“It’s about being ‘about right’ now, rather than perfectly right too late”

Dr Adam Kucharski, from the London School of Hygiene and Tropical Medicine, tells Tom Ireland how modelling has provided rapid insights into many aspects of the COVID-19 outbreak.

Can you give us an idea of what your research involved before the COVID-19 outbreak, and how it has changed since?

My work focuses on understanding infectious disease outbreaks. Prior to the current situation we were doing a lot of work with serological surveys and data on antibody dynamics, looking at things like flu, dengue, and Zika. We try to understand what this data tells us about infections in different groups, trying to understand how, over the course of a life when you get infected with multiple different viruses, what that does to your immune dynamics.

Alongside this we were also doing quite a work on understanding social behaviour. If you imagine a flu epidemic, its spread depends both on how people are interacting and the immunity in those groups, so we try to look at those trade-offs.

Obviously with COVID-19, being a respiratory infection, there's some similarities, and we've pivoted over to that alongside existing research. LHSTM has a huge number of people working on a range of different things, but certainly a large proportion of the modellers have moved over to COVID; a lot of us worked on outbreaks and real-time analysis projects already. And some of us who are fortunate to have Wellcome funding and these kind of things have a bit more flexibility to switch. There are other people who are maybe allocating 20% of their time to support ongoing research.

I was interested by your Twitter thread about all the different insights modelling studies have helped provide during this crisis – the idea that it's not just about predicting how the disease will spread through a population. Can you explain what modelling suggested about the virus that other studies could later confirm?

I think models can have two main functions – one is to make some predictions – either about specific scenarios, or more general forecasts, the other is just to help us understand something and explain what's going on. In the early stages of this outbreak models did a lot to actually give us clues about what was really happening.

One early model for example, looking at cases that had been picked up internationally outside Wuhan, suggested that there were probably hundreds if not a couple of thousand cases already, at a point when China had only reported about 40 pneumonia cases. The model was a signal that this was much bigger than we thought. There were also early models that showed this was probably human to human transmission rather than just a single point-source spillover event.

A lot of the work we've been doing isn't just these big medium-term scenario analyses, but actually just trying to have better numbers on things, for example, trying to estimate what proportion of people who get ill will die from it, what proportion who get infected will die. In the data that just came out of Germany today [09/04/20], from a much bigger serological survey, the estimate of fatality risk post infection is pretty consistent with the very rough estimate we came up with about six weeks ago. So I think modelling can be a very useful tool for making sense of all the different bits of data you have available.

We’re interested in how science works during a crisis – what have you noticed is being done differently owing to this being an emergency?

I think one thing with outbreaks that is different to traditional scientific culture is just the speed. It's a classic example of being about right now rather than perfectly right a year too late, and I think definitely in the work we've been doing we've been trying to get things out very quickly.

Even submitting to pre-prints now is arguably too slow, let alone traditional journals, because it takes quite a while to get them screened. We're now putting things on our website as well as preprint archives, to make it available quickly. We've added an internal peer review system now for all our content, because we realised that if we are sort of bypassing peer review in the short term, we want to make sure you are at least double-checking the work before it goes out.

I think the openness between teams has been really good. If you look at past outbreaks, like Ebola, there was often data coming out months or years later that nobody had any awareness of at the time. I think that culture has shifted a lot.

A lot of what we're doing we do knowing it has a very finite usefulness – if you get some rough data and get a severity estimate or a transmission estimate, in a month or so there'll be better data that will mean it's not needed anymore, but in that period of time your estimate could enable a lot of other scenario analysis that wouldn't be possible. The idea here is not to do a long project and get a paper that will define the field for the years to come. It's getting something that's immediately useful and available, with the knowledge that you're going to have to refine it over time.

What kind of tools are you using to communicate your work quickly and how are you keeping track of other data?

The deluge of things coming out is hard [to keep up with]. We've actually got Master’s students who are helping to triage a lot of papers coming out - particularly as there are now so many case reports and these kind of things that actually collating what is and isn't in those papers is really important, but very hard for a small group to do.

We have a dedicated repository set up, which means that we can make papers and interactive tools and code databases all available together in one place. And it means that we can just get things up as soon as they're ready.

Have you sought new funding or additional support for the work you're doing, and if so how easy was it to navigate the relevant systems to get that to where it's needed?

I think the funders have been better than perhaps in the past. We've got some from, for example, UKRI to build up those activities that we're doing and expand them, as well to go out and conduct things like social interaction surveys. Obviously the population is interacting very differently to how it was a month ago, so being able to quantify that is really important for trying to estimate transmission.

Through EU grants there are some consortia building up and we also got some funding to set up some sort of household surveillance in low and middle income countries to actually look at places where it’s harder to pick up cases and understand transmission.

Can you talk us through some of the challenges of working during these strange times? For example, keeping yourself and staff safe, getting hold of in-demand equipment and reagents, or the effect on other research and departmental business?

I think it's important to make sure that for junior researchers who have insecure funding, that we can make sure they're in a better position. There are many people across science who are on quite short contracts and are now taking time away from the research that would get them the traditional CV-friendly outputs, really giving up their time to work on this because they think it's important. That's great but equally we need to make sure that they're supported.

In terms of how we work, we made the shift to remote working as soon as that advice was changed. There are challenges around working with some sensitive datasets, making sure that we're treating that data securely but also making sure that people who need to access it can to do analysis on it. And I think as many people finding life is now a series of Zoom and Skype meetings.

I think for modelling, where often you want to just get a whiteboard and sketch a few things out and it'll take five minutes, things can actually take a bit longer. So I think we are making those adjustments.

In terms of equipment, we can do lots of things on a normal laptop. If you're running a dozen scenarios for the UK that might be a few hours on a laptop. If you're trying to do that for 50 countries, then we do have high performance computers and these kind of things that we can turn to and access remotely as well.

Broadly speaking, how do you feel the life sciences has responded to this crisis?

I think generally it seems to have been pretty open and collaborative. There’s obviously areas of disagreement over what to prioritise and what's happening where but I think we've seen a lot of good, useful work coming out very quickly. Especially as things like serological tests ramp up and we have more data coming in, I think a lot of the groups that I know who are involved in that will be doing really useful stuff.

Dr Adam Kucharski is an associate professor at the London School of Hygiene and Tropical Medicine (LSHTM). His research focuses on the dynamics of infectious diseases, in particular, how social behaviour and immunity shape disease transmission.