Background. Discrete choice experiments (DCEs) are increasingly used in the health care context to inform on patient preferences for health care services. In order for such experiments to provide useful and policy-relevant information, it is vital that the design includes those options that the respondent faces in the real-life situation. Whether to include opt-out, neither, or status quo alternatives has, however, received little attention in the DCE literature. We aim to investigate whether the use of different unforced choice formats affects DCE results in different settings: 1) opt-out versus neither in a health care market where there is no status quo and 2) including status quo in addition to opt-out in a health care market with a status quo. Design. A DCE on Dutch citizens’ preferences for personal health records served as our case, and 3189 respondents were allocated to the different unforced choice formats. We used mixed logit error component models to estimate preferences. Results. We found that the use of different unforced choice formats affects marginal utilities and welfare estimates and hence the conclusions that will be drawn from the DCE to inform health care decision making. Conclusions. To avoid biased estimates, we recommend that researchers are hesitant to use the neither option and consider including a status quo in addition to opt-out in settings where a status quo exists.
Empirical investigation of experimental design properties of discrete choice experiments in health care
