Schedule - Parallel Session 7 - Health-Related Decision Making

IMC Boardroom 2 - 09:00 - 10:30

Can Context Explain Violations of Procedural Invariance in Health State Valuation?

Danae Arroyos-Calvera; Rebecca McDonald; Graham Loomes; Andrea Isoni; Judith Covey; Jose Luis Pinto-Prades; Mike Jones-Lee


Some governmental bodies (such as the UK’s Department of Transport) use people’s stated preferences about their willingness to trade off their wealth for physical risk reduction to inform policies that have an impact upon society’s investment in safety. Several methods are used to elicit people’s values. According to standard economic theory, these values should be invariant to the method used to infer them. However, substantial failures of procedural invariance have been observed. Jones-Lee et al (1995), for example, obtained values for preventing injuries through willingness to pay questions (WTP) that were 3 to 10 times higher than those elicited through standard gambles (SG). One possibility is that such disparities arise because health states, as described in typical valuation studies, are unfamiliar and difficult to imagine, they may be highly dreaded, and are not an item people have experience in trading off for money. Such factors may be contributing to respondents’ failure to respond consistently. In addition, WTP and SG health valuation studies cannot be incentivized and typically confront participants with very small probabilities, which may be difficult for them to handle. Another possibility is that WTP and SG questions each prompt rather different mental processes and are liable to produce disparities irrespective of the items to which they are applied. To investigate this possibility, we apply WTP and SG methods to consumer goods (e.g. toasters) which, by contrast with health states, do not produce strong emotional responses, people are familiar with their characteristics and often own these items or have used, bought or sold them. Also, we use larger probabilities (the lowest is 5/100 in our study vs 1/1,000 in physical risk studies), and fully incentivize responses. In our study, participants see 10 objects. First, they rank them in order of preference, then value them by providing money equivalents. Finally, they undertake a series of standard gambles in which half of the gambles involve only the objects and the other half substitute in their money equivalents. If we find that disparities between WTP and SG persist, we hope to be able to identify the circumstances in which they are stronger or weaker, in order to better understand the nature of the procedural effects involved. If such disparities are largely eliminated when familiar goods and larger probabilities are used, this may suggest that anomalies in health and safety studies are primarily due to the use of unfamiliar outcomes and/or small probabilities. The experiments are now underway. By the time of the conference, we will have run the study and have an initial analysis of the data.

Danae Arroyos-Calvera

PhD Student, Warwick

Emotion-Based Health Beliefs and Behaviors

Elyès Jouini


There are in health beliefs and in health behaviors a certain number of well-known anomalies. (1) Subjective perceived health risk differs from objective health risk. There are systematic biases in the estimation of one’s health risk that vary among individuals and among illnesses. (2) There is also heterogeneity in the type of bias. Some individuals underestimate their health risk, but some others overestimate their risk. In the extreme, some individuals are even hypochondriac/ostriches. (3) Testing rates are too low even when there is no cost of testing and when the medical benefits are known and obvious. (4) Testing and preventive health care behavior do not necessarily increase with objective risk. Those who are rationally the most in need of medical services are not those who use them the most. (5) Subjective risk increases with objective risk, testing increases with subjective risk but testing does not necessarily increase with objective risk. (6) Some people overuse health services while others underuse health services (‘doctors avoiders’). (7) There are divergent anxiety responses to information: information reduces the anxiety of some individuals (the ‘want to knowers’) but raises the anxiety of others (the ‘avoiders’). (8) Confirmatory testing: some individuals who are certain to be ill test in order to prove what they already know. These health behaviors seem irrational at the individual level, but they also have severe consequences for the public health system: patients who should seek medical care stay inactive while the system is utilized heavily by those who are not in need of medical care. We propose a model in which health risk, or more precisely the risk of bad news (illness), generates emotions ex-ante, in the form of anxiety or worry but also ex-post in the form of possible disappointment. We consider these emotions in addition to the standard health consumption utility. Health beliefs and behaviors result from the optimal management of these emotions, i.e. realize the best trade-off between anxiety and anticipated disappointment. The perceived risk level can be seen as the choice of the optimal self-insurance level against emotional risk: it represents how much the individual is willing to sacrifice in terms of peace of mind at date 0 to reduce his future vulnerability to disappointment of bad news. Optimists choose less anxiety but more vulnerability to future bad news, whereas pessimists choose more anxiety but more psychological preparedness to the future bad news. Concerning preventive health behavior, the same trade-off is involved. Testing permits to avoid future bad news and associated disappointment but living with the bad news can be costly in terms of anxiety; not testing permits to keep one’s illusions but exposes more to future bad news. We show that our assumptions (1) are all supported by considerable evidence and (2) permit to explain all the mentioned anomalies.

Elyès Jouini

Professor, Université Paris-Dauphine

Measuring Uncertainty Preferences for Health

Olivier L’Haridon; Arthur Attema; Han Bleichrodt


It is well-known that expected utility has empirical deficiencies. Prospect theory has developed as an alternative with more descriptive validity. This study elicits utility of life duration in a framework that is robust to violations of expected utility. In addition, our method allows for a parameter-free measurement of loss aversion and probability weighting for both gains and losses. Finally, we are the first measure utility of life duration in the case of uncertainty. We conducted individual experiments where respondents had to make choices using health outcomes, which included a treatment with known probabilities and treatment with unknown probabilities. First, when comparing these treatments, we found evidence for uncertainty aversion. Despite this, the two treatments showed the same general pattern: concave utility for gains, convex utility for losses, and steeper utility for losses than for gains. Hence, our findings confirm the S-shaped utility that has often been observed for monetary outcomes. The amount of loss aversion was not significantly different between the treatments, with median loss aversion coefficients varying between 1.3 and 1.8 (depending on the specific definition used). The probability weighting functions showed the usual inverse S- shape, indicating overweighting of small probabilities and underweighting of large probabilities, both for gains and for losses. In particular, our results suggest duality of the probability weighting functions. In conclusion, our data are supportive of prospect theory and source dependence in the health domain. However, our finding of convex utility for lost life years contradicts results of previous studies that used a different elicitation method, and suggests this behaviour may be method-specific. More research is needed to settle this issue.

Olivier L'Haridon

Professor, University Rennes 1

The EQ-5D in Health Judgements and Choices

Rebecca McDonald; Timothy Mullett


The Quality Adjusted Life Year (QALY) is used in health economics and policy to enable comparisons between different health states. It has two components: duration and quality of life. To estimate the latter, the EuroQol EQ-5D descriptive system presents health states as a combination of scores along 5 dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Each dimension is assigned a score from 1 (no problems) to 5 (extreme problems). These can be converted into a single, overall quality weight using subjective judgements elicited through experiments and surveys. If preferences for different health states are inadequately captured in this way, then policy decisions made on their basis are likely to be suboptimal. The EQ-5D dimensions are heavily weighted towards physical as opposed to mental health as (with a ratio of 4:1). If survey respondents perceive the 4:1 split as a signal about the appropriate weight to be placed on physical and psychological health, this might influence their survey responses. This is a particular concern where there are doubts surrounding the stability of underlying preferences for health states, and hence for the consistency of health judgments elicited in different ways, or on the basis of different formats of information presentation. We present a series of three experiments that test the extent to which the EQ-5D presentation influences the decisions of healthy individuals making choices or judgments about fifty health states. In our first experiment, we demonstrate that naming the diseases significantly alters their judged severity, an effect explained by how much the health state is feared, or “dreaded”. This suggests the EQ-5D does not capture all health state aspects that concern individuals. Our second experiment elicits choices, instead of judgments. Again, we find a significant effect of the label. However, we find remarkable consistency between choice and valuation. Our first two experiments are supplemented by Eye Tracking technology to provide process-level data on attention to different information. In our third experiment we directly address the ratio of mental and physical health dimensions by altering the dimensions of the EQ-5D. We split the mental health dimension into two and combine mobility and usual-activities into one. This significantly alters the relationship between the dimensions and the judged health state severities and indicates that information presentation influences survey responses. Our research has implications for the robustness of EQ-5D-derived quality of life scores, raising questions for practitioners and academics about what factors should be allowed to influence the prioritisation of different health states in policy decisions.

Rebecca McDonald

Post Doc, University of Warwick