Do Framing and Attribute Number Influence Choice Consistency in a Preference-Based Value Assessment of Equity and Efficiency?
Author(s)
Mesfin Genie, PhD1, Surachat Ngorsuraches, PhD2.
1Newcastle Business School, The University of Newcastle, Newcastle, Australia, 2Auburn University, Aurbun, AL, USA.
1Newcastle Business School, The University of Newcastle, Newcastle, Australia, 2Auburn University, Aurbun, AL, USA.
OBJECTIVES: To investigate how attribute number and experimental manipulations (framing) affected choice consistency in a preference study of equity and efficiency, using health plans for newer Type 2 diabetes (T2D) treatments as a case study.
METHODS: A discrete choice experiment (DCE) was conducted involving 701 patients to elicit their preferences regarding health plan options for newer T2D treatments. The DCE included four experimental scenarios, i.e., Experiment 1: personal health outcomes only, which assessed preferences based on individual efficacy, risk, and cost; Experiment 2: equal health outcomes for self and others with poorer health; Experiment 3: unequal health outcomes between self and others with poorer health; and Experiment 4: equal health outcomes for self and others with poorer health, but incorporating an extra attribute. The study employed multinomial logit (MNL), heteroscedastic multinomial logit (HMNL), and heteroscedastic latent class logit (HLC) models to analyze the effects of these experimental conditions on choice consistency.
RESULTS: Both pooled baseline and separate MNL models revealed that higher efficacy and risk levels significantly increased and decreased the likelihood of selecting a health plan, respectively. Adding equal health outcomes for others (Experiment 2 VS 1) showed no significant impact on choice consistency (HMNL: coefficient = 0.0222, p = 0.6592). Having unequal health outcomes for self and others (Experiment 3 VS 1) resulted in a significant reduction in choice consistency (HMNL: coefficient = -0.2903, p < 0.001). Although the HMNL model showed no significant impact of an extra attribute (Experiment 4) on consistency relative to Experiment 2 (coefficient = -0.0824, p = 0.1210), the HLC model indicated reduced consistency for a majority class (67% class share; coefficient = -0.1468, p = 0.017).
CONCLUSIONS: Varying attribute numbers and framings significantly impacted choice consistency in a preference-based assessment of equity and efficiency attributes of health plans for newer T2D treatments.
METHODS: A discrete choice experiment (DCE) was conducted involving 701 patients to elicit their preferences regarding health plan options for newer T2D treatments. The DCE included four experimental scenarios, i.e., Experiment 1: personal health outcomes only, which assessed preferences based on individual efficacy, risk, and cost; Experiment 2: equal health outcomes for self and others with poorer health; Experiment 3: unequal health outcomes between self and others with poorer health; and Experiment 4: equal health outcomes for self and others with poorer health, but incorporating an extra attribute. The study employed multinomial logit (MNL), heteroscedastic multinomial logit (HMNL), and heteroscedastic latent class logit (HLC) models to analyze the effects of these experimental conditions on choice consistency.
RESULTS: Both pooled baseline and separate MNL models revealed that higher efficacy and risk levels significantly increased and decreased the likelihood of selecting a health plan, respectively. Adding equal health outcomes for others (Experiment 2 VS 1) showed no significant impact on choice consistency (HMNL: coefficient = 0.0222, p = 0.6592). Having unequal health outcomes for self and others (Experiment 3 VS 1) resulted in a significant reduction in choice consistency (HMNL: coefficient = -0.2903, p < 0.001). Although the HMNL model showed no significant impact of an extra attribute (Experiment 4) on consistency relative to Experiment 2 (coefficient = -0.0824, p = 0.1210), the HLC model indicated reduced consistency for a majority class (67% class share; coefficient = -0.1468, p = 0.017).
CONCLUSIONS: Varying attribute numbers and framings significantly impacted choice consistency in a preference-based assessment of equity and efficiency attributes of health plans for newer T2D treatments.
Conference/Value in Health Info
2025-05, ISPOR 2025, Montréal, Quebec, CA
Value in Health, Volume 28, Issue S1
Code
MSR145
Topic
Methodological & Statistical Research
Disease
SDC: Diabetes/Endocrine/Metabolic Disorders (including obesity)