Reckoning with Risk: Learning to Live with Uncertainty

Reckoning with Risk: Learning to Live with Uncertainty by Gerd Gigerenzer. Allen Lane, £14.99, pp 310. ISBN 0 713 99512 2. Rating: ★★★

In healthcare, information often comes in the form of numbers. Professionals are expected to make good quality judgments based on this evidence. Patients are often assumed to be less numerate than professionals and to have less ability to interpret healthcare evidence. As a result, many healthcare workers feel that one of their roles is to help people to understand the choices open to them by acting as interpreters of the evidence, particularly amid uncertainty.

There is some evidence to support the idea that the public has problems understanding risk. Research dating back to the 1960s has found that professionals' ideas of risk are often different from those of lay people. The public tends to overestimate the frequency of uncommon health risks and to underestimate the frequency of common risks. Many lay people have trouble recognising estimates of statistical uncertainty, and so it seems reasonable to conclude that healthcare staff should be able to offer support to people in understanding the consequences of treatment choices.

Gigerenzer, a scientist at the Max Planck Institute in Berlin, argues that while healthcare staff like to see this as part of their role, most of us struggle to understand numeric measures of risk and uncertainty. Researchers presented UK service commissioners with details of four cardiac surgery services and four breast cancer screening programmes. The effectiveness of each service and each screening programme was shown in a different form: as a relative risk reduction, an absolute risk reduction, the number needed to treat, or the number surviving treatment. Commissioners preferred the service and the screening programme in which relative risk was reported, as the advantages appeared largest. Only three of 140 commissioners realised that all of the cardiac surgery services and all of the breast screening programmes showed exactly the same results, presented in different ways.

Lest clinicians smile at the innumeracy of service commissioners, Gigerenzer reports numerous studies showing similar problems in clinical practice. In studies in the United States and Germany, experienced doctors struggled to work out the likelihood that a woman in her 40s who had a positive mammogram on routine screening but who had no other risk factors would prove to have breast cancer after further investigation. Supplied with the relevant statistical information, only two out of 48 German doctors and five out of 100 US doctors were able to work out the correct answer, so limiting their ability to provide advice to the woman.

Problems with this type of calculation are not confined to doctors. One of Gigerenzer's students embarked on a tour of 20 professional HIV counsellors, who seem to have done their best to answer his questions on personal risk, and on his likelihood of being infected with HIV, should he test positive. Many were very knowledgeable about HIV and AIDS, and well able to provide good clinical information. They struggled badly, however, when trying to answer his questions on risk. Most denied the existence of false positive tests, and only one came close to being able to tell him the likelihood of his being truly infected were he to test positive.

Gigerenzer suggests that special interest groups take advantage of our corporate numerical blindness by presenting results in certain ways. Pharmaceutical company advertisements and press releases from researchers frequently offer relative risk rather than absolute risk reduction. He quotes a leaflet prepared by doctors who favour hormone replacement therapy, which gave the advantages in the form of relative risks but the adverse effects as absolute risks, so guiding patients towards the decision the doctors thought best.

This makes gloomy reading for everyone who thinks professionals should understand the treatments they offer, and that patients should have the opportunity to give informed consent. Most of the staff quoted in the book wanted to understand numerical measures of risk, and they reported feelings of inadequacy at the difficulties they had in interpreting information for patients. Gigerenzer's answer is to teach professionals to think in terms of frequencies, rather than relative probabilities, and he offers reasonable evidence that professionals can learn these skills and apply them to new problems in their own clinical practice. Perhaps we should make a collective acknowledgement of the legacy of numerical anxiety that was often the residue of traditional undergraduate training, and learn what to do about it.