Protocol for a Health Preference Study on Digital Health Interventions: Integrating DCE and BWS to Capture Subject, Interaction, System, and Societal Value
Author(s)
Axel Christian Mühlbacher, PhD1, Andrew Sadler, M.Sc.2, Ann-Kathrin Fischer, M.Sc.1.
1Hochschule Neubrandenburg, Neubrandenburg, Germany, 2Gesellschaft für empirische Beratung mbH, Berlin, Germany.
1Hochschule Neubrandenburg, Neubrandenburg, Germany, 2Gesellschaft für empirische Beratung mbH, Berlin, Germany.
OBJECTIVES: The adoption of Digital Health Interventions (DHIs) is often hindered by fragmented evaluation approaches that overlook the multidimensional nature of value. This protocol outlines the design of a health preference study aimed at eliciting stakeholder preferences across four value dimensions (subject, interaction, system, and society) to inform comprehensive DHI assessment in Germany.
METHODS: The study applies prospective health preference design, combining Discrete Choice Experiments (DCEs) with Best-Worst Scaling (BWS) Type 1. Preferences will be elicited from two stakeholder groups: the general population (N = 4,000) and healthcare providers (N = 250-300). Attribute development was informed by a systematic review (97 studies), a review of systematic reviews (147 studies), qualitative interviews (N = 10), and expert consultations. This process yielded four decision models, each representing one value dimension: subject, interaction, system, and society. Across all models, 31 descriptive attributes were identified. In addition, two continuous attributes (out-of-pocket costs and individual daily time investment) were included in every model to ensure comparability across dimensions. Attributes were structured into four partial-profile DCEs, with each participant randomly assigned to one model. Each DCE includes nine choice tasks and one dominance test. A separate BWS1 exercise, based on a balanced incomplete block design, will assess attribute importance across the full set of 33 attributes. Data from both methods will be integrated using a joint likelihood estimation approach with shared anchor attributes to enable cross-model comparability.
RESULTS: The study will generate dimension-specific and cross-dimensional preference estimates, including trade-offs, willingness-to-pay, willingness-to-time-trade and subgroup differences. Analysis will use conditional logit, mixed logit, and latent class models.
CONCLUSIONS: This protocol offers a novel solution to model complex, multidimensional value structures in health preference research. It addresses key design and integration challenges in combining multiple DCEs with BWS for comprehensive DHI evaluation.
METHODS: The study applies prospective health preference design, combining Discrete Choice Experiments (DCEs) with Best-Worst Scaling (BWS) Type 1. Preferences will be elicited from two stakeholder groups: the general population (N = 4,000) and healthcare providers (N = 250-300). Attribute development was informed by a systematic review (97 studies), a review of systematic reviews (147 studies), qualitative interviews (N = 10), and expert consultations. This process yielded four decision models, each representing one value dimension: subject, interaction, system, and society. Across all models, 31 descriptive attributes were identified. In addition, two continuous attributes (out-of-pocket costs and individual daily time investment) were included in every model to ensure comparability across dimensions. Attributes were structured into four partial-profile DCEs, with each participant randomly assigned to one model. Each DCE includes nine choice tasks and one dominance test. A separate BWS1 exercise, based on a balanced incomplete block design, will assess attribute importance across the full set of 33 attributes. Data from both methods will be integrated using a joint likelihood estimation approach with shared anchor attributes to enable cross-model comparability.
RESULTS: The study will generate dimension-specific and cross-dimensional preference estimates, including trade-offs, willingness-to-pay, willingness-to-time-trade and subgroup differences. Analysis will use conditional logit, mixed logit, and latent class models.
CONCLUSIONS: This protocol offers a novel solution to model complex, multidimensional value structures in health preference research. It addresses key design and integration challenges in combining multiple DCEs with BWS for comprehensive DHI evaluation.
Conference/Value in Health Info
2025-11, ISPOR Europe 2025, Glasgow, Scotland
Value in Health, Volume 28, Issue S2
Code
PCR194
Topic
Patient-Centered Research
Disease
No Additional Disease & Conditions/Specialized Treatment Areas