Severity or Incidence: Labeling Side-Effect Attributes in Discrete Choice Experiments

INTRODUCTION: While medicine regulation in many countries requires prescription drug labels to classify side effects based on their incidence, when eliciting preferences of patients, most discrete choice experiments (DCEs) group side effects by severity. This mismatch between the presentation of side-effect attribute labels in DCEs and real-world choice contexts may impact patients' choices, bias risk preferences, and threaten external validity. We investigate whether using incidence compared with severity labels impacts preferences and DCE reliability. METHODS: 1105 Dutch respondents aged 60 years or older completed a survey including a DCE on influenza vaccination uptake, receiving either severity (i.e., mild or severe) or incidence (i.e., very rare or very common) attribute labels summarizing the included side effects. We randomized only the attribute labels across respondents, all else was held constant. We evaluated the effect of severity versus incidence labels on maximum acceptable risks (MAR), heterogeneity in side-effect risk preferences, reliability, choice consistency, and predicted uptake. Choice data were analyzed using mixed and heteroscedastic logit models. Other survey questions pertained to demographics, experiences, and attitudes towards the included mild and severe side effects. RESULTS: Respondents' perceptions of side effect severity did not always correspond to the clinical severity labels. Using severity compared with incidence labels increased the MAR and preference heterogeneity for severe/very rare side effects. No significant differences with regards to MAR and preference heterogeneity for mild/very common side effects, reliability or choice consistency were detected. DISCUSSION: Using severity compared with incidence-based attribute labels impacted MAR estimates and preference heterogeneity, suggesting potential priming effects. However, predicted uptake was not significantly impacted. Still, respondents' severity perception may not align with conventional severity labels, and in many countries, categorizing side effects using incidence-based attribute labels may represent patients' real-world choice contexts more accurately.