Modelling the relationship between the description and valuation of health states

A number of health state descriptive systems have been specifically designed for calculating a single index value for every state defined within the system. Since most descriptive systems define more health states than it is feasible to elicit direct valuations for in an empirical study, choices have to be made about how best to estimate values for all states from direct observations on a subset of those states. This paper discusses the relative merits of the decomposed and composite approaches to modelling health state valuations. The choice of approach to modelling depends partly on the extent to which there is a pre-commitment to a particular axiomatic model of preferences. The paper also considers various decisions that have to be made when attempting to generate valuation “tariffs” and looks at how such tariffs have been estimated for the HUI3 and EQ-5D descriptive systems. It is argued that a fruitful avenue for future research in this area would be to directly compare the predictive ability of the decomposed and composite approaches.