Dr Geoff Watts FMedSci, a freelance medical journalist, explores some of the issues around how best to communicate risk and statistics when writing about research.
Fresh from the Academy’s workshop on “Communicating evidence about medicines” I’m asked to offer a few thoughts on communication for the benefit of doctors and scientists in the early stages of a research career. So let me flag up one issue that should have been dealt its death blow decades ago but, on the evidence of yesterday, lingers on. It’s the potentially misleading use of relative, rather absolute, risk data.
Back in the Stone Age, during my own brief research career and before I crossed the fence and joined the media, I endured three badly taught courses on statistics. Each compounded the confusion left by its predecessor; but one thing I did grasp was the extent to which the perception of findings depends on the manner of their presentation. In fact I recall using this as a classic example of failed communication when I gave an inaugural talk following election to the Academy in 2003.
The example I used - still one of the most egregious - dates from well before that. It was the 1995 Pill scare when a large number of women stopped taking their contraceptives for wholly inadequate reasons. The then Committee on the Safety of Medicines had issued a statement reporting that certain brands were twice as likely to cause blood clots as others. Many women instantly stopped taking their pills - with predictable consequences for the birth rate nine month later.
The point that signally failed to be made clear was the low magnitude of the absolute risk: 1.5 cases/10,000 woman years (I think), and fewer than 2 deaths/100,000 woman years. Twice a very small risk is still a small risk. Twenty years later, and in spite of repeated suggestions that absolute as well as relative figures should always be given, the issue still creates confusion.
In fact I’d go further than that; I’d be inclined to offer only absolute figures. This will come as a surprise to you, but given two bits of information describing the same phenomenon, one of which appears more startling than the other, journalists show a statistically significant bias in their choice of which to use. Funny, that...
More insights, no doubt, in the Academy’s forthcoming #MedSciLife Twitter chat on Tuesday 14th June 12:30. Check it out.
- Dr Geoff Watts FMedSci