Probabilistic Attribute Presentation in Discrete Choice Experiments (DCEs): A Review of Current Practice
Author(s)
Gabriela Fernandez, MPH1, Saudamini Oke, MSc2, Matthew Quaife, PhD3.
1Patient-centered Research, Thermo Fisher Scientific, Wilmington, NC, USA, 2London School of Hygiene and Tropical Medicine, London, United Kingdom, 3Thermo Fisher Scientific, Hammersmith, United Kingdom.
1Patient-centered Research, Thermo Fisher Scientific, Wilmington, NC, USA, 2London School of Hygiene and Tropical Medicine, London, United Kingdom, 3Thermo Fisher Scientific, Hammersmith, United Kingdom.
Presentation Documents
OBJECTIVES: How probabilities are presented in discrete choice experiments (DCEs) has been shown to impact how participants interpret information. This study aimed to synthesize current practice in presenting probabilities in DCEs, and report the nature of attributes, framing, and visual presentation.
METHODS: A scoping review was performed to identify DCEs which included probabilistic attributes for health-related interventions from October 2022 through August 2024. Articles were identified from Medline, Embase, Web of Science, EconLit, and PsychINFO, with titles/abstracts and full-texts subsequently screened. Relevant studies were extracted using a pre-specified framework.
RESULTS: Data from 98 studies were extracted. DCEs spanned a range of therapeutic areas, most commonly oncology (30%), endocrinology (11%), and dermatology (7%). Most studies included patients (83%), were focused on assessing treatment preferences (81%), and reported some pre-testing (83%). The majority of DCEs included 5-7 attributes (70%) and 1-3 probabilistic attributes (77%).
More DCEs included probabilistic risk attributes (85%) than probabilistic benefit attributes (64%). Risks were most frequently presented as proportions alongside percentages (37%), and benefits most frequently as percentages only (30%). All risk attributes used absolute framing, whereas 16% of benefit attributes used relative framing. Graphics were used more often to present risks than benefits (59% v 46%); within these, arrays of person-like figures were used in 72% of risks and 71% of benefits. Exclusively male figures were used for 80% of benefits and 87% of risks. Blue (31%) and red (27%) colors were most used for risks, and blue (38%) and green (28%) for benefits.
CONCLUSIONS: Current practices of presenting probabilistic attributes were generally consistent in following good practice in pre-testing and using some visual representation. The variety of graphics used to display probabilistic risks and benefits, specifically gender/shape, and color, indicate that work to understand how participants interpret different presentations may assist critical DCE design decisions.
METHODS: A scoping review was performed to identify DCEs which included probabilistic attributes for health-related interventions from October 2022 through August 2024. Articles were identified from Medline, Embase, Web of Science, EconLit, and PsychINFO, with titles/abstracts and full-texts subsequently screened. Relevant studies were extracted using a pre-specified framework.
RESULTS: Data from 98 studies were extracted. DCEs spanned a range of therapeutic areas, most commonly oncology (30%), endocrinology (11%), and dermatology (7%). Most studies included patients (83%), were focused on assessing treatment preferences (81%), and reported some pre-testing (83%). The majority of DCEs included 5-7 attributes (70%) and 1-3 probabilistic attributes (77%).
More DCEs included probabilistic risk attributes (85%) than probabilistic benefit attributes (64%). Risks were most frequently presented as proportions alongside percentages (37%), and benefits most frequently as percentages only (30%). All risk attributes used absolute framing, whereas 16% of benefit attributes used relative framing. Graphics were used more often to present risks than benefits (59% v 46%); within these, arrays of person-like figures were used in 72% of risks and 71% of benefits. Exclusively male figures were used for 80% of benefits and 87% of risks. Blue (31%) and red (27%) colors were most used for risks, and blue (38%) and green (28%) for benefits.
CONCLUSIONS: Current practices of presenting probabilistic attributes were generally consistent in following good practice in pre-testing and using some visual representation. The variety of graphics used to display probabilistic risks and benefits, specifically gender/shape, and color, indicate that work to understand how participants interpret different presentations may assist critical DCE design decisions.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
P46
Topic
Patient-Centered Research
Disease
No Additional Disease & Conditions/Specialized Treatment Areas