Related Experiment Videos
Administration-order effects in head-to-head comparisons of the EQ-5D-3L and EQ-5D-5L descriptive systems: evidence
Ling Jie Cheng1,2, Xun Li3
1National Perinatal Epidemiology Unit, Nuffield Department of Women's & Reproductive Health, University of Oxford, New Radcliffe House, Radcliffe Observatory Quarter, Woodstock Road, Oxford, OX2 6GG, UK. Cheng.lingjie@nus.edu.sg.
Purpose:
The EQ-5D-3L and EQ-5D-5L are commonly compared within the same questionnaire, but answering one may influence responses to the other. We randomised administration order to estimate descriptive properties before exposure to the alternative version and quantify order effects.
Methods:
In Wave 11 of the Understanding Society Innovation Panel, respondents were randomised to the EQ-5D-5L-first condition (n = 952), the EQ-5D-3L-first condition (n = 961), or the EQ-5D-5L-only condition in the late questionnaire position (n = 983). Outcomes included the full-health ceiling, response distributions, and Shannon informativity.
Results:
When each version was answered in the early questionnaire position, the full-health ceiling was 52.1% for the EQ-5D-3L and 35.4% for the EQ-5D-5L (difference 16.7% points, 95% CI 12.2 to 21.3), and the EQ-5D-5L was more informative on every dimension. Moving the EQ-5D-5L to the late position without prior EQ-5D-3L exposure changed its ceiling by 2.5 points (95% CI -2.0 to 7.0), whereas prior EQ-5D-3L exposure at the same position raised it by 12.8 points (8.2 to 17.4). Among interviewer-present interviews, the EQ-5D-5L ceiling was 34.5% without prior EQ-5D-3L, 43.0% after the EQ-5D-3L without noticing repetition, and 53.2% among those who noticed.
Conclusion:
The EQ-5D-5L order effect arose predominantly from prior EQ-5D-3L exposure rather than questionnaire position and was stronger among respondents who noticed the repetition, consistent with response contamination. Studies comparing instruments within one questionnaire should randomise and report administration order.
