Combating Bovine TB – Statistician models Tuberculosis Cross-Infection between Cattle and Badgers


  • Author: Lillian Pierson P.E.
  • Date: 21 May 2014
  • Copyright: Image appears courtesy of iStock Photo

Bovine Tuberculosis (TB) is a huge and costly problem around the world, especially in Great Britain. The disease is caused by the bacterium Mycobacterium bovis (M. bovis). This strain of bacteria also causes TB infection in badgers, dogs, cats, goats, deer, pigs, and other mammals. Great Britain’s Department for Environment, Food and Rural Affairs (DEFRA) estimates that British taxpayers have spent £500 million to control the disease in the UK over the last 10 years. What's worse, the Department estimates that these same taxpayers will have to spend £1 billion to control TB in the UK over the next decade, if no further action is taken. But, what preventative actions can be taken?

thumbnail image: Combating Bovine TB – Statistician models Tuberculosis Cross-Infection between Cattle and Badgers

For one answer to this question, let's look at the work of Dr. Gabrielle Kelly of University College Dublin's School of Mathematical Science. Most recently, Dr. Kelly has been exploring the use of spatio-temporal statistics to model TB in cattle across Ireland. Although Ireland is experiencing a less severe problem with cattle TB, its badger population and cattle TB incidence make its problem very similar to that of Great Britain’s. While cattle TB is generally not widespread, it is quite severe in areas of infection. In these areas, it’s commonly thought that TB-infected badgers are the vectors of disease, but Dr. Kelly’s most recent work has shown otherwise.

The purpose of Dr. Kelly's spatio-temporal statistical modeling work is to establish the most effective scale and direction of selective badger killing in a geographical region that is showing a high incidence of cattle TB infection (commonly termed "reactive badger culling"). Her work will also be useful for the future of evaluating TB vaccines for badgers and cattle.

In an exclusive interview for Statistics Views, Dr. Gabrielle Kelly answered questions about her work as a statistician investigating spatio-temporal characteristics of Tuberculosis outbreaks in cattle.

1. As a statistician, when did you first experience a passion for statistics?

As an undergraduate I enjoyed probability problems and, when I conducted my first data analysis study for a medical unit, I realized the potential of statistics in many areas of application. I also realized the statistician on a project is the first to see the results and this is always very interesting.

2. What’s led your interest into the field of spatial statistics? What sorts of applications, programs, or programing languages do you favor when performing spatio-temporal statistical modeling?

Work on bovine TB spurred my interest in spatial statistics. Herds/setts in close proximity may be spatially correlated and such effects should be incorporated in modelling bovine TB incidence. Presently, I am applying spatial models in the area of forestry and tree growth. I use SAS for spatial modelling and also Diggle and Ribeiro's GeoR package with the R Language.

3. Can you tell me something about the statistics methods and techniques that you are using for the spatio-temporal modeling of TB in cattle? What key information do those methods provide you?

For spatio-temporal modelling of TB in cattle I use anisotropic generalized linear geostatistical models. These models allow me to test if spatial correlation is present, the direction of clustering, and the range of clustering. I have also used these models to model bovine TB in badgers and discovered badgers home ranges extended much further than previously thought. In a recent study I used these models to show cross-infection between badgers and cattle was quite limited.

Any evaluation of vaccines in the field will involve modelling TB incidence in badgers and cattle and thus spatial correlation will need to be (quantified). The range of spatial correlation present in unvaccinated populations will be a factor in the extent of (spatial) area in which the vaccine is administered.

4. What inspired you to use this statistical approach to solve problems related to TB in cattle? What other types of problems would you also like to explore using these methods?

Several authors had used nearest neighbor methods to examine spatial association of disease. However, such methods did not take covariates into account, so I proceeded to investigate how spatial correlation might be incorporated in fixed effects models. The availability of recently developed statistical software for spatial models facilitated this approach.

Presently, I am working in the area of forestry and developing spatial models (i.e. that account for tree proximity) for tree growth and timber volume where the stand may be subjected to different thinning regimes. Development of optimal thinning regimes is the ultimate goal in this work.

5. How will the work you are doing now be useful in evaluating TB vaccines for badgers and cattle?

Any evaluation of vaccines in the field will involve modelling TB incidence in badgers and cattle and thus spatial correlation will need to be (quantified). The range of spatial correlation present in unvaccinated populations will be a factor in the extent of (spatial) area in which the vaccine is administered.

6. What are a few examples of "best practices" for spatio-temporal statistical modeling? If “best practices” haven’t been formally defined yet, maybe you would define a few for us, please?

Initial estimates as to the scale of spatial association are necessary. It is not advisable to measure distances in metres if spatial correlation extends over several kilometres. In addition, if the scale of spatial correlation extends over the entire study area, estimates of spatial correlation that are infinite (or by reparameterization) or 0 are both consistent with the data, and the partial sill will be non-zero.

7. Are there any statistics resources that you find particularly useful when performing spatio-temporal statistical modeling?

I find the following references to be useful:

• Diggle P and Ribeiro P (2007). Model-based geostatistics. Springer
• Schabenberger O and Gotway C. (2005). Statistical Methods for Spatial Data analysis. Chapman and Hall/CRC

8. What is the biggest challenge you have faced in using spatio-temporal statistics models to quantify and describe incidence of TB in cattle?

My recent (big challenge was documented in my latest) paper on possible cross-infection between badgers and cattle. (The findings in this paper were) based on fitting numerous models and extensive comparisons. The final models represented those that provided good fits, but also had sound biological interpretation.

9. What are your top 3 predictions for how the practice of spatial statistics will affect positive changes in the world over the next 20 years?

This is a difficult question, but I would say:

1) It will help our understanding of the location of diseases and the spread of disease

2) It will become important in social science (where GIS coordinates of all household in a country become available) in understanding the geography of poverty, educational levels, etc. This will enable black spots to be more readily identified and targeted for remedial action.

3) We will be better able to determine whether infrastructure such as energy pylons, windmills, etc. have (distance-related) adverse (human) health effects.


(1) Kelly, Gabrielle E. : Spatio-Temporal Modelling of TB in Cattle Herds. Journal of Environmental Statistics, 3 (4) 2012-08, pp.1-9.

(2) Kelly, GE,McGrath, GE,More, SJ (2010). Estimating the extent of spatial association of Mycobacterium bovis infection in badgers in Ireland. Epidemiology and Infection, 138, 270:279.

(3) Kelly, G.E., More, S.J. (2011). Spatial clustering of TB-infected cattle herds prior to and following proactive badger removal. Epidemiology and Infection, 139, 220:229.

(4) Kelly, G. E. (2013). Joint Spatio-Temporal Modeling of Mycobacterium bovis Infections in Badgers and Cattle - Results from the Irish Four Area Project. Statistical Communications in Infectious Diseases, 5, 1:16.

Related Topics

Related Publications

Related Content

Site Footer


This website is provided by John Wiley & Sons Limited, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ (Company No: 00641132, VAT No: 376766987)

Published features on are checked for statistical accuracy by a panel from the European Network for Business and Industrial Statistics (ENBIS)   to whom Wiley and express their gratitude. This panel are: Ron Kenett, David Steinberg, Shirley Coleman, Irena Ograjenšek, Fabrizio Ruggeri, Rainer Göb, Philippe Castagliola, Xavier Tort-Martorell, Bart De Ketelaere, Antonio Pievatolo, Martina Vandebroek, Lance Mitchell, Gilbert Saporta, Helmut Waldl and Stelios Psarakis.