OBJECTIVE: To determine whether the source of preference scores has an impact on the cost-effectiveness of osteoporosis interventions. METHODS: Three groups of subjects aged > or =50 years–199 women without fractures and 183 women with osteoporotic fractures-were studied at two major medical centers. Medical history and comorbidity data were obtained from review of medical records. Health status was measured using the Medical Outcomes Study SF-36. Two preference-classification systems (i.e., quality of well-being scores estimated from SF-36 subscales and the Health Utilities Index) were also used. Preferences for current health and for hypothetical health states were assessed using a time tradeoff and implemented with a computer-based utility instrument (U-Titer). Wilcoxon’s rank-sum and signed-rank tests were used to compare preferences for current health among women with osteoporotic fractures with 1) directly assessed preferences for osteoporosis health states delineated by outcome descriptions and 2) preference scores obtained from the preference-classification systems. The potential impact of the source of the preference scores was estimated using a Markov state-transition model. RESULTS: The preference scores for hypothetical osteoporosis health states of the non-fracture subjects were approximately 50% lower than those of the women who had actually experienced the health state. Differences of this magnitude would change the estimated cost-effectiveness of a 15-year intervention (which for approximately $280 per year prevents hip fracture about as well as hormone-replacement therapy) from $25,000 per QALY gained when non-fracture subjects’ preferences were used to $94,000 per QALY gained when fracture subjects’ preferences were used. Preferences estimated using the Health Utilities Index and those directly measured in fracture subjects using the time tradeoff did not differ significantly. CONCLUSIONS: The Health Utilities Index preference-classification system may provide an efficient and inexpensive alternative to direct utility assessment in this patient group. However, there are important differences in the valuation of health states by women who have experienced osteoporotic fractures compared with women who have not. Cost-utility analyses based solely on fracture patients’ preferences for osteoporotic health states may undervalue prevention.