Would student social bubbles cut deaths from Covid-19?

Jim Dickinson looks at a working paper on bubbling and virus transmission - and isn't very reassured

Jim is an Associate Editor at Wonkhe

We’ve all done it. We’ve ventured out to the shops, seen a group of young people gathered somewhere (without appropriate social distance) and thought – what if the virus spread amongst each other – and then back to their (what like look more vulnerable) families?

And what if that was a group of students in September?

One of the things I highlighted in the long piece on SAGE and higher education was the relative (to the US) lack of academic work on the potential impacts of the “return to campus” of students in September.

A working paper from a Professor at Swansea University has emerged aiming to fill part of that gap – and its executive summary has quite an arresting opening bullet point:

Without strong controls, the return to universities would cause a minimum of 50,000 deaths.

And that’s a “rough but optimistic calculation”.

Alan Dix is Director of the Computational Foundry at Swansea, and what he’s done here is tried to model two things raised in that SAGE paperwork – reducing the spread within campuses, and limiting spread into wider society – treating the university as a large self-quarantine unit.

Of course, in an ideal world we would have modelling carried out by epidemiologists rather than computer scientists – but as yet very little has appeared in a UK context.

To get a handle on it all Dix tries to estimate the impact of where much of the sector is at right now – unrestrained campus re-opening (ie socially distanced, but not “bubbled”) and discusses (in informal terms) these two flows of infection, both within the student body and to the wider population.

On “bubbles” – groups of students housed and taught together that in theory contain infection and help efforts to track and trace once cases are discovered – he looks at groups of 10, 100, and 1000 to see what the impacts would be.

Then for interaction with the local community, he takes a punt at using the “normal” population level of contact. That’s not ideal, but as he says:

This could be an overestimate of cross-infection (especially for self-contained out-of-town campus universities) as students are largely interacting with one another, or might be an under-estimate (especially for city or multi-campus universities) as students will need to travel a lot on public transport between halls and campuses, eat out, and maybe be more “risky” in terms of night-life, etc.

He also notes potential “locality effect” that universities and local Directors of Public Health might need to consider. His modelling has to assume perfect mixing outside campus – where everyone in the population is equally likely to infect any other person. But in reality there are strong geographic effects – he notes for example that 100,000 university students in Manchester is 20%, in Aberystwyth students are a third of the term time population, and in St Andrews students outnumber the rest of the population of the town.

There’s plenty in the paper on the models used – but to cut a long story short, Dix concludes that to be effective, bubbles have to be small (approx a dozen) and very strictly maintained – probably more strictly maintained than is possible once you think about it for more than a few minutes.

He calculates that even a bubble size of 10 would increase overall population R by 10-20%, and if bubbles “leak” into wider groups of 50-100, this leads to larger scale outbreaks.

Essentially, Dix argues that a Covid-19 freshers’ flu-like surge is coming – and suggests that bespoke, on-campus testing and track and trace will need to be established very rapidly, as well as indicating that student to non-student transmission is also critical to protect communities. How viable the former is and how realistic the latter is remains to be seen – or, indeed, modelled.

Update: Since Dix’s working paper was published, both the paper and specifically its “headline deaths” figure have been subject to considerable criticism – see for example this discussion on it that featured on the BBC’s “More or Less” on 2nd September. This is of course the nature of working papers, but as we note above – this is arguably why the government’s SAGE pandemic modelling sub committee said back in July that proper modelling of UK campus reopening carried out by the right people was important, to identify both the level of risk and to determine important “commonalities and generalisable insights”. It’s certainly something we know has been done and published in relation to schools.

If fresh modelling has been done and supplied to government, we’ve seen very little of it in the open so far as of early September. As we note here, the last we heard was that DfE’s Chief Scientific Advisor was to establish a new science advisory group for higher and further education drawing on expertise from SAGE (and its subgroups) and PHE to provide advice – but we’ve heard nothing from it so far other than those SAGE minutes mentioning its creation.


Alan has since posted an explanatory blog on the work here.

3 responses to “Would student social bubbles cut deaths from Covid-19?

  1. You’re right that there’s not a lot of available academic work out in the UK on COVID and the return to university. I know quite a few people in the UK who are doing modelling in this area, and agree it would be good if some of that work made it to a public preprint server soon. In the meantime, there’s at least a working version of a collaborative report involving mathematicians (especially mathematical epidemiologists), epidemiologists, public health people, etc: https://gateway.newton.ac.uk/sites/default/files/asset/doc/2006/Unlocking%20Higher%20Education%20Spaces%20VSG_BRIEFING.pdf

  2. Not sure this is a wise blog post tbh, Jim. Plenty of other ways campaign for safety measure then use a very iffy model, not even an epidemiologists, and broadcast it. UCU now picked it up. its really not a great look at all.

Leave a Reply