American physicians refer patients to hospice too late. Christakis and Lamont's research shows that physicians are poor prognosticians and tend to overestimate how long a person who is terminally ill will live. Most of the patients (65%) in their study had cancer. This is similar to the proportion (60%) found in hospice patients in a survey carried out in 1995, despite that cancer is not the leading cause of death in the United States.1 Seven percent of the patients referred to these hospices died within hours of admission. This 11th-hour referral pattern is at least partly due to physicians not recognizing the nearness of death. A patient is eligible for hospice care if he or she has an estimated life expectancy of 6 months or less. As the authors point out, the actual length of stay is usually less than 6 weeks. Thus, most patients come to hospice during a period of rapid physical change and often in crisis. And they don't live long beyond the crisis. At times of crisis, the immediate management of symptoms and relieving the family overshadow the need to address the emotional and spiritual issues of remembering, forgiving, and bringing to closure the issues of a person's life. Providing a physically comfortable death is a worthy goal; it reduces regrets among survivors. Yet, more time provides the opportunity for the dying person to participate directly in the process of validating the past and planning for the future and gives the family the chance to relish or repair bonds with the dying person. The National Hospice Organization has tried to educate physicians about how to predict appropriate entry points to hospice for various conditions.2 These guidelines should be incorporated into the general education of physicians. The authors' suggestion that prognostication should be done by a “disinterested” experienced physician hits near a common thread of late hospice referrals. Physicians may be reluctant to acknowledge that patients they know well are close to death. This can be compounded by a patient's and family's preference to keep hoping for the patient to live longer. Those of us who know our patients longer often become attached to them. We, too, hate to admit that death is near. I remember a woman in her 60s I was treating for metastatic breast cancer. She was admitted to the hospital with gastric bleeding that was thought to be unrelated to her cancer. I talked to her and her husband and was optimistic about the reversibility of the problem. Because I was trying not to scare her, I did not discuss the issues of advanced directives and resuscitation. That night, she went into shock, required intubation, and went to the intensive care unit. Her husband was devastated and angry that she had had such treatment. The next day, he and I together decided that no additional treatment would be given to prevent her death. He sat with her for more than 24 hours before she died. My desire to be optimistic prolonged her dying and added anguish to her husband. Physicians often rail against the denial of patients and their families. Yet, we are not immune. Decisions at the end of life are not just guided by physicians. There is a complex interaction when a physician recognizes and acts on an accurate prognostication, what the physician tells the patient and family, and what the patient and family actually hear. Christakis and Lamont have begun to tease out the factors involved in end-of-life predictions. In oncology, the end of chemotherapy usually signals the terminal phase. In other specialties, prognosis is much less demarcated. Physicians should better define landmarks or turning points in prognosis and begin to acknowledge these to themselves and their patients. Only then can physicians adequately guide patients through the dying process.