Automatic External Defibrillators for Public Access Defibrillation: Recommendations for Specifying and Reporting Arrhythmia Analysis Algorithm Performance, Incorporating New Waveforms, and Enhancing Safety
A Statement for Health Professionals From the American Heart Association Task Force on Automatic External Defibrillation, Subcommittee on AED Safety and Efficacy
- cardiopulmonary resuscitation
- AHA Medical/Scientific Statements
- myocardial infarction
Automatic external defibrillators (AEDs) that accurately analyze cardiac rhythms and, if appropriate, advise/deliver an electric countershock were introduced in 1979. AEDs are widely used by trained emergency personnel (emergency medical technician [EMT]-paramedics, EMT-B’s, EMT-I’s, and first responders, such as firefighters and police personnel). In such hands, AEDs have proved accurate and effective and have become an essential link in the “chain of survival” as defined by the American Heart Association.1
A logical extension of the AED concept is “public access defibrillation” or widespread distribution and use of AEDs by nonmedical, minimally trained personnel (eg, security guards, spouses of cardiac patients).2 Public access defibrillation poses unique challenges. AEDs must be simple to operate, because in many cases the operator is a first-time user with minimal training. The device must accurately diagnose lethal arrhythmias under unfavorable conditions that may degrade performance. It could be misused, either inadvertently (eg, the patient is conscious and breathing) or deliberately. Safety must be emphasized, and the risk of injury to patient and rescuer minimized. An existing standard for AED construction and performance recognizes the challenges inherent in the various potential uses of AEDs.3
The purpose of this statement is to recommend strategies to the appropriate regulatory agencies to assist in evaluating
The accuracy of the arrhythmia analysis algorithms incorporated into AEDs
New or alternative defibrillation techniques, especially waveforms
The safety of AEDs when used by minimally trained lay rescuers (public access defibrillation).
This is a consensus document, reflecting the views of the members of the American Heart Association Task Force on Automatic External Defibrillation, its Subcommittee on AED Safety and Efficacy, and the AED Manufacturers’ Panel. This document is intended to supplement existing documents concerning AEDs, such as ANSI/Association for the Advancement of Medical Instrumentation (AAMI) DF39,3 the AHA Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiac Care,1 and the AHA Textbook of Advanced Cardiac Life Support.4 All AEDs, whether public access or not, should meet similar algorithm performance specifications.
Demonstrating Accuracy of the Arrhythmia Analysis Algorithm
An arrhythmia analysis algorithm should respond in one of two ways to an electrocardiographically recorded rhythm: it should advise (or in a fully automated system, deliver) a shock, or it should advise no shock (and not deliver a shock). An AED can also notify the operator of suspected artifact in the electrocardiographic (ECG) signal. Similarly, cardiac rhythm disturbances can be divided into three broad categories (Table 1⇓):
Shockable rhythms: lethal rhythms that terminate in the patient’s death unless defibrillation is delivered very quickly. These rhythms include coarse ventricular fibrillation (VF) and rapid ventricular tachycardia (VT) and are always (VF) or almost always (rapid VT) associated with a pulseless, unresponsive patient.
Nonshockable rhythms: benign (or even normal) rhythms that must not be shocked, especially in patients with a pulse, because no benefit will follow and deterioration in rhythm may result. Nonshockable rhythms include normal sinus rhythm, supraventricular tachycardias, sinus bradycardia, atrial fibrillation and flutter, heart block, idioventricular rhythms, premature ventricular contractions, and other rhythms accompanied by a palpable pulse and/or occurring in a conscious patient. To maximize safety in the event of misapplication of the device/electrodes, asystole is included in this group. The AHA Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiac Care discourage shocks for asystole.1
Intermediate rhythms: Rhythms for which the benefits of defibrillation are limited or uncertain. These include fine VF (associated with pulselessness and low survival rates) and VT that does not meet all criteria for inclusion in the shockable VT rhythm category.
Various cardiac rhythms are categorized in Table 1⇑.
Reporting Algorithm Performance
The task force divided arrhythmias into three categories: shockable, nonshockable, and intermediate (Table 2⇓). Patients with shockable rhythms (VF, rapid VT) potentially receive the greatest benefit (survival) from defibrillation at essentially no risk. High sensitivity for AED analysis (Table 3⇓) is required for this group. Patients with nonshockable rhythms derive no benefit from defibrillation and are at maximum risk. For reasons of safety, asystole is included in this group. High specificity is required. Patients with intermediate rhythms are unlikely to derive benefit or be at risk from defibrillation, making performance requirements inappropriate. Reporting arrhythmia analysis algorithm specificity or sensitivity is sufficient for this group.
Performance during developmental testing is an indication of what to expect during validation. During developmental testing of automatic rhythm analysis systems, the performance goal should be met or exceeded. This maximizes chances of equaling or exceeding the goal during validation testing, which should be performed with at least the minimum sample size per category given in Table 2⇑. (A sample consists of data required to make a single shock/no-shock decision.) The sizes selected in Table 2⇑ reflect a balance between reasonable confidence in performance and realistic limits on data available to demonstrate it. These are minimum sample sizes and may be exceeded. Data may be acquired from prehospital or in-hospital events. The size and geometry of the electrodes used to acquire the data should be reported.
For each category, the observed test results must equal or exceed the performance goal. For each rhythm category, the exact single-sided 90% lower confidence limit should be calculated, based on test results. This process will give a 90% probability that the actual performance is greater than the lower confidence limit calculated.
Table 2⇑ provides an example of calculation of lower confidence limit for observed performance equal to performance goals for each rhythm with specified performance goals.
Differences in ECG data acquisition preclude the development of a common (single) database against which every AED arrhythmia analysis algorithm could be tested. Therefore, the task force recommends that AED manufacturers report the performance of arrhythmia analysis algorithms of their own devices to the Food and Drug Administration (FDA), using the format in Table 2⇑. Appropriate electronic and/or hard copy documentation should be available for inspection on request.
The data used for algorithm development must be different from the data used for testing and validation. Validation of performance should be obtained in both the presence and absence of artifacts likely to be encountered in field use.
The signal characteristics of the data acquisition system used to gather the validation data set should be specified (bandwidth, phase characteristics, dynamic range).
The waveforms may include a discharge deflection and a postdischarge recovery period, making the timing of playback into a separate device critical, so that the device being tested is not required to analyze these discharge artifacts introduced during recording. If additional non-ECG signals are used (eg, respirometer, impedance detector), their acquisition characteristics should also be specified.
Algorithms may examine different rhythms recorded from the same patient. However, there can be only one sample of each specific rhythm from each patient.
Because many ECG rhythm segments may be classified differently by different physicians, the task force recommends that classification of segments as shockable, nonshockable, or intermediate require agreement among at least three qualified expert reviewers of cardiac arrest rhythms. Rhythm segments on which reviewers fail to reach 100% agreement can be classified, but the expert disagreement should be reported. The reviewers should use ECG criteria on which they have previously agreed. They should reach a consensus on the distinction between fine VF (an intermediate rhythm that should be shocked) and asystole (which should not be shocked) by employing the same criteria used by the AED being tested.
Effects of Artifacts
In real world situations in which AEDs are used, it is inevitable that artifacts will corrupt ECG data to varying degrees, potentially degrading specificity and sensitivity. Manufacturers should determine the effects of various artifacts, with emphasis on diagnosis of shockable and nonshockable rhythms. The effect of artifacts on diagnosis of intermediate rhythms is of less concern.
The most commonly encountered artifacts are motion artifacts, which are typically generated by cardiopulmonary resuscitation, agonal breathing or seizures, handling of the patient, and transport by stretcher and vehicle. Pacemaker stimuli can also interfere with algorithm performance. Static electric fields (commonly present in dry environments) exacerbate these artifacts.
Because there is no standard database of ECG signals or noise signals for testing AED algorithms, manufacturers should determine how to test their devices for reasonable performance in the presence of noise and specify in detail how this testing was done.
Alternative Waveforms for Defibrillation
The two presently accepted waveforms for transthoracic defibrillation in the United States are the damped sinusoidal waveform (Edmark, Lown, Pantridge) and the truncated exponential waveform. Alternative wave forms for transthoracic defibrillation such as biphasic waveforms, in clinical use in the former Soviet Union, have been introduced in the United States. Studies in animals have demonstrated the superiority of various alternative waveforms.5 6 7 More recently three studies in humans in the United States have reported comparisons of biphasic and monophasic waveforms.8 9 10 Patients undergoing provocative electrophysiological studies and implantation of an automatic implantable cardioverter-defibrillator received transthoracic biphasic waveform rescue shocks. These studies suggest that biphasic or other alternative waveforms may achieve equivalent shock success rates at substantially lower energies (or higher success rates at the same energies) when compared with damped sinusoidal waveforms. This in turn suggests the prospect of a reduction in size and weight of AEDs (an important consideration for public access) and/or higher success rates than can be presently achieved using available waveforms.
At present there are no published data on prehospital transthoracic defibrillation using alternative waveforms. The absolute success rate of any waveform for termination of VF will be lower in the prehospital setting (in which VF is often present for a prolonged period before shocks are administered) than in the rapid-shock environment of the electrophysiology laboratory or the coronary/intensive care unit. However, there is no a priori reason to suspect that the relative advantage of alternative waveforms over monophasic waveforms will not be maintained in prehospital use if such an advantage is demonstrated in hospital. In fact, in vitro studies by Jones et al11 and a study of intact dogs by Walcott et al12 have suggested that the superiority of biphasic waveforms may actually be increased over monophasic waveforms when shocks are delivered after longer durations of VF.
It is the consensus of the task force (with the exception of one manufacturer) that if alternative waveforms for transthoracic defibrillation are convincingly demonstrated to be equivalent or superior to standard waveforms in the electrophysiology laboratory or other hospital or prehospital settings, they should be provisionally approved for use in AEDs, pending acquisition of prehospital data. Performance of waveforms tested in the electrophysiology laboratory or other in-hospital or prehospital settings and incorporated into AEDs should be monitored as part of a postmarket surveillance program designed to carefully observe total system performance of these devices in their intended settings.
The task force recommends the following as a minimum standard for demonstrating equivalence of an alternative waveform versus standard waveforms: the upper boundary of the 90% confidence interval (with 5% in each tail) of the difference between standard and alternative waveform efficacy must be ≤10%, which permits a slight (5%) chance of acceptance of a waveform that is >10% less effective than the standard waveform. Similarly, the task force suggests that to demonstrate superiority of an alternative waveform over standard waveforms, the upper boundary of the 90% confidence interval of the difference between standard and alternative waveforms must be <0% (ie, alternative is greater than standard). If the standard waveform efficacy equals 90%, and the true (or hypothesized) alternative waveform efficacy is 95%, approximately 52 patients per group would be required to demonstrate equivalence, and 471 patients per group would be required to demonstrate superiority with a power of 0.9. These sample sizes are based on statistical tests of equivalence of new treatments described by Blackwelder.13
Postmarket surveillance should be maintained on any device introduced for in-hospital, emergency medical services, or public access defibrillation. It is important to document both failures and successes; reporting only problems or failures may give a distorted picture of performance. A well-designed postmarket surveillance study should allow observation of the total performance of an AED and its effectiveness in its intended environment. As part of such studies, the task force recommends that manufacturers obtain and submit to the FDA sufficient field data to demonstrate that AEDs incorporating an alternative waveform maintain satisfactory performance when used in the target population for AEDs. The performance reports should state the measured sensitivity and specificity for rhythm categories as well as the upper and lower bounds of the 80% confidence interval (10% per tail). This will allow accurate, prospective tracking of actual field performance.
Public access AEDs will be used by minimally trained personnel. The potential for misuse is high: use of AEDs is inappropriate in persons who are conscious and breathing or persons who are in true cardiac arrest but are receiving artifact-generating cardiopulmonary resuscitation during analysis of the rhythm. Deliberate misuse of an AED with an intent to cause harm may also be encountered.
To overcome these potential problems, the task force recommends that AEDs be specifically designed to prevent injury in the event of misuse. Innovative features that enhance safety are encouraged, such as voice chips that deliver a series of prompts to a rescuer who is opening or activating an AED (eg, “Shake the victim. If he or she groans or moves, do not attach the electrodes—call the emergency number.”). Alternatively, after the arrhythmia analysis algorithm has been satisfied, an AED might administer an unpleasant but low-strength “wake-up” shock; if the algorithm diagnosis was incorrect and the patient was not in cardiac arrest but merely in a deep sleep or intoxicated, such a preliminary shock would stimulate the patient to move or respond, alerting the rescuer not to deliver a defibrillation-strength shock. These suggestions are intended as examples only; other innovations/approaches may be even more effective.
The task force also encourages the design of devices that enhance rapid and effective deployment in conjunction with local emergency medical services, integrating AEDs into the AHA chain of survival. This can be accomplished through advanced communication technology. For example, AEDs could be designed to automatically activate the local emergency medical services system when the device is removed from its holder or its cover is opened. Other approaches and innovations may be even more effective.
These recommendations are presented to enhance the safety and efficacy of AEDs intended for public access. The task force recommends that manufacturers present developmental and validation data on their own devices, emphasizing high sensitivity for shockable rhythms and high specificity for nonshockable rhythms. Alternative defibrillation waveforms may reduce energy requirements, reducing the size and weight of the device. The highest levels of safety for public access defibrillation are needed. Safe and effective use of AEDs that are widely available and easily handled by nonmedical personnel has the potential to dramatically increase survival from cardiac arrest.
Automatic External Defibrillation Task Force
Myron L. Weisfeldt, MD, Chair
Richard E. Kerber, MD
R. Pat McGoldrick
Arthur J. Moss, MD
Graham Nichol, MD
Joseph P. Ornato, MD
David G. Palmer, Esq
Barbara Riegel, DNSc
Sidney C. Smith, Jr, MD
AED Safety and Efficacy Subcommittee
Richard E. Kerber, MD, Chair
Lance B. Becker, MD
Joseph D. Bourland, EE, PhD
Richard O. Cummins, MD, MPH
Bram D. Zuckerman, MD
Mary B. Michos, RN, Fire Chief
Joseph P. Ornato, MD
Roger D. White, MD
AED Research Subcommittee
Joseph P. Ornato, MD, Cochair
Barbara Riegel, DNSc, Cochair
Alfred P. Hallstrom, PhD
Graham Nichol, MD
AED Manufacturers Panel
Carlton B. Morgan, Heartstream, Inc.
William L. Post, Hewlett-Packard Company
John E. Kuphal, Laerdal Medical Corporation
Donald E. Brodnick, Marquette Electronics, Inc.
Robert A. Niskanen, Physio-Control Corporation
Kenneth F. Olson, SurVivaLink Corporation
Gary A. Freeman, Zoll Medical Corporation
The authors gratefully acknowledge the assistance of Patricia Bowser, AED Task Force Coordinator.
“Automatic External Defibrillators for Public Access Defibrillation: Recommendations for Specifying and Reporting Arrhythmia Analysis Algorithm Performance, Incorporating New Waveforms, and Enhancing Safety” was approved by the American Heart Association Science Advisory and Coordinating Committee in October 1996. It is being published concurrently with the Annals of Noninvasive Electrocardiology and Biomedical Instrumentation and Technology.
A single reprint is available by calling 800-242-8721 (US only) or writing the American Heart Association, Public Information, 7272 Greenville Avenue, Dallas, TX 75231-4596. Ask for reprint No. 71-0104. To purchase additional reprints: up to 999 copies, call 800-611-6083 (US only) or fax 413-665-2671; 1000 or more copies, call 214-706-1466, fax 214-691-6342, or To make photocopies for personal or educational use, call the Copyright Clearance Center, 508-750-8400.
- Copyright © 1997 by American Heart Association
Weisfeldt ML, Kerber RE, McGoldrick RP, Moss AJ, Nichol G, Ornato JP, Palmer DG, Riegel B, Smith SC Jr. American Heart Association Report on the Public Access Defibrillation Conference, December 8-10, 1994: Automatic External Defibrillation Task Force. Circulation. 1995;92:2740-2747.
Association for the Advancement of Medical Instrumentation. DF39, Automatic External Defibrillators and Remote-Control Defibrillators. Arlington, Va: AAMI; 1993.
American Heart Association. Textbook of Advanced Cardiac Life Support. Dallas, Tex: AHA; 1994.
Kerber RE, Spencer KT, Kallok MJ, Birkett C, Smith R, Yoerger D, Kieso RA. Overlapping sequential pulses: a new waveform for transthoracic defibrillation. Circulation.. 1994;89:2369-2379.
Bardy GH, Gliner BE, Kudenchuk PJ, Poole JE, Dolack GL, Jones GK, Anderson J, Troutman C, Johnson G. Truncated biphasic pulses for transthoracic defibrillation. Circulation. 1995;91:1768-1774.
Greene HL, DiMarco JP, Kudenchuk PJ, Scheinman MM, Tang ASL, Reiter MJ, Echt DS, Chapman PD, Jazayeri MR, Chapman FW, Ahmed M, Johnson JL, Niskanen RA, and the Biphasic Waveform Defibrillation Investigators. Comparison of monophasic and biphasic defibrillating pulse waveforms for transthoracic cardioversion. Am J Cardiol. 1995;75:1135-1139.
Bardy GH, Marchlinski FE, Sharma AD, Worley SJ, Luceri RM, Yee R, Halperin BD, Fellows CL, Ahern TS, Chilson DA, Packer DL, Wilber DJ, Mattioni TA, Reddy R, Kronmal RA, Lazzara R, for the Transthoracic Investigators. Multicenter comparison of truncated biphasic shocks and standard damped sine wave monophasic shocks for transthoracic ventricular defibrillation. Circulation. 1996;94:2507-2514.
Jones JL, Swartz JF, Jones RE, Fletcher R. Increasing fibrillation duration enhances relative asymmetrical biphasic versus monophasic defibrillator waveform efficacy. Circ Res. 1990;67:376-384.
Walcott GP, Melnick SB, Chapman FW, Jones JL, Ideker RE. Comparison of monophasic and biphasic waveforms for external defibrillation in an animal model of cardiac arrest and resuscitation. J Am Coll Cardiol. 1995;25:405A. Abstract.