Sunday, October 23, 2011
Grand Ballroom AB (Hyatt Regency Chicago)
Poster Board # 36
(DEC) Decision Psychology and Shared Decision Making

Brian J. Zikmund-Fisher, PhD1, Holly Witteman, PhD1, Mark Dickson, MA1 and Ethan A. Halm, MD, MPH2, (1)University of Michigan, Ann Arbor, MI, (2)University of Texas Southwestern, Dallas, TX

Purpose: Unlike printed materials based on the average patient, computer-based patient decision aids can tailor risk estimates to individual patient characteristics. We sought to empirically evaluate whether such personalization of risk estimates would affect the perceived relevance of the provided risk statistics.

Method: We surveyed 3315 people age 55 and older from an online panel and asked them to imagine being diagnosed with carotid artery stenosis. Participants viewed a 4 page segment of an online decision aid designed for patients considering carotid artery surgery to reduce stroke risk. The decision aid section provided estimates of the risk of stroke with medical therapy only from 30 days through 5 years in multiple pictograph formats and then compared these risks to the risks of stroke and death following carotid artery surgery. While the risks with medical therapy were consistent for all participants, we randomized participants to view estimates of the risks with surgery that were either (a) the risks of the average patient or (b) tailored based on 6 risk factors (including female gender). We then compared treatment intentions, perceived risk, and ratings of the personal applicability of the information between groups.

Result: Preferred treatment selections did not differ based on personalization, and perceptions of operative risks increased when more risk factors were present as expected. Surprisingly, however, we also observed significant variations in perceived information relevance based on risk factors. While those at average or higher risk (2 or more risk factors) rated the information as more applicable to them than the untailored group (Mean Rating=3.36 vs. 2.60 on a 7 point scale, p<0.001), respondents in the personalized group who were lower than average risk (0 or 1 risk factors) rated the risk information as significantly less personally applicable when compared to participants in the untailored condition (Mean Rating=2.32 vs. 2.60, p<0.001).

Conclusion: Tailoring risk estimates can increase perceived relevance among those who have multiple risk factors, but it can lower perceived relevance among lower-risk individuals. Participants responded as if the act of stating that no risk factors applied to them meant that the resulting risk estimates also were not personally applicable. This effect, if replicated in patient populations, casts doubt on the advisability of providing tailored risk estimates to low risk populations.