Multi-method approach to valuing health states: problems with meaning

In an earlier article in Health Economics, Salomon and Murray argue that by applying maximum likelihood techniques to predetermined functional forms and to a data set where a number of health states are valued by means of four standard valuation techniques, underlying 'pure' valuations of health may be teased out, together with estimates of parametric relationships between these 'pure' valuations and valuations based on the four standard techniques. We argue below that 'pure' valuations of health are ordinal rather than cardinal and that the 'pure' values that result from the multi-method approach give a false impression of being cardinal. They are therefore not usable as weights for life years. In the unlikely event that the authors should be able to demonstrate cardinality in 'pure' valuations of health, it must be possible to have subjects express these valuations directly, in which case there seems to be no need for the indirect multi-method approach.