calculate risk of pathogen introduction.

risk themes for four scenarios: scenario 1: 2% farm levelprevalence throughout the whole country; scenario 2: 5%farm level prevalence throughout the whole country; sce-narios 3 and 4 assume that there are disease-free (Cat I)zones within the country, which otherwise has a farm levelprevalence outside of these Cat I zones of 2 or 5% respec-


tively. Experts’ assessment of theme weights was thereforedependent on two factors: the overall prevalence of the dis-ease in the country, and whether or not the site was in adisease free zone. The expert estimates used in the VHSV

plied for farm score calculations based on expert consultation (Oidtmanneights for risk themes B-E from 1 (in brackets: value for theme A from

Scenario

3Free zone surrounded by anendemic zone with lowprevalence

4Free zone surrounded by anendemic zone with mediumprevalence

0.62 (0.50) 0.56 (0.40)0.14 0.160.04 0.0750.06 0.0750.14 0.13



4 B.C. Oidtmann et al. / Preventive Veterinary Medicine xxx (2014) xxx–xxx

Table 3Model for ranking freshwater fish farms, VHSV case study. Relative weights or likelihood estimates (for transmission of VHSV to a receiving rainbow troutfarm site within a 12 month period) for individual risk pathways within risk themes used for calculation of farm scores in VHSV case study. Parameterestimates are based on expert consultation of most likely values (Oidtmann et al., in press). Medians provided.

Theme Parameter

Live fish and eggmovements

‘Species risk’ (likelihood of transmission of VHSV with a movement of a singleconsignment of live fish of a given fish species from a site subclinically VHSV infectedto a receiving site)

Rainbow trout 0.80Brown Trout 0.30Brook Trout 0.08Grayling 0.375Pike 0.005

‘non-disinfection risk’ (likelihood of transmission of VHSV via eggs from asubclinically infected farm)

Eggs not disinfected 0.40Eggs disinfected 0.01

Exposure via water ‘Upstream source type risk’ (likelihood of farm becoming infected due to presence ofVHSV infected source within 5 km upstream)

Fish farm 0.325Wild fish 0.10Fishery 0.20Processor 0.16

On-site processing ‘Consignment risk processing’ (likelihood of farm becoming infected due to themovement of a single consignment of dead fish onto farm for the purpose ofprocessing)

Infection status of source site unknown 0.05Source site infected 0.275

Short distance mechanicaltransmission

‘Source type risk’a (relative risk of farm becoming infected depending on type ofsource)

Farms 1Fisheries 0.5Wild 0.17

‘Area risk’a (relative risk of farm becoming infected based on farm size)>5 ha 11–5 ha 0.85<1 ha 0.68

‘Distance risk’ (likelihood of farm becoming infected due to presence of VHSV infectedfishery or farm within a given distance)

<500 m 0.6500 m–2 km 0.25>2 km–5 km 0.15

Distance independentmechanical transmission

‘Relative risk of farm becoming infected by route of mechanical transmission’ a

Staff working on other fish farms (or staff from other fish farms working on the site) 0.11Using equipment belonging to other fish farms (or other farms using equipment

belonging to this site)0.20

Vehicles delivering live fish coming onto site 0.13Other vehicles coming onto site 0.03Fishery on site (anglers coming onto site using their own equipment)? 0.06Unauthorised people entering the site 0.03Unauthorised vehicles entering the site 0.02Insufficient disinfection of vehicles entering the site 0.11Insufficient disinfection of footwear (boots, shoes) when entering the site 0.11Receiving and storing fish waste (mortalities and processing waste) from other fish 0.20

ts.

farms

a Relative risks were derived from weight estimates provided by exper

case study are presented in Tables 2 and 3. Values for otherpathogens and a comprehensive description of how thesevalues were elicited from experts is provided in Oidtmannet al. (in press).

Scenario specific settings of the model were appliedto calculate a farm score for each scenario (resulting in



4 scores per farm—one per scenario). Furthermore, twodifferent settings within each scenario were applied: inthe ‘Simple’ (default) model, only presence or absenceof potential sources within a distance of 5 km is taken

into account for calculating risk via exposure via wateror short-distance mechanical transmission, whereas in the‘Distance-specific model’, parameter estimates obtainedfrom experts were used to provide a more advanced wayof calculating a risk score for these themes.

An overview of the data required is provided in Table 4.


Definitions of terms used in this paper are provided in theglossary (provided in Appendix A).

The model has been implemented in MS Excel, using thefree PopTools add-in (http://www.poptools.org) to provide


http://www.poptools.org/

Please

cite

this

article

in

p

ress

as:

Oid

tman

n,

B

.C.,

et

al.,

M

odel

for

ran

king

fresh

water

fi

sh

farms

accordin

g

to

their

risk

of

in

fection

and

illu

stration

for

viral

haem

orrhagic

sep

ticaemia.

PR

EVET

(2014),

http

://dx.d

oi.org/10.1016/j.prevetm

ed.2014.04.005

AR

TIC

LE

IN P

RE

SS

G M

odelPR

EVET-3567;

N

o. of

Pages 17

B.C. O

idtmann

et al.

/ Preventive

Veterinary

Medicine

xxx (2014)

xxx–xxx

5Table 4Model for ranking freshwater fish farms. Data requirements for site risk score calculation.

Risk factor theme Risk Data requirement (data needs to be collected forthe specific farm)

Additional data for more complexversion of model

Relevant expert estimate for VHSV

Live fish and eggmovements

Live fish movements • Number of live fish suppliers; N/A Likelihood of transmission frominfected to uninfected site viasubclinically infected fish (dependingon species)

• Number of consignments by fish suppliers• Species supplied• Infection status of suppliers

Live egg movements • Number of egg suppliers • Developmental stage of egg• Whether water used for incubationmight be contaminatedNo parameter estimates were availablefor this part of the model. Therefore,this part of the model has not beenused in calculations.

Likelihood of transmission frominfected to uninfected site via rainbowtrout (RBT) eggs with and withoutdisinfection

• Number of consignments by egg supplier• Infection status of suppliers• Species supplied• Whether egg disinfection is applied

Exposure via water • Whether farm uses water with potential forcontamination from upstream sources• Type of sources (farm, wild, fishery, processing)present within 5 km• Whether susceptible species are present at sourceInfection status of sources

• Number of sources by type• Distance of each source to site to beassessed

Likelihood of transmission from infectedsource (within 5 km upstream) touninfected site via water depending onsource type (farm, wild, fishery,processing)

On site processing • Whether on-site processing takes place• Whether biosecurity measures are in place toprevent waste transfer (liquid or solid) into farm• Number of suppliers (of dead fish for processing)• Number of consignments by fish suppliers• Species supplied• Infection status of suppliers

N/A Likelihood of transmission from infectedsource to uninfected site via dead RBTmovements for processing

Short-distancemechanicaltransmission

• Type of source (farm, fishery, wild) within 5 km• Species on nearby source• Infection status of source• Distance to source• Area (size) of farm site being assessed

No additional data required • Likelihood of transmission from infectedsource to uninfected site via short distancemechanical transmission (depending onsource type)• Likelihood of transmission depending ondistance of source• Likelihood of transmission depending onfarm size

Distance-independentmechanicaltransmission

Presence/absence of risk pathways• Personnel shared with other farms• Equipment shared with other farms• Unauthorised people can enter site• Unauthorised vehicles can enter site• Fish delivery vehicles can enter site• Other vehicles can enter site• Vehicles not disinfected• No disinfection of protective clothing (i.e. boots)• Receiving waste from other farms• Fishery on site

N/A Relative weight for transmission touninfected site via individual long-distancemechanical transmission routes

All Theme weight


ING Model

Veterina



stochastic simulation capabilities. Outputs can therefore beexpressed either in terms of the deterministic score basedonly on the means of experts’ most likely value for eachparameter, or as a distribution by running a Monte-Carlosimulation using the full parameter distributions (results ofthe expert consultation provided PERT distributions). Forthe VHS case study, calculations of farm site risk scoreswere done using the deterministic settings of the modelin MS Excel. The model is provided in the Supplementaryonline documents.

2.2. Trout farms and data collection in VHS case study

2.2.1. EnglandThe farm ranking model for viral haemorrhagic septi-

caemia was applied to all farms stocking salmonid speciesin three river catchments in England (Yorkshire Ouse, Testand Itchen, providing 19, 17 and 6 case sites respectively).The Yorkshire Ouse was included in the analysis because itcontains the only fish farm in England to have experiencedan outbreak of VHS to date (Stone et al., 2008). The othercatchments were chosen based on the presence of multi-ple farm types (e.g. hatcheries, table farms and restockingbusinesses).

Data on the farms’ activities and management prac-tices were extracted from the Cefas Live Fish Movementsdatabase (LFMD1) which contains all data collected bythe Fish Health Inspectorate (FHI) during routine statu-tory farm visits and information collected during a reviewof aquaculture biosecurity prior to fish farm authorisa-tion in 2009. All farms in England and Wales are coveredin the LFMD. A Geographic information system (GIS) wasused to assess the proximity of hazards to case farms(including upstream and downstream farms, fisheries andprocessors (ArcMap V9.3; ESRI Corp. Redlands, CA, USA)).The locations of farms and fisheries were provided by theLFMD, a list of seafood processors was compiled usingUK Food Standards Agency (FSA) data and companies’websites.

Data availability was good for the farms in England.Exact size of farm was not available, but farm size was esti-mated to within one of the three classes (<1 ha, 1–5 ha,>5 ha) by a fish health inspector with knowledge of thefarms. Live fish and egg movements, water sources, andlocations of upstream and nearby farms and fisheries andmost of the biosecurity information required by the model(e.g. whether fish delivery vehicles enter the farm) wasprovided by the LFMD. Movements of vehicles other thanfish delivery vehicles onto the farm were assumed to occurfor every farm. Disinfection of vehicles was assumed to



occur on all farms. No farms were known by the FishHealth Inspectorate to receive waste from other farms. Datawere not available for the locations of wild populations,but as wild salmonid populations are known to be both

1 Cefas developed and hosts the LMFD, a resource jointly owned byCefas, Environment Agency, Welsh Government and Defra, to manage alldata relating to Aquaculture Production Business authorisation and reg-istration, aquatic animal imports and exports, the rearing and holding ofnon-native species and statutory aquatic disease testing and controls.

PRESSry Medicine xxx (2014) xxx–xxx

widespread and common in England, the distance to wildpopulations was set to 100 m for every English farm.

2.2.2. ItalyThe study area was in the north-eastern part of Italy

and included the regions of Friuli Venezia Giulia, TrentinoAlto Adige and Veneto, where 65% of the Italian fresh-water salmonid production is based. Data were collectedfrom trout farms, the most commonly reared freshwa-ter salmonid species in Italy. In this area two hundredfarm facilities, mainly table farms and a few farms keep-ing their own broodstock for egg production, are equallydistributed across the 3 regions. To provide a meaningfulvalidation of the risk ranking model, it was important toinclude farms which had experienced an outbreak of VHS.Ten such farms were selected, five of which had been diag-nosed with VHSV infection recently (between 2010 and2012), although in some of those 5 farms, infection mayhave been present for several years, but not detected. Theremaining 22 farms were selected randomly, resulting ina total of 32 trout farms (12 in Trentino Alto Adige, 10 inVeneto and 10 in Friuli Venezia Giulia Regions) from thetwo hundred present in these three Italian regions. Sevenfarms were in a category I (VHSV free), 15 in Category III(unknown) and 10 in a category V (infected). Twenty-eightof these included facilities growing fish for human con-sumption, three had broodstock and one was a hatchery.Data required for input into the model were collected dur-ing on-site visits. The visits were planned with three farmveterinarians and took place between the end of July andmid November 2012. During the visits, the farmers wereinformed about the project objectives (each farmer wasprovided with a detailed leaflet explaining the project) andthe farm site was inspected during a visit. A standardisedquestionnaire, implemented and tested during the study,was used for data collection.

Information on disease status, species stocked, farmactivities undertaken, risk factors related to VHS intro-duction and spread (i.e.: Live fish and eggs movement,exposure via water, geographical factors relevant fordisease introduction and spread) were included in thequestionnaire together with data on management practisesand biosecurity measures applied at farm level. Informa-tion was provided by the fish farmer and augmented byknowledge of a veterinarian visiting the farms on a regu-lar basis. To assess distances from case farms to hazardsand upstream-downstream proximity to other farms, dataon the location of farms, stored in a regional web GIS, weremade available for Veneto and Friuli Venezia Giulia Region,while for the Trentino Alto Adige region spatial data werecollected using of Google Earth. The geographical coor-dinates of farms have been stored in a geodatabase andanalyzed by using ArcMap V9.3. Data were not availablefor the location of wild populations. Since wild salmonidpopulations are known to be present in the rivers, the dis-tance was estimated by measuring the lengths of farm inletchannels, using Google Earth.


2.2.3. SwitzerlandTwo large, commercially important facilities from can-

tons Vaud and Zurich were selected for the study.



B.C. Oidtmann et al. / Preventive Veterinary Medicine xxx (2014) xxx–xxx 7

Table 5Model for ranking freshwater fish farms, VHSV case study. Summary of exposure of case study farms to selected risk routes of pathogen introduction. A

number in brackets indicates number of assessed sites with potential upstream sources of infection if wild fish populations are ignored.

Country Total number of farmsin case study

Number of farms. . .

Received LFMs Received eggmovements

Had upstreamsourcesA

Processed fishon-site

Had nearbysourcesA

Italy 32 17 15 30 5 32

i

2

aftif(

wt

ssosofdds

mscA

3

s00sIw4sb

twctme

Switzerland 2 2

England 42 36

An overview of some of the exposure routes of farmsncluded in the case study is provided in Table 5.

.2.4. Model settings for VHS case studyIn the calculation of scores within Theme A (live fish

nd egg movements), the value used for ‘Source status’ forarms in Cat IV (in an eradication programme) or V (knowno be infected) was set to 1 (the farms were assumed to benfected). For a farm in Cat III, the value was set to 0.33. Forarms in Cat I and II, the setting depended on the scenariosee Appendix A.1.2).

For calculation of scores within Theme B (exposure viaater), the infection status of upstream sites was assumed

o be ‘free from VHSV’.Of the farms from England, one farm carried out on-

ite processing, and sourced fish for processing from otherites. Of the five Italian farms that carried out processingn-site, two processed fish from other sites. The infectiontatus of suppliers was known for the English farms butnly known for one of the two Italian farms. Infection statusor each consignment was set to be 1 (infected) only whereisease status of supplying farms was category IV or V; thisid not apply for any of the processing farms in the casetudy.

For the calculation of scores for Theme D (short-distanceechanical transmission), the infection status of nearby

ites was assumed to be ‘free from VHSV’ for all farms of thease study. The full description of the model is provided inppendix A.

. Results

Risk scores for the farms in the England case studyhowed great variation, with scores ranging from 0.002 to.254. The mean score across the models and scenarios was.080. The two versions of the model (distance-specific andimple) scored similarly (see Supplementary material, Fig.), while the four scenarios generated very different scores,

ith means of 0.055 for scenario 1, and 0.097 for scenario (Table 6 and Supplementary material, Fig. II). Nine farmscored above the upper quartile consistently for each com-ination of scenario and model.

For the Italian farms, the risk scores varied from 0.011o 0.778, with an overall mean of 0.130 (Table 6). Thereas great variation in the scores, with five farms scoring



onsiderably higher than the other farms. There was lit-le difference between the scores generated by the two

odels (Supplementary material, Fig. III); this was true forach scenario. As for the English farms, there was a large

0 1 2 22 42(31) 1 42(41)

difference between the four scenarios, with scenario 2giving on average the highest scores (Table 6, see Supple-mentary material, Fig. IV), and this was true for both modelversions. Across both models, the means for scenarios 1, 2,3 and 4 were 0.129, 0.139, 0.128 and 0.129, respectively.For the top five scoring farms, scenario 1 gave the high-est scores, and scenario 4 the lowest, while for the lowerscoring farms, scenario 4 often gave the highest score, andscenario 1 the lowest.

There were just two farms in the Swiss case study, anddue to lack of data, they were only run through the simplemodel. The two farms each received fairly similar scores(scores of 0.400 and 0.462 using scenario 1). Switzerlandhad the highest average scores of the three countries(Table 6, Supplementary material, Fig. V), although thesample size was very small (n = 2). The high values weremainly due to the fact that the disease status of all farmsin Switzerland for VHS is III (“unknown”), so all incomingconsignments to the farms had disease status of III whichled to the source status in the model being set as 0.33 foreach consignment, increasing the score for Theme A (livefish and egg movements). As Theme A has a much higherweighting than the other themes, live fish and egg move-ments are the most important determinant of the overallrisk score. On average, farms in England had low scores incomparison with Italy and Switzerland (Table 6), and thiswas because all farms in England were classed as CategoryI for VHS, and therefore all consignments received by thecase study farms have a infection status of 0 (“uninfected”)in the model. The Italian case studies are split between thefive that score very highly (over 0.3) for each combinationof scenario and model, and the remaining 27 that alwaysscore lower than 0.15. The very highly scoring farms arethose that receive consignments of live fish or eggs fromfarms of Category III–V. All other farms only received fromCategory I or II, or did not receive live fish or eggs. Italy wasthe only country in this case study with variation in thedisease status of its farms. The higher risk farms accord-ing to the model are a mixture of Category III and V farms(Supplementary material, Fig. VI). The mean score (scenario1, simple model) for Italian farms in Category I was 0.029(compared to 0.055 for English farms); for farms in Cate-gory III 0.091; and for farms in Category V 0.258. The higherscore for English farms in Category I (all farms in Englandare Cat I farms) was due to higher scores in Themes A


and B.The distribution of overall risk scores for the entire set of

case study farms (n = 76) was highly skewed towards lowervalues, with over 90% of farms scoring less than 0.15, and



8 B.C. Oidtmann et al. / Preventive Veterinary Medicine xxx (2014) xxx–xxx

Table 6Model for ranking freshwater fish farms, VHSV case study. Mean risk scores (and standard deviation) for each country generated using the simple versionof the model, overall and for each scenario separately.

England (42 farms) Italy (32 farms) Switzerland (2 farms)

Mean of scores using simple model(all scenarios combined)

0.080 (0.050) 0.130 (0.206) 0.402

Mean for simple model, scenario 1 0.055 (0.032) 0.129 (0.237) 0.4310.137 (0.222) 0.4310.128 (0.197) 0.3840.127 (0.177) 0.363

Mean for simple model, scenario 2 0.096 (0.061)



the remaining seven scoring over 0.4 (Fig. 3). The follow-ing discussion of the theme scores applies only to scoresobtained using the simple model, as this allows compari-son of the three countries (specific distance data were notavailable for the Swiss farms and therefore scores have onlybeen calculated for Swiss farms using the simple model).The scores for Theme A (live fish and egg movements) showthe same distribution (Fig. 4(a)) as the scores based on all5 themes, which is expected due to the high influence ofTheme A scores on overall risk scores due to its high weightfactor (Fig. 3). The maximum score for Theme A was 1.The gap between low values and high values reflect thatreceived live fish consignments were categorised as eitherlikely to be infected or not, according to the disease statusof the source farm. Nine farms score 0 for Theme A as aresult of not receiving any live fish or egg movements inthe year for which data were collected.

Please cite this article in press as: Oidtmann, B.C., et al., Model for ranking freshwater fish farmsaccording to their risk of infection and illustration for viral haemorrhagic septicaemia. PREVET (2014),http://dx.doi.org/10.1016/j.prevetmed.2014.04.005

Themes B (transmission via water) and D (short dis-tance transmission) have much lower scores than theother themes, with maximum scores of 0.0170 and 0.0199,respectively. These scores are low as the infection status

Fig. 3. Model for ranking freshwater fish farms, VHSV case study. His-togram of risk scores for 76 rainbow trout farms from England, Italy andSwitzerland using the simple model and scenario 1.

Fig. 4. Model for ranking freshwater fish farms, VHSV case study. Histograms of scores obtained for all farms (n = 76) using the simple model and scenario1 for (a) Theme A, (b) Theme B, (c) Theme C, (d) Theme D, and (e) Theme E. Note the scales of the x axes for (b) and (d) are 0 to 0.02 compared to 0 to 1 asfor (a), (c) and (e).


ING ModelP

Veterina

otntftlf(o

sma00

sb3f

4

AmefV(

4

bIvw

rrdfIwirstmsu

bhwrtf

ARTICLEREVET-3567; No. of Pages 17

B.C. Oidtmann et al. / Preventive

f upstream and nearby farms, fisheries and wild popula-ions was assumed to be ‘not infected’ in the model (albeitot using 0 as a multiplier in calculations, as explained inhe detailed model description (Appendix A)). Twenty fourarms scored zero for B; these were farms located suchhat there were no upstream farms, fisheries, wild popu-ations or processors within 5 km. Just one farm scored 0or Theme D. Only one farm scored above 0 for Theme Con-site processing); this farm scored the maximum valuef 1.

Theme E (Distance-independent mechanical transmis-ion) has a slightly skewed unimodal distribution. Theinimum and maximum possible scores for Theme E are 0

nd 1, but the case study farms scored between 0.0249 and.5325. The most common ranges of values for Theme E is.3–0.4 (Fig. 4).

The rankings remained largely the same across all 4cenarios (Supplementary material, Figs. VIII and IX). Theiggest differences are found between scenarios 2–3 and–4, respectively. This is more apparent for the Italianarms.

. Discussion

The model was developed to facilitate risk ranking ofPBs for a range of pathogens and aquaculture environ-ents (cyprinid and salmonid aquaculture) and expert

stimates to parameterise the model are available forour pathogens listed by EU Council Directive 2006/88/EC:HSV, IHNV, KHV and Infectious salmon anaemia virus

ISAV) (Oidtmann et al., in press).

.1. Validity of the model—The case study

The risk ranking model was applied to risk score 76 rain-ow trout farms in 3 countries (42 from England, 32 from

taly and 2 from Switzerland). These countries represent aariety of situations with regards to VHSV infection status,hich was reflected in the scorings results.

The results from both England and Italy provided aange of risk scores indicating a potential to discriminateisk in two settings. To validate the model an indepen-ent measure of risk is needed. Infection status of thearm is clearly one measure. All of the 10 farms fromtaly which were in Category V were diagnosed as infected

ith VHS at some point since the year 2000. Their scoresndicate that model results are consistent with a highisk of VHSV. Fuller validation requires a larger randomample of farms from a category V area (i.e. where infec-ion is known to be present). However, since the disease

ay have been introduced in the past when the riskcore was different, ideally a prospective study would bendertaken.

The purpose of risk ranking is to discriminate risk leveletween farms for risk-based sampling. Thresholds forigh risk (farm to be sampled) need to be established



ithin each country, on the basis of how different theisk-level between farms is, and how many farms needo be tested in order to achieve the target confidence inreedom.

PRESSry Medicine xxx (2014) xxx–xxx 9

4.2. Model assumptions and limitations

The rankings generated using the four scenarios tendedto be fairly similar. It therefore appears that the choice ofscenario is of minor influence. In the expert consultation,experts provided weights for the 5 themes for scenarios1–4, but there was no separate consultation for parameterestimates within each of the 5 themes for each scenario. Theestimates for parameters within the themes were elicitedfor scenario 1 only. Therefore, the only difference when cal-culating farm scores for each theme are the theme weights,which may explain to a large extent why farms were rankedsimilarly across the 4 scenarios.

However, full validation has not been completed (e.g.due to lack of data, infection status of nearby farms wasnot available for the Italian case study sites). Regions orcountries enrolling on a programme to demonstrate free-dom from disease would not have (knowingly) had recentoutbreaks of the disease, or if they did, they would havebeen successfully contained. The country (or area) maystart from a history of no surveillance, or passive surveil-lance only, or from a history of active surveillance of limitedgeographic coverage.

Given that there were differences in the theme weights,it would probably be most appropriate to use the scenarioweights that best fit the situation: where it is a countryembarking on a programme to demonstrate freedom fromdisease, either scenario 1 or 2 should be used. Where it isa zone, it should be scenario 3 or 4. Which of these are themore appropriate to choose would depend on the expecta-tion of the prevalence in the country or zone.

The analysis of important risk factors was hampered bydifficulties accessing detailed data:

(1) for geographical factors, distances (water and direct)had to be obtained farm by farm for several of nearbyhazards. At present, there are no readily availabledatabases that could provide the required information.

(2) Status of farms: The status of individual nearby farms(risk rank) has a relevant impact on the status of thefarm being analysed. In the model, crude single esti-mates of the risk of neighbouring farms being infectedwere used. In a national system, all farms should beranked and georeferenced. This would allow the riskposed by individual neighbouring farms to be calcu-lated more accurately.

(3) Fish and egg movements play an important role in themodel. Data were obtained for a 12 month period, butit is recognised that fish and egg movements can behighly variable. One option to address this would beto use rolling averages over several years to smooththis variability. However, a better option would be touse more precise data, so the actual current risk can becalculated. This would require an up-to-date record ofall movements between farms.

We applied the same value of 0.33 for the likelihood


of Cat III source farms to be infected when calculatingfarm risk scores for farms in Italy and Switzerland. Thisvalue is likely to be an overestimation for Cat III farmsin Switzerland. Switzerland has not categorised its fish


ING Model

Veterina



farms into disease Categories I–V, which is why all farmssupplying live fish or eggs to the two farms in the studywere classified as Cat III. However, because Switzerlandhas a strict eradication policy if VHSV was detected, thelikelihood of Cat III source farms to be infected is prob-ably significantly lower than 0.33. Using this high valuehas resulted in high farm scores for the Swiss farms.Likelihood estimates for Cat III farms (for likelihood ofbeing infected) need to be adjusted to reflect the circum-stances in the region or country to which the model isapplied.

4.3. Evidence supporting the structure of the model

In the process of designing the model, we reviewed thepublished scientific literature to inform the structure of themodel and identify suitable data to parameterise it. It wasfound that the published data supported the relevance ofthe 5 risk themes, but that they were insufficient to provideweights for these or probability estimates for pathwayswithin the themes.

Evidence found in the scientific literature that supportsthe 5 risk themes for VHSV is summarised below.

VHSV has a broad range of susceptible species, includingseveral salmonid species, as well as a broad range of wildand farmed freshwater and marine species belonging to arange of families (Meyers and Winton, 1995; Skall et al.,2005; EFSA, 2008).

There is clear evidence that infected rainbow troutexcrete virus via urine and ovarian fluids (Neukirch, 1992),contaminating the surrounding water. The virus is subse-quently taken up by other fish via the gills (Meier et al.,1994). There is also the possibility of transmission due todirect fish-to-fish contact as VHSV has been found to repli-cate in epidermal tissues of rainbow trout (Oncorhynchusmykiss) (Yamamoto et al., 1992). The above data clearlysupport transmission via live fish.

VHSV can survive for several months in freshwater, andfor weeks in saltwater. The length of survival increasesas water temperature decreases (Ahne, 1982a,b; Hawleyand Garver, 2008). Although some data on the levels ofvirus shed in the urine is available (up to 105 TCID50 ml−1;(Neukirch, 1992)), no data are available on the quantitiesof virus discharged from infected farms. Similarly, no pub-lished data exist on the minimum infectious dose (MID) ofVHSV genotype 1a. The above highlights the risk associ-ated with exposure of farms due to presence of upstreamfarms.

Several of the wild fish species that may be presentin the vicinity of salmonid trout farms are susceptible toVHSV infection (e.g. Atlantic salmon (Salmo salar), pike(Esox lucius), brown trout (Salmo trutta), and grayling (Thy-mallus thymallus)).

As VHSV can be transmitted through the water, it ispossible that virus released from infected wild fish couldinfect farmed fish. There is some evidence of VHSV beingtransmitted from wild pike and brown trout to rainbow



trout fry (Konrad et al., 1984; Enzmann et al., 1993, 1992,1987).

Although there is no true vertical transmission of VHS(Vestergård Jørgensen, 1970; Bovo et al., 2005), the virus


may be transmitted between farms via surface contami-nated eggs (Bovo et al., 2005). Egg disinfection proceduresare nowadays standard industry practice and are highlyeffective at rapidly inactivating VHSV (Joergensen, 1973;Tuttle-Lau et al., 2010), justifying different settings in themodel depending on whether eggs have or have not beendisinfected.

Fish processing can pose a relevant risk for pathogenrelease as fish tissues may carry substantial virus quanti-ties; various types of processing may be undertaken, withfilleting generating substantial quantities of both solid andliquid waste (Oidtmann et al., 2011a,b; Pearce et al., 2014).Risk of exposure of farmed fish populations depends onlevel of containment of any waste and safe waste disposal.

VHSV can survive for several days in mud or air (Ahne,1982b). VHSV can therefore survive on fomites (e.g. shoes,vehicles, gear) for long enough to spread infection betweensites if equipment is not thoroughly dried. Biosecuritypractices (such as drying or disinfecting equipment andclothing on leaving/arriving at sites) can mitigate the riskof pathogen spread in this way.

VHSV can be transferred by piscivorous birds as exter-nal mechanical vectors (Olesen and Vestergård Jørgensen,1982; Peters and Neukirch, 1986). Birds and mammals (e.g.otters, rats) are known pests of fish farms, and while meas-ures are taken by fish farmers to prevent animals enteringfarms, these are not 100% effective.

The literature clearly supports that the pathways forpathogen transmission captured in the 5 risk themes of themodel are relevant, but it is insufficient to provide weightsfor the 5 risk themes or probability estimates. At the sametime, the lack of data in the published literature also high-lights that there is scope for field and experimental studiesto fill some of these data gaps.

4.4. Future developments of model

Further improvements could be made to the model bytaking into account number of upstream sources in the sim-ple model, factoring in live fish and egg movements ontosite over a longer time period than 12 months and by mak-ing a more detailed assessment of risks arising from on-siteprocessing. Provisions have been made in the model to takeinto account stage of egg incubation (green or eyed) andwater (virus free or not) used, where no egg disinfectionwas applied. However, parameter estimates remain to beobtained from experts.

Future studies could be undertaken to test the appli-cation of the model to KHV and IHNV using parameterestimates obtained through expert consultation (Oidtmannet al., in press). With some modifications, the model canalso be applied to marine aquaculture sites to risk rank sitesfor ISAV.

The work presented here has been undertaken as a ret-rospective study. To further validate the model, infectionstatus of fish farms following risk scoring would be moni-tored in a prospective study—in countries with or without


a programme to achieve disease free status.The model presented focuses on risk of pathogen intro-

duction. Further extensions could be envisaged that takelikelihood of detection into account (e.g. due to factors


ING ModelP

Veterina

id

4

voRhandstbapfddhgr

4

apshin

C

A

pEchTm

A

A

A

tsfin



nfluencing disease expression and therefore likelihood ofetection if infection was present).

.5. Efficiency and cost effectiveness

Accuracy and currentness of the farm data are rele-ant to obtain valid farm scores. For this study, data werebtained through farms visits. To achieve savings throughBS (where visits are undertaken to high risk farms atigher frequency and low risk farms may not be visitedt all) will require that the collection of farm data is orga-ised in a cost saving manner. This could be achieved byeveloping a self-reporting system that facilitates submis-ion of farm data by farmers and private veterinarians (e.g.hrough web-based portals); the validity of farm data cane checked through visits to randomly selected sites. Ide-lly, such a system would be developed as a European widelatform and database, where it would provide a resourceor countries that cannot afford development of their ownata capture/management systems. Besides capturing theata required for risk-ranking, the database would alsoold useful real-time traceability data to respond to emer-ency disease outbreaks and allow to better understand theisk posed by movement patterns.

.6. Conclusions

The presented model provides a tool for risk rankingquaculture sites for risk of pathogen introduction. It hasotential to be applied to a range of pathogens. Substantialavings can be achieved by focussing surveillance effort onigh risk farms; however, to fully realise the potential sav-

ngs, more cost efficient ways of farm data collection areeeded.

onflict of interest

The authors state that there is no conflict of interest.

cknowledgements

The work was undertaken in a cooperation Art 36roject of the European Food Safety Authority (EFSA), theuropean Agency responsible for the assessment of andommunication on risks related to food and feed, plantealth, animal health and welfare in the European Union.his work was co-funded by the Department for Environ-ent, Food and Rural Affairs (defra) project FC1201.

ppendix A. Description of the model

.1. Theme risk score calculation

.1.1. Theme A—Live fish and egg movementsThis theme deals with the introduction of pathogen

hrough consignments of infected live fish (for on-growing,



tock enhancement or processing) and contaminated eggsrom other source sites, but not including pathogen presentn transport water or contaminated transport tanks, lorries,ets or packaging (see Theme E).


A.1.1.1. Live fish introductions (LFI). The factors taken intoaccount when estimating the risk associated with consign-ments of live fish from a particular source are: (i) species,(ii) source site status and (iii) number of consignmentsreceived.

The probability of introduction of infection with a sin-gle consignment is estimated as the species risk (expertestimate. For experts’ estimates for VHSV: see Table 3) mul-tiplied by the probability that the source site is infected(source status). If it is known to be infected, then theprobability is 1, otherwise a constant for source sites notknown to be infected was used (see Section A.1.2 of thisAppendix).

consignment riskfish = species risk × source status

Where multiple consignments were received from thesame source site, the risk over all consignments is calcu-lated as

source riskLFI =1 − (1 − species risk × source status)number of consignments

The overall risk for introduction through fish move-ments from multiple sources was calculated as

total riskLFI = 1 − ˘(1 − source riskLFI)

The result is a value between zero and one that could beinterpreted as a probability of introduction via this path-way.

A.1.1.2. Egg introductions. The probability estimates forpathogen introduction with a single consignment of disin-fected or non-disinfected eggs (non-disinfection risk) werebased on expert estimation. The consignment risk was cal-culated as:

consignment riskeggs = non-disinfection risk

× source status

The source risk was cumulative across the number ofconsignments and was calculated as:

source riskeggs = 1 − (1 − non-disinfection risk

× source status)number of consignments

The overall risk due to the introduction of eggs is calcu-lated as:

total riskeggs = 1 − ˘(1 − source riskeggs)

The overall risk due to introduction of both live fish andeggs is calculated on the assumption that each risk repre-sents a probability:

overall site riskLFI & eggs = 1 − ((1 − total riskLFI)

× (1 − total riskeggs))

During the expert consultation, some experts suggested


that for non-disinfected eggs, stage of incubation (green oreyed eggs) and water used during incubation (virus free ornot) were relevant to the risk of transmission. These arerelevant where no disinfection is applied.


ING Model

Veterina

s



To enable taking into account these additional factorsin future, a second option for calculating risk for the intro-duction of pathogen via eggs was created.

In the more comprehensive version, a matrix of risk esti-mates was developed to include the incubation status andthe water used, to produce a water/incubation risk. If thebatch was disinfected, the disinfection risk is used. If thebatch is not disinfected, the water/incubation risk is usedresulting in an overall egg risk. No parameter estimates forlikelihood of pathogen transmission depending on stageof incubation (green or eyed eggs) and water used dur-ing incubation (virus free or not) have been elicited. In themodel, temporary place-holder values have been used untilbetter estimates can be made. Using the more complex ver-sion for calculating risk of pathogen introduction via eggs(egg C) the consignment risk is calculated as

consignment riskeggs C = egg risk × source status

The source risk is calculated as:

source riskeggs C =

1 − (1 − egg risk × source status)number of consignment

In the default model, risk of pathogen introduction viaeggs is calculated using the simple version of the totalriskeggs.

A.1.2. Constants for sites not known to be infectedConstants were as follows: Even where potential source

sites were considered not to be infected, the likelihood ofbeing infected was set to 0.02 (2%) (instead of 0) whenthe model was run using settings for scenario 1 and 3 andto 0.05 (5%) when the model was using settings for sce-nario 2 and 4 to account for a potential (yet undetected)presence of VHSV in the country. The same constants wereused in Themes B–D for source sites considered not to beinfected.

A.1.3. Theme B—Exposure via waterThis theme deals with the direct introduction of

pathogen into sites’ supply water by susceptible fish popu-lations and activities upstream: i.e. other farmed and wildstocks (including fish released by farms for restocking orconservation purposes) and fish processing.

Classifications for upstream source types are: (i) farm,(ii) wild, (iii) fishery and (iv) processing.

Source type was interpreted as measure of probability.Distance is included in the model in two alternative

ways. In the simple (default) version, distance is consid-ered dichotomously as a rule out factor2: Sources within5 km (upstream water distance) of the site being assessedpose a risk, while sources beyond 5 km do not. The 5 km



threshold was based on experience of Cefas experts andused in questions presented to the experts in the expertconsultation.

2 Within a risk pathway, if there is any factor that completely eliminatesthe risk, the probability is set to zero.


A.1.3.1. Simple version. In the simple version, site risk (ofthe farm to be assessed) is calculated based on whetherwater source was safe (bore or spring, a rule-out fac-tor), whether the upstream source has susceptible speciespresent (a rule-out factor), the source status (infected ornot; using a constant for sources not known to be infected;see Section A.1.2 of this Appendix) and the upstream sourcetype risk based on expert estimates. The upstream sourcerisk (considering only sources within 5 km water distance)is calculated as:

upstream source risk = species susceptibility

× source status

× upstream source type risk

The total risk due to sources within 5 km upstream iscalculated as:

site risk(exposure via water) = source water unsafe

× [1 − ˘(1 − upstream source risk)]

Each site type only contributed once to the Siterisk(exposure via water) score (i.e. presence of additionalsites of the same type did not alter the score). If atleast one source of a given source type was infected,‘Upstream source risk’ was calculated assuming source sta-tus ‘infected’.

A.1.3.2. Distance-specific version. In the more preciseversion, distance is treated quantitatively. A simple expo-nential decay curve was fitted empirically to experts’ riskestimates using five data points, representing relative riskwith distance. Risk was set at 1 for a distance of zero metres,and zero for a distance of 5 km. Intermediate values werebased on experts’ estimates for the relative importance ofdistance in the spread of infection from nearby sources(<500 m, 500 m–2 km, 2–5 km) see Supplementary mate-rial Fig. X. The distance risk, which may be interpreted as arelative risk with an upper bound of 1 and a lower boundof zero, is calculated as:

distance risk = 1e distance (km)

Distance risk ranges between 1 for a distance of zeroand decreases asymptotically towards zero, with a value of0.007 at a distance of 5 km. Distance risk was interpreted asa probability. In this version, source type risk was convertedfrom a simple probability estimate to a relative risk (with anupper bound of 1), to be used as a modifier of the distancerisk.

The upstream source risk is calculated as:


× source status

× source type relative risk


× distance risk

In the distance-specific version, site risk was calculatedtaking into account all upstream sources (multiple sources


ING ModelP

Veterina

of

Ap

•

•

•

A

to

oftp

(sc

l(oVra

s

Ap

•

•



f a given type were taken into account) using the formulaor Site risk(exposure via water) shown above.

.1.3.3. Assumptions. A number of assumptions and sim-lifications have been made in this part of the model:

In the distance explicit version, the simple exponentialdecay model used was based on experts’ input for shortdistance mechanical spread, not through the water sys-tem. It may therefore not be entirely applicable to thismode of spread.Direct estimates of the risk based on a source type of‘processing plant’ were not included in the expert consul-tation risk questions, and was derived from other suitableresults of the expert consultation.In the expert consultation, there was a category of ‘other’for source of spread through water. The main sourceidentified by experts under this category was flooding,however this was assigned a very low risk so has beenignored in the model.

.1.4. Theme C—On-site processingThis theme deals with the introduction of pathogen

hrough infected dead fish sourced from other sources forn-farm processing.

Three rule-out factors have been identified: (i) Absencef on-site processing, (ii) Absence of movements of fishrom other sources to the farm for processing, (iii) Sys-ems to completely prevent water or solid wastes from therocessing coming into contact with farmed fish on the site.

The source risk is calculated on the basis of: (i) speciessusceptible to the disease or not; a rule-out factor), (ii)ource status (known infected or not), and (iii) number ofonsignments moved onto the farm for processing.

Expert consultations elicited risk estimates for like-ihood of disease transmission with a consignmentconsignment riskon site processing) depending on the statusf source (known infected or not (for experts’ estimates forHSV: see Table 3). The overall source risk is the calculatedisk based on the number of consignments moved onto sitend the susceptibility of the species:

source risk = species susceptibility

× (1 − (1 − consignment risk)number of consignments)

The overall theme risk is combined across sources:

ite riskprocessing = 1 − ˘(1 − source risk)



.1.4.1. Assumptions. The following assumptions and sim-lifications were made:

Risk accumulates per consignment, rather than per fishprocessed.Species are either susceptible or not, with no differentialfor likelihood of transmission


• Risk of contact between processing and fish on site isdichotomous (1 or 0)

A.1.5. Theme D—Short-distance mechanical transmissionThis theme deals with the introduction of pathogen

from sources in close proximity to the farm through routesincluding piscivorous birds or other animals.

Factors influencing the source risk include: (i) The typeof source (farm, fishery, wild), (ii) Infection status of source,(iii) Distance to source, (iv) Area (size) of farm site beingassessed and (v) Species on nearby source.

The absence of susceptible species at the nearby sourceis treated as a rule-out factor. The risk associated withdifferent types of nearby sources (source type risk) wasestimated by experts and converted to a relative risk withan upper bound of 1. The infection status of the source useda constant when the source was not known to be infected(see Section A.1.2 of this Appendix). The importance of thearea (size) of the current farm was estimated by expertsand was converted to a relative risk with an upper boundof 1.

The source risk was calculated as:

source risk = species susceptibility × source status

× source type risk × distance risk × area risk

where ‘distance risk’ can be either calculated based on acategorical or continuous approach, similar to the methoddescribed for Theme B. In the categorical approach, esti-mates provided by experts for risk of short-distancemechanical transmission for three distance ranges: <500 m,500–2000 m, > 2000–5000 m are used as the ‘distance risk’(for experts’ estimates for VHSV: see Table 3). In the con-tinuous distance version, a simple exponential decay modelfor the effect of distance is used, as previously described forTheme B.

The continuous approach requires detailed informationon the hydrological proximity of all sources to be effectivelyimplemented. The categorical approach is used by default.

The overall cumulative risk across nearby sites is calcu-lated in the same way as previously described:

site riskshort dist = 1 − ˘(1 − source risk)

A.1.6. Theme E—Distance-independent mechanicaltransmission

This theme deals with the introduction of pathogenvia contaminated fomites. Experts assigned weights (forexperts’ estimates for VHSV see Table 3) to a series ofdichotomous risk factors (present/absent), grouped undera number of headings:

Resources shared with other farms

• Shared personnel• Shared equipment


Visits to site

• Unauthorised people• Unauthorised vehicle


ING Model

Veterina



• Fish delivery vehicles• Other vehicles

Disinfection

• Of vehicles• Of people

Other factors

• Receiving waste from other farms• Fishery on site

Farm-level data consists of a yes or no for each risk factor(factor present or absent). This is coded as 1 (present) and0 (absent). The risk score for each factor is calculated bymultiplying the factor data (1 or 0) by the factor weight.The overall site risk for the theme is calculated by summingall the scores.

site riskdist indep = ˙(factor status × weight)

A.2. Site risk score calculation

The final risk score is then calculated as the weightedaverage of the theme scores (see Fig. 2), with weights esti-mated through expert consultation (for experts’ estimatesfor VHSV see Table 2):

final farm site risk score = ˙(weight × theme score)

A.3. Glossary

‘aquaculture production business’ (as defined by Coun-cil Directive 2006/88/EC) means any undertaking,whether for profit or not and whether public orprivate, carrying out any of the activities relatedto the rearing, keeping or cultivation of aquacul-ture animals.

‘Area risk’ relative risk of farm becoming infected basedon farm size. Parameter estimates based onexpert consultation.

‘Consignment’ all fish (or eggs) delivered to a site from asingle source on a single day

‘Consignment risk’ risk of the assessed farm becominginfected due to the movement of a single con-signment of live fish or eggs. It is calculated as

consignment risk = species risk × source status

‘Consignment risk processing’ likelihood of the assessedfarm becoming infected due to the movement ofa single consignment of dead fish onto farm forthe purpose of processing.

‘Distance risk’ relative risk of farm becoming infecteddue to presence of VHSV infected fishery or farm



within a given distance. Parameter estimatesbased on expert consultation.

‘Distance-specific model’ Scores for a given farm sitecan be calculated using simple (categorical)


or distance specific settings. In the distance-specific model, parameter estimates obtainedfrom experts were used to provide a moreadvanced way of calculating risk via exposure viawater or short-distance mechanical transmission(compared to the ‘simple model’).

‘Egg risk’ The risk of introduction of a respective pathogenwith a single consignment of eggs onto theassessed farm site. Uses information about pres-ence/absence of disinfection, infection status ofwater used for incubation and incubation stageof eggs. Is used for the calculation of sourceriskegg in the more comprehensive version of themodel.

‘farm’ (as defined by Council Directive 2006/88/EC)means any premises, enclosed area, or instal-lation operated by an aquaculture productionbusiness in which aquaculture animals are rearedwith a view to their being placed on the market,with the exception of those where wild aquaticanimals harvested or caught for the purpose ofhuman consumption are temporarily kept await-ing slaughter without being fed. A farm may bemade up of multiple ‘farm sites’.

‘farm site’ means a single premise, enclosed area, orinstallation operated by an aquaculture produc-tion business.

‘non-disinfection risk (eggs)’ Likelihood of pathogentransmission from a subclinically infected sourcesite to the receiving farm site with a singleconsignment of disinfected or non-disinfectedfish eggs. Parameter estimates obtained throughexpert consultation.

‘Risk’ relates to the probability of the assessed farmbecoming infected, but is not a quantitative like-lihood estimate. It is expected that as probabilityof infection increases, risk increases (and viceversa).

‘Risk factor’ Condition or property influencing the likeli-hood of a site becoming infected via a particularrisk pathway (e.g. species risk, source status. . .).

‘Risk’ (/transmission) pathway: a pathway for thetransmission of the hazard (here a fish pathogen).A risk pathway (opposed to risk theme) is aspecific route of disease transmission (e.g. usingequipment belonging to other fish farms; bring-ing fish onto site for processing).

‘Risk score’ The score calculated for a given fish farmsite either at ‘risk theme’ level or for all themescombined, based on location (relative to potentialsources), pathogen introduction pathways, andbiosecurity practices that apply for the farm site.

‘Risk theme’ A group of risk pathways for the transmis-sion of the hazard (here a fish pathogen) sharingfeatures or mode of transmission. The five riskthemes used are: (1) Live fish & egg movements;(2) Exposure via water; (3) On-site processing;


(4) Short distance mechanical transmission; and(5) Distance independent mechanical transmis-sion. The individual risk themes are defined inTable 1.


ING ModelP

Veterina

‘

‘

‘

S

‘

‘

‘

‘

•



Risk weight’ a relative significance assigned to a risktheme (assigned to a risk theme by experts).

Rule out factor’ Within a risk pathway, if there is anyfactor that completely eliminates the risk, theprobability is set to zero.

Scenario’ During the expert consultation, the expertswere consulted on four scenarios, characterisedby variations in farm level prevalence of infectionwithin the country and location of the farms con-sidered. Scenario 1: 2% farm site level prevalencethroughout the whole country; scenario 2: 5%farm site level prevalence throughout the wholecountry; scenarios 3 and 4 assumed that therewere approved disease-free (i.e. category I) zoneswithin the country, which otherwise had a farmsite level prevalence outside of these zones of 2or 5%, respectively.

imple (default) model where options exist (e.g. morecomprehensive way of calculating a score for apathway, e.g. for risk of pathogen introduction viaeggs, or water), the simple or default model usesthe simpler version of the options available.

Site risk’ is the probability of the farm site assessed tobecome infected via one of the 5 risk themes.Using the example of live fish movements, ‘siterisk’ (live fish movements) is the risk score forthe assessed farm site of becoming infected viaall live fish movements onto site.

Site risk score’ Final score for the assessed farm site beingassessed.

Source’ The potential origin of infection. For consign-ments of live fish a source will be the supplyingfarm. In other risk themes, sources also includeprocessors, wild fish populations, and fisheries.Use of the term ‘source’ is not consistent withan assumption that the source is actually infectedand therefore an actual potential source of infec-tion. However, the term ‘source’ is used todistinguish between potential sources of infec-tion to the farm site that is being assessed, andthe farm site itself. Unless specified otherwise,a source is always a single source (e.g. a singlefarm site upstream, a single fishery upstream)and not the combination of all sources within arisk theme.

Source risk’ The risk of the assessed farm site becominginfected due to the introduction of the pathogenin question from a single source. A source risk iscalculated (used) in risk themes (1) Live fish & eggmovements; (2) Exposure via water; (3) On-siteprocessing; and (4) Short distance mechanicaltransmission.

In Theme live fish (or egg) movements, source risk is therisk of the assessed farm site becoming infected due tothe introduction of one (or more) consignments from asingle supplier (source). It is calculated as



Source (supplier) risk = 1

− (1 − species risk × source status)number of consignments


• In Theme ‘Exposure via water’, source risk is referredto as ‘upstream source risk’. Upstream sources can befish farms, wild fish populations, fisheries and processingsites. Upstream source risk is the risk of the assessed farmsite becoming infected due to the presence of a specificpotentially infected source upstream. Upstream sourcerisk is calculated as


× source status

× upstream source type risk

× distance risk

• In Theme ‘On-site processing’, a ‘source’ is the supplier offish for processing. ‘Source risk’ is the risk of the assessedfarm site becoming infected due to the introduction ofone (or more) consignments of fish for processing froma single supplier (source) for processing. It is calculatedas

source (supplier) risk = species susceptibility

×(

1 − (1 − consignment risk)number of consignments)

• In Theme ‘Short-distance mechanical transmission’, riskassociated with nearby sources can be calculated using2 different methods. In the distance specific method,source risk is the risk of the assessed farm site becominginfected due to presence of a specific potentially infectedsource nearby. It is calculated as

source risk = species risk × source status × source type risk

× distance risk × area risk

Details on the individual components of the equationsare provided elsewhere in this document.

The source risk may equal 0, if no susceptible fish onpresent at the source.

‘Source status’ Infection status of a potential source ofinfection. The source can be a processor, wild fishpopulation, fishery or farm. The true infection sta-tus is often not known. In case of the potentialsource being a farm, the declared infection sta-tus of a farm (2006/88/EC)) may be used as aproxy. Infection status categories are: category I:approved pathogen free; category II: not declareddisease free, but subject to a surveillance pro-gramme to achieve disease free status; categoryIII: infection status is unknown; category IV: sub-ject to an eradication programme; category V:known to be infected. Since there are multiplenotifiable fish diseases, a single farm may be indifferent categories for different pathogens (e.g.in category I for VHS, and category IV for IHN).A preset constant is used for score calculations in


Themes A-D. For potential source sites considerednot to be infected, the constant was 0.02 when themodel was run using settings for scenario 1 and3 and to 0.05 when the model was using settings


ING Model

Veterina



for scenario 2 and 4 to account for a potential (yetundetected) presence of VHSV in the country.

‘Source type’ Source types are fish farms, processors, wildfish populations, and fisheries.

‘Source type risk’ The risk associated with the type ofsource, either a probability estimate or expressedas a relative risk associated with source type;parameter estimates obtained through expertconsultation.

‘Species risk’ In Theme ‘live fish & egg movements’,‘species risk’ is the probability that a movementof (a single consignment of) live fish of a givenfish species from a site subclinically infected withthe pathogen in question leads to infection at thereceiving site. In Themes ‘Exposure via water’,‘On-site processing’ and ‘Short distance mechan-ical transmission’, information about species isonly used as a rule-in or rule-out factor. Riskis only present if susceptible species are moved(Theme ‘On-site processing’) or present at thesource (Themes ‘Exposure via water’, and ‘Shortdistance mechanical transmission’). Parameterestimates obtained through expert consultation.

Appendix B. Supplementary data

Supplementary data associated with this article can befound, in the online version, at http://dx.doi.org/10.1016/j.prevetmed.2014.04.005.

References

Ahne, W., 1982a. Untersuchungen zur Tenazität der Fischviren. Fortschr.Vet. 35, 305–309.

Ahne, W., 1982b. Vergleichende Untersuchung über die Stabilität von vierfischpathogenen Viren (VHSV, PFR, SVCV, IPNV). Zentralbl. Veterin-armed. 29, 457–476.

Anon., 2006. Council Directive 2006/88/EC of 24 October 2006 on animalhealth requirements for aquaculture animals and products thereof,and on the prevention and control of certain diseases in aquatic ani-mals. Off. J. Eur Union L328, 14, 43.

Anonymous, 2013. 2013/706/EU: Commission Implementing Decision of29 November 2013 amending Annex I to Decision 2009/177/EC asregards the disease-free status of Denmark for viral haemorrhagicsepticaemia and of Ireland and Northern Ireland territory of TheUnited Kingdom for Koi herpes virus disease (notified under docu-ment C(2013) 8385) In: European Commission (Ed.), Official Journal L322, 03/12/2013 P. 0042-0043.

Bang Jensen, B., Kjær Ersbøll, A., Korsholm, H., Skall, H.F., Olesen, N.J., inpress. Spatio-temporal risk factors for viral haemorrhagic septicaemia(VHS) in Danish aquaculture. Dis. Aquat. Org.

Baruchelli, G., Zanin, E., Allegretti, M., Bovo, G., Ceschia, G., Giorgetti, G.,1990. Official plan for the cure for Viral Haemorrhagic Septicaemia ofsalmonids in Trento District: analysis of the results achieved in tenyears time. Rivista di idrobiologia. Monte del Lago sul Trasimeno 29,105–111.

Bovo, G., Hill, B., Husby, A., Hastein, T., Michel, C., Olesen, N.J., Storset, A.,Midtlyng, P.J., 2005. Pathogen survival outside the host, and suscep-tibility to disinfection. European Commission Fish Egg Trade report:work package 3. Project VESO-1601. VESO Danish Institute for Foodand Veterinary Research, Cefas, Inra, Veterinaerinstiuttet, IZS, Aqua-Gen, Oslo, Norway.



Diserens, N., Presi, P., Bernet, D., Schüpbach-Regula, G., Wahli, T., 2013.Risk assessment for the design of a risk-based surveillance programmefor fish farms in Switzerland (in accordance with Council Directive2006/88/EC of the European Union). Revue scientifique et technique(Int. Off. Epizootics) 32, 751–763.


EFSA, 2008. Scientific Opinion of the Panel on AHAW on a request fromthe European Commission on aquatic animal species susceptible todiseases listed in the Council Directive 2006/88/EC. EFSA J. 808, 1–27.

Enzmann, P.J., Konrad, M., Parey, K., 1993. VHS in wild living fish andexperimental transmission of the virus. Fish. Res. 17, 153–161.

Enzmann, P.J., Konrad, M., Parey, K., Wetzlar, H., 1987. NatürlichesWirtsspektrum des Virus der Viralen Hämorrhagischen Septikämieder Regenbogenforelle. Tierärztliche Umschau, 42.

Enzmann, P.J., Konrad, M., Rapp, J., 1992. Epizootiological studies on viralhaemorrhagic septicaemia in brown trout Salmo trutta fario. Dis.Aquat. Org. 12, 143–146.

Hawley, L., Garver, K., 2008. Stability of viral hemorrhagic septicemiavirus (VHSV) in freshwater and seawater at various temperatures. Dis.Aquat. Org. 82, 171–178.

Joergensen, P.E., 1973. Inactivation of IPN and Egtved virus. Rivista Italianadi Piscicultura e Ittiopatologia 8, 107–109.

Kleingeld, D.W., 2010. Census and risk analysis of Lower Saxony aquacul-ture production business against the background of fish epizooticslegislation. In: Fakultät für Agrarwissenschaften. Georg-August-Universität Göttingen, Göttingen, pp. 251.

Konrad, M., Rapp, J., Enzmann, P.-J., 1984. Epidemiological studies on VHSin a wild fish population. Bull. Eur. Assoc. Fish Pathol. 4, 69–70.

Lyngstad, T.M., Tavornpanich, S., Viljugrein, H., Bang Jensen, B., Hellberg,H., Brun, E., 2011. Evaluation of the surveillance programme for viralhaemorrhagic septicaemia (VHS) in Norwegian salmonid farms. Jour-nal Épidémiologie et Santé Animale 50/60, 108–110.

Meier, W., Schmitt, M., Wahli, T., 1994. Viral haemorrhagic septicaemia(VHS) of nonsalmonids. Annu. Rev. Fish 4, 359–373.

Meyers, T.R., Winton, J.R., 1995. Viral hemorrhagic septicemia virus inNorth America. Annu. Rev. Fish Dis. 5, 3–24.

Neukirch, M., 1992. VHSV-infected rainbow trout. Oncorhynchusmykiss—a model for the study of temperature dependant excretion ofviral agents. In: Shariff, M., Subasinghe, R.P., Arthur, J.R. (Eds.), Diseasesin Asian Aquaculture 1. Fish Health Section, Asian Fisheries Society,Manila (Philippines), Bali (Indonesia), pp. 273–280.

Oidtmann, B., Crane, C., Thrush, M., Peeler, E., Hill, B., 2009. A Model for RiskRanking Fish Farms to Inform Disease Risk-Based Surveillance. Soci-ety for Veterinary Epidemiology and Preventive Medicine (SVEPM),London, UK.

Oidtmann, B., Joiner, C., Reese, R.A., Stone, D., Dodge, M., Dixon, P.,2011a. Risks associated with commodity trade: transmission of ViralHaemorrhagic Septicaemia Virus (VHSV) to rainbow trout fry fromVHSV-carrying tissue-homogenates. Transboundary Emerging Dis.58, 224–231.

Oidtmann, B., Joiner, C., Stone, D., Dodge, M., Reese, R.A., Dixon, P., 2011b.Viral load of various tissues of rainbow trout challenged with viral hae-morrhagic septicaemia virus at various stages of disease. Dis. Aquat.Org. 93, 93–104.

Oidtmann, B., Peeler, E., Lyngstad, T., Brun, E., Jensen, B.B., Stark, K.D.C.,2013. Risk-based methods for fish and terrestrial animal diseasesurveillance. Prev. Veter. Med. 112, 13–26.

Oidtmann, B., Thrush, M., Peeler, E., 2008. A scenario tree evaluation ofsurveillance for Gyrodactylus salaris, a fish parasite exotic to the UK.In: Annual Meeting of the Society for Veterinary Epidemiology andPreventive Medicine, 26–28th March 2008, Liverpool.

Oidtmann, B.C., Crane, C.N., Thrush, M.A., Hill, B.J., Peeler, E.J., 2011c. Rank-ing freshwater fish farms for the risk of pathogen introduction andspread. Prev. Veter. Med. 102, 329–340.

Oidtmann, B.C., Peeler, E.J., Thrush, M.A., Cameron, A.R., Reese, R.A., Pearce,F.M., Dunn, P., Lyngstad, T.M., Tavornpanich, S., Brun, E., Stärk, K.D.C., inpress. Expert consultation on risk factors for introduction of infectiouspathogens into fish farms. Prev. Veter. Med.

OIE, 2012. Viral haemorrhagic septicaemia. In: Manual of DiagnosticTests for Aquatic Animals. World Organisation for Animal Health,Paris.

Olesen, N.J., Vestergård Jørgensen, P.E., 1982. Can and do herons serve asvectors for Egtved virus? Bull. Eur. Assoc. Fish Pathol. 2, 48.

Pearce, F.M., Oidtmann, B.C., Thrush, M.A., Dixon, P.F., Peeler, E.J., 2014.Do imports of rainbow trout carcasses risk introducing viral haemo-rrhagic septicaemia virus into England and Wales? TransboundaryEmerging Dis. 61, 247–257.

Peters, F., Neukirch, M., 1986. Transmission of some fishpathogenic viruses by the heron, Ardea cinerea. J. Fish Dis. 9,539–544.


Skall, H.F., Olesen, N.J., Mellergaard, S., 2005. Prevalence of viral haemo-rrhagic septicaemia virus in Danish marine fishes and its occurrencein new host species. Dis. Aquat. Org. 66, 145–151.

Smail, D.A., 1999. Viral haemorrhagic septicaemia. In: Woo,P.T.K., Bruno, D.W. (Eds.), Fish Diseases and Disorders.


http://dx.doi.org/10.1016/j.prevetmed.2014.04.005

http://dx.doi.org/10.1016/j.prevetmed.2014.04.005

http://refhub.elsevier.com/S0167-5877(14)00145-7/sbref0005



























































































































































































































































































































































































































































































































































































































































































































ING ModelP

Veterina

S

T



Viral, Bacterial and Fungal Infections, Vol. 3. CABI Publishing,Oxford, pp. 123–147.

tone, D.M., Ferguson, H.W., Tyson, P.A., Savage, J., Wood, G., Dodge,M.J., Woolford, G., Dixon, P.F., Feist, S.W., Way, K., 2008. The first



report of viral haemorrhagic septicaemia in farmed rainbow trout,Oncorhynchus mykiss (Walbaum), in the United Kingdom. J. Fish Dis.31, 775–784.

uttle-Lau, M.T., Phillips, K.A., Gaikowski, M.P., 2010. Evaluation of theefficacy of iodophors disinfection of walleye and northern pike eggs


to eliminate viral haemorrhagic septicaemia virus. In: US GeologicalSurvey Fact Sheet 2009–3107., pp. 4.

Vestergård Jørgensen, P.E., 1970. The survival of viral hemorrhagic sep-ticaemia (VHS) virus associated with trout eggs. Rivista Italiana di


Piscicultura e Ittiopatologia 5, 13–15.Yamamoto, T., Batts, W.N., Winton, J.R., 1992. In vitro infection of

salmonid epidermal tissues by infectious hematopoietic necrosis virusand viral hemorrhagic septicemia virus. J. Aquat. Anim. Health 4,231–239.























































































































Date post:	30-Dec-2016
Category:	Documents
Upload:	angus
View:	223 times
Download:	5 times

Model for ranking freshwater fish farms according to their risk of infection and illustration for...

Documents