Qualification of discordant responses in utility assessment

In many studies of utility assessment, the discordant response rate is significantly high. Discordant responses suggest inconsistency and, in turn, suggest inaccurate measurement of personal values that can lead to erroneous medical recommendations. The most common method of dealing with these responses is to exclude them from the sample statistics as incoherent or confused respondents. This paper proposes another perspective on discordant responses. In a recent study eliciting utility values for states of health that follow stroke, we observed a high rate of discordant responses. Closer examination of these discordant responses reveals that discordant responses are not all alike. Simple qualitative and quantitative views of these differences suggest that there may be information outside the concordant population of responses, which is lost by their exclusion. In an effort to understand the elevated discordant response rate, the effect of relaxing the defining boundaries of a discordant response was explored.