The first science lesson I remember involved keeping cups of tea warm. I was in Year 6, aged 10 or 11. We were seeing whether styrofoam cups kept their heat better than china ones, and whether lids kept the heat in, and some other things; God knows, it was almost 30 years ago.
Anyway. The crucial point that our science teacher drilled into us was: when you do science, you change one variable, and you keep the rest constant. You want to see whether lids help? Then do one with a lid and one without, but if you put one in a styrofoam cup, put the other in one too, and use the same volume and temperature of water. Want to see whether styrofoam is better? Then do one styrofoam and one china, but if you give one a lid, you have to give the other one too.
The reason I bring this up is that a few weeks ago, it was announced that there would be a limited, phased re-opening of schools starting on 1 June. I had assumed that we would get to 1 June, give it a couple of weeks, and then see what impact — if any — the reopening had on infection rates.
But, then, a raft of other new measures were announced, all starting on the same day. Gatherings of up to six people are now allowed, in private gardens as well as parks; the extremely vulnerable are now allowed to leave their homes for short periods. Non-essential shopping is starting to reopen in outdoor settings. The Prime Minister told people that they can “even have a barbecue”.
But won’t it make it more difficult to assess the impact of each of these measures, if we haven’t kept each of them separate, changed one variable at a time while holding the rest constant? Isn’t that like putting a lid on your styrofoam cup but not on your china one?
In short: yes. “The measures have been coming one after the other or at the same time, so it will be very difficult to unpick the effect of any one measure,” said Dr Sarah Lewis, a genetic epidemiologist at Bristol. “By opening all at the same time you’ll effectively mask the outcome,” says Dr Konstantin Blyuss, a mathematical modeller at Sussex: “you’ll never know what hit you.”
You can reduce the uncertainty: partly through statistical methods, but mainly by using old-fashioned epidemiological contact-tracing methods. By simply asking people questions — where have you been, who have you been in contact with — you can see that Person A caught the disease from Person B at the local farmers’ market or the school gate or whatever, and can avoid the need for the high-level, statistical assessment of what influenced what.
But you need thousands of contact tracers to do that. Estimates suggest that, in order to keep the disease suppressed as they come out of lockdown, the US will need between 100,000 and 300,000 contact tracers. That implies that Matt Hancock’s army of 25,000 is at the low end of what the UK needs, assuming that the UK and US situations are broadly comparable.
And it’s not a simple job. A family member is a (retired) consultant epidemiologist who did a lot of infectious disease work; she once spent an entire Christmas Day tracing the contacts of infected people in a local meningitis outbreak. She says it’s not that you have to be a consultant epidemiologist to do it, but that it definitely helps having a background in epidemiology or public health so you know the follow-up questions to ask, and having good interpersonal skills so that people are willing to share private information with you.
US contact tracers are being offered a free six-hour training online course on Coursera, although whether they’re required to do it differs from state to state; in the UK, the training seems minimal. One US epidemiologist I spoke to (who works with the government so could only speak off the record) said that “absolutely there are training and skills required” for contact tracing, and that it’s not clear that over there, at least, those skills are in place. I don’t get the impression that they are here, either, although we shall soon find out.
This is of more than academic interest. If relaxing some lockdown measure — say, reopening schools — drives the rate of infections up, then it’s important that we know, so that we can decide whether or not to close it down again. If we only change one thing, it will be easier to know.
That said, it would still not be easy. Blyuss points out that people’s behaviour has been changing significantly as well, at least partly independently of government instruction, so it’ll be hard to tease that out from any policy changes anyway. “It’s nice to talk about rules about groups of six or more, but the reality is very different,” he says: “Everyone is in groups. The weather doesn’t help.”
Dr Babak Javid, an infectious disease specialist at Cambridge and Tshinghua universities, adds another layer of difficulty. It’s absolutely true that holding all of your variables but one constant makes tracking the disease easier, he says. But doing so comes at a cost. Say you reopen schools, and you want to see whether it has an effect on some outcome. The very earliest you could start to see anything would be a week or so later, when the first people infected would be showing symptoms, but realistically it would probably be two weeks; if you were looking at hospitalisations or deaths, even longer.
So if you wanted to reintroduce all these measures one by one, in this careful, targeted way, it would take a while. Including schools, I’ve mentioned four different measures. We’d be talking about two months to introduce them, as a bare minimum, and there are lots of other things to change — other school years, for a start. “If you do everything very precisely and incrementally, we’ll open up in September,” Javid says.
But the lockdown comes at significant social, health and economic costs. He stresses that he’s not saying that we ought to do it all at once (“I don’t want to comment on the overall net cost or benefit. That’s a policy issue, not a scientific issue”), but merely that it’s not a straightforward decision; many lives will be lost and damaged if we carry on with lockdown, as well as if we come out of it. “A&E attendance is way down,” Javid says. “Can heart attacks really have dropped by 30%? We’re delaying cancer diagnoses and treatments. These are genuine, real risks to prolonging lockdown, and they’re medical and sociological, not just economic.”
There is another way to assess the risks and impacts of the various relaxations, and that is to remember that this isn’t happening in a vacuum. We can look at other countries which have done some reopening themselves. I had a quick look at the indispensable Our World in Data, and picked six European countries — Denmark, France, Germany, Italy, Norway, Spain — which partially relaxed some measures no later than 11 May, to give us a full three weeks.
So far, there’s no sign in any of them of any increase in deaths or cases apart from a very tiny wobble in case numbers on 29 and 31 May in France, which seems likely to be noise and doesn’t seem obviously connected to measures taken three weeks earlier. All the scientists I spoke to agreed that there’s no sign anywhere of any major second waves; Lewis mentioned the Netherlands and Switzerland as other examples.
It struck me that there are three possible explanations for this. One, which is moderately hopeful, is that the measures taken to relax lockdown in each country are sufficiently conservative and well-targeted to avoid sparking a new outbreak. The measures are all somewhat different, and that makes me think that our own first, cautious measures — which look roughly in line with the broad swath of others — might similarly avoid a new wave of infections.
The second explanation is less hopeful. We’ve all become aware of the concept of “R” recently, the reproduction rate of a pathogen. It’s the average number of people that one infected person will pass the disease on to.
The word “average” is key there: If the R is 2, that could mean that out of 100 people, every single one passes it on to two people; or it could mean that 99 people pass it on to nobody, and one person gives it to 200 people (or anything in between).
It’s recently become clear that Covid-19 is more like the latter of those two examples. Most people don’t pass it on at all, but every so often, one person will happen to give it to dozens or hundreds. And that means that the behaviour of the disease can look very different, as the LSHTM epidemiological modeller Adam Kucharski explains in this Twitter thread. It could be that the disease lies apparently quiescent for a long period, as most people fail to pass it on, and then when we think we’ve got it under control it explodes up again. It may be that the reason we haven’t seen more resurgence in Italy or Spain is simply that they’ve been lucky so far.
(Equally, it may be that the superspreader events are largely prevented by the sort of lockdown measures that remain in place, says Javid — things like keeping bars, sporting events and churches closed.)
My third possible explanation is that there’s something strange going on with immunity, so some people are differentially immune to the disease even before they get it. The serology data coming back has found that even countries that have had terrible outbreaks — like Spain and the UK — have only had about 5% of the population infected. If there’s some reason why a non-negligible percentage of the population were already entirely or partially immune, then that could be keeping the second wave somewhat suppressed.
That third option is extremely tempting to believe, and it’s not insane — Oxford’s Sunetra Gupta believes it, and all of the scientists I spoke to thought it was plausible, but that as yet there’s not much good evidence for it. They also all said that a combination of all three factors was possible. “I’ve read evidence pointing to all three of those,” said Lewis. “I saw something suggesting herd immunity had been reached in London because not everyone exposed would get the virus, which would explain why virus numbers have gone right down despite people using public transport.” It’s worth adding that, pace the neuroscientist Karl Friston, Javid said it seems very unlikely that this background immunity theory explains the difference between the UK and Germany.
It has a dark side, though: one more expert, who also wanted to remain off the record — it’s all very cloak-and-dagger, this epidemiology stuff — pointed out that there have been very different levels of infection in, say, London and the south-west. Something like one in six Londoners have had it, compared to one in 20 elsewhere in the country.
He thinks that as measures are relaxed, the danger is that we see London do relatively well — acquired immunity, on top of whatever this mysterious background immunity is, meaning that the disease finds it hard to spread — while places like the south-west see sudden upsurges, because the population is still mainly susceptible. If he’s right, and he said it was his ‘prediction’, then major second waves could be triggered. It’s worth remembering that the second wave of the Spanish flu killed more people than the first — but it’s also worth remembering that public health policies were very different back then.
I started by saying that in science, you hold all your variables constant except one: keep the lid on your styrofoam cup and your china cup. That’s true, and if we were doing pure science — if we only cared about finding out what lockdown measures worked and which didn’t — then it would be simple: introduce measures one at a time, wait and see, do it slowly.
But we’re not doing pure science. We’re also trying to make a country that works for its citizens, in conditions that change daily. “We’re trying to build a plane as we fly it,” my US epidemiologist told me. The most important thing, according to Javid, will be “nimbleness; being able to change policy in the light of new evidence”. If it turns out opening schools was wrong, then close them again. And we in the media need to be wary of shouting about mistakes and U-turns and instead say: when the facts change, you change your mind.