Coronavirus Lessons: A Scratch Analysis
What Britain can learn from other countries & whether its current strategy is sound
by Ethar Alali
I spend a lot of my working day just doing math in different contexts and building tech to crunch it. With COVID-19 here, that hasn’t changed and the fact the company is doing well, is because of decisions the math told us we should take 9 years ago, which I’ve seemingly annoyed a lot of folk by saying.
Here in the UK, I’ve been looking at whether or not our public health response has been any good and whether other countries have done things we probably should have, or at least can, learn from. I’ve looked at the COVID-19 responses as of yesterday and compiled a quick table of correlations.
The data covers 165 countries reporting their case numbers to the World Health Organisation. The WHO recommends that countries test, test, test. Yet, the UK has chosen to take a very different approach and only test when people present with relevant symptoms in hospital. With some virus carriers not presenting any symptoms, on the face of it, this doesn’t seem like a particularly sound plan.
Accounting for temporal differences (the virus didn’t start infecting everyone at the same time) the death rate in countries which don’t have such a programme are significantly higher than average, but crucially, it is the public health response that appears to matter, not the number of residents.
Test, Test, Test!
The WHO has been consistent in its message. Every country should undertake a programme of testing for Coronavirus and act accordingly. This is extremely solid advice from a statistical perspective. When crunching the data, a number of significant correlations are clear and these give us a theoretical strategy for dealing with the outbreak.
The number of deaths is not strongly correlated with the number of people in a population, nor is it evenly distributed across nations. Hinting that localised public health action is probably better than national. Partly due to the speed at which localised testing can occur, without centralised authorisation. NHS trusts and Clinical Commissioning Groups around the UK would do well to act first and not wait for NHS England confirmation. NHS Scotland has already acted and is ahead of the game in this respect.
Recapping Correlations
The correlations in the data are pretty strong. Correlations effectively tell us that the behaviour of one thing is pegged to the behaviour of another. It’s basically a measure between -1 and +1 which tells us how related two events are. Usually, a “size” of anything over 0.7 is regarded as very strong; 0.5 strong and anything under that, weak. A positive number shows two things move up and down together. A negative number shows the move in opposite directions. So if one thing gets bigger, the other gets smaller.
For example, a very strong positive correlation (0.7 or above) between someone’s shoe size and… err… height, tells us there is a relationship between height and shoe size and if someone is taller they’re more likely to have bigger feet. Similarly, shorter people have smaller feet.
This all sounds obvious. But being able to quantify this simple concept, when combined with a number of other techniques can provide some serious insight into the relationships between doing and not doing things, simply from data.
Some relationships are incredibly strong and may as well be the same variable. The correlations tell us that the number of critical cases are closely related to the number of deaths but a fair number of people recover, even in critical cases. However, the correlations are not equal. Meaning it weights more against recovery. If you get a lot of critical cases, more of them will die than recover. So this means you must treat cases before they become critical. An obvious statement.
However, the recovery and critical detection correlations, also tells us that you must be able to detect cases to prevent critical cases and give people the best chance of recovery from all cases.
But this is all detected cases right? We could work backwards and say that detection leads to death! Well, no. At least, it isn’t when extrapolating a key fact.
You might have heard the saying “correlation is not causation”. this is absolutely correct. However, causal relationships always correlate one way or another. You just need a single piece, like a jigsaw, to switch it. With COVID-19, that piece is the fact you can’t bring anyone back from the dead! That one single fact removes the confounding “direction” and means these correlations can only work one way. Turning it from a correlation to causality.
This unravels the argument and means that detection is the most crucial part of the process. You must ‘Test, test, test!’ and start testing early. The WHO were right.
Empirical Evidence
That’s the theory. Is there anywhere this works in practise?
- Italy has by far the greatest number of deaths per million members of the population. Italy’s response was initially slow. The full quarantine is a relatively new measure designed to stem transmission of the virus. However, testing in Italy was left quite late. A few towns deciding to ignore central government and test their citizens regardless. It has been those towns that have had significant success!
- Germany has a 30% larger population than the UK. However, their testing has been 50% higher and started earlier. Meaning the number of UK deaths is 2.7x higher than Germany, with the USA 4.1x higher.
- Countries which have flattened their curves started testing early. Greece and Denmark in particular have kept their death and infection rates relatively low.
- The UK, which started late, has a Case Fatality Rate of 3.6%, the average worldwide is 4.2%, but for developed countries, the average is 4.02% with the high detection rates bringing death-tolls down to under 1%. Testing and early detection has been critical in their fight.
This tells us that countries that have left it late, let the virus go undetected and the death rate is directly related to that action.
Epilogue
Scary stuff right? People look to national governments to keep them safe. UK Gov have received criticism from many angles for their handling of the outbreak. Given the performance of other nations around the world, and the time that the UK has had to get on board, the criticism is fair. The action should have come earlier and mass testing must have been a key part of that plan.
Here’s hoping they get with the programme and ensure they get the experts on it.
