Note: This article was written using numbers from the 24/03/2020 in the UK, if you are reading in the future the situation will almost certainly have changed. If there’s one thing certain about the COVID-19 pandemic, it’s that everything we say now will probably be wrong by next week.
In these uncertain times, any news relating to the novel coronavirus SARS-CoV-2 and the disease it causes COVID-19 is massive. Every piece of research that is being done on the virus is hitting the internet in a massive wave, which is understandable given that we are all panic-refreshing twitter for the slightest hint of an update virtually all the time these days.
So it’s no surprise that, when news of a new study published by Oxford University came out, it spread around the globe like the pandemic disease it was based on. Media sources from everywhere on the planet have reported that a study had proven that half of the United Kingdom had already been infected with coronavirus, which if true is pretty huge news.
If half of the country has already been infected, it means that all of the new restrictions are basically a waste of time, which would change the course of the pandemic response entirely. It would be a piece of truly wonderful news for us all.
Which makes it extremely unfortunate that it is almost certainly not true.
The study in question was a fairly simple modeling exercise that has not been peer-reviewed but was published as a pre-print by some academics in the UK. They took a fairly basic model of epidemic spread, called the Susceptible-Infected-Recovered (SIR) model, that gives you an idea of how many people are infected with a disease at a point in time based on whatever assumptions you enter in at the start. Based on their assumptions, they calculated that by the start of March, more than half of the entire population of the UK had been infected with SARS-CoV-2.
And while it all sounds very rosy — if most people are already infected, there’s no point in locking down anything — there was a huge, glaring flaw in the paper.
Whenever you read a model, the most important thing to look at is not the actual maths behind the graphs — although obviously this is important — it’s the numbers that have been plugged into the model from the start. If I assume that COVID-19 is a significantly milder disease than good old influenza, and model that curve, it will tell me that things are fine and dandy no matter what the situation looks like in reality.
Which, it turns out, is exactly what these scientists did.
The entire paper rests on a single assumption: that 1% or fewer of the population is susceptible to COVID-19. This rate is about half that of seasonal influenza, and would mean that the case-fatality rate of coronavirus would probably fall somewhere between 0.01%-0.1% overall. If you assume that the death rate is 1 in 100,000, then with about 400 deaths in the UK it would mean that 40 million people have been infected and voila! Good news.
The question then is really whether this assumption is well-supported or not.
I think you can guess what the answer to that would be.
Never Assume Anything
In short: it’s total nonsense. If you look at data from the World Health Organization, the US CDC, China, Italy, South Korea, and everywhere else in the world, it’s pretty clear that the case-fatality rate — the proportion of people who die per COVID-19 infection — is somewhere between 0.5–3%. The hospitalization rate per infection is unclear at this point, but the best estimates put it at 15–35%, depending on age group and where in the world you look.
If we use these much more realistic figures, our rough guess would put the maximum number of infections in the UK at below 100,000, with a more realistic estimate based on other modeling studies of around 40–50,000 cases total.
The problem is, as I’ve written before, that modeling is incredibly hard and you should only do it very carefully. The intention of the authors of this study was good — they were trying to get the government to start doing more tests on asymptomatic people — but the message that was put out was very different to that.
In times like these, it’s worth remembering the old statistical adage “All models are wrong. Some models are useful”. Predicting the ‘true’ number of cases in any country is not an easy thing to do, with the best models coming out well after the fact. If you are going to read a model, take extra special care to sense-check the assumptions — if they disagree with basically everyone else who’s written on a topic, it might be worth considering that they are wrong.
So no, it’s incredibly unlikely that half of the United Kingdom is infected with coronavirus right now. It is far more likely, based on the published evidence from across the world, that this pandemic will be with us for weeks or even months to come.
Be prepared, and stay safe.
Listen to Gid’s coronavirus special on Sensationalist Science: