Is risk attitude outcome specific within the health domain?
Introduction
Individuals’ risk attitude is an important concept within the health domain. Firstly, because medical decisions are generally made under the conditions of uncertainty, the optimal treatment from a patient's perspective will depend amongst other things on their risk attitude (see for example Fraenkel et al., 2003). Secondly, there is some evidence that more risk averse individuals are less likely to engage in unhealthy behaviour such as smoking and drug use (see for example Barsky et al., 1997). Thirdly, the role that risk attitude should play in the measurement of either dimension of the quality adjusted life year (QALY) is debated. Whilst the standard QALY model assumes risk neutrality with respect to life-years, a power QALY model which risk adjusts life years has been proposed (Pliskin et al., 1980). Several alternative methods for eliciting individuals’ preferences for quality of life exist and these vary in terms of whether or not they incorporate individuals’ risk attitude. It is often argued that individuals’ risk attitude should be incorporated because most medical decisions are made under conditions of uncertainty and that therefore the standard gamble (SG) method is superior. However, in SG exercises it is usually risk of immediate death that is traded off. As Dolan (2000) notes “most uncertainty is not of the stark ‘life or death’ form” and therefore the argument in favour of SG may not hold if risk attitude varies across health outcomes.
Economists generally assume that individuals exhibit a single risk preference which governs risk taking behaviour in all contexts. This has been debated within the psychology literature. There is some evidence that risk attitude varies across different outcomes and domains (Weber et al., 2002, MacCrimmon and Wehrung, 1990). Whether risk attitude varies across outcomes within the health domain is unclear because no previous studies were expressly designed to test this. Previous studies have either investigated risk attitude for life years (McNeil et al., 1981, Miyamoto and Eraker, 1985, Stiggelbout et al., 1994, Verhoef et al., 1994, Martin et al., 2000, van Osch et al., 2004), or risk attitude with respect to quality of life (Eraker and Sox, 1981, Breyer and Fuchs, 1982).
The main aim of this study is to investigate whether risk attitude varies across outcomes within the health domain. A within-sample design is used to test for differences in individuals’ risk attitude for the two dimensions of the QALY, namely life years and quality of life. Risk attitude is elicited from a sample of university students using the certainty equivalent method with even chance gambles. There is substantial evidence that framing effects are present when using stated preference methods (for an overview see McFadden, 1999). This study examines two different framing effects, namely an order effect and a sequence effect. The first framing effect investigates whether changing the order of the questions has an impact on the estimates of risk attitude. The sequence effect is tested with respect to the quality of life gambles. The certainty equivalent for the quality of life gambles consists of time spent in full health and time spent in ill-health. Risk attitude may vary by the sequence of ill-health and full health for two reasons. Firstly, evidence suggests that individuals tend to prefer improving sequences (Loewenstein, 1987, Loewenstein and Prelec, 1993, Roelofsma and Keren, 1995). Secondly, the present value of a sequence will vary depending on whether the ill-health occurs before the full health.
Health has a time aspect inextricably bound to it, and because of this individuals’ time preferences play a role in the measurement of risk attitude for health outcomes (Gafni and Torrance, 1984). Little is known about the interrelationship between time preferences and risk attitude. In the monetary domain, Anderhub et al. (2001) found a correlation between risk attitude and time preferences, while Barsky et al. (1997) found no relationship. This study investigates the interaction between risk attitude and time preferences in the health domain. Individuals’ implied time preference rates are elicited under the condition of certainty using additional stated preference questions.
Section snippets
Risk questions
Appendix 1 shows examples of the different questions used. Risk attitude is measured using the certainty equivalent method. Let (b,x) and (b,z) denote any two magnitudes of outcome b. (b,y*) is the certainty equivalent of the lottery [(b,x),p;(b,z),1 − p], if and only if: (b,y*) ∼ [(b,x),p;(b,z),1 − p]. Certainty equivalents less than, equal to, or greater than the expected outcome of the gamble indicate risk seeking, risk neutrality or risk aversion, respectively. Moreover, the higher (lower) the
Results
Fourteen respondents expressed indifference both with respect to the risk neutral certainty equivalent as well as with respect to the risk seeking and risk averse certainty equivalents in a total of twenty-nine gambles. This pattern of response indicates that the individual is not engaging with the exercise and the responses are therefore removed. Due to a further fourteen missing observations and respondents preferring a certainty equivalent that was equal to the worst outcome in three
Discussion
The main aim of this paper was to test whether individuals’ risk attitude for life years differ from their risk attitude to quality of life. There were five gambles: gamble between winning £200 and winning £50; gamble between immediate death and 5 life years; gamble between 5 life years and 15 life years; gamble between 10 years in full health and 10 years in moderate ill-health (EQ-5D state 22222); and gamble between 10 years in full health and 10 years in severe ill-health (EQ-5D state
Acknowledgements
The Chief Scientist Office of the Scottish Government Health Directorates funds HERU. The views expressed in this paper are those of the authors and not necessarily those of the funders. We would like to thank two anonymous referees for their helpful comments and suggestions.
References (29)
EuroQol: the current state of play
Health Policy
(1996)- et al.
A comparison of the discounted utility model and hyperbolic discounting models in the case of social and private intertemporal preferences for health
Journal of Economic Behavior and Organization
(2002) - et al.
On the interaction of risk and time preferences: an experimental study
German Economic Review
(2001) - et al.
Preference parameters and behavioral heterogeneity: an experimental approach in the health and retirement study
Quarterly Journal of Economics
(1997) - Breyer, F., Fuchs, V.R., 1982. Risk attitudes in health: an exploratory study. Working Paper 875. National Bureau of...
- et al.
The estimation of marginal time preference in a UK-wide sample (TEMPUS) project
Health Technology Assessment
(2000) Modeling valuations for EuroQol health states
Medical Care
(1997)The measurement of health-related quality of life for use in resource allocation decisions in health care
- et al.
Assessment of patients’ preferences for therapeutic outcomes
Medical Decision Making
(1981) Debiasing
Risk-attitude and patient treatment preferences
Lupus
Risk attitude and time preference in health
Management Science
Health state valuations from the general public using the Visual Analogue Scale
Quality of Life Research
Consistency of risk premium measures
Agricultural Economics
Cited by (34)
The stability of physicians’ risk attitudes across time and domains
2023, Social Science and MedicineA systematic review of the main anomalies in intertemporal choice
2023, Journal of Behavioral and Experimental EconomicsRisk preferences and training investments
2023, Journal of Economic Behavior and OrganizationOccupational hazards, health costs, and pesticide handling practices among vegetable growers in Pakistan
2021, Environmental ResearchRisk attitudes of people with ‘manageable’ chronic disease: An analysis under prospect theory
2018, Social Science and MedicineCitation Excerpt :However, the evidence with respect to risk attitudes regarding life years and quality of life is mixed, with other evidence suggesting risk aversion over quality of life (Attema et al., 2016). Furthermore, we expressed quality of life in percentages whereas van der Pol and Ruggeri (2008) used EQ-5D classifications, and our gambles did not include immediate death as opposed to the lifetime gambles of van der Pol and Ruggeri (2008). It is therefore not clear that their results also apply to our setting.