October 26, 2021

A family member went to get his booster jab the other day, at a drop-in centre. It involved queuing for two and a half hours, outdoors, in a car park. The other people in the queue were mainly in their eighties; several gave up and went home.

But after an autumn of absolutely pathetic numbers — for a while we were barely jabbing 40,000 a day, first and second doses combined, down from ten times that number and more in the spring and early summer — the programme does appear to have picked up a little. There were hundreds of thousands of booster jabs alone given over the weekend.

The Government has pinned its entire winter strategy on those boosters. Sajid Javid, the health secretary, has said Plan B, will not be introduced because of them. Plan B is a milquetoasty set of minor interventions, including greater use of vaccine passports, and legally requiring masks in some places. They may ask for people to work from home if things get worse. It’s not exactly Protect and Survive.

And we’re not going there. Yet.

But there is a problem, which is that Covid-19 in the UK has become really hard to predict. Last year, whatever you looked at — caseshospitalisationsdeaths — the numbers followed a simple pattern. They started going up slowly, then they went up quicker, then quicker, then quicker. Then the Government imposed a lockdown, and, a little while and a few thousand more dead people later, the numbers crashed back down again.

But this year, the pandemic is no longer doing nice straightforward exponential curves. If you look at hospitalisations over the last six months or so, you’ll see that they started a familiar-looking rise in about June, and they clambered up to about 1,000 a day in late July and then – they just stopped, and went back down a bit. Then they rose again, and back down. Now they’ve started going up again, but … will that carry on? Will the curve fizzle out again? Unlike last year, it’s not easy to tell what will happen next.

Adam Kucharski, a mathematical epidemiologist, explained why winter 2021/22 is trickier to model than 20/21. First, he says, in any epidemic, “When R is near 1, it’s a headache.” R is the number of people that each infected person goes on to infect, on average. So if your initial 100 cases lead to 200 subsidiary cases, then your R is 2. If R is above 1, your epidemic is growing exponentially; if it’s below 1, it’s dying away.

When R is well above 1, it’s fairly easy to predict what will happen: cases will go up, faster and faster, until it runs out of susceptible people or until policies and behaviours change so that R goes below 1 and the epidemic declines again. That’s how it was last year.

But now, thanks to vaccines and natural immunity, the number of susceptible people around is lower — we’re close to herd immunity — so even with essentially no restrictions, R is close to 1. And when it’s close to 1, smaller things can affect the course of the epidemic much more. “It bounces around,” says Kucharski. “Schools reopening, or closing for half term. The vaccine rollout and boosters. If the epidemic is increasing or decreasing, some small change can knock it the other way.”

Suggested reading
Sajid is right: we don't need more restrictions

By Amy Jones

One way to think of it is that if your R is 2.5, and, for instance, schools close for half term, then your R might go down to 2.3. It’s the difference between an epidemic that’s spreading quickly, and one that’s spreading not quite so quickly. But if R is 1.1 and you knock it down to 0.9, then it’s the difference between a pandemic that’s growing and one that’s shrinking.

It also means that individual behaviour becomes important. Around our way, it recently became very obvious that lots of schoolchildren (and their parents) were getting the disease. So when ours got the sniffles we were quicker to take them out of school, and more wary of going to shops. Presumably millions of people have behaved in similar ways. It wouldn’t have made much difference last year — the disease was spreading so rapidly — but it could be enough now to change things from an up-arrow to a down-arrow.

There’s a more interesting dynamic at play as well, says Kucharski. Imagine the first two waves of the pandemic as a forest fire in a virgin forest. Once a tree starts burning, it easily spreads to the trees around it, and, left to its own devices, the fire will carry on until the forest has burnt away. An immunologically naive population is the same. “When you have a fully susceptible population,” says Kucharski, “the simple exponential curve models will give you something sensible.”

But we no longer live in a virgin forest. Vaccinations and acquired immunity means that the large bulk of the population is less susceptible to infection. In our analogy, you could imagine that it’s a far sparser forest, with occasional clumps here and there representing groups with lower vaccination uptake. Now, if a tree catches fire, it might not spread to any others. Or it might be in a clump, and spread around that clump quickly, and then burn out. Or a spark might make it from the clump to a nearby clump. It’s much more dependent on luck and randomness.

“You have a much patchier network for the virus to spread in,” says Kucharski. “A lot more dead ends, and surges in younger groups and places with lower vaccination cover; you get pockets of susceptibility, and you get large outbreaks, but it has to work its way from one pocket to another. When you’ve got lots of these happening in parallel, you have a much bumpier dynamic, more conducive to a bobbling-along plateau than one that has clear growth or decline.”

All of which makes the epidemic much harder to predict. But there’s another layer of difficulty on top of that, which is that you’re not just predicting an epidemic: you’re predicting human behaviour. More specifically, you’re predicting the behaviour of the current government and parliament. There is not an algorithm which says “At 1,500 hospitalisations a day, we re-enter lockdown” or anything.

Other countries do have something like that — in Taiwan, for instance, there are four levels of restrictions, which are imposed on a specified set of conditions. Level 1 is imposed if there are cases imported from overseas that result in “isolated community transmission”; Level 2 is imposed if there are “domestically transmitted cases from unknown sources”, etc. If you could predict the behaviour of the epidemic in Taiwan, you’d have a pretty good basis for predicting the policy response to it.

But not in the UK. Plan B, or any further restrictions, will be imposed if the Government decides to impose them. I think that’s a bad system, and the Taiwanese method of removing the decision as far as possible from political pressure would be better. But that’s the system we have. So as well as predicting what the virus will do, we have to predict how 361 Tory MPs will respond to the virus, and how Boris Johnson and Sajid Javid will respond to them. So predicting whether we’ll end up in a lockdown is hard.

But perhaps there’s a more interesting question, which is: should we act as though we might have to, and start doing something about it now? Should we impose Plan B?

For all the current doom and gloom, we are in a better place than even the most optimistic of the government’s SPI-M-O models projections from early September. And those models now tend to forecast a decline in cases from later this month, even if we don’t impose Plan B. We’re not facing an immediate crisis.

Oliver Johnson, a mathematician at Bristol, says an autumn decline is plausible: “The key thing is it has been so crazy in schools, with 7% or 8% of schoolkids infected at any one time,” he says. “You can’t have that forever: you’ll run out of people. So it’s a case of working out when you’ll run out of people.” The little clumps of forest will burn down eventually.

The pandemic, though, has been consistently hard to predict and the great failings of the British Covid response have been trying to be too cleverdoing things too late, and not thinking about risk-benefit calculations or worst-case scenarios.

This feels very much like one of those situations. It’s true that, probably, if we leave it and do nothing, things won’t get that much worse, and may soon get better. But there’s a non-trivial chance that they will get worse. Given that the costs of imposing Plan B restrictions are minor, why not take simple, low-cost steps now to avoid the unlikely, but plausible and very bad, future shit show.

We have a suite of tools available to us – not just Plan B, but also things like increased rapid testing – that can avoid future disruption. But, also, if we impose them now, we can undo them later easily if they’re unnecessary, or tweak them if they’re not quite right.

For instance, there are concerns that vaccine passports will damage the entertainment industry and have only a small impact on the virus. If that’s true, we can reduce their use once that’s clear.

And we should be able to avoid stringent restrictions being imposed again, and another catastrophic Christmas lockdown. “Given the tools we now have available,” says Kucharski, “it would be completely absurd if we got to that point.”

I agree. We absolutely should not need to consider a Plan C, one in which we shut stuff down, stop people seeing their loved ones, close the pubs, ruin Christmas. And we can make it all but certain that we don’t have to do that if we take small, sensible precautions now. Hopefully enough elderly people will stand in cold car parks to get their boosters; but let’s not rely on it entirely.