THERAPY AND PREVENTION

DEFIBRILLATION

Efi of an automated external defibrillator in the

cacy
management of out-of-hospital cardiac arrest:
validation of the diagnostic algorithm and initial
clinical experience in a rural environment
KENNETH R. STULTS, B.S., PA-C, DONALD D. BROWN, M.D., AND RICHARD E. KERBER, M.D.

ABSTRACT Automatic external defibrillators (AEDs) may have advantages over manual defibrilla-
tion in managing prehospital cardiac arrest, particularly in rural communities. We conducted a two-part
evaluation of a commercially available AED. We first established the diagnostic accuracy of the AED's
rhythm recognition algorithm by challenging it with 205 cardiac arrest rhythms previously recorded
from actual patients in the field. The AED demonstrated 100% specificity and 92% sensitivity for
ventricular fibrillation (VF) in this nonclinical setting. We then compared the clinical efficacy of AEDs
in 18 small communities (study group) with that of manual defibrillation in 18 additional communities
(control group) of similar size. Ambulance technicians using manual defibrillators correctly diagnosed
VF more frequently than the AEDs (98% vs 83%; p < .025). Specificity for VF was similar in the two
groups (100% for AEDs vs 94% for technicians; p > .10). AEDs were able to deliver shocks more
quickly than was possible with the manual defibrillators (1.56 vs 2.77 min; p < .001). The ability of the
AEDs to terminate VF was excellent, converting VF in 28 of 29 (97%) patients to some other rhythm
compared with only 37 of 53 (70%) patients in the control group (p < .01). Hospital admission and
discharge rates were similar for the two groups. Ten of the 35 (29%) patients managed with AEDs
Downloaded from http://ahajournals.org by on October 12, 2021

achieved admission and six (17%) were ultimately discharged. In the control group 17 of 53 (33%)
patients with VF were admitted and seven (13%) were discharged (p > .75). AEDs are an effective
alternative to manual defibrillation in small communities.
Circulation 73, No. 4, 701-709, 1986.

CARDIOVASCULAR DISEASE continues to be the
leading cause of death in the United States, particularly
sudden cardiac death occurring outside of the hospital.
To effectively manage sudden cardiac death in the
community, providers of prehospital emergency care
must arrive within minutes of a victim's collapse and
competently offer an advanced level of care.1-4 One
widely explored response to the sudden cardiac death
problem involves training paramedic-level ambulance
technicians in all techniques of advanced cardiac life
support, according to the standards of the American
Heart Association.' Although paramedic services have
From the Emergency Medical Services Learning Resources Center,
University of Iowa Hospitals and Clinics, and the Cardiovascular Cen-
ter, Department of Internal Medicine, University of Iowa College of
Medicine, Iowa City.
Supported in part by grants from the Cardiac Resuscitator Corp.,
Wilsonville, OR, the U.S. Public Health Service (07 H 000374-01-0),
and the NHLBI (HL 14388).
Address for correspondence: Kenneth R. Stults, B..S., PA-C,
EMSLRC, University of Iowa Hospitals and Clinics, Iowa City, IA
52242.
Received Feb. 15, 1985; revision accepted Jan. 16, 1986.
Vol. 73, No. 4, April 1986
been effective in improving survival after out-of-hos-
pital cardiac death in a number of settings," 6,7 they do
not appear appropriate for most communities with pop-
ulations of 25,000 and under. Paramedic programs are
not only expensive,8 they also require that technicians
maintain competence in a wide range of complex
emergency skills. Maintenance of these skills demands
frequent use,9 which does not occur in medium- and
low-volume ambulance services.
An alternative strategy, which was designed to over-
come the time and expense associated with full para-
medic training as well as the problem of infrequent use
of skills, is to train emergency medical technicians
(EMTs) in a single component of advanced cardiac life
support - defibrillation. Controlled studies have dem-
onstrated that EMTs trained in 16 hr to recognize ven-
tricular fibrillation (VF) and defibrillate can improve
cardiac arrest survival in urban,'0 suburban," and ru-
ral4 communities over that possible with basic life sup-
port alone. However, maintaining even these relative-
701
 STULTS et al.
ly simple skills is not easy. Depending on the size of
the community, a defibrillation-trained technician
(EMT-D) may have the opportunity to use his or her
skills in actual patient care only once every several
years.4 12 Yet it is imperative that when these skills are
required, they be performed quickly and efficiently if
the patient is to have any chance for survival. As a
result, intense monitoring mechanisms must be in-
stalled to continuously assess the competency and,
when it occurs, the in-field performance of the EMT-
Ds. Regular (at least quarterly) mandatory skill re-
view sessions have become standard for EMT-D level
ambulance services. 3
Yet another alternative for providing advanced car-
diac care outside of the hospital would be to train
ambulance technicians in the use of an automated ex-
ternal defibrillator (AED). If such a device were suffi-
ciently simple to operate, initial training time could be
significantly reduced even beyond the 16 hr required
for manual defibrillator training. Furthermore, if it
were possible to relax the stringent rquirements for
maintenance of manual defibrillation skills with the
AED while at the same time maintaining an equivalent
survival-to-discharge rate, then the AED would offer
low-volume ambulance services an attractive alterna-
tive for providing effective emergency cardiac care.
We conducted both a preclinical and a clinical eval-
Downloaded from http://ahajournals.org by on October 12, 2021
uation of the AED. These evaluations were designed
(1) to evaluate the accuracy of the AED's algorithm for
recognizing VF and differentiating it from all other
rhythms by assessing its responses to a series of heart
rhythms previously recorded during out-of-hospital
cardiac arrest and (2) to compare the efficacy of the
AED with that of standard manual defibrillation in
small communities where defibrillation is the only ad-
vanced intervention available out of the hospital.
Methods
The particular AED evaluated in this study was a commer-
cially available device, the HeartAid 95, manufactured by the
Cardiac Resuscitator Corp., Wilsonville, OR. Although limited
clinical use of this device has been reported previously,"1'7 the
logic parameters evaluated in this study and currently in general
use are different from those used in the earlier trials. This
evaluation was conducted in two parts.
Algorithm validation. To validate the diagnostic accuracy
of this AED, a single unit was presented with a series of 205
actual patient rhythms previously recorded as part of a study of
cardiac arrest in rural Iowa.4 All rhythms were recorded by
defibrillation-trained ambulance technicians during actual re-
suscitations in the field. The technicians used manual cardiac
monitor/defibrillators equipped with two-track cassette record-
ers that continuously recorded both the patient's electrocardio-
gram (ECG) and the voices of the EMTs. All cassette recorders
(LifePak 5, PhysioControl Corp., 450SL, Medical Research
Laboratories) were calibrated at a standard 1 mV. Segments
702
were selected for presentation to the AED that were sufficiently
long and free of artifacts (including those due to cardiopulmo-
nary resuscitation [CPR]) to give the device a fair opportunity to
accurately assess and "treat." The minimum length that will
allow the AED's algorithm to recognize VF and commit to
delivering a shock is 8 to 10 sec. all rhythm segments presented
were at least 8 sec in length.
The algorithm for recognizing VF and committing to shock
delivery is as follows: (1) The AED will begin to charge if it
detects at least two "positive intervals" within a 9.6 sec period.
A positive interval is defined as six or more deflections that are
at least 0. 150 mV (1.5 mm trough-to-peak) in amplitude with a
maximum slope of 3.2 mV/sec (termed VF counts) occurring in
a 2.4 sec period (rate at least 150/min). (2) The AED will
discharge if it detects a third positive interval within 7.2 sec of
the start of capacitor charging. Thus it takes a minimum of 4.8
sec (2 x 2.4) to begin charging; it requires a minimum of 7.2
sec (3 x 2.4) to commit to a shock after being presented with
VF. (For field units it may take as long as 12 sec from the initial
recognition of VF to actually deliver the shock because of the
additional time required for capacitor charging.) If the device
has not detected three positive intervals within a 15 sec period,
even if it has begun the charging process, it will abort and begin
analyzing again. Therefore the rhythm segments presented
ranged from 8 to 15 sec in length. In addition, the device will
not charge, or the charge will be aborted, if QRS complexes
occur at a rate equal to or greater than one-half that of the VF
counts. The QRS detector actually has variable sensitivity based
on the previous QRS complex. However, the maximum sensi-
tivity is specified by a repetitive haversine wave signal with a
290 ,uV amplitude, a 100 msec duration, and a 1 Hz repetition
rate.
Because the voice/ECG recorders simultaneously recorded
the patient's ECG and the EMTs' voices, it was possible to
reliably identify segments where CPR was stopped and the
patient was left completely alone while the EMTs interpreted
the rhythm. Artifact was further reduced during the recording
process as a result of our protocol, which mandated immediate
application of chest leads and did not allow the use of paddle
monitoring by the ambulance technicians. The 205 ECG seg-
ments, each associated with a separate resuscitation attempt,
were divided into two general groups: those associated with
patients who were successfully resuscitated in the field by the
EMT-Ds and those associated with patients who could not be
resuscitated. For purposes of this study, successful resuscitation
meant achieving hospital discharge. They were further divided
within these two general groups based on our interpretation of
the rhythm as VF, asystole, or pulseless bradycardia.
The diagnostic decisions ("shock" if VF, "no shock" if not
VF) of a single AED, which was modified so that it did not
actually deliver a shock but did simulate the charge-up and
defibrillation process, were compared with decisions made at
the scene by EMT-Ds. Each ECG segment was presented to the
AED twice.
In addition, interdevice variability of 14 AEDs that had been
distributed to community ambulance services for the in-field
clinical trials was evaluated. A subset of the original 205 rhythm
segments was presented to each of these 14 devices once and to
five randomly selected AEDs twice. This subset was selected in
the following manner: 10 segments each of coarse, medium,
and fine VF and five segments of asystole were randomly select-
ed from the 195 rhythm segments to which the single unit had
responded appropriately. In addition, all 10 segments of fine VF
that were missed by the single AED were included. We subse-
quently eliminated from this subset all ECG segments that were
shorter than 10 sec in duration, even though they had been
appropriately "shocked" by the single AED. We did this to
CIRCULATION

provide the AEDs with a fair and realistic opportunity to cor-
rectly assess the rhythm segments. In field use, the protocol for
using the AED calls for allowing up to 15 sec for rhythm
recognition, charge-up, and shock delivery. For purposes of this
study, it was believed that segments of at least 10 sec duration
would be sufficient to reliably test the AED's ability to recog-
nize VF and to begin the charging process. Actual delivery of a
shock was not required to receive credit for correct perfor-
mance. (The 10 segments that had been missed by the single
AED ranged from 14 to 20 sec in length and all were included in
this part of the study.) When the segments that did not meet the
10 sec minimum were eliminated, the distribution of 34 rhythm
segments indicated in figure 1 resulted.
Rhythm segments were designated as coarse, medium, or
fine VF according to the following arbitrary criteria: The great-
est waveform amplitude occurring in a given ECG segment at a
rate of at least 1 00/min was determined. For rhythms designated
coarse VF, these amplitudes were greater than 0.8 mV; for
medium VF, the amplitude range was 0.4 to 0.8 mV; and for
fine VF, the amplitudes were less than 0.4 mV. Amplitudes less
than 0.1 mV were classified as asystole. In assessment of inter-
device variability, no comparison was made with the decisions
of EMTs made in the field at the time of the recording. It was
simply noted whether the devices responded appropriately to the
rhythms presented, i.e., did they "shock" VF and refrain from
"shocking" rhythms other than VF. Whether the recorded
rhythm actually was VF or not was determined by two of the
authors (R. E. K., D. D. B.) by consensus.
Sensitivity, specificity, and overall accuracy scores were de-
termined. For the comparison of two proportions, the chi-square
test appropriate for paired proportions was applied.
Clinical field trial. From May 1, 1984, through June 30,
1985, 36 rural cornmunities from all regions of Iowa participat-
ed in the defibrillation study in either the control group (18
communities) or the experimental group (18 communities).
Downloaded from http://ahajournals.org by on October 12, 2021
THERAPY AND PREVENTION-DEFIBRILLATION
TABLE 1
Clinical field trial: population comparisons of study and control
communities
Community
Control
Experimental (manual
Characteristic (AED) defibrillation)
Average population 7,690 8,000
Population range 1,300-31,000 1,400-34,000
community to form a matched pair based on equivalent popula-
tions (table 1).
All ambulance technicians in the experimental group were
trained to use portable AEDs (HeartAid 95). All AEDs used in
this study were equipped with two self-adhesive, dual-function
monitor/defibrillation pads in lieu of both standard monitoring
electrodes and hand-held defibrillation paddles. These pads
have been previously shown to be an effective alternative to
paddle defibrillation in the hospital setting. 18 One pad is placed
over the apex, the other on the upper chest just to the right of the
sternum. Excellent artifact-free ECG tracings resulted from this
pathway. The AEDs used in this study were modified to deliver
only 335 J shocks.
With rare exceptions, these EMTs had no prior training in the
use of defibrillators of any type, and in all cases the AED was
the only source of defibrillation out of the hospital in these
communities throughout their participation in the study. Ambu-
lance technicians in the control group continued to use standard
manual monitor/defibrillators (Lifepak 5) also equipped with
two-channel cassette recorders. The training program and skill

Communities in the experimental group were selected from a
larger list of applicants on the basis of the following criteria:
local physician support, EMT commitment to employ the AED
in all cardiac arrests according to a strict study protocol, and
community size. Control communities were assigned by the
study staff from a large pool of Iowa communities already
providing out-of-hospital defibrillation by otherwise basic level
EMTs. A control community was assigned to each experimental
maintenance requirements for the control group have been pre-
viously described.4 Briefly, the skills of cardiac monitoring,
recognition of rhythms associated with cardiac arrest, and defi-
brillation were taught in a 16 hr course. All EMT-Ds were
required to successfully manage a simulated cardiac arrest pa-
tient, demonstrating their ability to deliver a shock within 90 sec
of arrival at the "patient's" side while making no important
errors of omission or commission, to complete the course and

Subset of 34 Rhythm Segments

- 1 1 1
' .
1. Coarse Ventricular Fibrillation . . A
Amplitude Range*: >0.8mV A .-li J.. . ir, . 11 ( .- I k / . ._ J-.mv
A
irli hi
"...j
Frequency Range: 20040Oppm**

2. Medium Ventricular Fibrillation - 8

Amplitude Range: 0.4-0.8mV A
Frequency Range: 200-400ppm ___t___ ___-_, ____

3. Fine Ventricular Fibrillation 16 L

Amplitude Range: 0.10-.3OmV
Frequency Range: 100-40Oppm JC

fl~~~~~~~~~~~~~ .. .. .. I ... .. ..

4. Asystole .. ..
fl I i
.. ..
~~~..
....
.. ..
.. .. .. .. ..

Amplitude Range: r0.1OmV

Frequency Range: < 1 OOppm

*See METHODS for determination of VF waveform amplitude.

**ppm = peaks per minute
FIGURE 1. Classification of VF and asystole.

Vol. 73, No. 4, April 1986 703

 STULTS et al.
receive certification. After completion of the course, each
EMT-D was required to recertify according to these same stan-
dards every 30 days to maintain these critical but infrequently
used skills.
Training for the experimental group (n = 341 EMTs) consist-
ed of a 4 hr course conducted in the local community devoted
almost exclusively to the operation of the AED according to a
strict protocol. No attempt was made to teach any aspect of
cardiac rhythm assessment. At the completion of training a
practical examination was given limited to assessing the stu-
dent's ability to attach the device to a CPR mannikin and to
operate it according to the study protocol.
The EMTs in the control group (manual defibrillation, n =
324 EMTs) were authorized to deliver up to three 300 J shocks
for VF. Up to three additional shocks were authorized if a
patient's rhythm reverted to VF after previous conversion to an
organized, perfusing rhythm. EMT-Ds in this group were also
authorized to shock what they considered to be asystole; this
was done to maximize their sensitivity for fine VF, which might
be interpreted as asystole. The protocol encouraging shocks for
asystole had been implemented before the beginning of this
AED study and was simply continued.
In the experimental communities, the emergency technicians
attached self-adhesive monitor/defibrillation pads upon verifi-
cation of cardiac arrest, stopped CPR, and switched the device
to the automatic mode. The patient was left completely alone
and the EMTs counted to 15. If, during that count, the algorithm
for VF was satisfied, the device warned bystanders to stand
clear through a voice synthesizer and automatically delivered a
shock of 335 J. The study protocol called for up to eight shocks
to be delivered in pairs with 1 min of CPR between each pair. If.
after any count of 15, no shock was delivered and the patient
remained pulseless, CPR was resumed for 1 min and the AED
was given another opportunity to treat. If the AED did not
deliver a shock on the second opportunity, the patient was
Downloaded from http://ahajournals.org by on October 12, 2021
transported immediately to the hospital.
Data were collected from written ambulance run reports,
voice/ECG cassette tapes, and interviews with the EMTs. The
information retrieved from both groups included the ambulance
response time (the time from the call to the arrival of the ambu-
lance), whether the arrest was witnessed, whether CPR was
performed before arrival of the ambulance and by whom, the
proximity of the collapse to the local hospital (in miles), the
cardiac rhythms (initial rhythm and rhythm after each shock),
elapsed time from ambulance arrival to delivery of the first
shock, and elapsed time between subsequent shocks. Outcomes
were recorded for each patient: hospital admission or death
before admission, and discharge or death in the hospital. The
return of spontaneous pulse and blood pressure at any point,
with or without assisted ventilations, was also recorded. Ambu-
lance response times, time at the scene, and transport time were
determined from official dispatcher records. Time to initiation
TABLE 2
Algorithm validation: performance of AED vs manual defibrillation by EMTs
Group I Successfully resuscitated (n = 33)
VF
Asystole/pulseless bradycardia
Group Ii Nonsurvivors (n 172)
VF
Asystole/pulseless bradycardia
704 ~
~ ~ ~ ~
704
of CPR was estimated by the EMTs through questioning of
bystanders. Time from ambulance arrival to defibrillation was
measured directly from the cassette recordings made at the
scene of the arrest.
All cardiac arrest tapes from both the experimental and con-
trol communities were reviewed by the authors, and written
summaries and comments regarding performance of the proto-
col were submitted to the local ambulance service and medical
directors. Each rhythm segment upon which either the AED or
the EMT-Ds based a treat/no-treat decision was interpreted in-
dependently by two of the authors (R. E. K., D. D. B.) who, at
the time of rhythm interpretation, were blinded to the actual
treatment decisions made in the field.
Although the EMT-Ds in the control communities were re-
quired to perform the defibrillation protocol adequately in a
simulated cardiac arrest situation monthly (as previously de-
scribed), no such requirements were made of the technicians
using the AED. In fact, throughout the period of the study, no
formal retraining was conducted for the AED technicians after
the initial 4 hr training program.
Univariate statistical techniques, including chi-square tests,
Fisher's exact test, and t tests, were used to analyze differences
in the two groups of communities. Specifically, chi-square tests
and Fisher's exact test were used, as appropriate, when compar-
ing two proportions such as admission and discharge rates,
sensitivity and specificity, etc. Two sample t tests were used for
all comparisons of distances and elapsed time.
Results
Algorithm validation. The single AED correctly rec-
ognized and delivered a simulated shock to all 33 seg-
ments of VF recorded from patients who had been
successfully resuscitated in the field by EMT-Ds. It
also correctly recognized and "shocked" 79 of 89 seg-
ments of VF that had been recorded from patients who
could not be resuscitated in the field. The ambulance
technicians, on the other hand, had correctly recog-
nized and shocked 88 of these 89 rhythms. However,
the ambulance technicians also shocked nine of 83
rhythms interpreted by us to be asystole. The AED
never "shocked" a rhythm other than VF (table 2).
Overall, the AED demonstrated 92% sensitivity and
100% specificity in the diagnosis of VF compared with
99% sensitivity and 90% specificity for the ambulance
technicians. The AED and the technicians had identi-
cal accuracy scores of 95%. The AED was insensitive
Actually
"Shocked" shocked
by AED by EMTs
33/33 33/33
010 0/0
79/89 88/89 (p < .025)
0/83 9/83 (NS)
~ CIRCULATIO
CIRCULATION
 THERAPY AND PREVENTION-DEFIBRILLATION
only to very fine VF, consistent with the parameters of TABLE 4
its recognition algorithm (see Methods), which en- Clinical field trial: sensitivity and specificity of AED and EMT-Ds
(by patient)
sures 100% specificity at the expense of some sensitiv-
ity. This single AED was consistent on both trials with AEDA EMT-DsA
all ECG segments. Rhythm (correct/total) (correct/total) Significance
With respect to interdevice variability, each of 14 VF 29/35 53/54 p < .025
AEDs correctly recognized and "shocked" all five seg- Asystole 0/39 45/47 NA
ments of coarse VF and all eight segments of medium Organized 0/27 3/50 p > .10
VF. Of the 16 segments of fine VF, five were Sensitivity 83% 98%
Specificity 100%B 94%B
"shocked" by all 14 devices, seven were "shocked" by
some devices and not by others, and four were not NA = not applicable.
recognized as VF by any of the AEDs. None of the 14 AResults are reported as number of patients shocked/number of
units responded to any of the five segments of asystole patients with the indicated rhythm at any time during their management.
Therefore the totals are higher than the actual number of patients.
(table 3). When the 34 rhythm segments were present- BSpecificity for EMT-Ds using manual defibrillators was calculated
ed a second time to five randomly selected AEDs, 31 as the proportion of organized rhythms not shocked, since these were
of the 34 segments were correctly recognized by all the only rhythms the EMT-Ds were trained not to shock. For the AED,
five devices. Two of the devices responded consistent- specificity is the proportion of all non-VF rhythms not shocked.
ly on each trial to all 34 segments, while one was
inconsistent on three segments and two were inconsis- Of the remaining 114 cases with good documentation,
tent on one segment each. All inconsistencies occurred 53 (46%) patients had an initial rhythm of VF, 43
with rhythm segments demonstrating fine VF. (38%) had an initial rhythm of asystole, and 18 (16%)
Clinical field trial. Between May 1, 1984, and June presented with an organized rhythm.
30, 1985, AEDs were applied in 80 cases of cardiac Sensitivity and specificity. All AEDs exhibited 100%
arrest due to heart disease. Mechanical failures pre- specificity, i.e., no rhythm other than VF was shocked
cluding both rhythm documentation and AED treat- (table 4). They exhibited moderate sensitivity to VF
ment occurred in eight cases. In four others the AED whether considered by patient or by segment (tables 4
apparently functioned normally, but failure of the and 5). The AEDs were occasionally insensitive to a
Downloaded from http://ahajournals.org by on October 12, 2021

voice/ECG cassette recorder prevented documentation fairly wide amplitude range of VF segments (<0. 2 mV
of the cardiac rhythm and the events surrounding the to >0.8 mV), although all patients in coarse VF (>0.8
resuscitation. Of the remaining 68 cases, 35 (51%) mV) and all but one patient in medium VF (0.4 mV to
patients had an initial rhythm of VF, 21 (31%) had an 0.8 mV) received at least one shock. Five patients in
initial rhythm of asystole, and 12 (18%) had an orga- five VF (<0.4 mV) received no shocks (table 5).
nized initial rhythm. By comparison, EMT-Ds using manual defibrilla-
In the control communities, a total of 124 cardiac tors were more sensitive to VF of all amplitudes (98%
arrests occurred during the same period. No mechani- vs 83%; p < .025). The only patient in VF who was
cal failures preventing defibrillation occurred, al- not shocked in the control communities was in a coarse
though in 10 cases tape failure prevented documenta- VF interpreted by the EMT-Ds to be an organized
tion of the cardiac rhythm and the events at the scene. rhythm. This high degree of sensitivity for VF is due,
in part, to our protocol, which encourages EMT-Ds to
TABLE 3
Algorithm validation: interdevice variability: 14 AEDs presented
shock asystole as well as VF. Therefore EMT-Ds using
with 34 segments the manual defibrillators generally shock all but orga-
nized rhythms and do not attempt to distinguish be-
"Shocked" Not "shocked"
by all Variably by any TABLE 5
devices "shocked" devices Clinical field trial: sensitivity of AED to VF by amplitudeA
Coarse VF 5 0 0 Amplitude of VF By patient (%) By segment (%)
(5 segments)
Medium VF 8 0 0 Coarse (>0.8 mV) 8/8 (100) 15/22 (68)
(8 segments) Medium (0.4 - 0.8 mV) 16/17 (94) 43/48 (90)
Fine VF 5 7 4 Fine (0.1 - 0.3 mV) 5/10 (50) 10/21 (48)
(16 segments) Overall 29/35 (83) 68/91 (75)
Asystole 0 0 5
AResults are reported as number shocked/total VF of given ampli-
(5 segments) tude.

Vol. 73, No. 4, April 1986 705

 STULTS et al.

TABLE 6 The EMT-Ds delivered shocks to 53 patients in VF,

Clinical field trial: termination of VF
Manual
defibril-
AED lation
No. patients shocked 29 53
VF terminated 28 (97%) 37 (70%)
No. patients to
organized rhythm 16 (55%) 23 (43%)
tween fine VF and asystole. Specificity for the EMT-
Ds in the control communities, then, was calculated as
the proportion of organized rhythms that were not
shocked (94%), since these represented the only
rhythms they were trained not to shock. In no case was
a rhythm with a pulse shocked. In two cases the
shocked rhythms were low-voltage (0.2 mV), irregu-
lar, wide-QRS complex (>0.16 sec) bradycardias
(rates 20 to 25 complexes/min). One was interpreted
by the EMT-Ds as VF, the other as asystole. In the
third case a patient had been successfully converted
from coarse VF to an irregular, wide-QRS complex
(>0.16 sec) rhythm with a rate of approximately 75
complexes/min. There was considerable irregular elec-
trical activity obscuring the baseline (amplitude aver-
aging 0.3 mV, rate >300/min), probably representing
atrial fibrillation with superimposed movement arti-
Downloaded from http://ahajournals.org by on October 12, 2021
converting only 37 (70%; p < .01) to some other
rhythm. Twenty-three patients (43%) were converted
Signifi- to organized rhythms.
cance Speed of shock delivery. Elapsed time from ambulance
arrival until delivery of the first shock was determined
p < .01 in both groups of communities by timing directly from
the voice/ECG cassette tapes (table 7). The mean time
p > .10
from EMT arrival to delivery of the first shock by the
AEDs was significantly shorter than for the EMT-Ds
using manual defibrillators (1.56 vs. 2.77 min; p <
.001).
Admission and discharge rates. Ten of the 35 (29%)
patients found in VF in the experimental communities
(AED) were admitted to the hospital with a spontane-
ous pulse and blood pressure, with or without assisted
ventilations. Six (17%) patients were ultimately dis-
charged in good condition.
In the control group (manual defibrillation), 17 of
the 53 (33%) patients with an initial rhythm of VF were
admitted to the hospital, and seven (13%) were dis-
charged alive (figure 2).
Table 7 provides a detailed comparison of important
variables observed in the experimental and control
communities. The only significant difference observed
is the shorter time from technician arrival to delivery of
the first shock with the AED.

fact. The EMTs initially reported that a weak pulse

was present. The pulse was subsequently lost, and the
active baseline was incorrectly interpreted as a return
to VF, although in fact QRS complexes remained visi-
ble. A single shock was then delivered, which had no
effect on the patient's rhythm.
Conversion rates. A high proportion (28/29, 97%) of
patients in VF who were shocked by AEDs were con-
verted to some other rhythm, including asystole (table
6). Sixteen of these patients (55%) were converted to
organized rhythms.
Discussion
There are two principal findings of this study. First,
we found that the diagnostic algorithm of the particular
AED evaluated, HeartAid, was highly specific and
sensitive for the diagnosis of VF. Second, we showed
that the AED is an effective alternative to manual defi-
brillation for the management of out-of-hospital cardi-
ac arrest in rural communities served by minimally
trained ambulance technicians. The AED demonstrat-

TABLE 7
Clinical field trial: characteristics of patients with cardiac arrest who were initially in VF and the time elapsed before
emergency care

Manual
Characteristic AED defibrillation Significance
No. of cases 35 53
Mean age (yrs) 63+ 12 65+ 14 NS
No. of witnessed arrests 25 (78%) 47 (89%) NS
Average time to initiation of CPR (min) 5.28 +4.00 5.44+-- 3.88 NS
Distance of event from ambulance (miles) 3.47 4.20 3.43 + 5.08 NS
Total time from call to first shock (min) 7.43 7.94 NS
Ambulance arrival time (min) 5.87 3.92 5.17 + 3.40 NS
Time from arrival to first shock (min) 1.56 0.89 2.77 1.14 p < .001
Total time at scene (min) 10.03+4.73 10.93+4.62 NS
CIRCULATION~~~~~~~~~~~~.
706 CIRCULATION

40 r-
rX NS -n o Manual Defibrillation
32% * Automatic External
30 I- I1 29% Defibrillation
20 H rg~ NS---
17%
13%
10 H
1753
0 whether considered by patient or by segment, than
Admitted Discharged
Patients In Ventricular Fibrillation
FIGURE 2. Hospital admission and discharge comparisons for patients
with an initial rhythm of VF treated with either the AED or standard
manual defibrillation.
ed high sensitivity to medium and coarse VF, the car-
diac arrest rhythms most often associated with surviv-
al.19' 20 The overall survival-to-discharge rate in
communities using the AED was comparable to that
previously reported for small communities served by
EMT-Ds4 as well as to that observed in the manual
control group in this study.
This initial clinical trial of the efficacy of AEDs in
rural communities reaffirms the importance of early
defibrillation in improving patient outcome after out-
of-hospital cardiac arrest. Hospital discharge rates in
Downloaded from http://ahajournals.org by on October 12, 2021
both the AED and manual defibrillation communities
were considerably better than those reported for com-
munities with basic life support alone.4 21 The effec-
tiveness of an AED, however, must be judged in large
part by criteria other than hospital discharge rates.
Patient outcome after cardiac arrest is a function of
many variables and cannot be attributed solely to accu-
rate diagnosis and defibrillation of VF, whether by
human or machine. The discharge rates reported for
both groups in this series, for example, are somewhat
lower (although not significantly) than those previous-
ly reported for rural communities served by EMTs
trained to manually defibrillate4 and considerably low-
er than those reported for EMT-Ds in urban10 and sub-
urban11 settings. It is not possible to ascribe these dif-
ferences to a particular method of defibrillation,
however, until one first controls for such confounding
variables as mean ambulance response times, propor-
tion of arrests that were witnessed, mean time from
collapse to initiation of CPR, etc.
Rather, an AED must be judged by variables specifi-
cally related to its own performance: (1) accuracy of
rhythm diagnosis, both by segment and by patient, (2)
rapid shock delivery for VF, (3) ability to convert VF
to any other rhythm (which rhythm is not a function of
Vol. 73, No. 4, April 1986
THERAPY AND PREVENTION-DEFIBRILLATION
the defibrillator), and (4) mechanical failure rate.
When assessed according to these criteria, the per-
formance of an AED can be analyzed separately from
the emergency medical services system in which it was
used and more general observations can be made. For
the remainder of this discussion, we will focus on this
AED's performance in these four categories.
Accuracy of rhythm diagnosis. In the clinical trial, this
AED was somewhat less sensitive for VF (p < .025),
EMT-Ds using manual defibrillators. However, of the
five patients determined by the authors to have had an
initial rhythm of fine VF yet who were not shocked by
the AED, three were in such fine VF that the criteria of
the AED's algorithm clearly were not met. Such fine
VF rhythms are rarely associated with a positive out-
come even when defibrillated.22 Only two cases gave
cause for serious concern. One was a patient in medi-
um, the other in coarse VF. Both patients were eventu-
ally shocked by the AED, but only after delays of 6 and
7 min, respectively, resulting from the first several
segments of VF not being shocked. It has not been
possible in either case to determine the cause of these
delays, although mechanical failures such as defective
monitor/defibrillation pads or cables have been ruled
out. Although VF was eventually terminated by the
AED in both of these cases, neither patient survived to
hospital admission. A similar incident occurred during
the study in the control (EMT-D) group: defibrillation
of coarse VF was delayed in one case for 7 min in a
control community because of an EMT-D's uncertain-
ty concerning interpretation of the rhythm. This patient
also converted from VF but did not survive to hospital
admission.
It must be strongly emphasized that the high degree
of sensitivity for VF exhibited by the EMT-Ds using
manual defibrillators can be attributed in large part to
our EMT-D performance protocol, which allows "de-
fibrillation" of asystole in the interest of ensuring that
all VF, including fine VF, is shocked. The relative
insensitivity of the AED to fine VF, on the other hand,
is a deliberate design trade-off in favor of the observed
100% specificity. This is a reasonable approach con-
sidering the totally automatic functioning of the device
and the potential, otherwise, for inappropriate shocks
and possible harm. By the same token, the specificity
of the two approaches cannot be compared except as it
relates to shocking of organized rhythms. In this regard
the AED was, again, 100% specific for VF, whereas
the EMT-Ds manually shocked three pulseless but or-
ganized rhythms (specificity 94%; p > . 10).
Rapid shock delivery. The AEDs were consistently
707
 STULTS et al.
able to deliver shocks to patients in documented VF
more quickly than could EMT-Ds using manual defi-
brillators (table 7). Since the total time that a patient is
in VF before defibrillation is probably the most impor-
tant determinant of outcome.3" this consistent differ-
ence may have important clinical significance. We
have observed in another setting that when ambulance
response times are short (mean 4.0 min) and controlled
for, the elapsed time from EMT arrival to delivery of
the first shock is an important predictor of successful
resuscitation (Stults and Brown: unpublished data).
The explanation for the observed difference in the
time required to defibrillate is probably related to the
greater number of tasks that must be accomplished
when using a manual defibrillator. The elapsed time to
first and subsequent shocks using manual defibrillators
encompasses (I) assessment of the rhythm followed by
resumption of CPR while the defibrillator is prepared.
(2) application of conductive medium and charging of
the defibrillator, (3) placement of paddles and careful
assurance that nobody is in contact with the patient at
the time of defibrillator discharge. Once the self-adhe-
sive monitor/defibrillation pads used with the AED
have been applied and the device is switched to the
" auto" position, nothing remains to be done but ensure
that everyone is clear and allow 15 sec for the AED to
perform.
Downloaded from http://ahajournals.org by on October 12, 2021
Ability to terminate VF. The VF termination rate ob-
served with the AED (97%) was much better than that
observed with manual defibrillation (70%) (table 6). It
was not possible in this study to determine the reasons
for this difference. One hypothesis is that there is
greater variability and error in defibrillation technique
(paddle placement, paddle pressure, conductive medi-
um) using standard hand-held paddles than with self-
adhesive monitor/defibrillation pads. Another possible
explanation is the more rapid shock delivery with the
AED. These explanations are not mutually exclusive
and, in fact, both may contribute to the observed im-
provement. The difference in conversion rates prob-
ably cannot be ascribed to the AED self-selecting pa-
tients with more easily terminated coarse VF. The
mean amplitude of the VF to which the AED was
insensitive was <0.3 mV (fine VF), whereas the mean
amplitude of the VF that could not be converted by
EMT-Ds using manual defibrillators was 0.5 mV (me-
dium VF). Whatever the reason, the ability of this
AED to terminate VF was excellent.
Mechanical failure rate. The total failure rate, includ-
ing all AED malfunctions that prevented rhythm as-
sessment, shock delivery, and/or rhythm documenta-
tion, was high (12 of 80 or 15%; eight preventing
708
rhythm assessment and shock delivery) when com-
pared with EMT-Ds using manual defibrillators (10 of
124 or 8%; none prevented rhythm assessment or
shock delivery; p < .001). Each failure in this series
was evaluated as thoroughly as possible. In one case, a
broken pin on the patient cable connector prevented
both rhythm assessment and the opportunity for treat-
ment had the rhythm been VF, which could not be
determined. Rhythm assessment, and thus any possi-
bility of treatment, was also precluded in seven cases
where no malfunction of the AED or its patient cable
could be discovered.
In these cases we believe that the self-adhesive mon-
itor/defibrillation pads failed. The most likely causes
of pad failure are dried conductive gel and/or a broken
wire in the pad itself. Our practice early in the study of
storing the pads rolled or folded in the lid of the defi-
brillator was the most likely cause of at least some of
the failures. Steps to prevent such problems were sub-
sequently taken, including storing pads flat (never
folded). In 32 consecutive cases since taking these
steps, there has not been a malfunction that prevented
AED rhythm assessment or defibrillation.
Failure of the cassette voice/ECG recorder prevent-
ed rhythm and performance documentation in four
cases where the AED apparently functioned normally.
Cassette failure occurred in an additional five cases,
but continuous paper recordings allowed us to docu-
ment the presence or absence of VF and the AEDs'
treatment decisions.
If one excludes from consideration the failures that
resulted from monitor/defibrillation pad malfunction,
the rate of failures that prevented patient assessment
and treatment was quite small (1 of 80), even when
compared with the manual defibrillators that exper-
ienced no such failures.
Training requirements. The incidence of cardiac ar-
rest is low in small communities, whereas the average
number of volunteer EMTs per ambulance squad is
high. As a result, an individual EMT can be expected
to use his or her defibrillation skills in actual patient
care only once every several years.4 12 It has therefore
been found necessary to require EMT-Ds trained in 16
hr to manually defibrillate to recertify every 30 days
through a rigid practical examination conducted by the
local medical director. Such periodic recertification
has been determined to be an essential component of
manual EMT-defibrillation programs.'3 Not only was
initial training time shorter for the AED in this study (4
vs 16 hr), but formal periodic sessions for skill mainte-
nance were not required at all. Critical EMT perfor-
mance variables (elapsed time from arrival to shock
CIRCULATION

delivery, adherence to protocol, etc.) did not measur-
ably deteriorate with the passage of time in communi-
ties using the AED. We do not wish to imply, howev-
er, that no periodic activities for skill maintenance
were undertaken by the EMTs using the AED, but that
no formal sessions were required or monitored as must
be done with manual EMT-defibrillation. In fact, the
EMTs reported practicing on their own an average of
every 6 weeks. Such practice should be strongly en-
couraged. As these programs are implemented on an
on-going basis, it seems only prudent that a more for-
mal review of the operation of the AED and the ap-
proved patient care protocol should be conducted as
often as once each quarter. The observed reduction in
the time required to acquire and maintain defibrillation
skills with the AED may make this approach particu-
larly attractive to low-volume, volunteer ambulance
squads as well as first-responder units of larger tiered-
response ambulance systems.
The overall effectiveness of any defibrillator, in-
cluding an AED, will be limited by the characteristics
of the emergency medical services system in which it
is used. Particularly in rural areas staffed by part-time
volunteer ambulance technicians, for example, ways
must be found to reduce the average time from patient
collapse to arrival of a defibrillator-bearing EMT.
Even with the ambulance response times observed in
Downloaded from http://ahajournals.org by on October 12, 2021
this study, however, early defibrillation by basic
EMTs using manual or automatic defibrillators greatly
improves the odds of successful resuscitation from car-
diac arrest in small communities over that possible
with basic life support alone.
We thank William Clarke, Ph.D., Department of Preventive
Medicine, the University of Iowa College of Medicine, for
assisting with the statistical analysis. Bernice Morrison, R.N.,
and Tom Toycen, R.N., REMT-P, assisted with tape review
and data retrieval. We are also indebted to Alfred Aronson,
M.D., and Bruce Haggar, B.S., M.S.E.E., for their expertise
and enthusiastic support in the initial development of this proj-
ect. Our deepest gratitude goes to the hundreds of emergency
medical technicians and their medical directors throughout the
state of Iowa, especially the volunteers, who have committed
themselves to providing the best emergency care possible.
References
1. Eisenberg MS, Bergner L, Hallstrom A: Paramedic programs and
out-of-hospital cardiac arrest. I. Factors associated with successful
resuscitation. Am J Public Health 69: 30, 1979
Vol. 73, No. 4, April 1986
THERAPY AND PREVENTION-DEFIBRILLATION
2. Cummins RO, Eisenberg MS, Hallstrom AP, Litwin PE: Survival
of out-of-hospital cardiac arrest with early initiation of cardiopul-
monary resuscitation. Am J Emerg Med 3: 114, 1985
3. Eisenberg MS, Copass MK, Hallstrom A, Cobb LA, Bergner L:
Management of out-of-hospital cardiac arrest: failure of basic
emergency medical technician services. JAMA 243: 1049, 1980
4. Stults KR, Brown DD, Schug VL, Bean JA: Prehospital defibrilla-
tion performed by emergency medical technicians in rural commu-
nities. N Engl J Med 310: 219, 1984
5. McIntyre KM, Lewis AJ: Textbook of advanced cardiac life sup-
port, 1981. Dallas, 1983, American Heart Association, Inc.
6. Myerburg RJ, Conde CA, Sung RJ, Mayorga-Cortes A, Mallon
SM, Sheps DS, Appel RA, Castellanos A: Clinical electrophysio-
logic and hemodynamic profile of patients resuscitated from pre-
hospital cardiac arrest. Am J Med 68: 568, 1980
7. Lewis RP, Stang JKM, Fulkerson PK, Sampson KL, Scoles A,
Warren JV: Effectiveness of advanced paramedics in a mobile
coronary care system. JAMA 241: 1902, 1979
8. Urban N, Bergner L, Eisenberg MS: The costs of a suburban
paramedic program in reducing deaths due to cardiac arrest. Med
Care 19: 379, 1981
9. Stross JK: Maintaining competency in advanced cardiac life sup-
port skills. JAMA 249: 3339, 1983
10. Weaver WD, Copass MK, Bufi D, Ray R, Hallstrom AP, Cobb
LA: Improved neurologic recovery and survival after early defibril-
lation. Circulation 69: 943, 1984
11. Eisenberg MS, Copass MK, Hallstrom AP, Blake B, Bergner L,
Short FA, Cobb LA: Treatment of out-of-hospital cardiac arrests
with rapid defibrillation by emergency medical technicians. N Engl
J Med 302: 1379, 1980
12. Ornato JP, McNeill SE, Craren EJ, Nelson NM: Limitation of
effectiveness of rapid defibrillation by emergency medical techni-
cians in a rural setting. Ann Emerg Med 13: 1096, 1984
13. ACT Foundation: Statement on EMT-defibrillation. JEMS 8: 37,
1983
14. Diack AW, Welborn WS, Rullman RF, Walter CW, Wayne MA:
An automatic cardiac resuscitator for emergency treatment of cardi-
ac arrest. Med Instrum 13: 78, 1979
15. Jaggarao NSV, Heber M, Grainger R, Vincent R, Chamberlain
DA, Aronson AL: Use of an automated external defibrillator-pace-
maker by ambulance staff. Lancet 2: 73, 1982
16. Heber M: Out-of-hospital cardiac arrest using the "Heart-Aid," an
automated external defibrillator-pacemaker. Int J Cardiol 3: 456,
1983
17. Cummins RO, Eisenberg MS, Bergner L, Murray JA: Sensitivity,
accuracy, and safety of an automatic external defibrillator: report of
a field evaluation. Lancet 2: 318, 1984
18. Kerber RE, Martins JB, Kelly KJ, Ferguson DW, Kouba C, Jensen
SR, Newman B, Parke JD, Kieso R, Melton J: Self-adhesive preap-
plied electrode pads for defibrillation and cardioversion. J Am Coll
Cardiol 3: 815, 1984
19. Diamond NJ, Schofferman JS, Elliott JW: Factors in successful
resuscitation by paramedics. JACEP 6: 42, 1977
20. Roth R, Stewart RD, Rogers K, Cannon GM: Out-of-hospital car-
diac arrest: factors associated with survival. Ann Emerg Med 13:
237, 1984
21. Enns J, Tweed WA, Donen N: Prehospital cardiac rhythm deterio-
ration in a system providing only basic life support. Ann Emerg
Med 12: 478, 1983
22. Weaver WD, Cobb LA, Dennis D, Ray R, Hallstrom AP, Copass
MK: Amplitude of ventricular fibrillation waveform and outcome
after cardiac arrest. Ann Intern Med 102: 53, 1985
23. Eisenberg MS, Bergner L, Hallstrom A: Cardiac resuscitation in
the community: importance of rapid provision and implications for
program planning. JAMA 241: 1905, 1979
709