"I think that statistics is the ultimate transferable skill": An interview with Christl Donnelly CBE


  • Author: Alison Oliver, Statistics Views
  • Date: 28 Feb 2018
  • Copyright: Image appears courtesy of Professor Donnelly

Christl Donnelly is a statistician and epidemiologist studying the spread and control of infectious diseases, with a particular interest in outbreaks. She studied mathematics as an undergraduate at Oberlin College and biostatistics as a graduate student at Harvard School of Public Health, graduating in 1992. She worked in the Department of Mathematics and Statistics at University of Edinburgh from 1992 to 1995 and in the Wellcome Trust Centre for Epidemiology of Infectious Disease (Department of Zoology, University of Oxford) from 1995 to 2000.

Since 2000, she has worked at Imperial College London where she is Professor of Statistical Epidemiology. Christl has studied the Zika virus, Ebola, MERS, influenza, SARS, bovine TB, foot-and-mouth disease, rabies, cholera, dengue, BSE/vCJD, malaria and HIV/AIDS. She was a leading member of the WHO Ebola Response Team (2014-2016). She was also deputy chair of the Independent Scientific Group on Cattle TB (1998-2007) which designed, oversaw and analysed the Randomised Badger Culling Trial. In addition to epidemiology and disease control, she has applied interests in conservation and animal welfare.

She was awarded a CBE in this year's New Year’s Honours for her work on epidemiology and control of infectious diseases.

This summer, Professor Donnelly will take up the post of Chair of Applied Statistics at University of Oxford. Alison Oliver talks to Professor Donnelly about her career so far.

thumbnail image:

1. What was it that first introduced you to statistics as a discipline and what was it that led you to pursue epidemiology and biostatistics as a career?

I started at Oberlin College thinking I would major in biology. I had always enjoyed math and in my second year it became clear that I should shift to majoring in math (though I was never clear what – other than not being an accountant – I would do with it). I took a year of probability and statistics in my third year. Once I decided that actuarial science wasn’t for me (I didn’t want dollar signs in front of all my calculations), I found biostatistics would be a great way for me to combine mathematics and biology. I got my MSc and ScD at Harvard straight after my undergrad degree.

2. How do you feel the teaching of statistics has evolved and adapted to meet the changing needs of students since you began teaching?

I taught in Edinburgh, in the Department of Mathematics and Statistics, as my first job for three years, from ’92–’95, and it was where I met my husband as well. I was coming straight out of my doctorate, so I was quite innovative. Controversially, some of my colleagues thought I should not be changing things. For instance, I used to mark my students’ homework, and then I would ask some of them to go to the board to write and explain their answers. I chose them carefully, specifically those who did well and could give a good explanation. But there was some questioning on if that was appropriate and whether it put too much stress on students. To which my answer was: I did it in high school when I was 15, so I thought third-year undergraduates could do it. And the students were fine with it.

Another thing I did was typesetting the notes, because I hate blackboards. I just hate chalk. If there had been whiteboards then, I probably would have had a different teaching style. So, all my notes were pre-prepared and I would usually talk through them. I also printed them out for the students, because I thought it was crazy for them to just write down what I’ve already typed out. This way they could write down extra information: this turned out to be controversial as well. In the end the head of the department said I could do both of those things, even though they were different. Those were just some practical aspects I added to teaching.

After three years I moved into epidemiology, and there is an important statistical side to it. It is constantly changing: there is always a new disease outbreak that has come through; and also the computational power has increased so much that there are more and more simulations, resulting in complexities that could not be examined before. Genetics and GPS trackers have hugely changed the sort of data available. What we can do, and therefore what we can teach, has changed too.

3. Your main research concerns have centered on understanding outbreak epidemiology and control. Using your background in biostatistics, you have been collaborating with transmission modelers since joining Sir Roy Anderson’s group in 1995. Diseases of interest include: Ebola, MERS, SARS, influenza, bovine TB, rabies, foot-and-mouth disease, dengue, BSE/vCJD, malaria and HIV/AIDS. Please could you tell us more about this working collaboration and a typical day?

I have got two modes. One is the response mode, which is where I live, eat, and breathe whatever disease there is. I am doing that slightly more rationally now that I have children, because I cannot put everything on hold the way I used to. Before having children, that was all I did: getting very little sleep, just analyzing data, having meetings, getting feedback, changing things, problem-solving, etc. I would also give feedback to all sorts of people, whether it was the Ministry of Agriculture, Fisheries, and Food for BSE; or foot-and-mouth disease; or SARS, interacting with people in Hong Kong. For the influenza pandemic, we talked directly with the WHO as well as with the Health Protection Agency locally, and the CDC in the US.

Ebola is the dominant thing I worked on for over a year, but I have many other irons in the fire at the same time, which involve different meetings. I could be doing bovine TB, then dengue, then Ebola in great apes, or looking at Campylobacter in chickens, all in the same space. Sometimes I have a meeting scheduled and the person sits down and starts talking to me from where we left off the last time, and occasionally I have to say, “Just remind me, what disease we are talking about?” Because if it has been something I have not thought about for a while and someone else has been off doing work, then I must get re-oriented.

I think that statistics is the ultimate transferable skill, which is why I can work on so many different things. It is very important that health research has people who know the influenza virus inside and out, and that is what they work on. But I really like the fact that I can go and work with them on influenza, and then I can work with somebody else when Middle Eastern Respiratory Syndrome (MERS) is discovered. Statistics is very adaptable.

As for Zika, we undertook some modelling looking at possible scenarios that was published in 2016 in Science. There was a statistical element involved, due to the time series of case incidence, and we tried to look at what might happen in the future. We received some money from the UK Medical Research Council to do serological surveys in four areas of Columbia.

Other things I have been involved in included analyzing incidence data that have already been collected or experiments that somebody else has designed.

4. You have worked on a variety of diseases as mentioned above and also have interests in ecology, conservation and animal welfare. Are you working on anything in particular at the moment?

There is a lot going on. For example, I have nine PhD students right now, five of whom started last year. One of them is working on bovine TB, looking at badgers down in Cornwall. Originally, she was involved in a study where we put GPS collars on badgers and cattle, and then watched how they shared the same space: we found out that while badgers showed a preference for pasture in their home territories, they actively avoided contact with cattle themselves. It also suggested that environment could be one of the most important transmission routes between species or possibly within species, as well. This is a result of the collaboration with Rosie Woodroffe at the ZSL, the Zoological Society of London, with whom I have been working on badgers for over 20 years now.

Another one is a student from Nigeria who is doing fieldwork there, studying lymphatic filariasis, or what we used to call elephantiasis. He is interviewing people with different symptoms and looking at the impact on livelihoods and health spending for those who have the disease; he is also considering stigmatization, how it affects their lives. The student is also interested in mapping the lymphatic filariasis risk across Nigeria, because he is concerned there might be some high-risk areas that are not being currently addressed by the government’s control program, because they are not identified as risky areas.

A third student is also collaborating with the ZSL. He is looking at zoonoses that have a conservation risk, one of those being Ebola in great apes. Unlike bats, who do not seem affected (which is why they are such a good reservoir for Ebola), great apes get sick and die, much like people do. We are trying to understand why in some places the outbreaks are very big, whereas in others relatively small. Obviously small outbreaks and very big ones happened in humans, but that had to do with our control measures. Different determinants would be involved when it comes to wild animals; we have to look at the ecology. This is one strand. The other one is to look at African wild dogs and rabies. Rosie Woodroffe had put collars onto wild dogs as well as domestic ones, and looked at how they shared the space, because of the possibility of spreading pathogens between them including, but not limited to, rabies. African wild dogs are critically endangered. They have other threats as well, for example people killing them over concerns about their impact on livestock. But disease is a big threat. And obviously if you have domestic dogs that have rabies, then that is a threat to human health as well as to conservation. If canine rabies can be addressed in a way that protects both, it is going to be an ultimate win-win.

Ebola is the dominant thing I worked on for over a year, but I have many other irons in the fire at the same time, which involve different meetings. I could be doing bovine TB, then dengue, then Ebola in great apes, or looking at Campylobacter in chickens, all in the same space...I think that statistics is the ultimate transferable skill, which is why I can work on so many different things. It is very important that health research has people who know the influenza virus inside and out, and that is what they work on. But I really like the fact that I can go and work with them on influenza, and then I can work with somebody else when Middle Eastern Respiratory Syndrome (MERS) is discovered. Statistics is very adaptable.

5. You have authored many publications. What is the article that you are most proud of?

I would say our first Ebola paper, from September 2014. It represented the work of a huge number of people, 14 at Imperial, colleagues at WHO and from other countries. Many people, unnamed on the paper, collected data from all the Ebola cases they could; and of course, the families who gave the information also contributed. It summed up the work we had done at that point, an intensive month’s work of analysing all the data that had been provided to us by WHO.

If you had asked me before the West African Ebola epidemic, I probably would have said the first big SARS paper that we did; and before that, the foot-and-mouth disease ones; and before that, the BSE. So, these big outbreaks have sort of punctuated things, and are important to understanding how my work has gone and in which direction. It is rewarding, but also very stressful: I remember at one point, being completely exhausted and stressed out with Ebola, thinking back at the foot-and-mouth disease in 2001, that I was stressed out over a livestock disease that was not killing thousands of people. But it affected people indirectly, especially economically. I cannot help thinking that if the foot-and-mouth epidemic had happened at the same time as an Ebola outbreak, people’s perception of foot-and-mouth would have been so different. It is all about what epidemic is in front of you at a specific time, and you get immersed in it. Being able to be involved in these things and working in a responsive mode, helping to control transmission, is very rewarding.

6. You were active in the Independent Scientific Group on Cattle TB. Please could you tell us more about this work?

I was the Deputy Chair of the Independent Scientific Group on Cattle TB and in 1998 we started the Randomized Badger Culling Trial. It cost 50 million pounds and it was the largest ever randomized field trial. It randomized 30 areas, each of 100 square kilometres, to three different culling treatments: one was no culling; another was repeated proactive culling that was across an entire hundred square kilometre area at the same time annually; and the third was the culling of badgers in small localized culls. That was a huge amount of work.

The last culls happened in 2005. Based on interim analyses, we found that the localized culling was making TB worse in cattle, so that was stopped by the Minister at the time. The widespread repeated culling improved things for cattle inside the culled areas, but made things worse for cattle that were in farms up to two kilometres outside. It was a complex system where, if you did culling intensively and repeatedly over large areas, you could obtain a net benefit, because you would then get a large area of benefit with a relatively smaller area of detrimental effect outside.

This pushed the argument to more of an extreme: either you do not cull and avoid the adverse effects, or you cull very large areas. Which is why, politically, it has been a struggle ever since: the Labour Government decided not to cull, while the Coalition Government, and then subsequently the Conservative Government, went ahead with culls. However, they were so expensive compared to their impact, so the expense was moved over to the farmers so they became farmer-led culls. So, farmers organized and paid for culling that was licensed by the government, but then, as taxpayers, we bore a large expense due to policing. There is still debate about it, some of the early culls missed their cull targets, and much like the localized culling, you do not want to be in a situation where you’re only getting a partial cull compared to what we got in the trial. Then you could risk spending money both on doing the culling and killing badgers, and potentially making things worse for cattle, or at least not making them better. So, it must be a sort of all or nothing: either you do it in a really organized intensive way or you don’t cull.

It was a really large research program. There were additional analyses on bait-marking, which is where you put bait for the badgers with indigestible coloured beads outside different badger sets; and after the badger eats them, they then deposit these indigestible coloured beads around the environment. If you map the beads, you get a sense for which badgers are moving where, as you might have the blue beads in one area, the red beads somewhere else, etc. Bringing the numbers of badgers down seemed quite logical because it was expected there would be less badger-cattle contact, so there would be less transmission to cattle. But the badgers that remain in areas that have been subjected to culling pressure move more widely. So we have a tradeoff: even though badgers that remain move more widely, and are more likely to be positive individually, there are fewer of them. It is a complex system.

It is not clear what the future will hold. Over time, as we obtain more data, we will be able to pin that down. It is an ongoing epidemiological challenge.

7. You were the Campion Lecturer invited by the former RSS President Peter Diggle at the 2016 RSS conference. Please could you tell us more about your chosen topic for the lecture?

There are so many different things going on, and there are all these challenges so I wanted to offer insight into the collaborative intense response mode as a way of working. It is a very different under-the-spotlight way of working, and it can be very intense. Especially when there can be controversial aspects of it: as I mentioned the control of foot-and-mouth disease was controversial but, for the control options that policymakers had at the time, the culling had to be done. So, I wanted to try and give a feel of that, and hopefully, for some of the younger people who were and are still choosing areas, I was able to point out the interesting challenges that are in infectious disease epidemiology. The UK Medical Research Council, if you look at the budget compared to other research councils, has had strong governmental support. Disease control is very important, especially now that some of the international aid budgets are going through the research councils, and the medical questions fit very well with that. There is a huge need in many international settings for statistical and epidemiological inputs.

8. What has been the best book on statistics that you have ever read?

My nostalgia soft spot is Probability and Statistics by Morris DeGroot, because that was the book I used as an undergraduate. In the US, teaching is very book-based, so for the year I did a Probability and Statistics course, I had Morris DeGroot. We worked on it much like I had done in school: you choose a textbook and you work your way through it. I still have it on my shelf.

But if I put nostalgia to one side, I can think of two books for two different areas. One is The Theory of Design of Experiments by David Cox and Nancy Reid, because I think that designing an experiment is the first step in actually forcing yourself to think clearly. If you are muddled in how you do the design, then things are not likely to go well from then on. In fact, designing the analysis is sometimes what we do, because there are observational data, or routinely collected incidence data, and you still need to think through that step even though you have not designed the experiment. You go through that process and think, “How would I represent what has happened to bring these data to me?”

Roy Anderson and Bob May’s Infectious Diseases in Humans represents the bible of mathematical epidemiology. It has very little statistics and estimation because that was much less a part of mathematical epidemiology then; it tended to be qualitative comparisons between the data and a model. Of course epidemiological inference has changed considerably in recent years, but the fundamentals remain.

9. Who are the people that have been influential in your career?

I would say that both David Cox and Roy Anderson influenced me profoundly. It was 1995, after my PhD and 3 years’ teaching, when I moved and joined Roy Anderson’s group in Oxford, and later migrated to Imperial in 2000. That is the group I have been with for over 20 years now. At some point Roy said, “Ok, David Cox is going to evaluate you for your probation period,” and I was absolutely terrified. But David is so lovely, and I have worked closely with him since then. I got David involved with the badger and cattle project, so he was part of the Independent Scientific Group on Cattle TB, and I have worked with him on multiple things. We have written a book together, as well as many papers. He is so wise and calm, with such a wide range of experience. And in spite of the fact that he is such a god of statistics, he sees himself primarily as a scientist rather than a statistician.

I had done a little bit of mathematical modelling when I was a graduate student at Harvard. It was very much seen as an odd hobby, because the emphasis in that biostatistics department was not on mathematical modelling. So, when I joined Roy’s group, I was moving to be a postdoctoral statistician, and I was a bit concerned about whether or not it seemed foolish to be going from being a lecturer back to being a postdoc. I did seek advice from Harvard and they said, “Oh this sounds like a great opportunity, you should go.” I met with Roy, and I thought that if this is half as good as he says, then this should be great. It turned out to be even better, he gave me such freedom. He said that people would come and ask for advice, and that I could get involved in projects as little or as much as I wanted to. I was a bit skeptical of that; I thought pressure would be applied anyway. But I really was able to do that. And after six months, the identification of variant CJD and its link to BSE was announced in Parliament, and I went from doing a few different things to be suddenly plunged into this intensive way of working that I had never done before. I ended up working on BSE for years and wrote a Chapman & Hall book on statistical analysis on BSE and vCJD with Neil Ferguson.

Edinburgh was a tough place when I was there because they had not had a statistics professor for some years. One of the staff members, the first one I met, asked me, “So do you have relatives in the UK?” and I said, “No, I don’t know anyone on this continent.” He said, “Then why did you apply for this job?” Edinburgh statistics was really in a hard place back then. I did meet my husband there. He was a new lecturer, starting at the same time. The undergraduate teaching I did really changed the way I thought about inference. I studied it at Harvard, but actually having to teach it and give examples was hugely useful. I worked a lot, really hard, on the teaching. I did not get that much research done when I was in Edinburgh, but it put me in a really good place, to try and think about, “What can we develop to address this question?” Inference is what puts the data and the model together. Although I didn’t get a lot of research done in those three years at Edinburgh, the teaching that I did gave me a really good foundation that I used when I moved into epidemiology at Oxford.

10. If you had not got involved in the field of statistics, what do you think you would have done? (Is there another field that you could have seen yourself making an impact on?)

I debated with Peter Diggle over whether this was a cheat of an answer, because I said a mathematical biologist or ecologist, which is pretty close to what I did. Otherwise, probably environmental science. I am not sure I would now choose to go into environmental science, but certainly as a student both in school and in undergrad, I was really interested in studying the environmental impacts on human, plant and animal health.

One of the things I did in school was science fair experiments. So at one point, our conservatory was full of bean plants that I had planted. I watered them with my own acid rain, that I created by acidifying water at different levels of pH. By watering all these bean plants, I wanted to see how they grew and how I could affect them. I was doing that because I had an amazing biology teacher who inspired us to do science fair projects and actually be running these things at home.

For a long time, I was meant to do something biological and inquisitive. I also thought it would be cool to be a detective, which is an evidence-based and problem-solving profession, but I thought I would be too scared of the bad guys. It just seemed to me that if you were a criminal, it would be easy to figure out who the detectives were and where they lived; and the idea that they would be after you seemed too scary.

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 StatisticsViews.com are checked for statistical accuracy by a panel from the European Network for Business and Industrial Statistics (ENBIS)   to whom Wiley and StatisticsViews.com 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.