Probabilistic Attribute Presentation in Discrete Choice Experiments: A Scoping Review of Current Practice.

objectivesThere is growing interest in how probability information is framed and presented in health-related discrete choice experiments (DCEs), given evidence that presentation formats can influence respondents' choices. Recent regulatory and methodological initiatives call for clear and standardized probability presentation but the extent to which these practices have been adopted in DCEs is unclear. This scoping review sought to characterize changes in DCE practices over time compared with prior similar review studies, and to examine the extent to which current practices align with the latest evidence-based guidance.

methodsWe conducted a scoping review of 98 health-related DCE studies published between October 2022 and August 2024 that included numeric probabilistic attributes, extracting data on probability framing, use of visual aids, and alignment with best-practice guidance. Articles were identified from Medline, Embase, Web of Science, EconLit, and PsychINFO.

resultsA total of 583 attributes were presented across the 98 included studies, of which 249 were probabilistic attributes. Probabilistic risk attributes were more common than benefit attributes (present in 85% versus 65% of studies). All risk attributes were framed in absolute terms, and 16% of benefit attributes were framed in relative terms, indicating some deviation from best practices. Graphical aids were frequently used to convey probabilities, more often for risk attributes (59% of studies) versus benefit attributes (47%). Icon arrays were the predominant visual aid (86% of risk visuals, 72% of benefit visuals), but design choices varied widely. More than half of these icon arrays depicted only male figures, and color schemes were inconsistent.

conclusionsCompared with a decade ago, recent studies increasingly incorporated recommended practices (e.g., pretesting and visual aids). Nevertheless, variability persists in probability framing (especially for benefits) and visual design choices, highlighting the need for further methodological guidance to standardize probabilistic attribute presentation in DCEs.