Marion Danner1, Vera Vennedey2, Mickaël Hiligsmann3, Sascha Fauser4, Christian Gross2, Stephanie Stock2. 1. Institute for Health Economics and Clinical Epidemiology, University Hospital of Cologne (AöR), Cologne, Germany. Electronic address: marion.danner@uk-koeln.de. 2. Institute for Health Economics and Clinical Epidemiology, University Hospital of Cologne (AöR), Cologne, Germany. 3. Department of Health Services Research, CAPHRI School for Primary Care and Public Health, Maastricht University, Maastricht, The Netherlands. 4. Center for Ophthalmology, University Hospital of Cologne (AöR), Cologne, Germany.
Abstract
BACKGROUND: In this study, we conducted an analytic hierarchy process (AHP) and a discrete choice experiment (DCE) to elicit the preferences of patients with age-related macular degeneration using identical attributes and levels. OBJECTIVES: To compare preference-based weights for age-related macular degeneration treatment attributes and levels generated by two elicitation methods. The properties of both methods were assessed, including ease of instrument use. METHODS: A DCE and an AHP experiment were designed on the basis of five attributes. Preference-based weights were generated using the matrix multiplication method for attributes and levels in AHP and a mixed multinomial logit model for levels in the DCE. Attribute importance was further compared using coefficient (DCE) and weight (AHP) level ranges. The questionnaire difficulty was rated on a qualitative scale. Patients were asked to think aloud while providing their judgments. RESULTS: AHP and DCE generated similar results regarding levels, stressing a preference for visual improvement, frequent monitoring, on-demand and less frequent injection schemes, approved drugs, and mild side effects. Attribute weights derived on the basis of level ranges led to a ranking that was opposite to the AHP directly calculated attribute weights. For example, visual function ranked first in the AHP and last on the basis of level ranges. CONCLUSIONS: The results across the methods were similar, with one exception: the directly measured AHP attribute weights were different from the level-based interpretation of attribute importance in both DCE and AHP. The dependence/independence of attribute importance on level ranges in DCE and AHP, respectively, should be taken into account when choosing a method to support decision making.
BACKGROUND: In this study, we conducted an analytic hierarchy process (AHP) and a discrete choice experiment (DCE) to elicit the preferences of patients with age-related macular degeneration using identical attributes and levels. OBJECTIVES: To compare preference-based weights for age-related macular degeneration treatment attributes and levels generated by two elicitation methods. The properties of both methods were assessed, including ease of instrument use. METHODS: A DCE and an AHP experiment were designed on the basis of five attributes. Preference-based weights were generated using the matrix multiplication method for attributes and levels in AHP and a mixed multinomial logit model for levels in the DCE. Attribute importance was further compared using coefficient (DCE) and weight (AHP) level ranges. The questionnaire difficulty was rated on a qualitative scale. Patients were asked to think aloud while providing their judgments. RESULTS: AHP and DCE generated similar results regarding levels, stressing a preference for visual improvement, frequent monitoring, on-demand and less frequent injection schemes, approved drugs, and mild side effects. Attribute weights derived on the basis of level ranges led to a ranking that was opposite to the AHP directly calculated attribute weights. For example, visual function ranked first in the AHP and last on the basis of level ranges. CONCLUSIONS: The results across the methods were similar, with one exception: the directly measured AHP attribute weights were different from the level-based interpretation of attribute importance in both DCE and AHP. The dependence/independence of attribute importance on level ranges in DCE and AHP, respectively, should be taken into account when choosing a method to support decision making.
Authors: Darren E Stewart; Dallas W Wood; James B Alcorn; Erika D Lease; Michael Hayes; Brett Hauber; Rebecca E Goff Journal: BMC Med Inform Decis Mak Date: 2021-01-06 Impact factor: 2.796