A Framework for Instrument Development of a Choice Experiment: An Application to Type 2 Diabetes

Objective: Choice experiments are increasingly used to obtain patient preference information for regulatory benefit–risk assessments. Despite the importance of instrument design, there remains a paucity of literature applying good research principles. We applied a novel framework for instrument development of a choice experiment to measure type 2 diabetes mellitus treatment preferences. Methods: Applying the framework, we used evidence synthesis, expert consultation, stakeholder engagement, pretest interviews, and pilot testing to develop a best–worst scaling (BWS) and discrete choice experiment (DCE). We synthesized attributes from published DCEs for type 2 diabetes, consulted clinical experts, engaged a national advisory board, conducted local cognitive interviews, and pilot tested a national survey. Results: From published DCEs (n = 17), ten attribute categories were extracted with cost (n = 11) having the highest relative attribute importance (RAI) (range 6–10). Clinical consultation and stakeholder engagement identified six attributes for inclusion. Cognitive pretesting with local diabetes patients (n = 25) ensured comprehension of the choice experiment. Pilot testing with patients from a national sample (n = 50) identified nausea as most important (RAI for DCE: 10 [95 % CI 8.5–11.5]; RAI for BWS: 10 [95 % CI 8.9–11.1]). The developed choice experiment contained five attributes (A1c decrease, blood glucose stability, low blood glucose, nausea, additional medicine, and cost). Conclusion: The framework for instrument development of a choice experiment included five stages of development and incorporated multiple stakeholder perspectives. Further comparisons of instrument development approaches are needed to identify best practices. To facilitate comparisons, researchers need to be encouraged to publish or discuss their instrument development strategies and findings.