Framing effects of risk communication in health-related decision making. Learning from a discrete choice experiment

Background How to communicate uncertainty is a major concern in medicine and in health economics. We aimed at studying the framing effects of risk communication on stated preferences in a discrete choice experiment (DCE) performed to elicit women's preferences for Hormone Replacement Therapy. Methods Two versions of the questionnaire were randomly administered to respondents. Multiple risks were expressed as natural frequencies using either a constant reference class (Design 1) or variable reference classes (Design 2). We first tested whether Design 1 would impose a lower cognitive burden than Design 2. We then examined whether the two designs resulted in different utility model estimates. Results Design 1 improved consistency (monotonicity and stability). However, rates of dominance or intransitive responses did not differ across designs. Design 1 decreased women's sensitivity to the risk of fractures and increased their sensitivity to the risk of breast cancer as compared to all other attributes. Discussion Framing effects of risk communication on stated preferences may be a major problem in the design of DCEs. More research is needed to determine whether our findings are replicable and to further investigate the normative question of how to improve risk communication in health-related decision-making.