Drs. Mueller and Kisloff raise questions about the causes of regional variations in spending and intensity. Regional differences in lack of health insurance and consequent differences in illness burden at age 65 years cannot explain the differences in spending and intensity observed in our study because we grouped regions according to differences in spending that were independent of regional differences in health status. Dr. Kisloff interprets our study as an attack on academic medical centers and technological advances. On the contrary, our study shows that low-spending regions provide just as much of the proven technologically advanced care (such as percutaneous coronary interventions) as high-spending regions. We also showed that regional differences in spending are due to differences in the use of the hospital and the intensive care unit as sites of care and in the frequency of visits, consultations, and associated tests and procedures. We believe that these latter differences are in large part related to differences in the relative availability of beds and specialists for the population served by the specific hospital in which patients receive their care (1). The data in Table 2 show that teaching hospitals in the lowest-spending regions provide more conservative care but, as was shown in our papers, achieve equal quality and outcomes. These findings indicate that conservative practice is perfectly consistent with academic practice and technological innovation.

Dr. Kisloff also questioned whether it was reasonable to generalize results from a study of 3 conditions to the "vast panoply" of other conditions. We chose patients with hip fracture, colorectal cancer, and heart attack because they provide insight into diverse care systems (orthopedics, oncology, and cardiology). These patients, however, had many other chronic conditions, and because we followed patients for up to 5 years, our analyses are likely to reflect the benefit (or harm) from more aggressive care for these other conditions as well.

Dr. Levin and Mr. Fernandes focus on the policy implications of our study. Mr. Fernandes suggests that a prescription drug benefit could lead to improved outcomes for elderly persons. We agree but would argue that such a plan should require the creation of a prescription drug database that would support the kinds of population-based analyses we carried out. We also agree with Dr. Levin that even the lowest-spending quintile may not provide an appropriate benchmark and have argued (2), as he does, for a demonstration project that would engage selected medical centers in efforts to improve both the quality and efficiency of care and provide benchmarks for the rest of the United States.