OBJECTIVES: Discrete Choice Experiments (DCEs) are a popular method to elicit stated preferences of patients in healthcare. There exist different options how to design a DCE. However, little is known about how these options influence the outcomes of a DCE study.

METHODS: A cross-sectional survey of 122 students at a German University was conducted, investigating preferences of students with regard to a hypothetical health insurance bonus program (fitness club, professional tooth cleaning, BMI assessment, coaching, pay-back bonus). Each respondent completed a questionnaire with three sets of choice tasks with 8 different choices each (one basic design (BD) and two further sets with design modifications). Design options tested, based on similar content, were (M1) increasing the number of levels for one attribute, (M2) changing the order of the attributes on choice cards, (M3) adding an opt-out option. Three outcome comparisons, based on a generalized multinomial logit model, were made: BD versus M1 (56 respondents), BD versus M2 (55), and BD versus M3 (43).

RESULTS: Both the effect size and even the direction of the influence of the attributes on respondents’ utility varied between the design options. So, as an example, the most important attribute in the BD design was the mandatory tooth cleaning (relative importance: 63%-69%), but the relative importance decreased to 38% when M2 was applied. Applying M1 and M3 changed the ranking of the importance of all other attributes, e.g. fitness club membership was rated as the second most important in the BD design, but the most unimportant attribute in M3.

CONCLUSIONS: If DCEs are used to inform health policy decision makers, it is crucial that presented results are valid and robust. Obviously, DCE design decisions may substantially influence outcomes of DCE studies. We recommend to take this aspect into account when designing a DCE study.