Thinking about risk. Can doctors and patients talk the same language?

Abstract

Misselbrook D and Armstrong D. Thinking about risk. Can doctors and patients talk the same language? Family Practice 2002; 19: 1-2.Risk models are a powerful tool for assessing the biomedical significance of health problems and medical interventions. We know that if John Everyman is a smoker aged 70 with a BP of 152/85 mmHg and a normal cholesterol, then he has a 25-30% risk of a cardiovascular event (CVD event) over the next 5 years. 1 Medical treatment will reduce that risk by 9% over 5 years to a range of 22-27%. We also know that if John Everyman never saw a doctor, but simply stopped smoking, his risk would fall to 15-20%, a much more impressive health gain.Balancing the gains in risk reduction from different treatments is an important part of clinical practice, and doctors have three statistical representations to help them in this task.(i) Absolute risk: there is a 25-30% probability of John Everyman having a CVD event in 5 years. (ii) Relative risk: John Everyman is 1.6 times as likely to have a CVD event in the next 5 years than his identical but non-smoking twin brother. (iii) Number needed to treat (NNT): we would have to treat 11 John Everymans for 5 years to prevent one CVD event.However, this choice of statistical support is rather illusory as all derive from the same underlying (usually trial) evidence. The question is less how much would John Everyman gain from treatment, but more how can this gain best be expressed? Chatellier recommended the use of NNT when considering the needs of the individual patient 2 but, as there are not 11 John Everymans in most practices, the 'personal probability of benefit' (PPB) may make the probability as meaningful as possible to the individual patient. (%PPB = 1/NNT × 100, i.e. John Everyman would have a 9% chance of benefiting from treatment over the next 5 years.) Different mathematical expressions of risk are difficult enough for the doctor, but are likely to be harder for patients. Misselbrook and Armstrong showed that patients make very different choices about treatment depending on which of the above risk statistics they used as the basis of their judgement. 3 Rather than empowering patients, such risk models can therefore make them yet more dependent on their doctors. Mathematical models are designed for the world of the doctor and do not fit easily with the world of the patient. So how can we proceed?First we must recognize that patients have their own risk models. These bear little relation to the mathematical risk models used by doctors. Davison and his colleagues found that the lay classification of risk was based on a polarity model rather than the gradation of a continuing spectrum. 4 People saw themselves as either high risk or low risk. This model identified 'likely candidates' for illness. Thus a beer-swilling heavy smoking overweight man would (rightly) be seen as at high risk of a heart attack. However, if he did not have a heart attack and his healthy living neighbour did, Davison found that a second element in the lay risk model came into play....

Print Media Coverage of Antibiotic Resistance

“…Words and phrases such as "hazardous," "high risk," and "increasing prevalence" are examples of qualitative statements. Qualitative statements hold limited information because they are vague and mean different things to different people (Bogardus, Holmboe, and Jekel 1999;Calman 1996aCalman , 1996bMisselbrook and Armstrong 2002); they therefore contain a low level of contextual precision. Simple numerical information (e.g., "Ten people contracted antibiotic-resistant pneumonia") contains more precision, but lacks the context of a population value (e.g., a denominator) with which to put the risk in perspective.…”

How should public health professionals engage with lay epidemiology?

“…Here, we should avoid being distracted by the way in which probability facts can be presented in several different ways. 23 All such presentations refer to the same fact; as such, this point does not support the concern. Nor should we be distracted by disputes in probability theory between, for example, Bayesians and Frequentists.…”

An Uncertain Future: Infertility and Chlamydial Infection

“…A discussion of percentage probabilities that relates to a group of people has little relevance for the individual who wants to know specifically about their own life chances. People are generally considered to apply a bimodal model of risk in which they see themselves as either high or low risk which bears little relevance to the mathematical models used by doctors (Misselbrook and Armstrong 2002). In communicating risk, the challenge is to provide understandable information that has personal relevance and reflects the probable health effect in a comprehensible form.…”