A Think Aloud Study Comparing the Validity and Acceptability of Discrete Choice and Best Worst Scaling Methods

Abstract

OBJECTIVES: This study provides insights into the validity and acceptability of Discrete Choice Experiment (DCE) and profile-case Best Worst Scaling (BWS) methods for eliciting preferences for health care in a priority-setting context. METHODS: An adult sample (N = 24) undertook a traditional DCE and a BWS choice task as part of a wider survey on Health Technology Assessment decision criteria. A 'think aloud' protocol was applied, whereby participants verbalized their thinking while making choices. Internal validity and acceptability were assessed through a thematic analysis of the decision-making process emerging from the qualitative data and a repeated choice task. RESULTS: A thematic analysis of the decision-making process demonstrated clear evidence of 'trading' between multiple attribute/levels for the DCE, and to a lesser extent for the BWS task. Limited evidence consistent with a sequential decision-making model was observed for the BWS task. For the BWS task, some participants found choosing the worst attribute/level conceptually challenging. A desire to provide a complete ranking from best to worst was observed. The majority (18,75%) of participants indicated a preference for DCE, as they felt this enabled comparison of alternative full profiles. Those preferring BWS were averse to choosing an undesirable characteristic that was part of a 'package', or perceived BWS to be less ethically conflicting or burdensome. In a repeated choice task, more participants were consistent for the DCE (22,92%) than BWS (10,42%) (p = 0.002). CONCLUSIONS: This study supports the validity and acceptability of the traditional DCE format. Findings relating to the application of BWS profile methods are less definitive. Research avenues to further clarify the comparative merits of these preference elicitation methods are identified.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 83 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.