Cost-effectiveness analysis and other forms of decision analysis are becoming more common in the medical literature and are increasingly influential in the development of health policy. Nevertheless, many clinicians find it difficult to apply policies developed from these analyses to individual encounters with patients. We examine the assumptions behind these analyses and argue that the perspective they embody can make clinical strategies appear to be less risky in theory than they are at the bedside. We believe that this problem underlies the intuitive concern many physicians have about policy analyses and calls into question the value of these analyses in shaping clinical practice. These analyses aggregate the benefits and burdens of alternative interventions across different individual persons. Thus, overall population risk appears blunted, as it would in a diversified portfolio of stocks that react differently to financial forces or in a herd of cattle that react differently to veterinary interventions. The assumptions behind these analyses make sense if aggregate outcome is what matters, but not if one cares about each individual investment or animal. Because such aggregation tends to understate individual risk, when applied to human health policy, it may misrepresent the interests of patients and cannot be assumed to provide useful guidelines for decision making at the bedside.