The comparability and reliability of five health-state valuation methods

The objective of the study was to consider five methods for valuing health states with respect to their comparability (convergent validity, value functions) and reliability. Valuation tasks were performed by 104 student volunteers using five frequently used valuation methods: standard gamble (SG), time trade-off (TTO), rating scale (RS), willingness-to-pay (WTP) and the paired comparisons method (PC). Throughout the study, the EuroQol classification system was used to construct 13 health-state descriptions. Validity was investigated using the multitrait-multimethod (MTMM) methodology. The extent to which results of one method could be predicted by another was examined by transformations. Reliability of the methods was studied parametrically with Generalisability Theory (an ANOVA extension), as well as non-parametrically. Mean values for SG were slightly higher than TTO values. The RS could be distinguished from the other methods. After a simple power transformation, the RS values were found to be close to SG and TTO. Mean values of WTP were linearly related to SG and TTO, except at the extremes of the scale. However, the reliability of WTP was low and the number of inconsistencies substantial. Valuations made by the RS proved to be the most reliable. Paired comparisons did not provide stable results. In conclusion, the results of the parametric transformation function between RS and SG/TTO provide evidence to justify the current use of RS (with transformations) not only for reasons of feasibility and reliability but also for reasons of comparability. A definite judgement on PC requires data of a complete design. Due to the specific structure of the correlation matrix which is inherent in valuing health states, we believe that full MTMM is not applicable for the standard analysis of health-state valuations