Eliciting Preferences for Hypothetic Health States from Patients: A Comparison between Time Trade-Off and EQ-5D in Subjects With Diabetes

Abstract:     
INTRODUCTION: The valuation of health state scenarios that describe disease-specific detail may be a more feasible task for subjects with experience of that disease than for the general public. In that context, valuation using profile instruments with pre-scored tariffs, such as EQ-5D, is relatively simple to implement and free from framing effects, but may be less sensitive to subtle differences or changes than direct elicitation methods, such as time trade-off (TTO). We conducted a study to elicit preferences from people with diabetes for clinical scenarios involving inhaled human insulin (EXUBERA®) or injected insulin treatment alternatives. The objective of the additional analysis reported here was to compare values obtained using EQ-5D and TTO.

METHODS: Written descriptions were developed for five clinical scenarios in Type 1 and Type 2 diabetes (T1D and T2D). Two variants were described for each scenario: one with injectable-only insulin regimen and the other with inhaled insulin to replace or reduce the number of daily injections. Computer-assisted personal interviews were conducted at six UK locations with people with diabetes aged over 18 years. After demonstration of the inhalation and pen injection devices, 344 respondents (66% male), 132 (mean age 49 years) with T1D and 212 (mean age 63 years) with T2D, rated the injection and inhalation variants of each clinical scenario by TTO and EQ-5D. Self-state was also valued by these two methods. Principal component analysis (PCA) was used to explore correlations between EQ-5D and TTO ratings for self state and the scenarios. Multivariate regression analysis was used to evaluate the impact of EQ-5D self-state on scenario ratings.

RESULTS: The mean TTO scores for self-state (0.83 for both T1D and T2D patients) were greater than mean EQ-5D scores (0.75 for both T1D and T2D). TTO values for scenarios ranged from 0.82 to 0.93 and EQ-5D scores ranged from 0.74 to 0.83. Valuations for inhaled insulin were greater than their injectable insulin counterparts by 0.043-0.074 by TTO and 0.020-0.043 by the EQ-5D UK tariff. Correlation coefficients between EQ-5D and TTO values ranged from 0.07 to 0.31 (p<0.05 for 5 out of 10 scenarios). PCA demonstrated that TTO values were associated with ratings of the anxiety/depression, and less so with the pain/discomfort, dimensions of the EQ-5D. EQ-5D values were significantly correlated with age but TTO values were not. EQ-5D values for self-state and for scenarios were significantly correlated, with coefficients from 0.30 to 0.36 (p<0.0001 for all scenarios). However, there was no significant impact of EQ-5D self-state on TTO values after adjustment on EQ-5D values for corresponding scenarios (p=0.89 for T1D and 0.96 for T2D).

CONCLUSIONS: EQ-5D valuations of health state scenarios described to subjects with diabetes were systematically lower than the respective TTO ratings. Correlations between the two measures were weak to moderate and appear to be driven mainly by EQ-5D ratings of anxiety/depression. TTO values were found to be more sensitive to differences between scenarios, and less affected by respondents' age and reported self-state, than EQ-5D values.

Keywords: Time trade-off, EQ-5D, diabetes

Working Paper Series