Schneiderman and coworkers' [1] finding of the lack of effect of offering advance directives may be interpreted to mean that physicians should not bother discussing these issues with their patients. As the authors note, the sample size lacked adequate power to feel confident about a negative finding; the treatment cost question alone would have required a sample eight times greater. More important, both the experimental and control groups were cared for in hospitals where unusual efforts were made to elicit patient preferences regarding cardiopulmonary resuscitation. As clinical experience and at least one study indicate [2], the writing of a do-not-resuscitate (DNR) order in a hospital frequently is accompanied by discussions about withholding or withdrawing many other kinds of therapy, including antiarrhythmics, blood products, and even antibiotics. Because more than 90% of the patients in the study by Schneiderman and colleagues had documented DNR discussions, it is likely that both study groups were having their treatment preferences elicited regardless of the presence of a formal advance directive. Indeed, a significant number in both groups chose to die at home [1]. The proper conclusion might be that offering advance directives did not add anything to patient care where the physicians were already communicating adequately with their patients. This study reinforces the view that the legal documents known as advance directives should be seen as opportunities for discussion, not as end points in themselves.

Finally, even if sample size had been adequate, I would not have been surprised to see no difference in treatment costs between experimental and control groups. The only randomized trial of hospice care and conventional care showed no difference in expense [3]. Even in the absence of a financial savings, we should still be eliciting patient preferences and offering palliative care options in order to promote respect for patients and appropriate care for the dying.