Assessing the impact of excluded attributes on choice in a discrete choice experiment using a follow-up question

Abstract

Health researchers design discrete choice experiments (DCEs) to elicit preferences over attributes that define treatments. A DCE can accommodate a limited number of attributes selected by researchers based on numerous factors (e.g., respondent comprehension, cognitive burden, and sample size). For situations where researchers want information about the possible impact of an attribute excluded from the DCE, we propose a method to use a question after the DCE. This follow-up question includes the attributes in the DCE with fixed levels and an additional attribute originally excluded from the DCE. The DCE data can be used to predict the probability that respondents would select one treatment profile over another without the additional attribute. Comparing the prediction to the percentage of the sample who selected each profile when it includes the additional attribute provides information on the potential impact of the additional attribute. We provide an example using data from a DCE on treatments for chronic lymphocytic leukemia. Cost was excluded from the DCE, but the survey included a follow-up question with two fixed treatment profiles, similar to two treatments currently on the market, and a cost for each. Preferences were sensitive to modest changes in cost, highlighting the importance of gathering this information.