The paper by Parkin and Devlin (P&D) makes an important contribution to the ongoing debate regarding appropriate methods for valuing health states (Parkin and Devlin, 2006). It seems entirely right to question the prevailing view in the health economics literature, that VAS comes a poor third behind SG and TTO, and other 'choice-based techniques' such as PTO. While we accept many of their responses to prevailing criticisms of VAS, we feel there are some outstanding criticisms of VAS as a tool for valuing health states.
P&D rightly expose the mistake that values for deriving QALYs must be utilities. Drummond et al. (2005) repeatedly suggest that SG utilities are the only utilities, and so effectively rule out VAS, TTO, or any other means of valuing health. This narrow view is strongly associated with the HUI research group, but has no foundation in the broader economics literature, where the term utility has many different meanings (Richardson, 1994). Furthermore, we also agree that the term CUA for cost per QALY analysis is potentially misleading and so prefer to see it as a form of CEA.
Where we do differ from P&D is in their narrow interpretation of the term 'stated preference', which they appear to limit to monetary valuation. This seems to reject the widespread use of the term in applied economics, including transport and environment, where it simply refers to values elicited from respondents' hypothetical choices over future states of the world. It has long been recognised that this is based on the notion of a utility function and that different arguments in that function can be used in the valuation -an argument well articulated in the context of health by Buckingham (1993). This is important since it provides a key rationale for SG and TTO (i.e. that use risk of death or life years, respectively, to value health) and indeed P&D's later argument for VAS (though as we suggest later rather less convincingly).
The conventional argument for using for SG and TTO rests on the context of choice in which they are framed. VAS does not present such a context of choice in the valuation task, but P&D refute the claim that respondents are not making any choice since the VAS task requires respondents to trade-off between different dimensions of health. We identify two potential flaws with this interpretation of the VAS task.
The first concerns the fact that VAS asks for more than a trade-off between aspects of health, since this could be achieved by an ordinal task, as it involves rating each state. It is the properties of the rating scale that researchers have had problems with over the years. The notion of desirability along a health