INVESTIGATING THE FRAMING OF RISK ATTRIBUTES IN A DISCRETE CHOICE EXPERIMENT: AN APPLICATION OF EYE-TRACKING AND THINK ALOUD

Monday, October 20, 2014
Poster Board # PS2-33

Candidate for the Lee B. Lusted Student Prize Competition

Caroline Vass, BSc, MSc1, Dan Rigby, BSc, MSc, PhD2, Stephen Campbell, BA, MA, PhD3, Kelly Tate, BSc4, Andrew Stewart, BSc, PhD4 and Katherine Payne, BPharm, MSc, PhD1, (1)Manchester Centre for Health Economics, The University of Manchester, Manchester, United Kingdom, (2)Department of Economics, The University of Manchester, Manchester, United Kingdom, (3)Centre for Primary Care, The University of Manchester, Manchester, United Kingdom, (4)School of Psychological Sciences, The University of Manchester, Manchester, United Kingdom
Purpose: To understand how the communication of risk in a discrete choice experiment (DCE) affects respondents’ decision-making heuristics and strategies. 

Method: An on-line pilot DCE was designed to understand the preferences of a purposive sample of female members of the public (recruited by posters in local cafes) for a breast screening programme described by two risk attributes (probability of detecting a cancer and risk of unnecessary treatment) and an out-of-pocket cost attribute, each with four levels. Two survey versions were used that varied how the risk attributes were presented as: (1) a percentage or (2) a percentage and icon array. Two approaches were used to understand if, and how these risk communication methods affected respondents’ decision-making heuristics and strategies:  (1) eye-tracking (2) retrospective think aloud cognitive interviews. Eye-movements were recorded as a series of co-ordinates 1,000 times a second. Eye-tracking data were analysed in terms of direction of motion and total visual attention (dwell time) to pre-defined areas of interest using descriptive statistics.  Immediately after completing the last choice question, respondents were asked a series of debriefing questions. Qualitative data were analysed using thematic analysis. DCE data were analysed using a conditional logit model. 

Result: In total, 35 female members of the public completed the DCE, with fifteen respondents completing the eye-tracking experiment. Respondents gave significantly more visual attention, suggesting information processing, to both risk attributes when risk was communicated with an icon array compared with using a percentage to present the risk. The mean dwell times were 6316 and 5043 milliseconds, respectively. Respondents completing the DCE with the icon array version exhibited significantly more upwards and downwards eye-movements (43% v 38% of saccades) suggesting calculations were made in line with expected utility theory. The eye-tracking data confirmed the self-reported attribute non-attendance as stated by respondents when asked the de-briefing questions with significantly lower (by almost 70%) mean dwell times on these attributes. The conditional logit model indicated both the probability of detecting a cancer and the risk of unnecessary treatment were significant attributes affecting women’s stated preferences. 

Conclusion: This pilot study demonstrates that eye-tracking can be used as a method to further understand the responses to a DCE and highlights the impact that risk attribute framing can have on respondents’ decision-making heuristics and strategies.