Labeling Risk: Comparing Occurrence vs Severity-Based Risk Labels in Patient Preference Studies

OBJECTIVES: When eliciting preferences using Discrete Choice Experiments (DCEs) it is crucial that choice scenarios align with patients’ real-life choice contexts to generate valid market share predictions. While EU guidelines require prescription drug labels to classify adverse reactions (ARs) based on their occurrence rate; ARs are grouped by severity in most DCEs. However, previous studies suggest that AR labels influence respondents’ risk perception and evaluation. Both AR labels prime respondents; potentially altering their judgment of the perceived burden and probability of ARs. A mismatch between risk presentation in DCEs and real-life may impact patients’ choices, bias risk preference estimates, and threaten external validity.

METHODS: Respondents were recruited for a DCE study on the flu vaccine (N=1,000 Dutch population aged 60+) and received either severity (i.e., mild/severe) or occurrence-based (i.e., rare/common) AR attribute labels. It was evaluated whether their responses systematically differed in terms of their (1) evaluation of AR as mild or severe, (2) maximum acceptable risks (MAR), (3) heterogeneity in risk preferences for AEs, and (4) reliability. Data were analyzed using mixed-logit models.

RESULTS: Not all AR labeled as ‘mild’ were also perceived as mild by the respondents; some were perceived as severe. Occurrence-rate-based AR labels were associated with higher MAR and slightly decreased preference heterogeneity. While reliability was not impacted by the different AR attribute labels, occurrence-rate-based AR were associated with decreased choice consistency.

CONCLUSIONS: Researcher’s classification of AR severity does not fully align with patients’ severity perceptions. Since overall severity categories capture many respondents’ AR severity perceptions inaccurately, and occurrence-based labels are associated with higher MAR for all AR, they likely capture respondents’ risk preference more precisely. Despite being associated with decreased choice consistency, occurrence-based AR labels represent patients’ real-life choice situation more accurately and their use should be considered in future preference studies.