Data collection

All of the data, such as digital ECG and compression signals recorded by the feedback sensor of the corpuls^3® automated external defibrillator, were stored on CF memory cards. These data were matched manually to the patient care reports from the national German resuscitation registry, in which all resuscitations are documented.

Data analysis

The ECG data collected and the compression signals from the feedback sensor were analysed using the ‘CPRAnalysis’ software program, which was developed in collaboration with GS Elektromedizinische Geräte G. Stemple GmbH, Kaufering, Germany. The program is based on Matlab^® R2016b (MathWorks, Inc.) and makes it possible to view resuscitation data as a whole and to annotate landmarks on the ECG. The data were converted into European Data Format (EDF) files using the established corpulsweb.review^® (2.0.4 Columbus) program.

Compression artefacts on the ECG were tagged manually by a single investigator (F.L.). The software was able to calculate pauses in compression and compression frequency, as described in many previous studies [3, 7]. Compression artefacts result from changes in thoracic impedance during chest compression [8].

Nonspecific ECG signals were analysed together with a team of experienced emergency physicians (A.B., P.B., R.L.) and were excluded if a clear analysis was not possible.

Every interruption that lasted more than 1.5 s between two compressions was defined as a pause in compression [7]. The longest pause and the chest compression fraction were analysed for each patient. Chest compression fraction is defined as the percentage of time with compression during the cardiac arrest, except for parts with ‘return of spontaneous circulation’ (ROSC).

Defibrillations were tagged automatically. The pre- and post-shock time was calculated as the time interval between the last compression before and the first compression after the defibrillation. ECG segments showing a potentially perfusable heart rhythm were defined as ROSC and therefore excluded from analysis.

Compression frequencies were calculated by the time delta between compressions. Sequenced compression artefacts with no interruptions longer than 1.5 s were grouped together in compression episodes. The mean compression frequency was calculated for every episode. In addition, the percentage of compressions between 100 and 120 compressions/min was detected in order to evaluate the providers’ adherence to the guidelines.

Compression depth was measured using the feedback sensor’s accelerometer. Depth analysis was thus possible only in the sensor-only CPR group and in the sensor-feedback CPR group. The compression bars on the accelerometer were matched to the compression artefacts on the ECG, so that transportation artefacts in the accelerometer signal were identifiable and were excluded from the analysis.

As primary endpoint we defined chest compression quality (chest compression fraction, average compression frequency, average compression depth). Secondary end points were defined as pauses in compression (longest pause, average pre-shock time, longest pre-shock time, average post-shock time, longest post-shock time), compression frequency (maximum episodic frequency, minimum episodic frequency, percentage of compressions 100-120/min) and compression depth (percentage of compressions with a minimum depth of 5 cm, percentage of compressions 5–6 cm in depth). The patients’ short-term survival (any ROSC, ROSC at hospital admission) was defined as a secondary end point as well (Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Primary and secondary endpoints.

https://doi.org/10.1371/journal.pone.0229431.t001

The completed datasets were exported into Excel files for further statistical analysis. The patients’ short-term survival was analysed as a secondary end point. For this purpose, clinical data, including ‘any ROSC’ and ‘ROSC at hospital admission’, were taken from the data in the German resuscitation registry.

Statistical analysis

The sample size calculation and the power analysis were based on the available literature at this time (08/2015), especially on the clinical study “Quality of out-of-hospital cardiopulmonary resuscitation with real time automated feedback: A prospective interventional study (Kramer-Johansen J, et al.)” [3]. This study defines the parameters average compression depth, average compression frequency and chest compression fraction as chest compression quality. Those parameters were therefore chosen for our sample size calculation. The multiple significance level has been set at 5% and for the comparisons with this calculated number of patients for each group a power over 80% was achieved.

Associations between categorical and continuous end points were tested using the Mann–Whitney U test and the Kruskal–Wallis test (in this case with estimates plus 95% confidence intervals using the Hodges–Lehmann estimator), and if a corresponding linear model showed normally distributed residuals, a search was made for such associations using the Welch test and ANOVA. Negative binomial regression was used to analyse associations between categorical variables and numerical variables as end points. In addition, associations between two categorical variables were examined using Fisher’s exact test, and associations between categorical end points and time-to-event end points were calculated using the log-rank test and the Kaplan–Meier method.

The following software programs were used for statistical analysis: IBM SPSS Statistics for Windows, version 25.0 (IBM Corporation, Armonk, New York, 2017) and R version 3.1.1 (R Foundation for Statistical Computing, Vienna, Austria, 2014; http://www.R-project.org).

Pauses in compression

In the present study, a chest compression fraction of over 80% was observed in the first study group performing conventional CPR with no Feedback-sensor, showing that raising the issue of the quality of chest compression in recent years has had an effect. Nevertheless, the chest compression fraction achieved by an above-average EMS was significantly improved by the introduction of a feedback sensor. The presence of the sensor alone evidently changed the behaviour of the paramedical staff, reducing interruptions in compression. The compression frequency also showed improved compliance with the guidelines merely through the presence of the feedback sensor. This confounding factor has been described as the ‘Hawthorne effect’—behavioural change due to an awareness of being observed [12]—and has previously been detected in connection with hand hygiene in hospital settings, for example [13]. The compression frequency was further improved when real-time feedback was activated. It can be concluded that the behaviour of emergency medical staff in carrying out CPR partly changed solely due to the fact that they were being observed during the study, and that this still left scope for further improvement.

The study also shows that introducing a real-time feedback system led to a significant decrease in the longest pauses in the cases evaluated. Prolonged pauses have been shown to be associated with decreased survival [14].

Chest compressions have to be interrupted in order to allow interpretation of the ECG and carry out defibrillation. Cheskes et al. described longer peri-shock and pre-shock pauses as being an independent predictor for poor survival in out-of-hospital cardiac arrest [15]. In the present study, both peri-shock pauses and pre-shock pauses were significantly shortened when the feedback sensor was introduced, still leaving potential for further minimisation.

Compression frequency

Many studies have shown that chest compressions are often poorly performed [4, 16]. The present data show that the compressions applied in the first group with no-sensor CPR were too fast; this has previously been associated with poorer survival [17, 18]. Use of the feedback sensor improved the compression rate, with further improvement when real-time feedback was activated. The mean compression rate reached the guideline recommendation of 100–120/min [2] in the third group when sensor-feedback CPR was activated. The percentage of compressions within the target frequency range also increased.

It is recommended that the chest compression frequency should be kept within 100–120/min in order to restore spontaneous circulation [19, 20]. In the present study, activation of the feedback system effectively reduced high compression rates, while inadequately slow compressions were not observed in any of the groups.

These results show that the configurations of real-time feedback systems, which focus mainly on compression depth, should be modified in order to avoid excessive compression frequencies. Previous studies have reported that there is a noticeable decrease in compression depth when compression rates increase [21]. In an earlier study, our own group showed that compression depth has a significant influence on the patient’s chances of survival [22], is highly accepted by health care providers and various other studies have shown that real-time feedback helps to achieve high-quality CPR [23, 24].

Compression depth

Our study showed that using a feedback sensor alone, without the feedback being activated, led to an adequate mean compression depth, which did not improve further when the system was activated. Kampmeier et al. previously found that the compression depth did not increase in the majority of cases even when real-time CPR feedback was used to achieve new guideline (2010) recommendation’s compression depth of 5–6 cm [25]. Other authors have concluded that variations in the anatomy of the chest, for example, may lead to incorrect compression depths [26, 27].

However, the proportion of adequately applied compressions increased substantially after activation of real-time feedback during the sensor-feedback CPR phase. More compressions within the explicit guideline recommendation of 5–6 cm were also detected with feedback, confirming the usefulness of real-time feedback in ensuring high-quality chest compressions. These findings are in accordance with those of previous simulator studies that have detected improvements in compression depth [28].

Rittenberger et al. have also shown that the quality of chest compressions declines in proportion to increasing complexity in the resuscitation scenario [29]. Supporting medical staff by providing a CPR feedback system to ensure high-quality chest compression improved the quality of CPR in the present study.

CPR feedback systems should be regarded as a tool for promoting adherence to the guideline and can be successfully used as such [30]. Real-time feedback systems are admittedly only used rarely in clinical practice, but it is recognised that assessing the quality of chest compression is important, even for purposes of comparing resuscitation data [31].

Limitations

In this study feedback sensor was only used in approximately 50% of cases. The main reason for this was that application of the sensor was omitted in the resuscitation scenario. This can be attributed to the large number of emergency units in the Osnabrück district and the fact that the feedback sensor had to be stored separately from the defibrillation electrodes. When a feedback system is being introduced, intensive motivational work is required to encourage medical staff to accept it. No differences in compliance were noted between the various emergency units, and there were no differences in overall compliance between sensor-only CPR and sensor-feedback CPR. Due to the design of the study it cannot be excluded that chest compression quality improved over time due to the fact that rescuers were aware of being part of a trial on resuscitation quality.

The quality of chest compressions has proven to be outcome-relevant [3, 32]. Despite differences in chest compression quality this study did not show significant differences between the three groups with regard to outcome. This can be attributet to the relatively small number of patients included in the, study which was not designed to show differences in survival. Since many factors play a role in survival of a cardiac arrest (e.g. gender, pre-existing medical condition, underlying cause, witnessed collapse) larger studies are needed to proof the clinical relevance of these findings [33].