Cost-Benefit Analysis of Badger and Cattle Management
SE3117
Report of Phase II: June 2005 to May 2007
Authors:
GC Smith, D Wilkinson(CSL)
S Rushton, M Shirley(Newcastle University)
RM Bennett, ID McFarlane(Reading University)
Table of Contents
1. Summary
1.1 Executive Summary
1.2 Summary
2. Introduction
3. Model
3.1 Economics
3.1.1 Introduction
3.1.2 Costs of TB if no action taken
3.1.3 Pre-movement testing
3.1.4 Costs of implementing culling or vaccination strategies
3.1.5 Cattle costs
3.1.6 VLA costs
3.1.7 Savings resulting from culling or vaccination strategies
3.2 Model Descriptions
3.2.1 Purpose
3.2.2 State variables and scales
3.2.3 Process overview and scheduling
3.2.4 Design concepts
3.2.5 Initialisation
3.2.6. Input
3.2.7. Submodels
3.2.8 Perturbation Processes
4. Verification
4.1 Badger-to-Cattle Transmission Rates
4.2 Cattle Mortality Rates
4.3 Cattle Herd Breakdown Rate
4.4 Sensitivity Analyses
4.5 Cross-Model Validation
4.6 Verification Overview
5. Validation
5.1 Badger Validation
5.2 Cattle Validation
5.2.1 Tests and reactors
5.2.2 Reactors per Cattle Herd Breakdown
5.3 Economic Validation
5.3.1 Costs of TB breakdowns
5.3.2 Annual costs of TB breakdowns
5.4 RBCT Validation
6. Methods
6.1 Control and Management Options Tested
6.1.1 Default settings and assumptions
6.1.2 Main Scenarios
7. Results
7.1 Sensitivity Analysis
7.1.1 Cattle skin test sensitivity
7.1.2 Cattle test-interval
7.1.3 Disease transmission rates
7.1.4 Pre-movement testing – sensitivity to scale
7.1.5 Perturbation
7.2 Population Management
7.2.1 Badger population
7.2.2 Prevalence of TB in badgers
7.2.3 Number of TB-infected badgers
7.2.4 Parish types
7.2.5 Cattle herd prevalence
7.2.6 Cattle herd breakdown rate
7.2.7 Varying control area
7.2.8 Varying farm compliance
7.2.9 Varying control rates
7.2.10 No-Immigration
7.2.11 Gamma-Interferon
7.3 Economic Modelling
7.3.1 Pre-movement testing – economic analysis
7.3.2 Economic benefit distribution – scenario 1
7.3.3 Economic comparison of different scenarios
7.3.4 No Immigration – economic analysis
7.3.5 Gamma-Interferon testing – economic analysis
7.4 Sensitivity Analysis of Economic Output
7.5 Perturbation – Alternative Options
8. Discussion
8.1 Verification and Validation
8.2 Cattle Testing
8.3 Badger Culling
8.4 Badger Vaccination
8.5 Economics
9. Conclusions
10. References
Appendix A – Economic Parameters
Appendix B – Farm Costs
Appendix C – Cattle Mortality
Appendix D – Cattle Herd Breakdown Rates (historic)
Appendix E – Parameter Values (CSL model)
E.1 Default Settings and Parameter Values used in the CSL model
E.1.1 Temporal Settings:
E.1.2 Spatial Settings:
E.1.3 Badger Settings:
E.1.4 Cattle Settings:
Appendix F – Parameter Bounds (CSL model Sensitivity Analysis)
Appendix G – Sensitivity Analysis Output
Appendix H – Cattle Testing Flowchart
Appendix I – Test Interval Algorithm
Appendix J – Badger Movement Flowchart
Appendix K – Model Assumptions
K.1 Model Assumptions
K.1.1 Temporal Assumptions
K.1.2 Spatial Assumptions
K.1.3 Badger Assumptions
K.1.4 Cattle Assumptions
K.1.5 Economic Assumptions
K.1.6 Badger Control Assumptions
Appendix L – Glossary
List of Figures and Tables
Figure 3.1.1. A diagrammatic representation of the cost-benefit analysis (CBA) incorporated into the CSL bovine tuberculosis (TB) model
Figure 4.5.1 Cross-validation of the models.
Figure 5.2.1.1. Number of reactors plotted against number of annual cattle TB tests.
Figure 5.2.2.1. Frequency distributions of number of reactors per CHB. A comparison of real and simulated data.
Figure 5.3.1.1. Scatter distribution of cost of each individual CHB in 2005, plotted against the number of confirmed new incidents (CNIs).
Figure 5.3.2.1. Distribution of the simulated annual cost (£) of all CHBs, multiplied up from the modelled grid area to the whole T1 and T2 areas.
Figure 5.4.1. RBCT validation of the NU model.
Table 5.4.1 A comparison of the RBCT CHB rate with simulated data.
Table 6.1.2. Details of the 7 badger control scenarios
Figure 7.1.1.1 Cattle skin test sensitivity.
Figure 7.1.2.1 Effect of changing parish test-interval dependent on CHB rate.
Figure 7.1.2.2 Effect of changing the routine cattle-test-interval.
Figure 7.1.3.1 Badger contribution to the CHB rate.
Figure 7.1.4.1 Effects of extending the area over which PrMT is applied.
Table 7.1.5.1. Change in simulated CHB rate following control.
Table 7.1.5.2. Change in simulated CHB rate following control.
Figure 7.1.5.1 Illustration of the spatial effect of culling in the NU model.
Figure 7.2.1.1. The percentage of badger groups subjected to culling for different rates of farm compliance.
Figure 7.2.1.2 Scenario 1: badger social group size.
Figure 7.2.2.1 Scenario 1: badger TB prevalence.
Figure 7.2.2.2 Example output from the NU model.
Figure 7.2.3.1 Scenario 1: Number of infected badgers.
Figure 7.2.4.1 Scenario 1: Parish test frequency.
Figure 7.2.5.1 Scenario 1: Cattle herd prevalence.
Figure 7.2.6.1 Scenario 1: CHB rate.
Table 7.2.7.1 Varying control area: Badger social group size.
Table 7.2.7.2 Varying control area: Number of infected badgers.
Table 7.2.7.3 Varying control area: Badger prevalence.
Figure 7.2.7.1 Varying control area: Badger prevalence.
Table 7.2.7.4 Varying control area: Mean CHB rate.
Figure 7.2.7.2 Varying control area: Mean CHB rate.
Table 7.2.8.1 Varying compliance: Badger social group size.
Table 7.2.8.2 Varying compliance: Number of infected badgers.
Table 7.2.8.3 Varying compliance: Badger prevalence.
Figure 7.2.8.1 Varying compliance: Badger prevalence.
Table 7.2.8.4 Varying compliance: CHB rate.
Table 7.2.9.1 Varying control rates: Badger social group size.
Table 7.2.9.2 Varying control rates: Number of infected badgers.
Figure 7.2.9.1 Varying control rates: Number of infected badgers.
Table 7.2.9.3 Varying control rates: Badger prevalence.
Table 7.2.9.4 Varying control rates: CHB rate
Figure 7.2.9.2 Varying control rates: CHB rate.
Figure 7.2.10.1 No immigration: Badger social group size.
Figure 7.2.10.2 No immigration: CHB rate.
Figure 7.2.11.1 Gamma-interferon: CHB rate.
Figure 7.3.1. Cash flow by method of badger control.
Table 7.3.1. NPV for badger control.
Figure 7.3.2. Cash flow to Defra for badger control.
Figure 7.3.3. Cash flow to industry for badger control.
Figure 7.3.2.1 Net Benefit: Scenario 1.
Table 7.3.3.1 Percentage chance of economic loss.
Figure 7.3.3.1 Discounted Net Benefit: shooting and vaccination.
Figure 7.3.3.2 Discounted Net Benefits: Defra and Industry.
Figure 7.3.4.1 Discounted Net Benefit: No immigration.
Figure 7.3.5.1 Discounted Net Benefit: gamma-interferon.
Table 7.4.1. Economic sensitivity to culling and vaccination.
Figure 7.5.1. Badger prevalence.
Table 7.5.1 Discounted net benefits of control with reduced perturbation.
Table 7.5.2 Percentage economic loss.
1. Summary
1.1 Executive Summary
- The aim of this project was to construct a simulation model of badgers, cattle and bovine tuberculosis epidemiology, and to evaluate the costs and benefits of selected badger and cattle management scenarios, as an aid to policy development.
- This involved constructing a generic spatial model and running two comparative simulations (one with the management scenario(s) and one without). The economic costs of both simulations are then compared and the Net Present Value recorded. This is repeated 100 times for each scenario, the results are then ordered and presented in histogram form.
- In addition, a GIS version of the model was produced. Both models produced similar output and both were validated against the results of the RBCT. The GIS version was also created to allow simulations in specific landscapes (e.g. using local farm structure and badger density) for future analysis of management scenarios in specific geographical areas.
- Badger culling is known to disrupt the behaviour of surviving badgers and this has been the hypothesised reason for an increase in cattle herd breakdowns in the area surrounding the badger culling. The changes to badger movement and disease transmission, as a result of culling, were simulated in both models using the best data available. This led to a predicted increase in herd breakdowns following badger culling, and this must be regarded as additional evidence for the effect of social perturbation on increased incidence of bovine tuberculosis in cattle.
- The models predicted that badger culling does not result in an overall decrease in the number of herd breakdowns, due to social perturbation. Economic analysis of the outputs therefore suggested that badger culling is not economically viable unless badger immigration can be prevented.
- The model predictions suggest that the observed differences between the RBCT and the Irish Four Areas Study are a consequence of badger immigration.
- These analyses were undertaken with the best available data. Further analysis may improve the accuracy of the model predictions, particularly with regard to temporal changes in badger social perturbation, but are unlikely to fundamentally change them.
- Simulations of cattle management suggest that Pre-Movement Testing may be economically viable, that routine skin testing could be further optimised in terms of area and frequency, and the any reduction in the cattle skin test sensitivity (as a result of poor implementation) could dramatically increase herd breakdown rates.
- We therefore recommend that further fieldwork be conducted to improve the parameterisation of badger social perturbation. We also recommend that further simulation be performed on cattle testing and badger vaccination.
1.2 Summary
This is a substantial and comprehensive technical project report, detailing the development, construction, verification and validation of a large complex, stochastic simulation model, with the potential to inform Defra policy on the likely effects of different management strategies on the incidence of bovine TB in cattle. The research project was overseen by a steering group, who commented and advised as the work progressed. Much of the detailed report deals with comments raised during these discussions, and a full description of the model and the parameterisation is given to allow a thorough peer review of the process. Whilst the full report would be very useful to a reviewer or another modeller, it would make more sense for others to read particular sections of interest.
The overall aim of this project was to construct a simulation model of badgers, cattle and bovine tuberculosis epidemiology, and to evaluate the costs and benefits associated with different badger and cattle management scenarios to produce a research model capable of answering policy questions related to badger and cattle management. This involved extending a previously constructed and published spatial model of badger TB epidemiology, constructing a cattle layer and adding economic components. The model was run using two comparative simulations (one with the management scenario(s) and one without). The economic costs of both simulated scenarios were then compared and the discounted Net Present Value recorded. This was repeated 100 times for each scenario and the results were ordered and presented in histogram form. The economic cost of each compartment of the model was evaluated on an individual animal basis. For example, the cost of a herd breakdown was calculated by assigning a cost to each cow slaughtered (using research results from a previous study of the costs associated with TB breakdowns on cattle farms), where the number of cattle slaughtered depends upon the detection of infection in the simulated herd.
A GIS version of the model was produced by Newcastle University to allow simulations in specific landscapes (e.g. using local farm structure and badger density) for future analysis of management scenarios in specific geographical areas. Both models produced similar output and both were validated against field results, including the results of the RBCT. Model verification and validation is a vital component of producing a useful model. Since the two models produced similar output and a full and independent sensitivity analysis gave very similar results we have a strong evidence base that the models are ‘correct’. Both models were then subjected to validation against field data (e.g. badger social group size, overall costs, number of cattle per herd breakdown, spatial herd breakdown rates seen in the RBCT, etc). This validation procedure allowed us to adjust the models to closely simulate reality by adjusting parameters for which there were limited data (e.g. the spatial extent of social perturbation). The models performed very well against the validation data sets, suggesting that they provide an adequate although simplified representation of the underlying biological and spatial processes in the badger population and its interaction with livestock.
The models constructed here are based on a vast amount of data and expertise, and the similarity of the output from the two models is extremely encouraging. The authors believe that the constructed models are the most accurate available to date (although there is always room for improvement). The output from these models will be presented at scientific meetings and published in peer-reviewed journals to ensure the models and results are subjected to scrutiny.
A variety of badger and cattle management scenarios were simulated, as suggested by Defra and the overall economic output is summarised below. For further detail on how these scenarios affect disease epidemiology in the badger, it is necessary to read Section 7. Pre-movement testing (PrMT) of cattle was more likely than not to give an economic benefit, and all other forms of TB management assumed that PrMT was already in place. All badger control methods are costly to implement. As a result of social perturbation (particularly that caused by badgers immigrating into the culled area) no method of badger culling was reliably economically beneficial. Badger vaccination, as simulated here, had the highest probability of achieving a net economic benefit. The benefit or loss with vaccination was generally much lower than with any badger culling strategy. The most important factor leading to the economic loss of badger culling was the immigration of individual badgers (infected or otherwise) from outside the culled area. In one example (for gassing) the presence of perturbation led to only 2% of simulations giving an economic benefit, while preventing immigration completely led to 95% of simulations achieving a benefit. Thus, reducing immigration and ensuring compliance is above some (as yet unknown) minimum threshold are both necessary to ensure that badger culling is potentially viable.
Limited investigations of cattle management suggest there is some potential for further improvement in TB management. Current cattle testing may be improved by optimising the test frequency against herd breakdown rate, optimising the geographical area over which test frequency is decided and limited use of the skin test combined with the gamma interferon test. The sensitivity of the current skin test should definitely not be reduced in field use as this can lead to a substantial increase in cattle TB. Approaches to ensure that the test is applied rigorously and to the highest standards are advised.
This model has been verified and validated to available data at the time of construction. It is capable of simulating most badger management strategies, many cattle management strategies and combined management strategies. However, without further data and programming, it is not yet able to simulate specific farm types (e.g. finishing farms) or simulate changes to on-farm biosecurity. For a full list of assumptions used in the model, see Appendix K Whilst the model is validated to the RBCT research results at the end of the trial, further analysis is required to take account of changes in herd breakdown rate following the cessation of culling. It is noted that, from analysis of the RBCT results after the completion of this model (Jenkins, Woodroffe & Donnelly, 2008), the assumptions used in the model on the temporal aspects of social perturbation no longer produce herd breakdown rates consistent with the available data, and the model will be further updated to take account of this.
2. Introduction
This report summarises more than two years of work in model construction, model verification and validation and the output of suggested scenarios for badger culling and some cattle management. The modelling is building on a series of previously published scientific papers, with the main development in this project of building a cattle-farming simulation in addition to the badger simulation, and using the latest available data on the badger’s behavioural responses to culling. The report is detailed and quite technical in parts, but the main conclusions are based on the overall costs and benefits of different management scenarios. These economic outputs are presented in Sections 7.3.5 to 7.3.9. Model validation, by simulating the effect of the RBCT is presented in Section 5.4. These two sections and the discussion/conclusions are the most important sections to read if you wish to avoid the technical aspects of the modelling.
The number of cattle herd bovine tuberculosis (TB) breakdowns has been increasing in recent years. Bovine tuberculosis, caused by Mycobacterium bovis, is a serious disease of cattle, and can act as a zoonotic disease (Smith et al., 2004). The European badger is often infected with TB (Delahay et al., 1998), and there is now conclusive evidence that they are responsible for transmission of the disease to cattle (IndependentScientificGroup, 2007). Previous strategies have all involved badger culling to reduce the incidence of cattle-herd breakdown (CHB) and such culling had not previously been scientifically shown to be beneficial (Krebs et al., 1998). Theory suggested that some culling strategies may actually increase the incidence of cattle TB (Swinton et al., 1997), through a process known as social perturbation. As this project was being performed, results were being published from the Randomised Badger Culling Trial (RBCT), which began in 1998. Results from the RBCT indicated that reactive badger culling leads to an increase in cattle herd breakdowns (Donnelly et al., 2003), and that large-scale proactive badger culling can reduce CHB in the area culled, but increase it in the immediately surrounding area (Donnelly et al., 2006). Both these effects are believed to be caused by social perturbation of the badger (Woodroffe et al., 2006).
The government is, increasingly, relying on evidence-based policies. For some policies, such as badger and cattle management, it is very difficult, politically complex, expensive and time consuming to obtain field-based evidence. This project modelled a variety of different badger and cattle management strategies and determined their overall cost or benefit to society, in relation to the consequent change in cattle herd breakdown rates. These outputs, along with their intrinsic uncertainties and assumptions, can be used to help inform policy decisions on methods of badger and cattle management to be employed, potential spatial configuration for mixed strategies and the costs and benefits associated with these strategies.
The CSL previously produced a simulation model of TB in the badger (Smith et al., 2001a; Smith et al., 2001b; Wilkinson et al., 2004), which included a simple representation of cattle herds. This original model was the first to give realistic levels of disease (spatial and temporal) without serious population suppression (Smith et al., 1995), and it can be used to simulate any badger control policy: culling, vaccination, and proactive or reactive removal. The University of Reading completed a project to assess the economic impacts of TB and alternative control polices (Defra report SE3112). The CSL model was used to simulate the badger culling patterns seen in the RBCT, and the economics was bolted onto the output in Phase I of the project (Smith et al., 2007). Here, the Reading CBA model was integrated into the CSL model, so that output from the CSL model included disease prevalence, cattle herd breakdowns and the overall costs and benefits. This approach ensured that the full amount of uncertainty and variability within the simulation model, and within the CBA can be included. All assumptions and limiting constraints are reported below, and badger and cattle management strategies were evaluated to demonstrate the applicability of this combined system.