Visualizing risk: Risk graphics' impact on patient understanding and choices in discrete choice experiments

Eliciting risk preferences is crucial in health research, yet little empirical evidence exists on which graphic types best support respondents' understanding of risks in stated preference studies, such as Discrete Choice Experiments (DCE). This study uses a mixed-methods approach to identify the most suitable risk graphics for patient preference studies using DCEs. For this purpose, a literature research, thirteen structured cognitive interviews, a randomized-controlled DCE with three study arms (N = 1,669), and a randomized-controlled DCE with two study arms (N = 630) linked to revealed preference data (N = 604) were conducted. Using the flu vaccine as a case study, we recruited Dutch respondents eligible for flu vaccination via a vendor company and their local GP. We assessed the impact of different graphic types on respondents' gist and verbatim understanding, internal validity, reliability, choice outcomes, prediction accuracy, and heuristics use. Cognitive interviews indicated that Icon Arrays with 100 icons, Column Charts and Stacked Column Charts were better understood than Icon Arrays with 10 icons, (Stacked) Bar Charts, and Risk Ladders. The randomized-controlled DCE findings suggested that Icon Arrays and Stacked Column Charts are associated with better comprehension, validity and reliability compared to Column Charts, particularly in certain subgroups. However, the graphic types were not associated with clinically relevant differences in their produced outcomes. Furthermore, no relevant differences in prediction accuracy (stated vs. revealed preferences) were detected between Icon Arrays and Stacked Column Charts. Our findings suggest that Icon Arrays and Stacked Column Charts perform equally well in DCEs and produce similar outcomes. These results support their use in future DCEs to improve risk communication and preference elicitation.