OBJECTIVES: Framing attributes in terms of mortality or survival heavily impacts Discrete Choice Experiment (DCE) outcomes, however very little is known about framing effects of other (less dominant) attributes. This study aimed to test how framing of a treatment effectiveness attribute in a DCE affects respondents’ decision-making behavior and their preferences.

METHODS: Three DCE surveys were randomly distributed among a representative sample of the Swedish population aged 18-65 years (n=1024). The DCEs consisted of 5 attributes related to antibiotic treatment. The effectiveness attribute was framed either positively, negatively (treatment effectiveness or failure) or using both frames. Bayesian D-efficient designs with priors from a pilot study were created and optimized separately for each of the three DCEs. All designs contained 48 unique choice tasks divided over three blocks. Latent class analysis was performed to estimate the attribute level estimates and preference heterogeneity, after which flat-lining and relative importance scores (RIS) were determined.

RESULTS: Respondents of the three DCEs did not differ with respect to their demographics, perceptions regarding the length and difficulty of the DCE, or flat-lining (P>0.05). Two-class models were fitted for the three DCEs separately which resulted in differences concerning respondents’ preferences, class-adjusted RIS and preference heterogeneity. While negative framing resulted in high preference heterogeneity, minimal heterogeneity was found when using both frames. When the attribute was framed positively it received a relatively higher RIS compared to when the attribute was framed as treatment failure or used both frames.

CONCLUSIONS: Framing an attribute to report treatment effectiveness or failure influences DCE outcomes. Showing both frames mirrored the ‘treatment failure’ results, implying that respondents likely focused on this information. Although the framing effects in this study are less pronounced as those reported for survival-mortality attributes; researchers ought to be careful in selecting, framing and presenting all attributes in their DCE.