This is the first study to investigate the no-flow fraction, perishock pause, and other objective quality parameters during BLS using eight different available public access defibrillators in a simulator setting.
As expected, our study found higher NFF values in BLS scenarios that used an AED compared to those that did not. Furthermore, we found extremely high NFF values in these scenarios with a high variability between the different models. In all cases, the perishock pause was always much longer than advised by the current resuscitation guidelines. However, when the rescuer optimized the sequence of actions and acted immediately before or after the voice prompt started (Tables 2 and 4), the NFF and the perishock pause both were significantly reduced, with minor variability.
The chest compression fraction has been identified as a cardiac resuscitation quality parameter that is associated with the survival rate [7],[8]. A small study from the PAD trial evaluated the quality of BLS and found very high values for NFF (0.52 to .83) when lay rescuers performed BLS and used an AED[9]. Two studies have shown that chest compressions during prehospital life support are interrupted in approximately half of the time [10],[11]. The current international resuscitation guidelines demand high quality chest compressions with minimal interruptions. Well trained teams can provide life support with NFFs as low as 0.24 [12]. In these cases, a defibrillator capable of manual defibrillation was used. However, an observational study showed that the NFF during in-hospital BLS with the use of an AED can be as low as 0.30 when nurses repetitively perform the BLS/ AED algorithm in a BLS training course [13].
The NFF during scenario 2 (asystole) was comparable to that of scenario 1 and was significantly higher than the value for BLS without an AED for all AEDs. Recent studies have found that only between 24 and 40 % of patients with OHCA are in a shockable rhythm when the ambulance arrives [14]. At the time of collapse, the proportion of patients who are in a shockable rhythm is higher, but there is still a considerable number of patients who are in a non-shockable rhythm. For those patients, the use of an AED resulting in a high NFF is harmful and may increase mortality.
The NFF was lower for six devices (VF) and 5 devices (asystole) during the scenarios with an optimized sequence of actions. The NFF was similar in all devices and was below .40, except for one AED in asystole. This set of scenarios can be compared with trained lay rescuers or healthcare professionals after practice with a specific AED. The first set of scenarios involves untrained lay rescuers using a public access defibrillator while strictly adhering to the guiding voice prompts. It is concerning that the use of commonly available AEDs, which are designed to be used by lay rescuers, might lead to poor quality life support, with chest compressions being delivered less than 30 % of the time. The wide difference between the standard use and the optimized sequence of actions implies that the human-machine interface of the devices is a major problem; manufacturers should expend considerably more effort on the development of optimal user guidance to ensure the best possible performance, even when used by lay rescuers who have never before observed a patient in cardiac arrest and who perhaps have never received BLS training.

Perishock pause
Survival to hospital discharge is substantially more likely when the first documented rhythm is shockable rather than nonshockable [15]. When a defibrillator is used, the perishock pause has been identified as an independent predictor of survival [16]. A substantial delay in providing chest compressions has been observed in the post-shock period with AED-equipped first responders [17]. As Cheskes et al. reported, a pre-shock pause of ≥ 20 seconds and a perishock pause of ≥ 40 seconds are predictors of worse survival after shockable cardiac arrest. Based on a log-linear model, any five seconds increase in the perishock pause interval decreases the rate of patient survival to hospital discharge by 14 % [16].
Therefore, the international guidelines for resuscitation advise a perishock pause of less than 5 seconds. This recommendation was not fulfilled in any of the AEDs tested in the first set of scenarios with strict voice prompt adherence. We noted that some AEDs do not permit chest compressions during the charging period, resulting in an almost fourfold variation in the perishock pause between the different devices. This variation in duration may be responsible for the variable success in patient resuscitation. In 2004, Snyder et al. alluded to the problem of the AED-determined variation in hands-off times[18]. However, the manufacturers may have implemented an algorithm, which does not prompt the rescuer to provide chest compressions for some seconds during charge and then take hands off for the shock as this may confuse lay persons. We need further studies to find out whether real lay rescuers are able to follow such an algorithm with minimised perishock pause.
The perishock pause was reduced in nearly all of the tested AEDs when the optimized sequence of actions (in the second set of scenarios) was used. Two devices met the demands of the current guidelines, implying that there is significant potential for improving user guidance. The necessary changes are small and include optimization of the device algorithms and voice prompts. Determining the best practices for AED voice prompting will be the subject of a follow-up analysis of the present dataset.
Time until first chest compression
A lone lay rescuer following the AHA guidelines for BLS would call an ambulance, fetch an AED (if available), and return to the patient to start life support [2]. However, when switching on the AED and following the voice prompts, the beginning of chest compressions is delayed for up to nearly 2.5 minutes. This stands in sharp contrast to what is taught in courses on cardiac resuscitation around the world: the check – call – compress paradigm, in which chest compressions are started as early as possible, and any unnecessary interruptions are avoided.
A retrospective analysis of OHCA cases found no benefit of bystander CPR for patients receiving the first shock between 1 and 5 minutes after collapse, whereas survival was much higher for patients receiving bystander CPR when the first shock was applied after more than 11 minutes [19]. However, even for the patients receiving the first shock within 5 minutes, this study does not answer the question whether shock first (and delayed start of chest compressions) or CPR first (and later application of the shock) is superior with regard to the survival rate. However, modern AED devices should guide the lay rescuer to soonest possible defibrillation, while assuring a short time until the first chest compression.
Duration of rhythm analysis
The duration of rhythm analysis depends solely on the technical specifications of the respective device. This value varied by about 100 % between the devices. However, the absolute difference between the fastest and slowest analysis duration was only 8 seconds. One potential future improvement is hands-on rhythm analysis during chest compressions. Making a serious effort to develop optimized devices may lead to shorter time intervals. However, a much larger effect on shortening the NFF among layperson rescuers could be achieved by inventing devices with improved usability and voice prompts.
Wide variation of CPR quality between different AED models
When a lay rescuer provides BLS and uses an AED model with which he or she is not familiar, the CPR quality depends critically on the AED device. We found extensive variations in important parameters such as the NFF and the perishock pause. Most clinical studies investigating the effects of the use of AED were not stratified by the AED model used. Future studies, especially large resuscitation registries, should always consider the defibrillator model. Even the firmware version might provide important additional information, as different versions may differ in terms of user guidance and thus lead to variations in the quality of BLS.
In this study, a trained person used the AEDs. However, lay rescuers use these machines in real cardiac arrest situations. As they are under high stress and may be afraid of using an AED, the voice prompts should be clear and precise. Some models provide more detailed voice prompts than others (see Additional file 1). We do not know whether more details result in better performance and less mistakes when the AED is used by untrained real lay rescuers and whether this may compensate the longer and more frequent interruptions. The intention of very precise instructions may be to assure a shock when used by a heterogeneous group of lay rescuers. However, if a lone rescuer places an emergency call and fetches an AED before returning to the patient, 2 minutes or more may already have been expired. At this time, soonest beginning of chest compressions is important and this should be taken into account when developing AEDs and when elaborating user guidance. Further research should be done to study the quality of BLS when lay rescuers use AEDs.
Limitations
This study was performed in a simulator centre, and a BLS instructor evaluated the devices. The quality of BLS in real cases may vary from that in this study.
With some AEDs, lay rescuers may not understand some voice prompts or they could experience usability problems that could not been found in the present study but may result in even worse BLS performance.