Modelling Coronavirus - What's Next?
About a month ago, I wrote an article about whether we should be scared about coronavirus or not. Flippantly, I said no. I pretty much thought it would all blow over, or that we wouldn’t get hit as hard as Italy or China. Clearly, I was wrong. Very, very wrong.
It’s probably not an exaggeration to say that the coronavirus crisis will be the most severe worldwide event since the Second World War. I hope that anyone reading this is keeping safe and coping with these ridiculously trying times.
I’m not here to say what the world will look like after all this because I have absolutely no idea - be very sceptical of anyone that does know it. But I will say this. We will get through it. Pretty much the entire world is working on trying to find a way out of this pandemic. That has to count for something! Also, hopefully, at the end of it, we’ll stop taking the NHS for granted and fund them to an absurdly high degree, because they fucking deserve it.
In the short term, one month into lockdown so far, I guess we’re all just starting to get into a routine. At which point, it’s pretty natural to look around and think “What next?”. When the government is going to the lockdown, and how are they going to do it? What’s the best way of getting life back to normal, without overstraining the NHS?
These are incredibly complex questions, so you’ll probably be pretty glad that a bunch of politicians aren’t the ones trying to answer them. They rely on experts to advise them who spend their lives trying to predict the spread of diseases and the behaviour of the public in particular situations.
Here I want to give you a bit of a flavour for the models these experts use to make their decisions. Perhaps, that’ll show you that we are not entirely flying blind. There is strong scientific backing for the steps we are taking for coronavirus, and hopefully, it will be translated into a good plan (although that’s really the job for the politicians).
One of the most challenging parts to handle about coronavirus is time. In fact, it takes on average two weeks for an infected person to start showing symptoms, which means that any changes the government makes will take two weeks to have any effect. So more than ever, we are forced to rely on mathematical modelling to try to get ahead of the coronavirus.
I don’t want to bore everyone into a stupor, so skip the next paragraph if you’re not interested at all (I won’t mind, I promise).
Mathematical modelling is pretty much exactly like it sounds. Writing out extremely complicated sets of equations to describe a situation, and then solving them (usually with computers). These might be equations to describe the spread of a virus through the population (generally very effective), or other ones to try to model things like drug interactions (these usually don’t work well). These models can be deterministic (only one exact outcome) or stochastic (random). Economists and social scientists have an entire branch of models to try to predict people’ behaviour within specific situations. But are also useful for understanding how best applying social distancing measures, for example.
The model that appears to be informing most of the government’s work at the moment is known as an “Agent-Based Model” (ABM). ABM is a computational model that takes actual data about the communities and movement of people and uses them to simulate the spread of a disease through the population by resembling the actions of individual people and collective entities such as organisations and groups. By feeding in a different range of parameters, we can try to predict different scenarios.
It was the mathematical model based on the 1918 Spanish Flu pandemic that drove most of the early response to coronavirus. In particular, it was the reason behind the controversial decision to not go into a full lockdown earlier. Models suggested that doing this would simply postpone the disease and we would get a surge as soon as the lockdown is lifted.
It was a separate modelling work which led to reverse this decision later on predicting that it was going to be unmanageable for the NHS.
In essence, I cannot over-emphasise how much I love modelling. And yes, I know this makes me a huge nerd.
However, these models would be completely pointless if there wasn’t a massive effort to understand the real data of how the disease is spreading through the population. Here is where epidemiologists and statisticians come in. Data can help to validate mathematical models, as well as giving actual feedback on how we’re doing.
So, how is this done? What kind of data do we need to care about?
Probably the most important quantity for understanding how a disease spreads is the basic reproduction number (R0). R0 is the average number of cases that will be caused by a single infected person, in an otherwise healthy population. Thus, R0 = 2 means that an infected person could potentially infect two more people. The R0 is the critical quantity in responding to a pandemic. If the R0 is greater than 1, then the disease is spreading, and the number of cases will grow exponentially. If the R0 can be reduced below 1, then the disease will ultimately die out. The R0 essentially dictates how much action has to be taken; an infection with a large R0 will quickly spiral out of control. Social distancing, lockdown, and ultimately vaccination, are some of the ways that R0 can be controlled. The R0 can be fed into mathematical models to predict the spread of the disease.
We are stuck in this lockdown because coronavirus has an exceptional high R0 (which is estimated to be somewhere between 2 and 4). Some epidemiologists are using infection data from around the world in an effort to constrain the R0 value of coronavirus. Which, in turn, informs governments on how to respond to the pandemic.
As usual, though, this process isn’t perfect because it relies on a large amount of good quality infection data, which unfortunately we do not have. Due, in part, to the difficulty of obtaining reliable information from some countries (as we have seen recently with China in the news). Another problem is that it’s hard to tell how many people have been infected by coronavirus, since the majority of them only present minor symptoms, so they will never go to a hospital and contribute to the reported cases. Until we can test more people in the population, the reported number of cases will always be underestimated, and our estimations for the R0 value will likely not be correct. Probably this is something that will improve over time which is one of the reasons why there is such a push to develop and implement tests. The R0 is not the only important parameter for understanding coronavirus. Still, the combination of statistics and modelling will hopefully let us estimate the others and try to understand this disease a little better.
So where does that leave us?
We all know the reason why we are in lockdown. The NHS simply can’t handle the influx of patients if we weren’t. And obviously, we can’t stay in lockdown forever. It’s even worse than that because if no one in the population has been exposed to the coronavirus, then no one is going to build some immunity and we’ll be vulnerable to another surge in cases. It seems like an impossible set of things to balance and to inform our decision we’re currently relying on epidemiologists, modellers and statisticians - hopefully, this now seems slightly less of mystical art.
For the first time in human history, you can literally save the world by sitting inside and doing nothing. Then do it, listen to the government, trust that they’re making this decision based on some good evidence, and please please stay safe.
Written by Griffin farrow
I’m Griffin – hi! I’m a PhD student in Physics in London, and yep, that makes me about as boring as it sounds. In my spare time, I try to work out what to do with myself. I’m new to this writing thing, so please be sympathetic!