Incorporating Best-Worst Scaling into Attribute Selection for Stated Preference Research: A Study Based on Preference for Choosing Antidepressants
Author(s)
Ying Tao, Bachelor1, Shimeng Liu, Ph.D.1, Yanfeng Ren, Master1, Fei Xie, Master2, Yingyao Chen, Ph.D.1;
1Fudan University, Shanghai, China, 2Shanghai Mental Health Center, Shanghai, China
1Fudan University, Shanghai, China, 2Shanghai Mental Health Center, Shanghai, China
Presentation Documents
OBJECTIVES: To explore incorporating the Best-Worst Scaling object case (BWS-1) into attribute selection process of stated preference research, and to inform researchers on attribute selection methodology.
METHODS: A BWS-1 questionnaire was generated using a balanced incomplete block design. Taking depressed and depression-prone groups as the research object, this study conducted an online survey to elicit preferences for choosing antidepressant drugs. Data analysis was performed using counting analysis and conditional logit model to judge attribute importance. The heterogeneity of preference was also explored.
RESULTS: The BWS-1 results yielded by the counting and modelling approach showed high consistency. Among the 13 attributes, the top three attributes influencing the preference for choosing antidepressants in depressed patients were liver/kidney injury, headache/dizziness, and somnolence/insomnia, while the last three attributes were decreased appetite, sexual dysfunction, and weight change, respectively. The top three attributes influencing the preference for choosing antidepressants in depression-prone respondents were liver/kidney injury, recurrence rate, and headache/dizziness, while the last three were decreased appetite, weight change, and duration of medication, respectively. Combining qualitative research and BWS-1 results, six attributes were ultimately included in the subsequent stated preference study as follows: the risk of liver/kidney injury, the risk of headache/dizziness, the risk of gastrointestinal adverse effects, sleep disturbances (somnolence/insomnia), remission rate, and monthly out-of-pocket costs.
CONCLUSIONS: BWS-1 can provide valid and reliable evidence for attribute selection for stated preference research. However, it is not recommended to solely rely on BWS-1 results to determine the importance of attributes. Research questions, decision-making environment and stakeholder opinions should be taken into account, to improve the rigour and transparency of study design.
METHODS: A BWS-1 questionnaire was generated using a balanced incomplete block design. Taking depressed and depression-prone groups as the research object, this study conducted an online survey to elicit preferences for choosing antidepressant drugs. Data analysis was performed using counting analysis and conditional logit model to judge attribute importance. The heterogeneity of preference was also explored.
RESULTS: The BWS-1 results yielded by the counting and modelling approach showed high consistency. Among the 13 attributes, the top three attributes influencing the preference for choosing antidepressants in depressed patients were liver/kidney injury, headache/dizziness, and somnolence/insomnia, while the last three attributes were decreased appetite, sexual dysfunction, and weight change, respectively. The top three attributes influencing the preference for choosing antidepressants in depression-prone respondents were liver/kidney injury, recurrence rate, and headache/dizziness, while the last three were decreased appetite, weight change, and duration of medication, respectively. Combining qualitative research and BWS-1 results, six attributes were ultimately included in the subsequent stated preference study as follows: the risk of liver/kidney injury, the risk of headache/dizziness, the risk of gastrointestinal adverse effects, sleep disturbances (somnolence/insomnia), remission rate, and monthly out-of-pocket costs.
CONCLUSIONS: BWS-1 can provide valid and reliable evidence for attribute selection for stated preference research. However, it is not recommended to solely rely on BWS-1 results to determine the importance of attributes. Research questions, decision-making environment and stakeholder opinions should be taken into account, to improve the rigour and transparency of study design.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
PCR206
Topic
Patient-Centered Research
Disease
SDC: Mental Health (including addiction), STA: Personalized & Precision Medicine