2018 Conference Presentation
Objectives: Studies using best-worst scaling (BWS) have increasingly been used in health economic evaluations for eliciting preferences for health-related states as they are generally seen as less burdensome than discrete choice experiments (DCEs). We present the application of the BWS method for establishing preference weights for the German version of the ASCOT-S (Adult Social Care Outcomes Toolkit, Service User version).
Methods: We conducted a profile-case best-worst experiment with a representative sample of the Austrian population (n=1,000). Utilities for each domain were estimated using a multinomial logit model taking into account the level of the attribute presented. This basic model was then extended to include both additive respondent-level covariates to investigate differences in specific preferences between groups and multiplicative scale factors to account for overall differences in consistency between groups.
Results: We found significant positioning effects for ‘best’ choices, with statements at the top being picked more often than those further down in the list. In terms of inter-group preference differences, we find that certain socioeconomic groups (based on age, education, gender) differ significantly in their preferences for certain domains or levels. For consistency of choices, factors related to survey completion (self-assessed understanding of the tasks and survey completion time) were shown to have the greatest effect.
Discussion: The findings provide insights into preferences for long-term care outcomes and the workings of BWS experiments. Investigating inter-group differences in taste can help to better understand factors influencing preferences at the personal level, while looking into factors affecting scale heterogeneity provides insight into the decision-making process in the context of the BWS tasks.