Why use a DCE for health preference research?
Health preference research requires a method that produces defensible trade-off estimates. DCEs are that method.
Why DCEs have become the standard method for health preference research and what makes them preferable to simpler alternatives.
Knowledge Base -> Foundations -> Health
Ben White, 07.07.2026
Why health preference research needs DCEs
Health preference research has a fundamental measurement problem. Simple rating or ranking questions tell you what patients prefer but not how much they prefer it or what they would sacrifice to get it. Without trade-off estimates, preference data cannot be used in benefit-risk frameworks or cost-effectiveness models.
DCEs solve this by forcing explicit trade-offs. Each choice set presents patients with realistic alternatives - better efficacy but more side effects, or a more convenient dosing schedule but higher cost. The pattern of choices across many such tasks reveals the underlying preference structure.
What regulators and HTA bodies expect
Regulators require preference methods that produce statistically valid, reproducible results. DCEs generate WTP estimates with confidence intervals, are grounded in random utility theory, and have a large peer-reviewed literature supporting their validity in health contexts.
The FDA's patient-focused drug development guidance specifically references quantitative preference methods. HTA bodies including NICE, IQWIG, and the Canadian Drug Agency all accept DCE evidence. Studies using simpler methods increasingly face questions about the validity of their estimates.
Setting up a health preference DCE in SurveyEngine
Step 1: Define the treatment attributes. Start with qualitative research to identify which treatment characteristics matter most to patients. Attributes must be clinically meaningful, understandable to patients, and variable enough to drive trade-offs.
Step 2: Define levels. Each attribute needs at least two levels spanning the realistic range of outcomes. Too narrow a range produces insufficient variation; too wide produces unrealistic scenarios that respondents dismiss.
Step 3: Generate a D-efficient design. The experimental design determines which combinations of attribute levels appear in each choice set. D-efficient designs minimise parameter uncertainty for a given number of choice sets.
Step 4: Pilot with target patients. Cognitive interviewing with 5-10 patients before main fieldwork identifies comprehension problems, unrealistic scenarios, and attribute framing issues that would otherwise contaminate the data.
Worked example - oncology patient preference for FDA submission
A phase III trial in a rare oncology indication produces a treatment with improved progression-free survival but higher rates of nausea and fatigue. The clinical development team needs to know whether patients consider the efficacy gain worth the tolerability cost.
A DCE with five attributes - progression-free survival, nausea incidence, fatigue severity, administration route, and monitoring frequency - is conducted with 200 patients. The results show the majority accept the tolerability costs for the observed efficacy gain, with significant heterogeneity by disease severity. The data is submitted to the FDA and incorporated into the label.
References
FDA (2019). Patient Preference Information - Voluntary Submission, Review in Premarket Approval Applications, Humanitarian Device Exemption Applications, and De Novo Requests, and Inclusion in Decision Summaries and Device Labeling. Guidance for Industry.
Bridges, J.F.P. et al. (2011). Conjoint analysis applications in health - a checklist: a report of the ISPOR Good Research Practices for Conjoint Analysis Task Force. Value in Health, 14(4), 403–413.
Planning a patient preference study? Contact SurveyEngine's health research team to discuss your study design and regulatory requirements.
Or Contact us at support@surveyengine.com — we're glad to help.