Validating the Dutch SF-6D and EQ-5D Using Pairwise Comparisons and Best-Worst Scaling > EQ-5D

EQ-5D

59.2

In April 2011, the LISS panel completed the EQ-5D questionnaire. The questionnaire serves to validate the Dutch EQ-5D (classification for describing health) and to understand the valuation of health in the Netherlands. The questionnaire also examines the best way of posing the questions. For this purpose, pairwise comparisons and best-worst scaling are used (the evaluation of three health states simultaneously). 

Respondents were first presented a number of example questions, so that they could practice with the setup and layout of the multiple choice questions. This was followed by 12 multiple choice questions. For each decision-making situation, respondents were asked to imagine they only had a few years left to live. For every question three options were given, indicating in what health condition the respondents could spend the last years of their life. Respondents then indicated which option they preferred.  

Depending on the group to which the respondent belonged, (some) questions consisted of pairwise comparisons; each question then comprised a part A and part B. In part A, a choice was offered between options A and B. These options offered an equal life expectancy but different health conditions: a better, equal of worse score on the 5 health aspects of “problems with walking”, “problems with washing or dressing”, “problems with daily activities”, “pain or discomfort” and “anxiety or depression”. In part B, exactly the same options were displayed as in part A, but here the choice was between option B, to live longer in poorer health, and C, to live shorter in perfect health. Also depending on the group to which the respondent belonged, (some) questions consisted of evaluations of the three health conditions simultaneously (best-worst scale distribution).  

The groups were as follows:
Group 1: 12 pairwise comparisons (questions eo11a019-eo11a042), followed by evaluation.
Group 2: 12 best-worst questions (questions eo11a043-eo11a066), followed by evaluation.
Group 3: 6 pairwise comparisons (questions eo11a067-eo11a078), followed by evaluation, and 6 best-worst questions (questions eo11a084-eo11a095), followed by evaluation.  

The remaining life expectancies and health conditions differed per question and per version (note that option C always represented perfect health, but with a shorter life expectancy than options A and B). Respondents were randomly assigned to one of the eight versions. The health aspects were displayed on the screen in random order (this order was maintained throughout the rest of the questionnaire).  

Colors were used for the scores on health aspects. The colors depended on the group to which the respondent belonged: Group 1: light purple=good health, dark purple=poor health. Group 2: light purple=poor health, dark purple=good health.   This was done to check in how far the colors contribute to the relatively large weight of the poor health conditions.  

The variables for the version, the question, the scores on the 5 health aspects, and the remaining life expectancies were captured.  

Single Wave Study

04-04-2011

27-04-2011

Health and well-being (LISS)

Survey methodology (LISS)

Jonker, M., Donkers, B., De Bekker-Grob, E.

CentERdata

© 2012 CentERdata

CentERdata/MESS Project

https://doi.org/10.17026/dans-zmb-2ec5