Automated External Defibrillators Do Not Recommend False Positive Shocks Under the Influence of Electromagnetic Fields Present at Public Locations

Abstract

Electromagnetic fields (EMF) reduce the signal quality of electrocardiograms and may lead to the misinterpretation by automated external defibrillators (AED). We designed this investigation as a prospective study, with a randomized sequence of AED applications on healthy volunteers. We chose busy public places where public access defibrillation was possible as test locations. Strong EMF were sought and found at train stations next to accelerating and decelerating trains. The primary outcome variable was the absolute number of shocks advised in the presence of sinus rhythm by five commonly used AED in Austria. For data analysis, the statistician was blinded in regard to the AED models tested. Data analysis was based on a per protocol evaluation. Of 390 tests run, 0 cases of false positive results occurred (95% CI: 0–0.77). AED can be regarded as safe, even with the interference of EMF present at train stations.

In Brief

IMPLICATIONS: Our study clarified that electromagnetic interference at public train stations does not cause mock shockable heart rhythms in the presence of sinus rhythm.

Automated external defibrillators (AED) have been designed for use even by medically untrained bystanders in cases of cardiac arrest. Life-saving measures can be provided by either delivering electric shocks or by audible and/or visible prompts from the AED that help with the initiation of cardiopulmonary resuscitation. Once the patient has been connected to the device via self-adhesive gel electrodes, the AED analyzes the patient’s heart rhythm. It decides whether shockable heart rhythms are present, and recommends for or against defibrillation. The sensitivity to detect shockable heart rhythms correctly varies from 95% to 100%. Nonshockable rhythms are determined with a specificity of 98%–100% (^1–3).

The most important source for errors during the rhythm analysis is motion of the patient, but electromagnetic fields (EMF) have also been identified as weakening signal quality (^4,5). However, findings in the literature are contradictory. Previous studies have reported no significant impairment of AED performance in the presence of strong magnetic fields (⁶) and high-frequency EMF (⁷), whereas a previous trial by our study group (⁸) reported that AED gave false positive decisions in the presence of strong EMF at locations such as transformer stations. Interference did not differ significantly in parallel or perpendicular positioning toward the source of EMF (⁸). However, these locations have no public access, and therefore, the practical consequences are limited. There are no further data obtained in a publicly accessible setting using a human experimental model. We therefore tested the hypothesis that EMF in publicly accessible settings, such as train stations, have an influence on the performance of AED. The results of this study will have a number of implications for Public Access Defibrillation (PAD) programs in general.

METHODS

This investigation was designed as a prospective study with a randomized sequence of AED applications and was performed according to the Good Scientific Practice Guidelines.

Data collection and procedures were approved by the local ethics committee. Before enrollment, written informed consent was obtained from each of the subjects.

Study Sites

Studies have identified significant electromagnetic interference (EMI) with electrocardiogram (ECG) analysis of AED by static EMF at 16.7 Hz and alternating frequencies independent of the angle of the devices and the source of EMI (parallel or perpendicular) (^4,5,9). EMI present at train stations both above and below ground were considered to be especially relevant for PAD. The characteristics in means and strengths of the electric and magnetic fields at the selected study sites were quantified by repetitive measurements using an EMF analyzer (EFA-300™, Narda, Long Island, NY; measurement uncertainty ±3%; data output: peak values).

Study sites and EMI characteristics are listed here:

1. Train Station “Westbahnhof,” 15 kV AC, 16.7 Hz power supply;
Test 1: Without train present (predominantly electrical field)
Test 2: Accelerating loaded passenger train (electric locomotives 1016, 1116, 1047, and 1044; Siemens AG) (predominantly magnetic field)

2. Technical Support and Service Centre of the Vienna Transit Authority (Wiener Stadtwerke-Verkehrsbetriebe), 750 V DC power supply;
Test 3: Accelerating unloaded underground train (low-floor underground motor car “T” (Bombardier Transportation), which was provided with a direct pulse inverter generating a three-phase current (range 20–300 Hz) (predominantly magnetic field)
Test 4: Decelerating unloaded underground train (predominantly magnetic field).

Materials

To obtain a representative sample for Austria, we invited six AED manufacturers and distributors currently active in the Austrian market to support this postmarket study. Five manufacturers joined the study and provided AED, batteries, and electrodes. The AED tested are listed alphabetically:

1. AccessAED®, Access Cardio Systems, Concord, MA (after the study had been started Access Cardio Systems Inc. recalled AEDs of all models by November 3, 2004 and discontinued business)
2. CR+®, Medtronic, Minneapolis, MN
3. Fred Easy®, Schiller, Baar, Switzerland
4. HS1®, Philips Medical Systems, Amsterdam, Netherlands
5. Responder®, GE Healthcare, Chalfont St. Giles, United Kingdom

All AED tested were semiautomatic, CE (the “CE” mark certifies conformity with European Union standards; the sign has to be printed on the specific product for marketing) certified and licensed according to the medical product laws in Austria. Participating manufacturers were asked to provide routinely produced devices fully functioning, except without the ability to deliver an electrical shock or a plastic cover over the shock button. Details on safety have been presented elsewhere (⁸).

For simplicity, AED models will only be described with their numbers according to the list above. In compliance with good scientific practice, it is understood that the results of our investigation would be published whether positive or negative for individual companies. All source data remained with the principal investigator, and no permission of the sponsors had to be sought before manuscript submission.

Subject Selection

For participant recruitment, a notice was posted at the Medical University of Vienna inviting healthy people between 18 and 85 yr of age to take part in the study, and 25 subjects were invited to participate. The exclusion criterion for the subjects was a severe underlying illness and arrhythmic heart failure. All volunteers underwent a medical examination, including ECG analysis, before the investigation.

Randomization and Blinding

Every participant received a set of five opaque envelopes containing the names of the different devices. Before every test, randomization was performed by drawing one of the envelopes carried by each participant.

Because it was not possible to ensure correct blinding, participants and investigators were not blinded to the actual device tested, which seemed acceptable considering the aim of the study and the possible outcomes (shock advised/no shock advised). Statisticians only received AED model numbers and were blinded to the specific manufacturer and model.

Study Procedure

All investigators were trained to attach pads according to the current European Resuscitation Council’s and the manufacturers’ guidelines (^10,11). Subjects were asked to lie down in a supine position, parallel to the railway tracks. All further actions, including analyzing the underlying heart rhythm, were performed as if in a real cardiac arrest situation. The study procedure was conducted as reported in our previous publication (¹²).

Data Management and Outcomes

The following data were documented on the case report form: number of test, model of AED, underlying ECG rhythm, difficulties with ECG interpretation or abortion of analysis, and the final decision of the AED to deliver a shock or not. The principal investigator and one additional supervisor monitored correct data documentation on site.

The primary outcome variable was the absolute number of shocks advised in the presence of sinus rhythm. The secondary outcome was the number of impaired analyses caused by participants’ movements or electrode failure.

Statistical Analysis

Continuous data are described with median and interquartile range (IQR) because of nonnormal distributions. Categorical data are described with absolute frequencies and percentages. Corresponding 95% confidence intervals (CI) are given under the assumption of independent measurements, either two-sided for percentages unequal to zero or one-sided for percentages equal to zero. Statistical calculations were performed using the statistical package SAS (SAS Version 9, SAS Institute, Cary, NC). P values ≤0.05 were considered statistically significant.

RESULTS

There were 20 participants for study site “A” and 19 participants for study site “B” on the 2 study days. There were no dropouts before or during the investigation.

The 20 participants included at Site A consisted of 9 men and 11 women. The participants’ median age was 23.8 yr (IQR 21.8–25.1 yr), median height 175 cm (IQR 170–180 cm), median weight 67 kg (IQR 63–79 kg), and median body mass index was documented at 23 (IQR 21–25). Thoracic impedance measurements were all within the acceptable range provided by the AED manufacturers. Therefore, 100 tests could be performed at this location, resulting in 200 tests performed at Site A.

The 19 participants at Site B consisted of 8 men and 11 women. The participants’ median age was 24.6 yr (IQR 22.2–25.6 yr), median height 174 cm (IQR 165–179 cm), median weight 66 kg (IQR 59–78 kg), and median body mass index was documented at 22 (IQR 20–24). Thoracic impedance measurements were all within the acceptable range provided by the AED manufacturers. All 19 participants were tested in a parallel position to the source of EMI situations with five different AED. Therefore, 95 tests could be performed at this location, resulting in 190 tests performed at Site B.

Measurements of electric and magnetic fields using the EMF analyzer at Site A are presented in Table 1, and measurements of Site B are presented in Table 2.

Table 1: Site A: Characteristics of Electric and Magnetic Fields

Table 2: Site B: Characteristics of Electric and Magnetic Fields

Our primary outcome variable was the absolute number of shocks advised in the presence of sinus rhythm. Of 390 tests run, 0 cases of false positive results occurred (95% CI: 0–0.77), versus 390 true negative results.

As a secondary outcome, we recorded the number of impaired analyses caused by participants’ movements detected or electrode failure. Of 390 tests run, no electrode failure was detected, and motion was detected twice (0.51%; 95% CI: 0.06–1.84) by an AED in the presence of EMI. AED 1, AED 3, and AED 5 (n = 78 each) did not detect participant motion. AED 2 and AED 4 (n = 78 each) detected participant motion in one (1.28%) case each.

DISCUSSION

None of the five AED tested recommended shocks in the presence of EMI at our test locations. Furthermore, no electrode failures were reported. Mock participant movements, which could theoretically impair correct ECG analysis, were rarely reported. The available data were inconclusive. One study reported no negative interference inside a coal-burning or steam-generated electrical power plant (⁶) during the use of three AED models and an ECG-simulator. However, magnetic flux density was clearly lower, at a maximum of 1.558 μT, than during the underlying investigation. Kanz et al. (⁵) reported poor sensitivity and specificity of some AED models which showed higher magnetic flux densities varying from 0.7 to 3.7 μT. They concluded that most AED models are susceptible to EMI, especially in terminals with 15 kV 16 2/3 AC power supplies. A previous investigation published by our group (⁸) reported relevant impairment of proper ECG analysis at magnetic flux densities from 1.5 to 158 μT. Again, susceptibility of the device was mainly dependent on EMI frequency and the model itself.

Although Kanz et al. (⁵) reported minimization of interference in parallel positioning, we were not able to confirm their finding, and found no significant difference between parallel and perpendicular positioning (⁸). This is why we chose not to test in both positions.

As frequent electrode errors or mock patient movement have been reported by Schlimp et al. (⁴), we also expected those errors to occur in our test locations. However, only a few such events occurred during our investigation.

All these differences may be explained by the absence of confirmation of data derived from ECG simulators. Schlimp et al. (⁴) found human ECGs differently receptive to EMI in contrast to simulated ECG signals. Our data, collected in volunteers and in a simulated cardiac arrest situation, seem to be in clear contrast to previous findings. Major differences in susceptibility to EMI are dependent on the source of signal. This may explain some of our findings.

We also estimate that artificially generated ECG rhythms, such as those provided by training dummies, may not be suitable for high-quality interference testing. Furthermore, previous studies (⁵) used much older AED models with significantly older hardware and software technology than in the current devices. Schlimp et al. (⁴) tested AED made for professionals that may also be of a significantly different design. Furthermore, investigators chose to lift the AED devices to hold them in the direction of the electrical source. We tried to simulate cardiac arrest situations as close to reality as possible.

We chose AED models that were commercially available in Austria. Our study locations were selected because our previous findings indicated that some of these AED models are susceptible to EMI, especially in the presence of 16.7 Hz fields (^5,8). Furthermore, modern railways, including underground trains, use alternating frequencies to control acceleration and deceleration. Their generators produce strong magnetic fields and are also regarded as potentially harmful to ECG signal quality.

However, under the environmental conditions of our investigation, none of the tested AED was prompted to deliver a shock in the presence of sinus rhythm. We found that magnetic and electric interference at these publicly accessible sites much weaker when compared with our findings in a transformer station with restricted public access. As we used the same AED models as in our previous study without any modifications, it seems obvious that the nature of EMI, especially field strength, was the crucial factor of proper AED functioning (⁸).

This study has some limitations. Our investigation was designed in a human model, as no literature is available indicating the possibility of extrapolation of experimental data of ECG simulators. Therefore, we were not able to investigate anything other than false positive decisions of AED in the presence of sinus rhythm including possible electrode failure or participant movement. Blinding on the level of the active investigation was not possible, but investigators and participants were not able to influence the primary outcome.

We conclude that EMI at train stations does not interfere with proper ECG analysis of AED to simulate shockable heart rhythms. However, participant movements were reported at times, which may have caused a delay of shock admission and cardiopulmonary resuscitation initiation. In regard to the large potential benefit of PAD, these adverse events do not seem to be of clinical relevance. It remains unclear if false negative shock decisions could be a consequence of EMI.

ACKNOWLEDGMENTS

We are indebted to the Federal Institute of Technology and to the contributors Wolfgang Nitsche, PhD, MSc Engineering and Guenther Gamperl, MSc Engineering, who kindly supported us with technical expertise and measurements on electromagnetic field characteristics during data collection.