STUDY OBJECTIVE: To document the characteristics and effectiveness of cardiopulmonary resuscitation (CPR) at non-fatal heroin overdose events in Melbourne, Australia. METHODS: A retrospective analysis of a computerised database of ambulance attendance records at non-fatal heroin overdose cases for the period 1/12/1998 to 31/7/2000 was undertaken. MAIN OUTCOME MEASURES: The main outcome measure was the rate of patient hospitalisation. The rate of CPR administration at heroin overdose cases was also examined, along with characteristics of the attendance, such as the age and sex of the overdose case, the relationship of person providing CPR to the overdose case as well as the location, time and date of the event. RESULTS: CPR was administered prior to ambulance arrival in 579 heroin overdose cases (9.4% of total heroin overdose cases attended) between 1/12/98 and 31/7/2000. A greater proportion of female overdose cases were administered CPR than males and CPR administrations were evenly distributed across attendances occurring in private and public locations. Bystander administration of CPR prior to ambulance attendance resulted in a significantly lower rate of heroin user hospitalisation (14.5%) compared to cases where bystander CPR was not administered (18.8%). CONCLUSIONS: While CPR administration prior to ambulance attendance at heroin overdose events is relatively uncommon (especially compared to out-of-hospital cardiac arrest), such administration was associated with a statistically significant improvement in clinical outcomes in cases of non-fatal heroin overdose. These findings suggest that the provision of CPR training to people likely to come into contact with heroin overdose events may be an effective strategy at minimising consequent overdose-related harm.