The risk of accuracy

Watch those percentages… (image by NavBack)

I love reading about risk. As someone who finds maths a bit of a challenge, it always becomes more real to me when the numbers relate to the real world. Using numbers to explain the risk of something happening seems like an excellent way to harness the power of maths – numbers are objective, concrete, and not vague.

But as always in the messy real world, it’s not quite that simple. That’s why I was really inspired by this blog post from the Risk Science Center at the University of Michigan, explaining how extra levels of ‘accuracy’ – more numbers – can actually make risks harder to understand.

This might sound like a slightly niche topic – luckily most of us don’t have to cope with communicating complex health risks in our everyday lives. But we do have to deal with literally endless stories in the media about the relative ‘risks’ of everything from an ‘antibiotic apocalypse’ to going blind from swimming in your contact lenses.

These stories can be simply emotive and number-free – making them rather unhelpful for understanding what the risk actually is. The antibiotic apocalypse story is a good example – a genuine risk that is (perhaps justifiably) being hyped up. It shouldn’t be news to very many people that antibiotic resistance is a huge problem – we’re running out of effective antibiotic drugs because the bugs they’re designed to kill are evolving. But does using the word ‘apocalypse’ really help?

To me this seems like it would push readers to one of two extremes – either thinking ‘oh god not another disaster prediction’ and flicking to another story, or becoming terrified by the prospect of dying in a dirty hospital ward and so perhaps stockpiling antibiotics to fend off the impending apocalypse. Neither of these reactions is useful.

Then there’s the alarmingly accurate risk figure. On the face of it, this recent story about how a vegetarian diet can cut your risk of heart disease looked great (I’m a veggie so I was pleased). But as the Risk Sense blog post explained very clearly, being told that vegetarians are 32% less likely to die or need hospital treatment as a result of heart disease is actually meaningless in the context of an individual person like me.

I don’t know what my baseline risk of heart disease is. I’ve got low blood pressure, but I’ve never had my cholesterol levels checked. Although I eat a lot of fruit and veg, I don’t always get my five a day and as a vegetarian in need of protein, I eat a lot of cheese. So who knows what my heart disease risk profile looks like?

This is that old chestnut of absolute versus relative risk – sure, my risk of heart disease might be lower, but lower than what? If I don’t know what risk I started out with, I don’t know how a risk reduction of (perhaps) 32% affects me. I might still have a high risk overall.

As Brian Zikmund-Fisher writes on the Risk Sense blog, ‘we need to acknowledge that we often provide risk data to people who need risk understanding and personally relevant meaning without actually translating the former into the latter.’

To be fair, he’s talking about health professionals here – but you could argue that journalists writing about health risks are performing a public service for interested readers. The only reason this story about vegetarians is newsworthy is because heart disease affects a lot of people – people who want to understand their risk and how to manage it. People who should get helpful information, not almost meaningless numbers that are relative to other knowledge they don’t have.

The worst kind of risk story in the media is the completely overhyped one. This one about a woman who says she became blind in one eye as a result of going swimming in her contact lenses is awful. It’s sad, because she became blind, but it’s also ludicrously bad communication because it strongly implies that swimming in your contact lenses carries a significant risk of blindness.

It conveniently ignores the fact that this particular woman may not have got the infection through swimming in her lenses. It doesn’t give a direct quote from an expert or any figures for whether swimming in lenses is actually that dangerous. But because (a) wearing lenses and (b) swimming are both fairly common activities, the risk of doing both at the same time and going blind suddenly seems quite large.

So what’s the purpose of this particular piece of risk (in)communication? There’s no clear message, apart from ‘go to the doctor if your eyes feel funny’. I’ve previously been told my opticians not to shower or swim while wearing lenses, but none of them explained why or how risky this actually was… so I ignored them. Not a great outcome for that bit of risk communication then.

Brian again: ‘it is the responsibility of every risk communicator to have a specific purpose in mind at the time of a communication AND to select the risk format that is most congruent with the recipient’s informational needs.’

Here’s how not to do it. In Susan Coolidge’s book ‘What Katy Did’, Katy’s aunt tells her not to go on the swing in the woodshed. Katy asks why, and her aunt tells her to obey her elders. Katy ignores her aunt, the swing breaks, and Katy is seriously injured. This is not great risk communication – the aim was to protect Katy, so the message totally failed.

A better example: my family doctor when I was a kid was a lovely lady called Dr Jolly. Despite her name she was actually quite a no-nonsense woman. My sister was once prescribed some steroid cream, and told not to put it on her face. She asked why. Dr Jolly said threateningly, ‘because it thins the skin and you could end up with a sore that will never heal.’ What was the message here? Don’t use the cream on your face. What was the outcome? Mission accomplished. Dr Jolly might have slightly exaggerated the risk – but in a way that worked for her patient.

I like this idea that risk communication doesn’t have to be scrupulously accurate to be effective. In fact, it might be all the better for it – sometimes details just get in the way. The point is – what’s the purpose of the communication? What is the outcome meant to be? Risk communication should help the recipient, not cover the ass or demonstrate the intellect of the communicator.