We read with interest Hanson and colleagues review [1] on clinical interventions and care at the end of life. It is encouraging to see that many studies have increased the documentation of advance directives through both patient education and physician education. It is distressing, however, that despite our proven ability to document a patient's wishes, little or no proven benefit is enjoyed by a patient in his or her last days of life.

One possible explanation for our inability to change clinical outcomes through better documentation of advance directives has been the language in which the documents are written. Despite painstaking efforts to reduce hospitals' do-not-resuscitate policies to their bare minimums, housestaff and attending physicians alike often find the policies cumbersome and difficult to interpret. These policies are made more obtuse by language that attempts to cover every contingency but is too vague to serve as a guide in any particular situation. Appropriate patient care remains an issue that often frustrates housestaff and leads to unwanted and costly treatments.

A possible solution to this discordance between the patients' wishes and the care they actually receive is to shift the emphasis from do-not-resuscitate orders to a "values history" [2] or "goals of treatment" form. Our current focus on the technical aspects of medical care (such as intubation and defibrillation) sidesteps the larger issue of the patient's diagnosis and prognosis in the context of treatment options, spiritual values, and quality of life. Documentation of such a discussion could provide a more meaningful and useful guide to managing the patient's end-of-life care. This documentation may help the patient better understand his or her disease, help the physician better understand the patient's wishes and desires, and lead to greater concordance between patient wishes and the care received.

Hanson and colleagues also point out the lack of effect on pain and suffering, use of life-sustaining treatments, and overall costs despite improved documentation of patient preferences. Authors seem to assume that patients would prefer fewer life-sustaining treatments and less pain and suffering and that if physicians adhered to their patients' end-of-life preferences spending in the final months of life would be more prudent. Perhaps we are already meeting patient preferences for goals of therapy. Perhaps patients, when given the choice, will opt for some pain and suffering if given the possibility of extending life. This may require the use of certain life-sustaining treatments, even though the physician may personally disagree with the wisdom of providing such care.