OBJECTIVES: Previous studies showed that framing of mortality or survival attributes heavily impacts Discrete Choice Experiment (DCE) outcomes. However, little is known about framing effects of less dominant attributes often included in health related DCE studies. The aim of this study was to investigate how the framing of attributes impacts respondent choice behavior and DCE outcomes regarding attribute level estimates, relative importance scores (RIS) and preference heterogeneity.

METHODS: Hypothetical antibiotic treatments were described using five attributes. Four attributes were static: antibiotic resistance contribution, treatment duration, side-effect risk, and out-of-pocket costs. A fifth treatment attribute was framed in three ways: effectiveness (positive), failure rate (negative) or a combination of both. Three Bayesian D-efficient designed DCE surveys framed positively, negatively, or both were developed using pilot study priors (n=129). The DCE surveys were randomly distributed among a representative sample of the Swedish population aged 18-65 years (n=1024). Latent class analysis was used to retrieve attribute level estimates, RIS, and preference heterogeneity. Model outcomes were compared between the datasets. Potential scale differences were tested with a Swait and Louviere test.

RESULTS: Significant differences were found between different framings for RIS, preference heterogeneity, and preferences. No significant differences were found between the three DCEs regarding respondent demographics, perceptions of DCE difficulty or length, or flat-lining. Two significantly different classes were identified in all three framing format data sets: prioritization of costs or prioritization of medical aspects. The Swait and Louviere test showed no scale differences between the datasets but significant differences in attribute level estimates.

CONCLUSIONS: Framing attributes as either effectiveness or failure rate influences DCE outcomes. Showing both framing formats more closely mirrors ‘effective rate’ framing, implying that respondents likely focus on positive information. Researchers should be careful in framing and presenting attributes in their DCE as this can change the preference outcomes.