Valentine KD, Shaffer V, Hauber B. Eliciting preferences for cancer screening tests: comparison of a discrete choice experiment, a threshold technique experiment, and an individual difference measure. Presentation to be given at the 41st Annual Meeting of the Society for Medical Decision Making (SMDM); October 21, 2019. Portland, OR.

PURPOSE: To extend our understanding of the association between three methods of preference elicitation—discrete choice experiments (DCE), threshold techniques (TT), and a new individual difference measure—and to evaluate their ability to predict desire for cancer screening with an unbeneficial test.

METHODS: 300 undergraduate students completed the DCE, the TT, and the items for the Evaluating Screening Tests for Cancer: Attributes People Emphasize (ESCAPE) scale. The ESCAPE scale measures the importance of accuracy (false positives, false negatives), benefits (live saved), familiarity, cost, and personal preference matching for a cancer screening test. The measures of accuracy, benefit, familiarity, and cost were translated into attributes for participants to consider within the DCE and TT (See Table below). Next, participants were told about a hypothetical test screening for breast (women) or prostate (men) cancer with no mortality benefit and were asked if they wanted this test.

RESULTS: Risk threshold values for DCE and TT were largely uncorrelated (max r=.23 between DCE false negative threshold and TT false negative threshold). The largest association was between the ESCAPE cost subscale and the TT cost threshold (r=-.35, indicating that those who reported cost as being a more important consideration were more likely to not be willing to pay more for an increasingly beneficial test). Additionally, 54.8% of our sample stated they wanted to receive the unbeneficial screening test but none of the preference measures successfully predicted this choice (all ps>.14).

CONCLUSIONS: These three methods of measuring preference resulted in categorically different preference estimates for the same participant. Further, none of these preference elicitation methods were capable of predicting screening test choice. These results suggest a violation of the principle of procedure invariance that underlies models of rational choice. This calls into question the assumption underlying these techniques—that patients have stable, measurable preferences regarding attributes of health states previously unexperienced. This study suggests that we need to explore whether these potential differences can be explained by differences in the assumptions underlying the methods as decision making on the individual and aggregate levels may be informed by these and similar types of preference elicitation methods.

Share on: