A Head-On Comparison of EQ-VT- and Crosswalk-Based EQ-5D-5L Value Sets

BackgroundNo systematic country-level comparison has been undertaken between crosswalk- and EQ-VT-derived EQ-5D-5L value sets. Crosswalk values can differ from EQ-VT-based EQ-5D-5L value sets owing to valuation protocols, changes in societal preferences over time, and a change in the label of the highest level on mobility in moving from EQ-5D-3L to EQ-5D-5L. This study aimed to compare the five-level (5L) crosswalk and EQ-VT value sets to explore differences between them at the country level.MethodsFrom the countries with both time trade-off (TTO)- or discrete choice experiment (DCE) + TTO-based EQ-5D-3L value sets and EQ-VT-based EQ-5D-5L value sets, 19 pairs of EQ-5D-3L/EQ-5D-5L sets were found. For each of these EQ-5D-3L value sets, 5L crosswalk sets were developed and compared with the corresponding national EQ-5D-5L valuation set using correlation analysis, ranges, values of specific states, Bland-Altman plots, and scatter plots. Three of the countries have EQ-5D-3L and EQ-5D-5L valuation data for the same set of respondents. These three cases were analyzed separately, as they provide a "true" test of the differences between the two value sets.ResultsSpearman correlation between the crosswalk and valuation sets ranged from 0.831 to 0.989, being below 0.9 in 11 pairs of value sets. The difference in the percentage of negative values ranged from +22.5 to -18.8%, and the difference in the ranges within each pair of value sets ranged from +42.7 to -18.4%. The average mean absolute difference of values (crosswalk versus EQ-VT) was 0.149. This was below 0.1 in only 5 of the 19 EQ-VT/crosswalk set pairs. For the states comprising one level 5 and four level 1s, no country preserved its ranking of importance of the five dimensions in moving from crosswalk to EQ-VT values. Most of the Bland-Altman plots and scatterplots revealed a pattern that placed states with the highest level on mobility as a separate band from other states.DiscussionAll of the criteria showed poor agreement between the crosswalk- and EQ-VT-based value sets. The differences in labels for the most extreme response option for the mobility dimension leads to substantial differences in values between these value sets.ConclusionsCrosswalk and EQ-VT value sets should not be used interchangeably, except under circumstances where it is not possible or feasible to conduct a direct EQ-5D-5L valuation study.