From July until September 2004, all deaths were registered prospectively in all departments of Freiburg University Hospital, Germany, a large teaching hospital with approximately 55,000 inpatient admissions per year. A retrospective chart review was done for all patients who died during this time period using a tool validated in two American and Australian projects. Main outcome measures were patients' identification as dying by medical staff, Do-Not-Resuscitate (DNR) orders, and the presence of comfort care plans. The cohort comprised 226 consecutive death events. Seven percent of patients had a written advance directive. DNR orders were available for 65% of patients and were entered into the charts on average 5.9 days prior to death. Thirty-eight percent of charts had evidence that staff recognized that the patients were dying. This prognosis was noted on average 3.8 days prior to death. According to chart notes, clinicians documented cancer patients as dying more frequently than patients with cardiovascular disease (P=0.029). In the chart entries, comfort care plans were completed fully for 14% and partially for 27% of patients. On average, comfort care plans were put in place nine days prior to death. Cancer patients had significantly more frequent comfort care plans than patients with cardiovascular diseases (P<0.001). In 59% of medical charts, there was no evidence of a comfort plan. Approximately one-third of dying patients received active life-sustaining treatment at time of death. These data highlight the need for systematic strategies to monitor patients' needs and to improve quality of care, especially during the last four days before death.