The R-number – and the danger of false certainty

11 May 2020

5:48 AM

11 May 2020

5:48 AM

Not much about Boris Johnson’s Sunday night television address was clear. The one definite new measure – one which will shape coverage for weeks to come – is the UK’s new ‘COVID Alert Level’, a five-stage measure that the prime minister said would be determined primarily by ‘the R’ – the rate at which the virus grows.

There’s just one problem with that. It’s a figure few people think we have any real ability to track day-to-day. In making its decisions, the government has very little scope for error: it knows lockdown comes at a huge economic toll. But it also knows that spotting a rise in cases just a few days too late could end in calamity, if it lifts restrictions too early. The R number was presented as a solution to that, something whose precise measurement will determine government policy. The Prime Minister said he’ll be driven by data – but is he being honest about how reliable the data is?

At the moment, we hear a lot of very precise numbers at government daily. We hear an exact figure in the tens of thousands about how many new tests have been conducted. We hear a number of new cases in the thousands. And we hear a number of deaths, still in the hundreds. These appear definitive. They’re being obtained by dint of great effort and announced as official government figures by senior ministers every day. But one person left largely unimpressed by them is Professor Sir David Spiegelhalter – and you might expect him to be onside, given he’s the professor of public understanding of risk at the University of Cambridge.

Spiegelhalter’s concern, shared over the weekend with the BBC, is that the level of precision being given is a false one. The tests total handed out includes a variety of measures that might not meet a normal person’s definition of a test being ‘done’. The totals are preliminary estimates at best, not exact counts. Even the death tolls – the count you think would be the most immutable – are substantial undercounts, still being updated for particular days after several weeks, or even a month or more, has passed. They will rise further still when, months later, a proper reckoning of excess mortality can be made.

His issue with this is that it doesn’t respect the public: it is false precision and thus grants false authority – by not admitting the uncertainty around all the numbers being used, it gives false reassurance. We could communicate more honestly.

This seems an entirely reasonable proposition. But the UK government is basing its rationale on loosening lockdown with all the authority of science and of certainty – there will be a five-level system, akin to that of terror alerts, to let the public know our current level of risk. These levels will change as ‘the R’ rises or falls, and with them the severity of the restrictions to our everyday lives will adjust.

Another warning of the dangers of ‘false precision’ has been made by Dr Adam Kucharski of the London School of Hygiene and Tropical Medicine, an advisor on the UK’s SAGE committee, and the author of ‘The Rules of Contagion’. He warns that there are lags built in at every stage of collecting data – hospitalisations happen a few days after infections, deaths may not be recorded for days. As testing criteria change, the rate at which we find out about new cases might change, and the case totals themselves may become incomparable. Furthermore, the models designed to estimate Re have to contend with complexities in transmission – if new cases shift from mostly being in the community, to in care homes, and back again, the figures tell us very little.

The danger becomes that we build not so much a statistical house of cards as an intricate and elaborate statistical factory – models taking in too-precise figures and estimates from other models, generating their own results, which are fed into yet more models, all of which then disappear into a closed box, out of which pops whichever answer the government has decided we want today.

The government has been handed a task – managing the Coronavirus outbreak – it never asked for, and which is difficult beyond the ability of most of us to comprehend. Even the advisors just trying to calculate figures upon which those decisions can be based are likely doing the most difficult and high-stake work of their careers.

It’s high stakes. And despite their greatest hopes, there’s absolutely no way to know how likely it is to work. There is a real risk that with the false level of certainty they’re projecting, Boris Johnson’s government is talking out of its Rs.

Got something to add? Join the discussion and comment below.

Show comments