We appreciate these respondents' thoughtful comments on our report of what families remember about how loved ones died. Dying is more difficult, and what constitutes optimal care more uncertain, that most have acknowledged. Our findings illuminate potential targets for improvement, but we are all still learning how to achieve that improvement.

Reform efforts must test the extent to which citizens want every chance at survival, which Drs. Ainslie, Steinberg, and Palmer suggest may be a strong force, and the extent to which this desire can be tempered by the wish for a peaceful death and a meaningful dying. Dr. Steinberg is correct that some symptoms, as we measured them, may be unavoidable, and that some treatments, such as tube feeding, may be justified. At this early stage of this research field, our data are more descriptive than evaluative.

Some of our recent reports [1-3] show that prognoses often support hopefulness, even during what turns out to be the last few days of life. On the day before death, most hospitalized patients who died with congestive heart failure in SUPPORT had a prognosis better than 50-50 to live another 2 months. When all nine diagnoses are considered together, most patients in SUPPORT had prognoses at least that hopeful 1 week before death. As Dr. Steinberg points out, the comfortable conception that patients have a transition from "cure" to "care" does not fit the dying that most people actually face. Perhaps we need a "mixed management" model in which patients and families are helped to face dying even while pursuing every reasonable opportunity to live longer with their serious illnesses. Whether such a model best serves patients ought to be rigorously tested, a process that should also apply to reform suggestions, such as those offered by Mr. Hahn.

In fact, health care for the last phase of life urgently needs trials of innovations, probably trials that primarily use the methods of targeting, evaluation, and dissemination developed for continuous quality improvement. Because three quarters of persons in the United States die while receiving care that is financed by Medicare, the Health Care Financing Administration is in a uniquely pivotal position to shape, lead, and finance a serious research effort to learn how to improve care for the last phase of life.

For far too long, dying has been largely undescribed; for most dying patients, no one is really held accountable for developing care systems that work well. Our findings might stimulate innovations to shape systems of care that persons in the United States could count on in the last phase of life [4].