Objectives
To empirically test to what extent the layout of choice tasks (i.e., displayed in words or graphics) in a Discrete Choice Experiment (DCE) influences the attribute estimates and the conclusions drawn from the DCE.
Methods
A DCE questionnaire was sent to the parents of 2500 newborn babies aged 6 weeks at maximum. Each questionnaire contained two times the same 9 choice tasks, ones words were used to describe the attributes and levels and ones graphics were used. The DCE consisted of five attributes related to the decision of parents to vaccinate their newborn baby against the rotavirus (vaccine effectiveness, frequency of severe side effects, protection duration, location, costs). Mixed logit models were conducted to estimate the relative importance of the attributes.
Results
Preliminary results are based on 279 observations from 31 parents. In February 2013 data collection will be completed and analyzed. When comparing the choices of every respondent per choice tasks, 58% chose inconsistent at 1 or more choice tasks and 35% chose inconsistent in two or more choice tasks. In both datasets (layout in words and graphics), vaccine effectiveness (βeffects code 1=0.64 and βeffects code 1=1.00, βeffects code 2=0.67 and βeffects code 2=0.01), frequency of severe side effects (βeffects code 1=0.26 and βeffects code 1=0.41, βeffects code 2=1.22 and βeffects code 2=0.89) protection duration (β=0.37, β=0.17) and costs (β=-0.10, β=-0.11) showed significant attribute estimates (P<.05). However, the relative importance of these attributes differed between both datasets.
Conclusions
For now it can be concluded that the presentation of the choice sets (by either using words or graphics) in a DCE influences study outcomes. Besides extensive pilot testing to ensure the choice tasks are understood and interpreted as intended, it might be worthwhile to include discussions about the layout of the choice tasks in the focus group stage of the DCE designing process