Is best-worst scaling suitable for patient preference elicitation? A nationwide representative comparison with discrete choice experiments among patients with type 2 diabetes in China

Objective: Incorporating patient perspectives is essential in health technology assessments. Selecting appropriate tools, such as discrete choice experiment (DCE) or best-worst scaling (BWS), is critical for generating high-quality evidence. This study aims to compare preference estimates and response consistency between DCE and BWS profile case (BWS-2) and provide empirical guidance on the optimal method for patient preference elicitation. Methods: Multistage stratified cluster sampling was used to recruit 3327 patients with type 2 diabetes mellitus (T2DM) from 50 health institutions for face-to-face interviews conducted from September to December 2022. Participants completed DCE and BWS-2 tasks involving hypothetical antihyperglycaemic medication profiles defined by seven attributes: efficacy, hypoglycaemia risk, cardiovascular benefits, gastrointestinal adverse events, weight change, mode of administration and out-of-pocket cost. Each participant completed six DCE and six BWS-2 tasks in randomised order (DCE first followed by BWS, or BWS first followed by DCE). Responses were analysed using mixed logit and generalised multinomial logit models. Method concordance was assessed with Spearman's correlation, and response consistency was evaluated through scale heterogeneity. Results: Preference weights from DCE and BWS-2 showed moderate concordance (Spearman's r>0.70). DCE exhibited greater scale heterogeneity (τ=0.931, p