Methods: Our sample consisted of 343 African American and 374 Caucasian veterans ages 50-79 receiving primary care in the Pittsburgh VA Health Care System or Philadelphia VA Medical Center. Patients were eligible to enroll if they indicated moderate or severe hip or knee pain, stiffness, and functional impairment as measured by the Western Ontario and McMaster Universities Osteoarthritis Index. To measure health risk preferences, subjects were asked to choose between two medications to treat knee pain. The first drug was a “lottery” drug, i.e., it had some probability of completely eliminating knee pain and some probability of having no effect. In the six choice scenarios, the probabilities that this drug worked were, in the order they were presented, 50%, 99%, 75%, 10%, 25% and 90%. The second drug involved no uncertainty; it always reduced knee pain by 50%. By eliciting responses to multiple scenarios with varying probabilities that the first drug was effective, we can assess whether subjects' responses were consistent. Subjects' risk preferences were classified according to the percentage of scenarios in which the lottery drug was chosen.
Results: Eighty-nine percent of respondents gave consistent responses to the health risk preferences instrument. Risk preferences varied across respondents; 18% were extremely risk averse, 16% were somewhat risk averse, 39% were risk neutral, 21% were somewhat risk loving and 6% were extremely risk loving. However, risk preferences did not differ by race (Jonchkheere-Terpstra Z=-0.6028, p= .55), income (Z=0.64, p= .52), education (Z=-0.44, p=.66) or age (Z=1.17, p=.24).
Conclusion: Surprisingly, given the prior findings among non-veterans, we found no variation in risk preferences by race or education. We conclude that variation in risk preferences does not appear to explain racial disparities in joint replacement surgery among veterans. The difference between veteran and non-veteran patient populations emphasizes that understanding of the role that risk preferences play in medical decision making remains incomplete.