How the Oxford University algorithm decides who should be shielding

Around 1.7 million more people will be added to the shielding list in England, after being found to be at potential serious risk from COVID-19.

They have been identified by a new algorithm that looks at multiple factors, and will be sent a letter from the NHS in the coming days.

Very simply, this algorithm works by using the details of your medical records to assess how likely you are to catch COVID-19 and die.

It sifts through this information, then gives you a score out of 100: a high score means you’re more likely to get COVID-19 and die, a low score means you’re less likely.

The people with the highest scores are added to the shielding list and prioritised for vaccination.

Whether you get one of the higher scores depends on your personal characteristics such as your age, gender or ethnicity, as well as your weight compared to your height.

Older people are known to be more vulnerable to COVID-19. Men are more vulnerable, so they will score higher. The same applies to ethnicity.

Other factors which will increase your score include whether you smoke, whether you are homeless, or, crucially, whether you are taking certain kinds of medication.

Some cancer patients, for instance, weren’t included on the first shielding list. This algorithm can run through the NHS database, pick out those people, then flag them up for vaccination.

Poverty is known to be linked to more severe outcomes from COVID-19 and this algorithm includes a measure of deprivation, based on your postcode.

It’s an algorithm too, so it works in the same way.

It takes a whole series of measures, including unemployment, level of car and home ownership, and level of household overcrowding, then uses that to calculate poverty in your area.

It gives you a score for this, which feeds into your final score when it is calculated at the end.

You might be wondering exactly how these scores are calculated. That is a much bigger question: suffice to say it involves some extremely difficult maths.

Fortunately, all the details of the model will be published so people who understand the maths can go through it. If that’s not you, you can get a sense of it yourself by using this online calculator which uses the algorithm to estimate your score based on the information you give it.

There are of course concerns, but before we get into those, it’s worth pausing just to reflect on how mind-bogglingly complicated this is.

The algorithm isn’t just trying to predict who will die from COVID-19 if they catch it, something we now unfortunately know all too much about. It is trying to predict who will catch COVID-19 in the first place.

Unless the University of Oxford academics who developed it invented time travel at the same time, the results will inevitably be imprecise, particularly because it does not include some factors which clearly put you at greater risk of catching COVID-19, such as occupation.

Subscribe to the Daily podcast on Apple Podcasts, Google Podcasts, Spotify, Spreaker

As a result, it is quite likely that some people who are actually low risk will end up jumping the queue.

After the chaos caused by the attempt to mark last year’s A-levels with an algorithm, the idea of an imprecise algorithm picking people to be vaccinated might send shivers down your spine.

But if we just focus on its downsides we might miss its upsides: its ability to run through more data than a human ever could.

Algorithms aren’t inherently bad. They’re just tools – and, like any tools, they can be used poorly or well. This one could help the NHS pick vulnerable individual out of the population, but it is as only as good as the data that goes into it.

Perhaps the biggest challenge will not be running it, but explaining it to patients who get (or don’t get) an invitation as a result.

Read Entire Article