Interview with Dr Gwenan Knight, assistant professor at the the London School of Hygiene & Tropical Medicine.
What was the aim of your initial work on COVID-19 clusters for the studies you brought out in May and June, and how did you go about it?
Stepping back pre-COVID, one of the things I’m interested in is where antimicrobial resistance is selected for – is it in the hospital or the community? Similarly, for COVID-19 we wanted to know where people were getting infected. We did a review of the scientific literature in PubMed and used that as a standard baseline. Then we did online searches for media articles reporting clusters.
To gather this information, we made a spreadsheet database of peer-supported data. It’s not a comprehensive list of all clusters ever – but I think for the types of setting in which transmission has been reported to happen, we are covering many of our bases.
What did you discover about clusters during your work earlier this year?
We were expecting cluster settings like cruise ships, healthcare and elderly care locations – which is what we found. In addition, we found that many of the settings were indoors, which seems to be supported by ongoing evidence that there is less transmission outdoors.
We would also have expected lots of school clusters, so it struck us as odd when we didn’t initially find that – but not so much with the knowledge we now have. Kids seem to be much more asymptomatic, and we aren’t doing universal screening for infections, so we’re not necessarily seeing school clusters coming up. That’s something I would really like to look into.
At the moment, we’re trying to get data from Public Health England to explore the cluster data they have – and trying to say what would you expect and what do you see, given the knowledge we currently have of transmission?
We also tried to look for the first hundred cases in a country and where they got infected because we thought countries would definitely do it for the first 100, but we really didn’t find much data on that. We found very little open-access systematic contact-tracing data; a lot of it was media reports or one-offs rather than a systematic understanding of where transmission was occurring.
There are also many biases in the information that’s out there. For instance, there’s recall bias – you’re more likely to remember the big events that you’ve attended than the transport there or back. Media reports can also be massively biased, but we really tried to pin down reports with detailed numbers of people who were actually confirmed to be infected in that setting.
Is the database an ongoing thing and, if so, how do you plan to improve and expand it? Also, how can you overcome the data challenges you mention?
Yes, it’s ongoing. There are other databases out there, and we now have an amalgamation with another two that includes over 1,500 cluster events. The plan is to make that open access, updating ours with a web platform that people can contribute to, download and edit. Then the idea would be to maintain that, because what we’re going to see now as we’re coming out of lockdowns is the number of clusters starting to increase again.
What will be really interesting is to monitor whether the settings for transmission have changed and how they might have changed pre and post lockdown or pre and post intervention measures. We’re also extracting more data to try and say something about how transmission occurs.
But again, it’s going to be hard to pull all that information out. I would hope that at a country level, analyses will start to emerge of what settings are important. I do know of countries where they have been collecting data and they’re starting to make it available to researchers. Also, Singapore, for example, has had an amazing COVID dashboard which lists all the clusters they have, the number of cases there and things like that.
What do you hope will be the main use of this data?
As we want the database to be open access, I would say its main use would be as an evidence base and for determining exit strategies – which places we can open and which we can’t, and which are the most risky. Also, in which settings should you monitor individuals’ attendance?
What I would really hope for is for certain countries to release their data for us to use as case studies to say, look, this is what you can do when you produce this – it can help understand where to close risky settings or prevent risky behaviour, as well as aid increased interaction without increased transmission.
But if we think back to where we were at the start of this, we really had no idea where this was transmitting. For example, the lack of symptomatic clusters linked to schools was very different to what we would have expected. This database is another piece of evidence on understanding this.