Discrete Choice Experiment (DCE) vs Technical Threshold (TT) Analyses: How to Choose?

OBJECTIVES: Patient preferences are crucial in informing appropriate decision making, hence the importance of robust preference elicitation methodologies. While Discrete Choice Experiments (DCE) and the Threshold Technique (TT) are prominent methods, there is a lack of comprehensive guidelines for their practical application in health technology assessment. This study aims to establish criteria for selecting the most appropriate method according to the clinical context.

METHODS: We conducted a targeted literature review in Pubmed to identify studies employing DCE or TT to understand patient preferences in healthcare. Our objective was to compare these methods in terms of methodology, results, and practical use. Data extraction encompassed method phases, sample sizes, cognitive burden, and outcomes.

RESULTS: DCE employs complex experimental designs to capture patient preferences, requiring larger samples (> 100 patients) and questionnaire of longer duration. It is recommended limiting choice tasks to 8-16 with 6-7 attributes per task and 2-4 levels per attribute due to the varied treatment profiles from one choice task to another (Bridges et al. 2011, De Bekker-Grob et al. 2015). Conversely, TT involves simpler tasks where participants choose between two product profiles, systematically varying the level of one attribute until the preference switches to the alternative product, identifying trade-offs directly. TT is faster, suitable for smaller patient groups (e.g., rare diseases), with studies typically involving 100 or fewer respondents (Sung et al. 2004, Steures et al. 2005).

CONCLUSIONS: Both DCE and TT are recognized methods to elicit patient preferences, but their application depends on specific study requirements and contexts. Considerations of patient recruitment and cognitive burden are crucial. Clearer guidelines on sample sizes and methodological choices are needed to optimize their application in diverse clinical contexts.