End-of-life communication is crucial because most U.S. hospitals implement cardiopulmonary resuscitation (CPR) in the absence of do-not-resuscitate directives (DNRs). Despite this, there is little DNR utilization data to guide the design of communication-training programs. The objective of this study was to determine DNR utilization patterns and whether their use is increasing.

A retrospective database analysis (2000–2005) of DNR data for 206,437 patients, the entire patient population at Memorial Sloan-Kettering Cancer Center (MSKCC), was performed.

The hospital recorded, on average, 4,167 deaths/year. In 2005, 86% of inpatient deaths had a DNR, a 3% increase since 2000 (p < .01). For patients who died outside the institution (e.g., hospice), 52% had a DNR, a 24% increase over 6 years (p < .00001). Adult inpatients signed 53% of DNRs but 34% were signed by surrogates. The median time between signing and death was 0 days, that is, the day of death. Only 5.5% of inpatient deaths had previously signed an outpatient DNR. Here, the median time between signing and death was 30 days.

Although DNR directives are commonly utilized and their use has increased significantly over the past 6 years, most cancer patients/surrogates sign the directives on the day of death. The proximity between signing and death may be a marker of delayed end-of-life palliative care and suboptimal doctor–patient communication. These data underscore the importance of communication-training research tailored to improve end-of-life decision making.