Purpose: The purpose of this study was to investigate whether discrete choice (binary) questions yield different health discount rates than the ping-pong method for assessing average group responses.
Methods: Using a Web-based time-preference survey, we compared the two methods of elicitation for two health vignettes. The survey presented two hypothetical health states: a migraine scenario and an ill-health scenario derived from an EQ-5D health state. We recruited graduate students and outpatients at the University of California, San Diego to complete the survey. Consenting graduate students were emailed a link to the survey Web address where they completed the survey from their personal computers. Outpatients completed the survey on a computer in the waiting room of a clinic. The survey was randomized such that each participant received either one ping-pong vignette or two discrete choice vignettes to ensure an approximately equal response burden for all subjects. All participants completed demographic questions at the end of the survey. We piloted the survey prior to commencing recruitment to evaluate the feasibility and effectiveness of the Web-based application. We calculated a median response for each group using the individual responses from the pingpong approach and a regression from the discrete choice approach. We used these medians and the discount rate formula to determine discount rates for each method and each health vignette presented. We compared methods using 95% confidence intervals for the discount rates calculated.
Results: 161 students and 86 patients completed the survey. Discount rates for the ping-pong method of elicitation were 74% (95% CI 62%-91%) and -8% (95% CI -16% to -2%) for the migraine and EQ-5D ill-health scenarios respectively. Discrete choice questions yielded discount rates of 41% (95% CI 18%-72%) and 0.4% (95% CI 6%-10%) for the two respective scenarios. The study showed a statistically significant difference in discount rates between health vignettes, but did not show a difference in discount rates calculated from the two different elicitation methods for either vignette.
Conclusions: These findings suggest that further time preference research can be efficiently conducted using the discrete choice survey method and that a Web-based application can provide an effective means of conducting time preference research. Results are consistent with literature stating that subjects have different discount rates for different health states.
Candidate for the Lee B. Lusted Student Prize Competition