During a one-year period all patients with cardiac arrest (CA) taken care of by three ambulances were studied. An incidence of 110 cardiac arrests/100,000 inhabitants/year was found. The majority of CAs affected the elderly and occurred during the day in their homes. The majority of CAs were witnessed but cardiopulmonary resuscitation (CPR) had been initiated by bystanders in only a few cases. The ambulance arrived within a mean time of 7.7 +/- 4.0 min. Forty-eight per cent of the CA patients showed ventricular tachycardia or ventricular fibrillation (VT/VF) on ambulance arrival. Patients with a prolonged ambulance delay showed a lower incidence of VT/VF than patients with a short delay. Patients in whom CPR had been initiated by bystanders showed a significantly higher incidence of VT/VF (67%) than unattended patients (45%). Bystander CPR was furthermore associated with an increased incidence of VT/VF in patients with prolonged ambulance delay. VT/VF was present at the time when the ambulance arrived in 86% of the CA patients who had received CPR from a bystander and were reached within 8 min by the ambulance.