Why are the papers by Spiegelhalter and Walker et al. from the May 2014 issue of BJOG such easy reads? That is not what we usually expect from scientific papers. Is it the use of natural numbers? Presenting perinatal mortality as x out of 1000 is more understandable than ratios and percentages (Hoffrage et al. Science 2000;290:2261–2). Or is it the use of the micromort, a friendly unit of the probability of death which allows comparisons of acute risks? Probably for both these reasons: the cumulative effect of clearly presented statistics is to help us read on to the end.
At Sense About Science we work with researchers and the public to improve the communication of evidence, particularly on difficult issues. People come to us with their questions. Pregnant women coming across a headline claiming hair dye causes stillbirth might ask, ‘Should I be worried, or is this just another scare story?’ At the heart of their questioning, people are trying to get to grips with the benefits and harms. Women are often not given useful analogies (something Spiegelhalter recommends) and risks are seldom presented using natural numbers. From speaking to clinicians, we hear that many have never been given tips on how to communicate risks, yet it is their role to share the harms and benefits for different options. Here are a few pointers to help.
When it comes to health, it can be better to use the word ‘chance’ instead of ‘risk’ because ‘chance’ is not associated with any idea that the outcome is good or bad in people's minds. Also, think about how you frame the numbers; give both the bad and the good outcome, so as well as giving the chance of 1 in 1000 of the baby dying, you can also say there is a 999 in 1000 chance of the baby not dying.
Spiegelhalter reminds us that this study looked at population level, so looked at the average. This may tell you little or nothing about the individual risk—the risk faced by the woman you are talking to might be very different.
Medical research is often misrepresented when the context of the research is missing. To know if a change in risk actually matters, people need to know what the risk was to begin with (absolute risk). The difference between stating relative or stating absolute risk reductions can influence both doctors’ and patients’ perceptions of a risk (Skolbekken BMJ 1998;316:1956–8; Malenka J Gen Intern Med 1993;8:543–8). Always give the absolute risk alongside the relative risk.
We are starting to see journal editors and journal press officers treating omission of absolute risk as a cardinal sin. Including an orphaned relative risk in the headline (‘Stress doubles risk of stillbirth’) is precisely what leads to news stories becoming sensationalised. When traced back this might not originate with the journalist but in the press release or, digging deeper, sometimes even the original research paper failed to include the absolute risks. The Association of British Pharmaceutical Industry code of practice (2011) bans the use of relative risk without also giving the absolute risk; is it time for the International Committee for Medical Journal Editors to include this in its recommendations too?
As clinicians, whether you are talking to patients about treatment options, writing for an academic journal, or talking to journalists, communicating the statistics clearly means that everyone is better informed about the decisions they make.
Disclosure of interests
The author was the Managing Editor of BJOG between 2006 and 2011.