OBJECTIVES: To quantify the use of do-not-resuscitate (DNR) orders in a tertiary-care children’s hospital and to characterise the circumstances in which such orders are written. DESIGN: Retrospective study conducted in a 500-bed children’s hospital in Taiwan. PATIENTS: The course of 101 patients who died between January 2002 and December 2005 was reviewed. The following data were collected: age at death, gender, disease and its status, place of death and survival. There were 59 males and 42 females with a median age of 103 months (range 1-263 months). 50 children had leukaemias, and 51 had malignancies other than leukaemia. The t test and the chi(2) test were applied as appropriate. RESULTS: The study found that 44% of patient deaths occurred in the paediatric oncology ward; 29% of patient deaths occurred in the intensive care unit; and 28% of patients died in their home or at another hospital. Other findings included the following: 46 of 101 (46%) patients died after attempted cardiopulmonary resuscitation and 55 (54%) died with a DNR order in effect. The mean age at death was 9.8 years in both groups with or without DNR orders. CONCLUSIONS: From the study of patient deaths in this tertiary-care children’s hospital, it was concluded that an explicit DNR order is now the rule rather than the exception, with more DNR orders being written for patients who have been ill longer, who have solid tumours, who are not in remission and who are in the ward.