Comparison of physicians' decisions regarding estrogen replacement therapy for menopausal women and decisions derived from a decision analytic model.

Decisions regarding estrogen replacement therapy were obtained from 50 physicians for 12 cases representing menopausal women with systematically varying levels of cancer risk, fracture risk, and symptom severity. Their decisions were compared with a decision analytic model for which each physician provided needed quantities--subjective probabilities, utilities of various outcomes, and weightings of the importance of the outcome categories. The majority of observed decisions were not to treat. By contrast, the decision analysis based on physician-provided estimates indicated that the optimal strategy was either to treat or a toss-up. Sensitivity analysis showed that these conclusions would hold over all possible utilities, over all plausible probabilities of cancer, and so long as symptom relief and fracture prevention were also considered as treatment objectives. The increased probability of early detection of cancer by regular follow-up was systematically incorporated into the decision analysis but apparently neglected in unaided clinical judgment, which follows the principle of minimizing the most important risk, regardless of its probability.