Chart review of patients who died in the hospital was used to describe the pattern of end-of-life decision-making and care for hospitalized dying patients and to propose a structured process of assessing the suitability of patients for palliative care. The setting was a large urban academic medical center, and the sample comprised 200 of 205 consecutive adult deaths during the first 4 months of 1996. The main outcome measures were identification of the patient as dying, do-not-resuscitate (DNR) orders, and comfort care plans. Charts of 72% of the patients had evidence that they were considered dying. DNR orders were in place for 77% of all the patients, and 46% had comfort care plans. Presence of a health care proxy was significantly associated with DNR orders and comfort care plans (P < 0.001). On average, comfort care plans were put in place 15 days after admission, as compared with an overall mean length of stay of 17 days. Substantial proportions of patients with comfort care plans continued to receive antibiotics (41%) and blood draws (30%). Only 13% of the patients on mechanical ventilation and 19% of those on artificial nutrition and hydration underwent withdrawal of these interventions prior to death. These findings suggest opportunities and challenges for improving practice patterns for hospitalized dying patients. We recommend several measurable objectives for evaluating end-of-life decision-making and care and propose the development of a goals of care assessment tool to guide appropriate transitions from life-sustaining treatment to comfort care and plan palliative services.