Risk in Health: More Information and More Uncertainty

Risk is a major factor in health, with a strong focus on minimising risk wherever possible. The mathematical starting point is probability. Reliable or relevant data is often missing or hard to get. Moreover, the results of studies are all too easily interpreted wrongly – even by medical experts. Usually it is seen as useful to have more information in making decisions. As we show below this is not always true; we will use the exemplar of breast cancer and screening as an illustration throughout this paper to explain circumstances where there is ‘more information and more uncertainty’ following Knight and type 2 errors. We identify the different stakeholders and parts of their internal criteria that form their ‘rationality’, which may well be idiosyncratic. The intention is to pave the way for a ‘shared’ decision which is best for individuals and for society simultaneously. Statistics educators will find an important field of research and teaching. INTRODUCTION The focus of this paper is on health and the fact that, not least with the increasing access to information via the Internet, there are advantages as well as dangers. We begin with some general comments before assessing the quality of information available. We analyse the different types of risk and we note the different interest groups involved and that the view of risk depends on one‟s role in the situation. The key example we will use in this paper is mammography screening for breast cancer. Many (but not all) patients want to know about factors affecting their health. In the debate on informed consent, there is a growing tendency by doctors to explain all potential side effects of a suggested treatment. Shared decisions and responsibilities should assist those involved to be effective; it does not always work that way. Indeed it may lead to a decrease in accountability of professionals. Well-informed patients can certainly improve the communication between medical doctors and themselves. For example, a well-informed patient can disclose and discuss wider issues relating to the health problem being discussed. But poor communication may sometimes increase the risk of incorrect treatment. More information may also reduce the mutual trust between doctor and patient: as Voltaire said, “medicine is what we do while nature takes its course” (the placebo effect), so trust is crucial. Hence there can be dangers in being too open, especially in discussing (Knightian) uncertainty, based on lack of knowledge rather than risk based on sound information. It is not always easy to separate the two aspects. Sound information in health is usually based on statistical methods and the decisions rely on probability models. The methods involved are complex and mathematical concepts are rarely well understood, especially when conditional probabilities are involved, as shown by recent research by Huerta (2009). He has classified the complexity of such problems and shown that students at school and university are not able to apply ideas learnt in a new context. Gigerenzer (2002) has developed methods, which are directly comprehensible. One example for his approach is his focus on natural frequencies, to replace the corresponding conditional probabilities. However, Gigerenzer‟s approach is bound to the standard model, which presupposes probabilities of the diagnosing procedure to err in both cases of a woman with and without cancer of the breast and a probability (prevalence) of this cancer among women of such an age, despite strong arguments against such an averaging view (is the subgroup to which a patient is attributed, really relevant?). It also (wrongly) assumes that the mathematical result is equally relevant for all stakeholders, including the individual woman tested, the doctor who investigates her, the hospital that has installed the equipment, and the health institutions that have promoted the screening scheme. Our analysis below shows that the stakeholders differ in their role so much that no common rationality is left as a unifying element. Furthermore, in their decisions people use ideas beyond what is usually perceived as rational. Notwithstanding that, pseudo-statistical figures or unreliable statistics are used in the discourse. Thus, to support people to understand the statistics is only one step in the right direction. A holistic view of the situation is required to improve the management of risk. In what follows a systems analysis approach is chosen to extend the perspective of a decision that finally has to be shared between unequal stakeholders. THE QUALITY OF INFORMATION People have developed a deeply causal image of life (see Kapadia, 2010). If anything is wrong we simply have to find a suitable remedy. The health sector, too, feels a strong pressure, as do doctors to find immediate cures. In health, emotions and desires also play an important role. There can be various differing scenarios of which two are noted below. 1. There are severe symptoms requiring action. The need is to heal directly such as with a broken leg, or relief of severe pain. The expectation is that the treatment helps immediately. Effectiveness of the treatment is relatively easy to judge using related measurements linked to the symptoms in the short or long term. 2. There are indications that dangers await in the future. Actions are taken which either safeguard the patient or can be fear-mongering. An evaluation of effectiveness is hypothetical as the possible disease (such as cancer) may be benign. What information is relevant for judging the situation? To get valid information has always been difficult; the change with the Internet is an explosion in the availability of information but no new ways of judging validity (Miller, 2004). More important, good information is often hard to understand and may involve mathematical concepts and statements on basis of models or scenarios, which are bound to (hidden) restrictions. The validity of information can be judged by its source and the role of the stakeholder in the „game‟, or the specific health situation. In any case, information has to be interpreted and judged for quality. Where does the data come from? Does the information depend on restricted models? Is the information merely opinions or anecdotal? Ways to judge information are multi-faceted and subjective (even inside science) and may depend on minor issues like whether an opinion leader, a prominent actor, an attractive woman, or an „expert‟ transmitted the information. With regard to the role of the stakeholder there are huge differences in their criteria:  Doctors: A doctor has to follow the state of the art and avoid risk. If (s)he decides for a scheme deviating from treatment lege artis, (s)he is liable for adverse consequences.  Experts: A scientist seeking funds has to make the case important, perhaps exaggerating potential negative consequences in the future. Such a stakeholder bears no risks.  The pharmacology industry: Such enterprises have an economic interest in promoting specific treatments or schemes of diagnosis. Within certain limits, they have a liability and may be sued for compensation (Khamsi, 2005), but they also need to make a profit.  The media and politicians: Media are interested in sensations and hot air. Their speculations on future (adverse) developments bear no risk or liability. Politicians have to be seen to be good at making decisions, and bear the risk of their personal career, which might end immediately if their decisions judged (even incorrectly) by the media to be wrong.  The patient: (S)he wants to improve his (her) health and has to bear short and long-term risks and consequences and may not base decisions only on rationality. An example of a conflicting view is measles and how the risk of infection is reduced by vaccination. There is a clash between an individual and a societal perspective. Information from studies might induce an individual to avoid the jab (because of its potential side effects); on the other side, for society – once the information from the studies is judged to be (at least slightly) in favour of the vaccination – there is a strong interest that all individuals take the vaccination to prevent an epidemic. Moreover, an individual‟s decision against vaccination is only successful if enough others have already taken the vaccination („free-rider effect‟, see May & Silverman, 2005). Thus different criteria form the basis of rationality, for the research community in medicine, for the health sector (hospitals and doctors), for the pharmaceutical industry, and, by no means least, for the patients; all have differing expectations, risks and responsibilities.