Abstract Survival from out-of-hospital resuscitation depends on the strength of each component of the chain of survival. We studied, on the scene, witnessed, nontraumatic resuscitations of patients older than 17 years. The influence of the chain of survival and potential predictors on survival was analyzed by logistic regression modeling. From 1030 patients, 139 survived to hospital discharge. Three prediction models of survival were developed from the perspective of the different contributors active in out-of-hospital resuscitation: model I, bystanders; model II, first responders; and model III, paramedics. Predictors for survival (with odds ratio) were: in model I (bystanders): emergency medical service (EMS) witnessed arrest (0.50), delay to basic cardiopulmonary resuscitation (CPR) (0.74/min) and delay to EMS arrival (0.87/min); in model II (first responders): initial recorded heart rhythm (0.02 for nonshockable rhythm), delay to basic CPR (0.71/min and 0.87/min for shockable and nonshockable rhythms) and to defibrillation (0.89/min), and in model III (paramedics): need for advanced CPR (4.74 for advanced CPR not-needed), initial recorded heart rhythm (0.05 for nonshockable rhythm), and delay to basic CPR (0.77/min and 0.72/min for shockable and nonshockable rhythms), to defibrillation and to advanced CPR for shockable rhythms (0.85/min), and to advanced CPR for nonshockable rhythm (0.85/min). The area under the receiver–operator characteristic curve for model I was 0.763, for model II was 0.848, and for model III was 0.896. Of survivors, 50% had restoration of circulation without need for advanced CPR. Three survival models for witnessed nontraumatic out-of-hospital resuscitation based on the information known by bystanders, first responders and paramedics explained survival with increasing precision. Early defibrillation can restore circulation without the need for advanced CPR. When advanced CPR is needed, its delay leads to a markedly reduced survival.