Meeting Brochure and registration form      SMDM Homepage

Monday, 16 October 2006


Tamara Sims, MA1, Alan Garber, MD, PhD2, and Mary K. Goldstein, MD, MS2. (1) Stanford University, Stanford, CA, (2) Stanford University, VA Palo Alto Health Care System, Stanford, CA

An inherent issue with the Standard Gamble (SG) method is that it may not accurately capture health preferences of those who avoid taking risks or have difficulty understanding the SG. Previous studies employing the SG have found some respondents unwilling to accept any risk of death, regardless of the health state, or rating all health states the same. It is unclear whether this invariance is a reflection of their true preferences.

578 older adults completed FLAIR, a computerized utility assessment program, providing SG utilities for hypothetical health states of dependency in 6 Activities of Daily Living (ADLs) and a seriously ill condition (coma). Participants rating all ADL health states identically were considered invariant. Research assistants (RAs) recorded participants' explanatory comments.

117 (20.2%) participants gave invariant responses, of which 62 (10.7%) were unwilling to accept any risk of death (utility=1) for all ADL dependencies. 57 of these participants rated the seriously ill health condition lower, while 5 rated it 1.. Participants' explanations included willingness to accept any health condition (n=2) and unwillingness to risk death (n=2), while one expressed confusion during the task. 55 (9.5%) participants gave invariant ratings other than 1, of which 43 rated seriously ill lower than the ADL dependencies. 2 participants gave higher ratings for seriously ill, and 7 gave equal ratings (3 did not rate seriously ill). Participant comments included willingness to accept conditions (n=2), dislike for questions (n=1), and preference for death over the health states (n=2); 2 participants expressed confusion about this rating.

Of those unwilling to accept any risk of death for any ADL dependency, most took some risk for the severe health condition. Nevertheless, a few participants would not take any risk involving death suggesting that the SG method may not be an appropriate measure of preference for this small subset. Of those giving the same rating other than 1 for the ADL dependencies, most rated the severe health state as lower. Researchers using the SG method should consider including a severe health state rating and recording participant explanations to evaluate whether invariant responses represent actual preferences.

See more of Poster Session II
See more of The 28th Annual Meeting of the Society for Medical Decision Making (October 15-18, 2006)