Duke 2008 | ASHE | Equity and Efficiency in Health and Healthcare

Background: Stated-choice studies increasingly are used to estimate willingness to pay (WTP) for healthcare interventions and outcomes. WTP estimates only are valid and unbiased if respondents actually are willing to pay the amount indicated by their choices. In situations where there is uncertainty about choices, the likelihood that subjects will choose any of the constructed alternatives must be taken into account in estimating societal WTP. Aquiescent responses, or so-called 'yea-saying', may occur where subjects' responses overstate their actual willingness to accept the offered alternatives relative to the status quo. A possible explanation for such behavior is the availability heuristic. The information provided in the survey and possible framing effects may make the intervention appear more salient and attractive than it would appear in an actual market decision. Alternatively, nay-saying may occur when patients are unwilling to consider alternatives to their current treatment.

Objective: The purpose of this study is to test whether yea-saying bias exists for colorectal cancer (CRC) screening tests and to estimate the size of the implied bias in WTP estimates by comparing uptake estimates for patients with those of physician proxies and with observed clinical levels.

Methods: A web-enabled stated-choice survey of patients and physicians in Canada and the United States elicited preferences for attributes of CRC screening tests. Each CRC test was described by nine attributes: process, pain, preparation, frequency, follow-up, complication risk, sensitivity, specificity, and out-of-pocket cost. Respondents answered a forced-choice question for two constructed alternatives, followed by choice between the alterative selected in the first question and the alternative of no screening. 1087 US and 501 Canadian respondents participated and 100 physicians responded in both countries. Physicians were asked to indicate their patients' preferences. Using bivariate probit regression, responses were modeled using main effects for the scenario attributes and interaction terms with the optout dummy. Uptake and uptake-adjusted societal WTP was calculated for common CRC screening tests.

Results: For all groups the most important attribute was sensitivity. Other key attributes were those related to test performance and the testing process. Predicted patient CRC screening uptake estimates exceeded both those of physician proxies and observed levels in the US and Canada. Thus the upward bias in patient uptake likelihood resulted in patient WTP estimates that overstate societal values.

Conclusion: Because societal WTP values account for both the effects of perceived utility and the probability of testing, they are strongly influenced by yea-saying. Designing stated-choice surveys that accurately elicit tradeoffs between costs and health attributes is a serious research challenge. Yea-saying may result in unrealistic measures of uptake that results in overstated societal WTP. Designing surveys that allow for a realistic opt-out or no-purchase choice, minimizing inadvertent framing effects that might encourage acquiescent responses, modeling opt-out choices carefully, and comparing results to revealed behavior can help minimize potential bias in such studies.