The accuracy of decision-making of a semi-automatic defibrillator during cardiac arrest.

The accuracy of decision-making of a semi-automatic defibrillator was assessed at 57 cardiac arrests in 55 patients. The initial rhythm was ventricular fibrillation at 40 arrests and was correctly identified in 37 (sensitivity 92.5%). In the other 17 arrests initial rhythms were not ventricular fibrillation and 16 were correctly identified (specificity 94%). Continuous electrocardiographic recordings of the arrests were analysed in 8-12 s segments: of 432 segments showing ventricular fibrillation, 352 were correctly identified (sensitivity 81%); 3895 of 4139 segments showing rhythms other than ventricular fibrillation were correctly identified (specificity 94%). The percentage accuracy of detection of ventricular fibrillation varied from 0-100% (mean 91%) and for non-ventricular fibrillation from 51-100% (mean 92%). When 104 segments of ventricular fibrillation and 470 segments of non-ventricular fibrillation with cardiopulmonary resuscitation artefact were excluded from the analysis, a sensitivity of 91% and specificity of 96% were obtained. Thus the detection of cardiac arrest rhythms using a microprocessor based detection system for patients with cardiac arrest has a high sensitivity for ventricular fibrillation and specificity for non-ventricular fibrillation.