OBJECTIVES: The Canadian EQ-5D-5L valuation study used time-trade-off (TTO) responses for 86 health states of a possible 3,125 defined by the instrument. It is not uncommon for some health states to have equivalent TTO values (i.e., ties) at the individual level. In contrast, the final value set identifies differences in the public’s preferences for any pair of health states. However, the difference can be small, raising the question, what difference is minimally detectable according to individual respondents?

METHODS: The TTO responses of each respondent were transformed into pairwise differences in TTO values. The explanatory variable used pairwise differences in EQ-5D-5L index score. Three logistic regression models were used to construct curves representing the probability that a respondent considers a pairwise comparison between health states with a particular difference in EQ-5D-5L index score to be a tie, worse, or better difference. Receiver operating characteristic curves were constructed to find optimal thresholds for the difference in index score that discriminates a worse/better health state based on the observed TTO values of individuals.

RESULTS: Data were from 1,073 respondents from four Canadian cities (Vancouver, Edmonton, Hamilton and Montreal). The respondents provided a total of 48,285 pairwise TTO comparisons of which 20.2% were ties, 40.1% worse, and 39.8% better. Probability curves show that index score differences near 0 have as low as 34.7% (95% Confidence Interval: 34.2-35.2%) probability of representing a worse/better difference in health state, and as much as 30.6% probability of representing a tie. EQ-5D-5L index scores of -0.054 (-0.067 to -0.045) and 0.047 (0.037 to 0.066) maximized the sensitivity (0.79: 0.78-0.80; 0.80: 0.78-0.81) and specificity (0.79: 0.78-0.80; 0.78: 0.77-0.80) of discriminating a worse/better health state.

CONCLUSIONS: Ties in the Canadian EQ-5D-5L valuation study produce a range of small differences in index score (-0.067 to 0.066), which may represent ambiguous differences according to individual respondents.