Posted 28 April 2020
In today’s dose of ‘Things COVID’, I am trying something a little different.
The challenge of daily writing is a huge learning curve for me and as part of that I am keen to explore different types of post. Today’s is a lesson in statistics but please, give it a go as I have tried to make it as ‘untechy’ as I can and yet still convey the key point of relevance to so many of the daily COVID headlines.
All too often, the wrong interpretation of data can lead to very wrong courses of action. And all too often, those deciding on action are falling into essentially the same data traps time after time.
What I want to try today is to guide you through what I believe is one of the most common traps. I believe it is a worthwhile journey because if you can understand the data trap here, I think you will become much more adept at spotting the same trap elsewhere.
It is very different article to yesterday’s, but if you can follow my analogical reasoning involving shoe sizes and reading ages, you will get valuable insights in the world of all data, not just COVID data.
Let me start with the headline:
“CHILDREN FALLING ILL WITH INFLAMMATION SYNDROME LINKED TO CORONAVIRUS”
A very troubling headline from yesterday. Reported cases of COVID-19 in children have so far been extremely rare – either children do not become infected or if they do, they do not show significant symptoms. A few small scale studies suggest that they are not particularly key players in the transmission of the virus either.
So yesterday’s headline suggesting that perhaps this has changed and now children are getting infected could represent a significant evolution in the course of COVID19.
I want to offer a bit of balance on this and offer an alternative hypothesis which, though still worrying is less to do with COVID19 and more to do with lockdown.
If children are showing ‘inflammation syndrome’ and if inflammation is a reaction of their developing immune systems, we could indeed be seeing the results of coronavirus attacking their immune systems. But alternatively, might we be instead seeing this inflammatory syndrome because of changes to the children’s immune systems rather than COVID infection?
Let me try and explain this by looking at shoe size!
Imagine we went out and measured the shoe size of a large group of children and also noted the reading ability of those same children. We could then plot those data on a graph.
What you might see is something like Graph 1 attached to this post. As expected, as a child’s reading ability goes up, so they have bigger feet. We know this makes sense because an older child will have bigger feet and an older child will also, usually, have more developed reading skills.
The problem is how someone might interpret the graph. If they ignored ‘Age’ as the hidden factor here, they might erroneously suggest:
“If you want your child to have bigger feet, work on their reading ability”
You can clearly see this would be a nonsense. By ignoring the factor ‘age’ and with it the appreciation that age drives both reading ability and shoe size, we are only left with the correlation between reading ability and shoe size. And that wrongly leads to an assumed causation between the two.
Bear with me, I will get back to COVID19.
I can represent the relationships between age, shoe size and ability to read in a diagram (Graph 2). In this picture I am showing my three quantities of interest: Age, Shoe Size and Reading Ability. The arrow between Age and Shoe Size indicates the trend that as Age increases so does Shoe Size. The arrow between Age and Reading Ability represents a similar trend between those factors. The dotted line between Shoe Size and Reading Ability shows that there will indeed be a relationship between these two things but it is a consequence of Age and not because Shoe Size drives Reading Ability (or the other way round).
Graph 3 shows the perspective of someone who suggested that to get bigger feet, just work on reading age – they have ignored the now hidden variable of Age, only leaving that one dotted arrow between Shoe Size and Reading Ability. Missing Age from the diagram leads to the spurious conclusion.
Back to Inflammatory Syndrome in Children: Look at Graph 4. A bit more complex because I have four factors rather than three, but the story is very similar to the shoe size story above.
The dotted arrow (Arrow 4) between COVID19 and Inflammatory Syndrome (in children) is the story the newspaper headline is pointing us to: Is COVID19 causing this illness in children?. The other arrows represent my alternative hypothesis:
– COVID19 caused lockdown. (Arrow 1)
– Lockdown causes a reduction in children’s immune systems (Arrow 2)
– Diminished immune systems in children make them more prone to acute infection (Arrow 3)
If arrows 1 to 3 are correct, we have created an apparent link between COVID19 and Inflammatory Syndrome (Arrow 4)
Arrow 2 is the key point in my pathway. Children’s immune systems are still developing. Key in that development process is continued exposure to pathogens – playing in the dirt, mixing with other children in the classroom, engaging in playground sports and so on. That continuous low level exposure boosts the adaptive immune system of the children. Some papers suggest it also plays a role in developing their innate immune systems too.
If during development of those childhood immune systems, exposure to pathogens significantly reduces, atrophy of the immune system can be quick and significant and more so than in a fully developed adult immune system.
So there is my alternative hypothesis. Lockdown is causing the vulnerably of the children to infections and the presentation of inflammatory syndrome increases. Please note this is an hypothesis but I believe not an unrealistic one.
Graph 5 is the scheme of things with the Lockdown causation route faded-out (a bit like our shoe size model without age). The newspaper headline is looking at the world through Graph 5 rather than Graph 4.
And here is the danger.
If we use Graph 5 to manage the Inflammatory Syndrome in children we may take the decision to embed ourselves deeper in lockdown (more lockdown, less spread of coronavirus). If we assume Arrow 4 is the only connection, we would expect to see fewer cases of the inflammatory syndrome in children.
However, this would be a disaster if my alternative model (Graph 4) is true. Increasing lockdown (Arrow 2), increases depletion of children’s immune systems (Arrow 3) leading to more inflammatory syndrome (Arrow 4).
This would be far more serious that just wrongly believing that getting your children to read better will give them bigger feet!
If you got this far, I hope I have not bogged you down with Graphs and Arrows but hopefully I have shown how important it is to keep rational in an ever changing and irrational world.