Feasibility, validity and test-retest reliability of scaling methods for health states: the visual analogue scale and the time trade-off.

Abstract

The feasibility, validity and reliability of the Time Trade-Off (TTO) and Visual Analogue Scale (VAS) methods in obtaining preference values for health states were compared in a random sample of the Spanish population (n = 294). Respondents valued 43 EuroQol-5D health states in face-to-face interviews. Convergent validity was assessed by examining the relationship between values, and the effect of sociodemographic and health variables on values was used as a means of assessing construct validity. Test-retest reliability was analysed in a subgroup of 50 respondents, using the intraclass correlation coefficient (ICC) and generalisability theory. Rates of non-response and missing data were low on both methods, though the VAS took considerably less time to administer. VAS and TTO values correlated highly (r = 0.92), though there were differences in the ordering of health states between methods, and in the number of health states rated worse than death. VAS values were compressed into a considerably smaller valuation space than TTO values. Respondents in higher educational categories assigned higher TTO values to 12 health states. Mean ICCs (95% CI) at individual level were 0.90 (0.88-0.92) and 0.84 (0.81-0.87) for the VAS and TTO, respectively. Generalisability analysis showed variance due to time to be 0 for both methods. In conclusion, the VAS was more feasible and slightly more reliable than the TTO, whilst doubt can be cast on the degree of convergent validity existing between the two methods. The compression of VAS values means that the TTO is likely to discriminate better between health states, and it may have greater construct validity if results from larger samples confirm that there are genuine differences between sociodemographic subgroups.