How should scientists and journalists report risk? The way in which a statistic is presented can entirely change how alarming it sounds. And too often, both newspapers and scientific journals choose the most alarming, but least informative, way.
For instance, according to a study published in the 听this month, if you father children in your fifties, your children are more likely to suffer various health issues, including seizures. Specifically, if you are aged 45 to 54 when you become a father, your children are 18听per cent more likely to suffer seizures than if you are 25-34.
Advertisement
Since many people these days听have children in midlife, that is an alarming statistic. But it is misleadingly so. It is presented as what is known as 鈥渞elative risk鈥�: how likely one group is to have seizures compared to the other. But it doesn鈥檛 tell you how likely your child actually is to have seizures.
That figure is known as 鈥渁bsolute risk鈥�, and it is both more revealing and more reassuring. Your child鈥檚 absolute risk of suffering seizures if you have a child when you are 30 is 0.024听per cent: that is, 24 out of every 100,000. If you have a child at 50, it is 0.028听per cent. , a statistician at the University of Oxford, says: 鈥淎n 18听per cent increase sounds shocking, but in reality, it will affect four people in every 100,000.鈥�
It鈥檚 all relative
The difference is vital. But too often, when scientific stories are reported in the news, we read about relative, rather than absolute, risk. New 杏吧原创听has not been immune to the tendency, and the 鈥渙lder fathers鈥� story is not the only one in recent weeks: the BBC Radio 4 Today programme discussed a conference paper finding a 鈥�100听per cent increase鈥� in breast cancer risk among women who prefer to sleep and stay up late, compared to those who get up and go to bed early. But, as the much better on the BBC website made clear, the absolute risk in a given eight-year window was 2 in 100 for 鈥渘ight owls鈥�, compared to 1 in 100 for 鈥渓arks鈥�.
鈥淩elative risk is fine for scientific inference,鈥� says , professor of the public understanding of risk at the University of Cambridge. So if you鈥檙e interested, from a purely scientific point of view, in whether two things are associated 鈥� whether your risk of cancer is linked to how much meat you eat 鈥� then it鈥檚 helpful. But if you want to help people make decisions about their life, 鈥渋t鈥檚 useless,鈥� he says. 鈥淚t鈥檚 totally the wrong measure. You cannot decide what鈥檚 an appropriate action without absolute risk.鈥�
This is not just an issue for the media: journals and scientists often fail to report absolute risk prominently. The BMJ study above reported all its findings in relative risk, hiding the absolute risk away in a table, despite their saying 鈥淲henever possible, state absolute rather than relative risks鈥�. And in August, the Lancet published a on the health impacts of drinking which concluded that there was 鈥渘o safe level鈥�. But it, too, gave relative risks, also against its own .
Insist on absolute
The Lancet press office, admirably, managed to get the absolute risks from the authors for its . But if scientists don鈥檛 make it easy to find, then journalists, who are often not especially numerate, 鈥渃an only report the relative risk鈥�, says Rogers.
If the study or story has any recommendations about individual action or policy, then scientists and journalists should both insist on reporting it in absolute risk, says Spiegelhalter. 鈥淚t鈥檚 so irritating that they don鈥檛. We know from both experience and research that absolute risk, and especially the expected number of people 鈥� what does it mean per 1,000 people or per 100 鈥� is such a clear way of communicating.鈥�
He added: 鈥淥ne must suspect that the reason it鈥檚 sometimes not done is because when it鈥檚 not put in those terms, the effects don鈥檛 seem very important.鈥�
听
