Risk Communication for Discrete Choice Experiments
Weissler E1, Wallace MJ2, Yang JC2, Brotzman LE3, Corriere MA3, Secemsky EA4, Sutphin J2, Johnson R2, Gonzalez J1, Tarver ME5, Saha A5, Gebben D5, Chen AL5, Malone M5, Farb A5, Zikmund-Fisher BJ3, Reed S2
1Duke University, Durham, NC, USA, 2Duke Clinical Research Institute, Durham, NC, USA, 3University of Michigan, Ann Arbor, MI, USA, 4Beth Israel Deaconess Medical Center, Boston, MA, USA, 5U.S. Food and Drug Administration, Silver Spring, MD, USA
OBJECTIVES: Respondents’ understanding of probabilistic risk levels in discrete-choice experiments (DCE) is essential to elicit valid benefit-risk preferences. Few risk communication studies have investigated visual risk-communication approaches in DCE choice tasks wherein it is often necessary to consider multiple risks between options. We sought to compare risk-communication approaches in a DCE context and to evaluate whether performance varies by subjective and objective numeracy and graph literacy.
METHODS: To convey two independent probabilistic treatment-related outcomes and two time points (4 attributes) in a DCE, we designed six choice-task layouts incorporating combinations of separate or integrated event risk information, use or omission of graphical icon arrays, and use of horizontal or vertical formats. Respondents ages 40-75 were recruited by a survey vendor with oversampling of Black/African Americans based on the target population of interest, peripheral artery disease. The primary outcome was the percent correct of 12 training questions that required identifying and comparing risk information.
RESULTS: There were 2,242 respondents (mean age 59.8±10.4; 21.9% Black/African American). The primary outcome differed significantly among the six risk-communication approaches, from 70.0% correct with vertical integrated icon arrays to 78.3% with separate horizontal icon arrays (p<0.001). Separate rather than integrated risk levels performed best regardless of vertical (76.7%) or horizontal (78.3%) format, or omission of icon arrays (76.7%). In multivariable analysis, integrated risk information was associated with 0.63 fewer correct responses out of 12 (p<0.001; 95%CI: -0.82, -0.45). When graphical icon arrays were omitted, performance was worse among respondents with lower graph literacy and better among respondents with higher graph literacy (interaction, p=0.007).
CONCLUSIONS: These results support use of separate icon arrays to optimize understanding of risk information, particularly for individuals with low graph literacy. To the best of our knowledge, this is one of few assessments of how the principles of risk communication apply in the DCE context.
Conference/Value in Health Info
Value in Health, Volume 25, Issue 6, S1 (June 2022)
Stated Preference & Patient Satisfaction
Cardiovascular Disorders, Medical Devices, Surgery