Pitfalls in communicating risk
Dr Mike Campbell offers tips on conveying medical risks to the patient
If an average 40-year-old buys one Lottery ticket twice per week for a year their chances of dying each week before the draw are greater than their chances of winning the jackpot. This is not generally understood. Similarly, the recent scares over HRT and MMR reflect a long and troubled relationship between risk suppliers and risk consumers.
It is the small, involuntary risks that might be avoided that cause the most argument (see table on page 52). This is particularly true of the unknown risks, since it is impossible to prove zero risk. The public has much less tolerance of involuntary risks than ones perceived to be under individual control.
The Government requires much higher safety levels, and will spend much more money per death saved, on rail transport than on road transport. Parents will take the much greater (and known) risk from measles, mumps and rubella to avoid what they perceive to be the unknown risks of MMR.
Why accept any risk, however small, of a terrible affliction such as autism, when a zero risk option is available? The outcomes of measles or autism are not comparable.
Risks are also subject to fashion. The risk of death from meningitis has not changed recently, but it is now customary for newspapers to report deaths of students over the winter.
The fear of child abduction has instituted major social changes in the way children are brought to school and in their social activities, despite the lack of evidence that it is any more dangerous for children out of doors now than 50 years ago. Other fashions have come and gone, without noticeable change in the underlying risk, such as listeria and necrotising fasciitis.
GPs have to interpret information coming from government agencies and from drug companies and therefore need to know how risk is measured. Both the BMA and the Royal Statistical Society have devoted recent issues of their journals to risk communication1, 2.
Communicating with professionals
Three concepts are essential. These are:
lRelative Risk (RR)
lAbsolute Risk Reduction (ARR)
lNumber Needed to Treat (or Harm) (NNT or NNH).
These are most easily explained with reference to the box on the right.
Imagine that we have a cohort of people, some of whom over time are exposed to a hazard and some of whom are not. The risk of an event, such as death, is expressed as the number of events occurring in a group, divided by the length of time the cohort has been followed up.
The relative risk is the ratio of the risk in the exposed group to the risk in the unexposed group.
For example, the risk of a venous thrombotic event in third-generation oral contraceptives (low-dose) is 30 per 100,000 women years, and for second-generation oral contraceptives it is 15 per 100,000 women years. Thus the relative risk of the third-generation Pill compared with the second is 30/15=2.
The absolute risk reduction (increase) is the difference between the two risks and is 30-15=15 per 100,000 women years.
The number needed to harm (NNH) is the inverse of the ARR, so NNH=1/ ARR=6,700 women years. This means 6,700 women would have to each take the third-generation Pill for one year for one additional woman to suffer a DVT.
In an ideal world one would attach confidence intervals to these estimates to indicate how precisely they have been estimated since these risk might have been estimated from small numbers. A confidence interval is a range of values that indicate where the true value of a parameter is likely to lie.
In general, to convey information about risk adequately one needs to know both the absolute risk and the relative risk. For women aged over 50, for example, mammography reduces their risk of dying from breast cancer by 25 per cent.
This may sound impressive but given that the absolute risk is low, what it means is that over 10 years out of 1,000 unscreened women about four will die from breast cancer, whereas out of 1,000 women who have regular mammography three will die from breast cancer.
The consequence of the CSM warning on third-generation Pills in 1995 was an increase in the number of legal abortions, from 39,000 in the quarter before the warning to 45,000 in the succeeding quarter. Pregnancy carries a much higher risk of a thrombotic event (or DVT) than being on the low-dose third-generation Pill.
Communicating with the patient
Many authors suggest it is best to communicate risk by relating it to the risks of daily activities.
Examples of these are the Calman Chart3 and the Paling Perspective Scale4. Risks from many medical procedures, such as vaccination, are tiny and it may be informative for the patient to realise that often the greatest risk they face is the journey by car to and from the surgery5.
Pictorial representations such as bar charts are useful, where the bars are proportional to the probabilities. A chart such as the Paling Palette can help people understand small probabilities. It has 1,000 figures drawn on it, and the doctor or counsellor can fill in the relevant risks by highlighting a number of the figures while sitting in front of the patient4.
Avoid 'single event probabilities'. One might tell a patient on prescribing Prozac that they have a 30 to 50 per cent chance of developing a sexual problem. Some patients will assume that 30 to 50 per cent of their attempts at sex will go wrong6.
Further tips for improved risk communication are given in the box on the left.
No medical procedure is completely safe and patients need to be informed in a clear and non-alarmist way about the level of risk. There should be an ongoing debate about medical risks and work done to elaborate these for many medical procedures.
Risks associated with the COC3
Venous thrombotic episodes per 100,000 women per year:
No use 5
Pill (2nd generation) 15
(3rd generation) 30
lRelative risk of a VTE in
relative to 2nd =
30/15 = 2
lIncreased risk of a VTE in
relative to 2nd =
15 per 100,000 women
Top tips on communicating risk4
lAvoid using purely descriptive terms (such as 'high' or 'moderate'). These reflect the speaker's perspective and the patient often understands the risk to be a completely different order of magnitude.
lDo not use probabilities, but use whole numbers with a consistent denominator, so that an incidence of 0.04 becomes 'in any year, out of 1,000 patients 40 will develop the condition'. Avoid 'single event probabilities', but make the denominator clear.
lGive both positive and negative outcomes. Edwards relates a situation where a mother who had a baby with spina bifida asked about the chance of recurrence in a subsequent birth. The risk was stated to be one in 10. One might as legitimately state that out of 10 mothers in a similar condition, nine of them had had normal babies.
lUse absolute risks rather than relative risks.
lUse visual aids for probabilities.
1 BMJ 327, September 2003
2 Journal of the Royal Statistical Society Series A 166, part 3, 2003
3 Calman KC. Cancer: science and society and the communication of risk.
4 Paling J. Strategies to help patients understand risks.
5 Edwards A. Communicating risks through analogies.
6 Gigerenzer G, Edwards A. Simple tools for
understanding risks: from innumeracy to insight.
Mike Campbell is professor of medical statistics at the University of Sheffield