It is increasingly being acknowledged that expected utility has substantial empirical deficiencies. Prospect theory has developed as an important alternative, with more descriptive validity. In particular, several applications of prospect theory have recently been performed in health economics, but to date no quantification of prospect theory’s full functional had been performed in this domain. This quantification is important though, since it will facilitate a better description of health-related behavior, an improvement of policy recommendations, and a reduction of biases in utility elicitation methods, such as the popular time tradeoff method. This paper is therefore the first to do so, by simultaneously measuring the utility function for life duration and probability weighting for both gains and losses, and a loss aversion index. An adaptation of a semi-parametric method introduced for the monetary domain is used for this measurement.

I find a significant amount of loss aversion, comparable in size to estimates in the monetary domain. Furthermore, universal risk aversion is observed, which, for gains, can be completely attributed to underweighting of probability 1/2. That is, the utility function for gained life years becomes close to linear. For losses, however, there is much less probability weighting, and concavity is still substantial even under prospect theory. The latter result contrasts with the common finding in the monetary field of convex utility for losses. An explanation may be that utility of life duration is confounded with time preference; if the latter is positive, i.e., individuals give less weight to the future, it will cause utility of life duration to be more concave, and, hence, may be erroneously interpreted as more risk averse behavior. However, the degree of implied time preference is higher for losses than for gains, which may be due to hyperbolic discounting.

The results suggest that the QALY model is indeed not valid if expected utility is assumed, justifying the use of non-expected utility QALY models. This first attempt to quantify its complete functional seems promising, but more research is required to refine estimation methods, and to further investigate the distorting consequences of time preference and the peculiarities found in the loss domain.