"Tell Me a Story": A Randomized Study Comparing Tabulated and Narrative Presentations of a Discrete Choice Experiment
Author(s)
Liz Morrell, PhD1, James Buchanan, MA, DPhil2, Laurence Roope, PhD1, Koen Pouwels, PhD1, Julie Robotham, PhD3, Sarah Walker, PhD1, Sarah Wordsworth, BSc, MSc, PhD1.
1University of Oxford, Oxford, United Kingdom, 2Queen Mary University of London, London, United Kingdom, 3UK Health Security Agency, London, United Kingdom.
1University of Oxford, Oxford, United Kingdom, 2Queen Mary University of London, London, United Kingdom, 3UK Health Security Agency, London, United Kingdom.
OBJECTIVES: Discrete choice experiments (DCEs) are frequently used to evaluate the preferences of the public, patients, and healthcare professionals. Choice alternatives in DCEs are usually presented in table format. However, our pilot work suggested some respondents might prefer a narrative format, presented as a patient story. Given limited information on respondent preferences for narrative formats, we randomised respondents to see either table or narrative formats of a DCE on antibiotic prescriptions, among the UK general public.
METHODS: The online DCE used 12 choice questions, described by seven attributes such as symptom severity and risk of complications. The narrative version provided the same information as the table, but structured as a short paragraph. We assessed respondent experience of the two formats by examining completion time, drop-out rate, self-reported difficulty, and use of short-cuts (speeding, same choice for all questions, nonsense free-text responses). Respondents’ choices were modelled using heteroskedastic logistic regression to identify any differences in variability of responses, and mixed-effect logistic regression with an interaction term for format.
RESULTS: 1608 adults completed the survey (802 table, 806 narrative). We saw no significant differences in survey experience measures (all p>0.05). Models for the subsamples were consistent with the pooled model, but were a better fit than the pooled model after allowing for differences of variability (likelihood ratio test p<0.001), indicating differences in preferences and/or variability. Three of the attributes showed a small effect of format (interaction p-values<0.05). When shown both, 58% preferred the table format.
CONCLUSIONS: Although some differences in preferences were observed, these were minor, leaving the main DCE conclusions unchanged. These are reassuring findings, suggesting either format can be used. However, there may be stronger effects in groups typically under-represented in online DCEs (e.g. lower educational attainment), for whom the tabular format may even be a deterrent, leaving their voices missing in preference measures.
METHODS: The online DCE used 12 choice questions, described by seven attributes such as symptom severity and risk of complications. The narrative version provided the same information as the table, but structured as a short paragraph. We assessed respondent experience of the two formats by examining completion time, drop-out rate, self-reported difficulty, and use of short-cuts (speeding, same choice for all questions, nonsense free-text responses). Respondents’ choices were modelled using heteroskedastic logistic regression to identify any differences in variability of responses, and mixed-effect logistic regression with an interaction term for format.
RESULTS: 1608 adults completed the survey (802 table, 806 narrative). We saw no significant differences in survey experience measures (all p>0.05). Models for the subsamples were consistent with the pooled model, but were a better fit than the pooled model after allowing for differences of variability (likelihood ratio test p<0.001), indicating differences in preferences and/or variability. Three of the attributes showed a small effect of format (interaction p-values<0.05). When shown both, 58% preferred the table format.
CONCLUSIONS: Although some differences in preferences were observed, these were minor, leaving the main DCE conclusions unchanged. These are reassuring findings, suggesting either format can be used. However, there may be stronger effects in groups typically under-represented in online DCEs (e.g. lower educational attainment), for whom the tabular format may even be a deterrent, leaving their voices missing in preference measures.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
P38
Topic
Methodological & Statistical Research, Patient-Centered Research
Disease
No Additional Disease & Conditions/Specialized Treatment Areas