The use of 'do not resuscitate' (DNR) orders was analysed on a Dutch geriatric ward for a 4-month period. Of 148 admissions, 68 (58%) received a written DNR order. The use of DNR orders was significantly influenced by age (> 83 years) and the pre-arrest morbidity (PAM) index. PAM > 4 almost always resulted in a DNR order, and PAM score was the only significant contribution to DNR orders under regression analysis. The issuing of DNR orders by geriatric residents was compared with independent assessments by the other two health-care team members. In 50% of cases where significant comorbidity was zero, at least one of the team suggested a reason for a DNR order, with a mean of 1.9 reasons. The most commonly cited reasons were age (24%), depression (20%) and poor prognosis (18%). To evaluate non-patient-related factors involved in DNR decisions, we studied the involvement of patient or family in the decision, and the extent of agreement between health-care team members. Only 3% of patients and 24% of families were involved in the DNR decision. Disagreement with the residents' decisions was 20% for staff nurses and 17% for consultants. Physicians use factors besides comorbidity to make DNR decisions, and further study of such factors is necessary for the development of standardized DNR policies.