You might remember that back in late July papers from SAGE started to emerge on universities and the steps that they and government ought to take to minimise risk.
We discovered that in early July, SAGE had seen two sets of modelling on and from UK universities – stuff from Bristol which we looked at here when it was published in September, and one looking at “transmission and network dynamics” within the University of Warwick.
Over three months later, a more developed version of that Warwick work has now been uploaded, and guess what – reducing risk requires a bespoke testing strategy and good adherence by students to the rules on getting a test and self-isolating.
As ever the usual caveats on preprints apply – medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.
There’s a quite interesting “network approach” being used here, using four “layers” of contacts that students might have. In its “household contact” layer researchers constructed on-campus accommodation units to match that of a representative campus university. Then, within a household, they assumed each individual could transmit infection to each other person within their household – and might randomly make contact with students outside their direct household on the same floor or within the same block.
In what it calls the “study” layer researchers divvied up students into cohorts by department and stage of study, and assumed contact occurred between study friendship groups outside of any face-to-face classes. In other words, they have assumed that “Covid secure” teaching spaces minimise transmission risk – it’s meeting up afterwards that matters. And in its “societies and sports clubs” layer, researchers applied a probability of forming a contact with each other individual in the group, and in its “dynamic social contacts” layer a bunch of random, dynamic contacts made each day with other students were modelled.
They then ran the simulation with a student population of 25,000, with 7,155 students resident on-campus and the remainder off-campus, and ran it for welcome week plus the ten weeks academic autumn term.
If that sounds awfully like Warwick, you’d be right – although at least in theory that means one might be able to run the model again with different institutional parameters to see the results. The problem may well be that – just as with the Bristol model – the model doesn’t seem to take much account of commuter students or “hybrid” students that regularly travel between university and home during term time.
Anyway, they then assessed the situation with three big variables. First, they analysed how the strength of adherence to guidance on isolation and engagement with test and trace affected case burden and accumulated isolation time (ie the total number of student isolation days over the term). Next, they considered adoption of a policy of strict room isolation for on-campus residents displaying symptoms. Third they analysed a collection of scenarios involving the mass testing of students to study the impact on overall case load, the expected time spent in isolation per adhering student and the prevalence of infection at the conclusion of the autumn academic term.
By now this is probably unsurprising, but the model shows a vast difference between students adhering to test and trace rules and not doing so – we’ve an end of term student population prevalence here of 15% v 2%, along with a sizeable number of non-symptomatic cases at the end of the term. We’ve written before about support for and costs of supporting students to get a test and self-isolate, all of which appears to be being left to universities to handle.
In comparison, isolation rooms – where on-campus students reporting symptoms and testing positive are rehoused to prevent further transmission to housemates – doesn’t appear to make much difference. And mass testing made a difference, but not a huge one – an early mass test covering all students, for example, resulted in the lowest estimate for the proportion infected, but only as long as students played ball with test and trace rules.
There’s plenty of limitations to this Warwick model. They haven’t included university staff members, or infection to and from the local community, or the presence of other respiratory infections. And as we said above, the assumptions tumbling into the development of the model are all a bit Warwick, and only some of that problem is fixable by playing with the variables.
What’s also fascinating is the assumption that the big problem comes at the end of term when students return home for Christmas, a conclusion we know went on to influence formal SAGE guidance to SAFGE and is now eating policy bandwidth inside DfE and nations counterparts. As infection rates look like they are falling here in the middle of term in university towns and cities, have we got a fault in the predictive powers of the models or a big difference between on and off campus willingness to get a test?
Crucially, what this preprint seems to tell us is that testing isn’t the panacea some seem to think it is. It’s persuading symptomatic students to get a test and then persuading their household to self-isolate that makes the difference. And despite advice pointing us in this direction, we may not know how unsuccessful we’re being with students living off campus until it’s too late.