An economic decision support tool for simulating paratuberculosiscontrol strategies in a UK suckler beef herd

Richard Bennett *, Isobel McClement, Ian McFarlane

Department of Agricultural and Food Economics, University of Reading, Earley Gate, Whiteknights Road, Reading RG6 6AR, United Kingdom

Preventive Veterinary Medicine 93 (2010) 286–293

A R T I C L E I N F O

Keywords:

Johne’s

Beef suckler herds

Costs and benefits

Economic

Deterministic decision support tool

A B S T R A C T

A dynamic, deterministic, economic simulation model was developed to estimate the costs

and benefits of controlling Mycobacterium avium subsp. paratuberculosis (Johne’s disease) in

a suckler beef herd. The model is intended as a demonstration tool for veterinarians to use

with farmers. The model design process involved user consultation and participation and the

model is freely accessible on a dedicated website. The ‘user-friendly’ model interface allows

the input of key assumptions and farm specific parameters enabling model simulations to be

tailored to individual farm circumstances. The model simulates the effect of Johne’s disease

and various measures for its control in terms of herd prevalence and the shedding states of

animals within the herd, the financial costs of the disease and of any control measures and

the likely benefits of control of Johne’s disease for the beef suckler herd over a 10-year period.

The model thus helps to make more transparent the ‘hidden costs’ of Johne’s in a herd and the

likely benefits to be gained from controlling the disease. The control strategies considered

within the model are ‘no control’, ‘testing and culling of diagnosed animals’, ‘improving

management measures’ or a dual strategy of ‘testing and culling in association with

improving management measures’. An example ‘run’ of the model shows that the strategy

‘improving management measures’, which reduces infection routes during the early stages,

results in a marked fall in herd prevalence and total costs. Testing and culling does little to

reduce prevalence and does not reduce total costs over the 10-year period.

� 2009 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Preventive Veterinary Medicine

journal homepage: www.elsevier.com/locate/prevetmed

1. Introduction

Johne’s disease is caused by the agent Mycobacterium

avium subsp. paratuberculosis (MAP) and is characterisedby gradual wasting and chronic enteritis, resulting indecreased production, reduced fertility, increased replace-ment rates (Chiodini et al., 1984) and loss of cull valuebecause diagnosed Johne’s diseased animals are notallowed to enter the UK food chain (FSA, 2001). Therefore,infection with Johne’s disease implies significant economiclosses to the herd (Kudahl et al., 2007; Ott et al., 1999).Johne’s is typically transmitted from dam to young via thefaecal–oral route or via contaminated colostrum and milk

* Corresponding author. Tel.: +44 0 1183786478.

E-mail address: r.m.bennett@reading.ac.uk (R. Bennett).

0167-5877/$ – see front matter � 2009 Elsevier B.V. All rights reserved.

doi:10.1016/j.prevetmed.2009.11.006

(Begg and Whittington, 2007). There is a long, progressivesubclinical phase with a proportion of infected animalsbecoming clinical cases between 2 and 6 years old, thoughthe range can be 4 months to 15 years (Caldow et al., 2001).The disease can be split into four stages according to theseverity of clinical signs and the potential for sheddingorganisms via faeces. For every clinical case there may beover 25 subclinical cases, some of which may be highlyinfectious (Whitlock and Buergelt, 1996).

There have been many approaches to modelling andmanaging the cost-effective control of Johne’s in dairycattle but far fewer in beef cattle and even less that haveattempted to provide simulated information to aiddecision making at farm level. The few model builderswho have attempted to model the impacts of Johne’s on asuckler herd have either adapted dairy models orapproached the epidemiology from a different perspective

R. Bennett et al. / Preventive Veterinary Medicine 93 (2010) 286–293 287

(Humphrey et al., 2006). The lack of models that simulateJohne’s in beef cattle could be due to the lack ofquantitative data about the disease in beef cattle (Hum-phrey et al., 2006).

The model presented here aims to provide a user-friendly, decision supporting, deterministic, simulationmodel to use with suckler beef producers to demonstratethe likely costs of Johne’s in their herds and the potentialcosts and benefits of different measures to control thedisease. The need for the model was identified by astakeholder group of cattle farmers and veterinarians inthe UK. This paper describes the modelling approach andmethods, provides some example model outputs andreports the feedback from veterinary advisers and othersas to how useful they found the model.

2. Materials and methods

2.1. Approach

A participative approach was used in the developmentof the model. This means that intended users andstakeholders (beef industry representatives, farmers,veterinarians and others) were closely involved andconsulted in an iterative design and development process.This helped to ensure that the resultant model met theneeds of users, that users had a sense of ‘ownership’ of themodels and that uptake of the model was more likely thanif it had been developed without appropriate consultation.

A three-stage approach was used in the development ofthe model. The first stage involved construction of aconceptual, diagrammatic model that charted the effects ofJohne’s disease and measures for its control on the sucklerbeef production system. This was used as a point ofdiscussion with veterinarians to ensure that all relevanteconomic impacts of Johne’s were included. Moreover, thisdiagrammatic model helped to communicate to veterinar-ians and users the basic concepts that lay behind theresultant computerised simulation model. The next stageinvolved transforming the conceptual model into acomputerised systems simulation model using Stella 91

(www.iseesystems.com) software. The Stella software usesa modelling layer in which stocks or populations aredefined, together with transition steps between popula-tions at preset time intervals and an equations layer inwhich initial values and transition rules are defined. Thefinal stage involved the design of a graphical user interface(GUI) to interact with the modelling layer, to enable usersto enter input values into the model that then generate anumber of different outputs as required by users. The GUIwas tested with users in workshops and modified in thelight of their comments.

2.2. Model specification

2.2.1. The suckler cow system

The suckler cow system modelled is assumed to bebroadly representative of most UK systems. All the animalsrun as one unit, cows have a default calving percentage of90% per annum (that can be changed by the user), calvesare kept in the herd until weaning at 6–7 months, a

proportion are then sold for finishing elsewhere and aproportion of heifer calves are kept for replacements (themodel adjusts the proportion kept based on the controlmeasures chosen). If there are not enough replacementsdue to heavy culling the model will cost in the price ofpurchasing in-calf certified Johne’s-free replacementheifers.

The model is deterministic and dynamic, and integratesthe epidemiological pathway of disease under differentcontrol scenarios with the economic consequences ofthose different scenarios.

2.2.2. Epidemiology

The epidemiological part of the model is herd based andso simulates the proportion of animals in sets of age groupsand disease states, as opposed to modelling the diseasestate of individual animals.

The course of Johne’s infection is defined by six mainstates (see Fig. 1); susceptible (those animals younger than1 year and not yet infected), resistant (animals over 1-year-old and not infected are assumed resistant), latent (earlyasymptomatic stage of the disease), low shedding (asymp-tomatic stage of the disease), high shedding (potentiallysymptomatic) and clinical (displaying severe symptoms).The key mechanism driving animals through each diseasestate is based on the current disease pressure (dictated bythe number of shedding animals present in the herd) andthe age of exposure. To simulate the presence of infectionin the model, the initial premise is that 10% of the animalsin the herd are in the low shedding state.

Input variables to the model to describe herd composi-tion, herd management, acquisition and disposal valuesand veterinary charges for diagnosis and treatment aresupplied by the user. The probability of infection iscalculated from data obtained from both the researchliterature and expert opinion.

At each time step, the number of animals to betransferred between groups is calculated, based on apredefined set of transition probabilities. The transitionsare calculated in a preset order, with appropriate timesteps to simulate the progression of disease prevalencewith time (the user can see a graphical depiction of this).

There are some 13 states and 29 transitions within themodel but the main ones relating to the course of Johne’sinfection are the states and transitions of (i) susceptibleanimals to infected/latent state, (ii) from latent to lowshedding, (iii) from low shedding to high shedding and (iv)from high shedding to clinical cases. These four transitionsare characterised by the following expressions:

It ¼ Pi� St�1 (1)

where It = number of animals becoming latently infected intime period t, Pi = probability of infection of susceptibleanimals and St�1 = number of susceptible animals in theprevious time period t� 1;

LSt ¼ ðPlsþ PdcÞ � It�1 (2)

where LSt = number of low shedding animals at time t,Pls = the probability of latently infected animals becominglow shedders and Pdc = the probability of a cow progres-

Fig. 1. Johne’s disease states and progression in a suckler herd.

R. Bennett et al. / Preventive Veterinary Medicine 93 (2010) 286–293288

sing to the next infection state at calving which is constantwithin the model at 0.32 (from Kudahl et al., 2007);

HSt ¼ ðPhsþ PdcÞ � ðLSt�1 � LSDt�1Þ (3)

where HSt = number of high shedding animals at time t,Phs = the probability of animals moving from low sheddingto high shedding state and LSDt�1 is the number of lowshedding animals that are culled from the herd in t� 1;and

CLt ¼ ðPclþ PdcÞ � ðHSt�1 �HSDt�1Þ (4)

where CLt = the number of clinical cases at time t, Pcl = theprobability of high shedding animals becoming clinicalcases and HSDt�1 is the number of high shedding animalsculled from the herd in t� 1.

Probabilities of moving from latent to low shedding tohigh shedding to clinical states are determined by thedisease pressure within the herd at any one time whichdepends on the weighted numbers of animals in each state(giving greater weight to clinical and high sheddinganimals compared to low shedding ones) where diseasepressure at time t (DPt) is defined as:

DPt ¼ðLSt=POPÞ þ ðHSt=POPÞ0:5 þ ðCLt=POPÞ0:25

3(5)

where POP = the total population of cows in the herd. It isalso assumed that Pls = DPt/4, Phs = DPt/2 and Pcl = DPt.

Transition probabilities are also influenced by herdmanagement choices, actual or hypothetical, suggested bythe user. Three infection paths are modelled, each with anassociated probability: (1) infection passed directly frominfected dam; (2) infection passed by faeces in the calvingarea; and (3) infection passed in colostrum and milk.Table 1 shows the probabilities associated with eachinfection route and their source. Parameter estimates arederived from the research literature.

In addition, transmission and progression of the diseaseuses six-weekly time steps from birth to 19 weeks (Kudahlet al., 2007) to take account of the rate at which theprobability of becoming infected diminishes during theyear after birth. After 19 weeks the model updates herddemographics and a proportion of the herd will move intoeach shedding state, or not if considered resistant (i.e. over1-year-old), every 6 months (Humphrey et al., 2006). Thismaintains a constant population size over time. It ispossible to use the model to simulate the effects of anumber of different control strategies, including combina-tions of measures.

2.2.3. Control strategies

Control strategies incorporated into the model includethe following.

2.2.3.1. Testing and culling. Test accuracy is known to varywith parity and infection state. Mean test sensitivities are

Table 1

Parameter values for Johne’s beef suckler model infection routes.

Infection route Probabilities Reference

Foetal infection (infected dams) 0.20 Groenendaal et al. (2002)

Foetal infection (clinical dams) 0.70 Groenendaal et al. (2002)

Perinatal infection (weeks 1–6)a f(MR) Adjusted Kudahl et al. (2007)

Perinatal infection (weeks 1–19)a,b f(MR + ER) Adjusted Kudahl et al. (2007)

Year 1 infection (weeks 19–52) f(ER) Adjusted Kudahl et al. (2007)

Test sensitivity—low shedding 0.15 ELISA serum test sensitivities derived from Dargatz et al. (2001) and Wells et al. (2006)

Test sensitivity—high shedding 0.80a MR is the ‘milk risk’ associated with dam’s shedding state contaminated milk and colostrum and is determined by disease pressure and calf rearing

practices.b ER is the ‘environment risk’ associated with the numbers and shedding states of animals in the herd at that time and is determined by both disease

pressure and farm environment.

R. Bennett et al. / Preventive Veterinary Medicine 93 (2010) 286–293 289

set within the model to reflect the difference between lowand high shedding animals (see Table 1).

2.2.3.2. ‘Test and cull mild’. This involves the annualtargeted testing of high shedding, higher parity animals.Positive animals are confirmed and then culled. Therelationship is given as:

CUðHSÞ ¼ RC�HS (6)

Only high shedders get culled in this ‘mild’ strategywhere CU(HS) is the number of high shedding animals thatare culled, RC is the rate of cull (a function of testsensitivity) and HS is the number of high shedding animalsat the time of testing.

2.2.3.3. ‘Test and cull severe’. This is similar to ‘test and cullmild’ but also includes the annual targeted testing andremoval of low shedding animals. The relationship is givenas:

CUðLSþHSÞ ¼ RC� ðLSþHSÞ (7)

Low and high shedders get culled in this ‘severe’strategy where CU(LS + HS) is the number of low and highshedding animals that are culled and LS + HS is the numberof low and high shedding animals at the time of testing.

2.2.3.4. ‘Improve management’. This aims to reduce thenumber of infected animals entering the herd and involvescalf hygiene and clean environment husbandry measuresaimed at reducing the milk and faecal routes of infection tothe neonatal and young calf. These measures are assumedto reduce all risks of infection by 50% (based on Kudahlet al. (2007) and Pennsylvania State University (2007)).This is achieved by measures including; ensuring adesignated clean calving area, rearing calves in separateage groups, washing machinery and other feed/water/bedding/waste equipment between young animals andolder cows and careful waste and slurry spreading andpasture management, which includes not allowing youngstock on recently spread fields. The relationship is given as:

ICt ¼ ICt�1 þ ðPf � CEPÞ (8)

where ICt is the number of infected calves at time t, Pf is theprobability of faecal infection and CEP is the number ofuninfected calves less than 6 weeks old.

2.2.3.5. ‘Improve management and test and cull severe’. Thisinvolves husbandry measures to reduce the milk and faecalroutes of infection and the annual targeted testing andremoval of low and high shedding animals. A reduction of50% in the environmental risk of infection is assumed.

2.2.4. Economic model structure

Quantifying the economic impact of Johne’s disease isbased on McInerney’s (1996) broad definition of the costsof disease, later defined by Bennett (2003) as direct diseasecost (C) where;

C ¼ ðLþ RÞ þ T þ P (9)

where L is the value of loss in output due to disease, R is theincreased expenditure on farm resources such as feed andlabour, T is the cost of increased inputs to treat disease andP is the cost of disease prevention.

The model uses a set of discounted financial valuesrepresenting costs and savings associated with the diseaseand the options for treatment during a period of 10 years.The main economic impacts of Johne’s included in themodel are (i) reduction in sales output (i.e. due to lowerweights of cull cows and calves), (ii) cow replacement costs(rearing and purchase of additional heifer replacements),(iii) mortality/fallen stock costs, (iv) additional veterinaryand farm labour costs and (v) control costs (costs of testing,improved management measures, etc.). The financialmodel represents animals as saleable up to the stagewhere risk of mortality is significant. The input parameterssuch as herd size, cull rate, calving percentage, in-calfheifer replacement value, calf value at weaning, cull cowvalue and national fallen stock charges (NFSS) havebaseline default values or the user can enter herd detailsthemselves (see Fig. 2).

The model output is split between information showingestimated numbers of animals in each disease state andinformation showing disease impacts. The number ofanimals in each disease state is dictated by the controlmeasure chosen which in turn drives the total costcalculations. At each time step, the financial effects areupdated to account for changes in valuations, includingcumulative (discounted) costs associated with theoption(s) selected regarding control strategy, given thecost parameters provided by the user for each simulation.

Supplementary on-farm labour costs due to Johne’s areincluded (on a scale from 10 h per month for up to 10

Fig. 2. The model’s graphical user interface.

R. Bennett et al. / Preventive Veterinary Medicine 93 (2010) 286–293290

animals in shedding states to a maximum of 50 h permonth for 250 animals in shedding states), as well asadditional veterinary charges resulting from the presenceof Johne’s disease in the herd. All these costs were scaled toincrease with herd size, but not in direct proportion. Forexample, some economies of scale apply to veterinarycharges when many animals are tested together. Test anddiagnostic charges respond to scale so that the moreanimals tested the cheaper the cost per animal—althoughthis is highly farm and handling system specific. Defaultcharges in the model shown are derived from interviewswith practising veterinarians.

Default estimation of the costs associated with man-agement measures to control Johne’s are included, basedon interviews with farmers and veterinarians to estimateper animal per week costs of implementing these changes.The result of the interviews suggested that a nominal costcould be £0.05 per animal per week. However, a verystrong caveat was attached to this estimate, namely thatthe figure entered into this box is likely to be highly farmspecific (and thus can be changed by the user by inputtingan alternative value).

2.2.5. Graphical user interface (GUI)

The development of the Johne’s suckler beef modelfocused not only on epidemiological and economicrelationships but also on managing the human interactionsand end users’ requirements. This influenced the design ofthe model’s GUI (Fig. 2).

The screen shot shows herd details that the user canchange in the top left hand side with veterinary inputs,including choice of disease control measure, shown below

which the user can also change. The model outputs areshown on the right side of the page with numbers ofaffected animals in different disease states shown at thetop and the associated economic costs below.

It is difficult to communicate to herd managers the likelyimpacts of Johne’s disease because test accuracy is depen-dent on the shedding state of the animal and productionlosses during the asymptomatic phase are difficult toquantify. Therefore, a core advisory group of veterinarianswith experience of Johne’s disease in suckler cows wasinvolved with the output format for the model. The GUIoutputs focused on showing herd prevalence details overthe 10 years in terms of populations in each shedding state,so that the ‘iceberg’ effect could be appreciated, breaking thecosts and losses down into reduced value of sales (reducedcalving percentage, reduced cull cow and calf weights),replacement costs, mortality, additional resources costs(veterinary and labour costs associated with diseasedanimals) and expenditure on control. The disease outcomespresented on the output part of the GUI are herd prevalence,performance parameters such as output loss and total herdcost in terms of the sum of production losses and controlexpenditures.

In addition, each output box has a ‘pop up’ informationbox which helps to ensure correct interpretation of themodel’s outputs. A graphical function is included as alinked page showing disease prevalence details over 10years. Users also highlighted important GUI features. Theseincluded that the input screen must be entirely visiblewithout scrolling down or needing to move along the pagebecause (a) this took too long and (b) users tended to forgetthat there were more inputs to complete, and that the

Table 2

Cumulative Johne’s shedding cattle populations by Year 10 (default values).

Shedding populations No control Test and

cull mild

Test and

cull severe

Improve

management

Improve management

and test and cull severe

Clinical 1 0 0 1 0

High 7 4 3 5 1

Low 14 14 14 7 8

Latent 46 47 47 24 30

Example relates to a 100 cow suckler herd with an initial Year 1 assumption of 10% of cattle being in a low shedding state.

Table 3

Cumulative Johne’s costs in the beef suckler herd over 10 years (discounted £s at 2008 prices)a.

Costs per control strategy No control T&Cb m T&Cc s Improve management Improve management with T&C s

Reduced sale value 34,151 32,191 31,926 16,480 19,053

Replacement cost 3,349 7,824 6,914 1,982 5,549

Fallen stock cost 1,147 0 0 991 0

Addn. vet and labour cost 13,459 12,746 12,746 9,919 9,994

Control costs 0 856 856 318 2,342

Total cost over 10 years 52,106 53,617 52,442 29,690 36,938a 3% discount rate.b Test and cull mild.c Test and cull severe.

R. Bennett et al. / Preventive Veterinary Medicine 93 (2010) 286–293 291

whole process from data input to data output must nottake more than a few minutes because neither veterinar-ians nor their clients are likely to be prepared to spendlonger than this (and hence might otherwise not use themodel).

2.3. Model delivery

Delivery of the model is via a dedicated website(www.fhpmodels.reading.ac.uk), where it can be freelydownloaded, with a link to install the free viewer version ofthe Stella software (ISEE Player). Anyone can access themodels (though the modelling layer is locked tounauthorised access) and user details are not collected.

3. Results

The main model output is the tabulation of the number ofanimals in each disease state and the associated disease andtreatment costs, with a link to graphical output. Someexample output from the model is presented here and isbased on herd and management default figures set at; 100cow suckler herd with a calving rate (calves reared) of 90%, acull rate of 17%, an in-calf (Johne’s-free) replacement heifervalue of £800, calf value at weaning of £500, a cull cow valueof £650 and NFSS charges of £100 per animal. Veterinarydefault details and figures are set at; test and diagnostics£110 per hour, nominal, farm specific, per animal, per weekcharge for implementing management changes £0.05 andmean weight loss due to disease 15%. Default values areeasily adjusted by the model user (see Fig. 2).

Results from running the model (taken from the outputtables of the GUI) using the above default values aresummarised in Tables 2 and 3. Table 2 shows the numbersof animals in the different shedding states after 10 yearsand Table 3 shows the discounted cumulative costs overthe same time period (with a discount rate of 3%), bothaccording to the disease control measure used.

This example output shows that by Year 10 the ‘test andcull mild’ strategy reduces the number of high sheddinganimals and removes the clinical case, whilst the ‘test andcull severe’ further reduces the number of higher sheddinganimals but does not reduce the number of low sheddinganimals, which is largely due to the (assumed) lowsensitivity of the test in detecting these animals. The‘improve management’ strategy reduces the number ofanimals in all disease states, apart from the clinical case,whilst the combination strategy of ‘improve managementand test and cull severe’ removes the clinical case and mostof the high shedding animals and substantially reduces thenumbers of animals in the low shedding and latent states.In terms of costs, Table 3 shows that the ‘improvemanagement’ strategy appears to result in the lowestnet disease cost (saving over £22,000 over 10 yearscompared to doing nothing), followed by the combined‘improve management and test and cull severe’ strategy.

Better hygiene management and targeted testing andmonitoring of herd health benefit the control of diseasesother than Johne’s. However, these benefits have not beenincluded in the economic calculations.

4. Discussion

The need for a Johne’s suckler beef cost-benefit modelwas identified by a farm health planning cattle stakeholdergroup consisting of farmers, veterinarians and industryrepresentatives. The aim of the group was to encourageproducers to introduce proactive measures to help managedisease and improve herd performance. The Johne’ssuckler beef model is one of a number of disease controlmodels developed by the researchers as part of anindustry-government farm health planning initiative(see www.fhpmodels.reading.ac.uk).

The paucity of previous models that simulate Johne’s inbeef cattle is, in part, due to the lack of quantitative dataregarding the disease in beef cattle (Humphrey et al.,

R. Bennett et al. / Preventive Veterinary Medicine 93 (2010) 286–293292

2006). It is clear from examples of modelling Johne’simpacts in dairy herds that a number of factors hinder themodelling of Johne’s in the beef suckler herd. Firstly,although methods do exist for measuring output forsuckling animals, such as weighing animals before andafter suckling (Coates and Penning, 2000), there does notappear to be any data on this for Johne’s infected sucklerherds. Therefore assumptions concerning disease effectssuch as reduced calf growth rates and reduced weaningweights must be made. Secondly, reliable prevalence dataare sparse due to the limitations of the diagnostic tests,making predictive modelling in beef systems difficult tovalidate (Humphrey et al., 2006). Thirdly, the optimal calfmanagement control strategies applied to dairy herds,such as ‘snatch calving’ and feeding MAP-free milk (seeKudahl et al., 2007) cannot be applied to the suckled calf as,by definition, the dam and calf are not separated until thecalf is weaned. Therefore assumptions must be made as tothe efficacy of control strategies that are more easilyapplied to the suckler beef system. In order to give usersconfidence in the outcomes, the assumptions are madeclear to them and where appropriate users can introducealternative assumptions into the model.

The model does not currently include a vaccinationstrategy. This is because vaccination only reduces andprolongs emergence of the number of clinical animalswithin the herd and considerable disease pressure stillexists due to the ‘iceberg’ effect of Johne’s disease withinthe herd (Whitlock and Buergelt, 1996). Furthermore, asJohne’s cannot be cured, control strategies focus onbreaking transmission routes of infection and/or a reduc-tion of infection pressure by testing and culling infectiousanimals (Kennedy et al., 2001).

Results from an example run of the model show that testand cull strategies have a limited ability to control Johne’s ina beef suckler herd and may not reduce costs due to thedisease. This is mainly due to the assumed low sensitivity ofthe test to detect latent and low shedders compared to amuch higher test sensitivity for high shedders and clinicalcases. Thus, test and cull strategies remove the latter butleave the vast majority of latent and low shedders in theherd, which may then progress to high shedding and clinicalstages. The results show improved management measuresto control Johne’s as being more effective because there is anassumption that these measures reduce the risks associatedwith disease spread within the herd (by 50%) and thusreduce the total number of infected animals over time.Clearly, if a higher test sensitivity or a lower level of riskreduction due to management improvements is assumedthen the results could be very different is assumed and/or alower level of risk reduction due to management improve-ments the results could be very different.

Testing and validation for the Johne’s suckler beefmodel was difficult due to the paucity of data. However, itwas considered appropriate to use two measures of‘representativeness’; (i) that the number of animals withineach shedding state satisfied the ‘iceberg effect’ (Whitlockand Buergelt, 1996); and (ii) that the numbers of infectedanimals and rates of infection and the estimated costs ofdisease concurred with empirical observations of research-ers and practising veterinarians in the field. Results

generated by the model appear to be consistent withempirical data reported in the research literature. Forexample, the example results presented above are consis-tent with those presented by Kudahl et al. (2007) whoclaimed a similar benefit from combining testing andculling with improved management in terms of thereduction of within-herd prevalence of Johne’s.

Many decision support systems (DSS) and decisionsupport tools (DST) that are developed are not widelyadopted by users. McCown (2002) suggests that a keyfactor in uptake success of DST is credibility of theinformation provided. The comments from users sug-gested that references to published data sources regardingparameter values contained within the model enhanced itscredibility. Credibility was further helped by using a moreparticipatory approach (involving stakeholders, end usersand funding providers) to model development. Veterinaryadvisers (30) were asked at the 2008 UK Dairy andLivestock Event if they considered the model to be usefulwhen discussing Johne’s or wider disease issues with theirclients. Fifty-five percent of those asked considered themodel ‘very useful’ and 45% considered the model ‘quiteuseful’ (on a five point scale ranging from ‘not at all useful’to ‘very useful’) with others adding that they liked the waythe model allows the simulation of a combined strategy.Veterinarians reported a favourable response from theirclients largely through using the Johne’s model as more of adiscussion facilitator. Some reported that the very processof thinking about disease in their herd and entering theirown farm data into the model was as useful as specificoutputs produced by the model.

5. Conclusion

The model output clearly shows an underlying issue ofJohne’s control—which is the large number of subclinicalcases in a herd which, individually, remain largelyunidentified due to available tests having low sensitivity.This means that test and cull strategies are very limited intheir effectiveness to control Johne’s. Improved herdmanagement to reduce the risk of disease spread withina herd may be more cost-effective, although there is a lackof data on the efficacy of such measures within a sucklerbeef herd. The model helps users to explore the issuesconcerning Johne’s in a herd and the likely benefits andcosts of different control strategies.

One of the main measures of success of a decisionsupport model is whether intended users find the modeluseful and whether they believe it actually helps them tomake better decisions. Feedback from users shows that themodel presented here is used and valued by veterinariansand their clients. In addition, cattle industry groups,pharmaceutical companies and veterinary/agriculturalcolleges have also been using the model. The challengeis to continue to use a participative approach to developand update the model so that it continues to meet theneeds of its users.

Conflict of interest

None declared.

R. Bennett et al. / Preventive Veterinary Medicine 93 (2010) 286–293 293

Acknowledgements

The authors would like to thank the Department ofEnvironment, Food and Rural Affairs of Great Britain forfunding the study and the Defra Farm Health PlanningTeam, John Sumner and the Farm Health Planning CattleGroup, Keith Cutler of Endell Veterinary Group, Matt Dobbsof Westpoint Veterinary Group, Peter Orpin of ParkVeterinary Group, Neil Blake of Charter Veterinary Group,George Caldow and Alistair Stott of the Scottish Agricul-tural College and members of the Johne’s Initiative Groupfor their help and support with the development of theJohne’s suckler beef model. Any errors or omissions are theresponsibility of the authors.

References

Begg, D.J., Whittington, R.J., 2007. Experimental animal infection modelsfor Johne’s disease, an infectious enteropathy caused by Mycobacteri-um avium subsp. paratuberculosis. Vet. J. 176, 129–145.

Bennett, R.M., 2003. The ‘direct costs’ of livestock disease: the develop-ment of a system of models for the analysis of 30 endemic livestockdiseases in Great Britain. J. Agric. Econ. 54, 55–71.

Caldow, G., Greig, A., Gunn, G.J., Humphry, R., Low, C.J., Ashworth, S.W.,Jones, G.M., Stott, A.W., Cranwell, M.V., Sharp, J.M., Stevenson, K.,Henderson, D.C., 2001. Assessment of surveillance and control ofJohne’s disease in farm animals in GB. SAC, DEFRA, 67–82, http://www.defra.gov.uk/animalh/diseases/zoonoses/zoonoses_reports/sac2.pdf.

Chiodini, R.J., van Kruiningen, H.J., Merkal, R.S., 1984. Ruminant para-tuberculosis (Johne’s disease): the current status and future pro-spects. Cornell Vet. 74, 218–262.

Coates, D.B., Penning, P., 2000. Measuring animal performance. In: Man-netje, L., Jones, R.M. (Eds.), Field and Laboratory Methods for Grass-

land and Animal Production Research. CAB International,Wallingford, UK, pp. 353–402.

Dargatz, D.A., Byrum, B.A., Barber, L.K., Sweeney, R.W., Whitlock, R.H.,Shulaw, W.P., Jacobson, R.H., Stabel, J.R., 2001. Evaluation of a com-mercial ELISA for diagnosis of paratuberculosis in cattle. JAVMA 218(7), 1163–1166.

FSA (UK Food Standards Agency), 2001, http://www.johnes.org/handouts/files/Review%20of%20evidence.pdf.

Groenendaal, H., Nielen, M., Jalvingh, A.W., Horst, S.H., Galligan, D.T.,Hesselink, J.W., 2002. A simulation of Johne’s disease control. Prev.Vet. Med. 54, 225–245.

Humphrey, R.W., Stott, A.W., Adams, C., Gunn, G.J., 2006. A modelof the relationship between the epidemiology of Johne’s diseaseand the environment in suckler-beef herds. Vet. J. 172, 432–445.

Kennedy, D., Holstrom, A., Plym Forshell, K., Vindel, E., Suarez Fernandez,G., 2001. On farm management of Paratuberculosis (Johne’s disease)in dairy herds. Bull. Int. Dairy Federation 362, 18–31.

Kudahl, A., Ostergaard, S., Sorensen, J.T., Nielsen, S.S., 2007. A stochasticmodel simulating paratuberculosis in a dairy herd. Prev. Vet. Med. 78,97–111.

McCown, R.L., 2002. Locating agricultural decision support systems in thetroubled past and socio-technical complexity of ‘models for manage-ment’. Agric. Syst. 74, 11–25.

McInerney, J., 1996. Old economics for new problems—livestock disease:presidential address. J. Agric. Econ. 47, 295–314.

Ott, S.L., Wells, S.J., Wagner, B.A., 1999. Herd-level economic lossesassociated with Johne’s disease on US dairy operations. Prev. Vet.Med. 40, 179–192.

Pennsylvania State University, 2007. Pennsylvania State University VetExtension Service, http://vetextension.psu.edu/biosecurity/.

Wells, S.J., Collins, M.T., Faaberg, K.S., Wees, C., Tavornpanich, S., Petrini,K.R., Collins, J.E., Cernicchiaro, N., Whitlock, R.H., 2006. Evaluation of arapid fecal PCR test for detection of Mycobacterium avium subsp.paratuberculosis in dairy cattle. Clin. Vaccine Immunol. 13 (10),1125–1130.

Whitlock, R.H., Buergelt, C.D., 1996. Preclinical and clinical manifestationof paratuberculosis (including pathology). Vet. Clin. N. Am-Food 12,245–356.