Customize your JAMA Network experience by selecting one or more topics from the list below.
Wik L, Kramer-Johansen J, Myklebust H, et al. Quality of Cardiopulmonary Resuscitation During Out-of-Hospital Cardiac Arrest. JAMA. 2005;293(3):299–304. doi:10.1001/jama.293.3.299
Context Cardiopulmonary resuscitation (CPR) guidelines recommend target values
for compressions, ventilations, and CPR-free intervals allowed for rhythm
analysis and defibrillation. There is little information on adherence to these
guidelines during advanced cardiac life support in the field.
Objective To measure the quality of out-of-hospital CPR performed by ambulance
personnel, as measured by adherence to CPR guidelines.
Design and Setting Case series of 176 adult patients with out-of-hospital cardiac arrest
treated by paramedics and nurse anesthetists in Stockholm, Sweden, London,
England, and Akershus, Norway, between March 2002 and October 2003. The defibrillators
recorded chest compressions via a sternal pad fitted with an accelerometer
and ventilations by changes in thoracic impedance between the defibrillator
pads, in addition to standard event and electrocardiographic recordings.
Main Outcome Measure Adherence to international guidelines for CPR.
Results Chest compressions were not given 48% (95% CI, 45%-51%) of the time
without spontaneous circulation; this percentage was 38% (95% CI, 36%-41%)
when subtracting the time necessary for electrocardiographic analysis and
defibrillation. Combining these data with a mean compression rate of 121/min
(95% CI, 118-124/min) when compressions were given resulted in a mean compression
rate of 64/min (95% CI, 61-67/min). Mean compression depth was 34 mm (95%
CI, 33-35 mm), 28% (95% CI, 24%-32%) of the compressions had a depth of 38
mm to 51 mm (guidelines recommendation), and the compression part of the duty
cycle was 42% (95% CI, 41%-42%). A mean of 11 (95% CI, 11-12) ventilations
were given per minute. Sixty-one patients (35%) had return of spontaneous
circulation, and 5 of 6 patients discharged alive from the hospital had normal
Conclusions In this study of CPR during out-of-hospital cardiac arrest, chest compressions
were not delivered half of the time, and most compressions were too shallow.
Electrocardiographic analysis and defibrillation accounted for only small
parts of intervals without chest compressions.
Since the first standards and guidelines for cardiopulmonary resuscitation
(CPR) were published 30 years ago1 (with the
latest update in 20002,3) health
care professionals in and out of the hospital have been trained accordingly
around the world. The importance of CPR, defined as chest compressions and
ventilation, for survival of cardiac arrest patients has been demonstrated,4 and there are indications that the quality of CPR
performance influences the outcome.5-7
When tested on mannequins, CPR quality performed by lay rescuers and
health care professionals tends to deteriorate significantly within a few
months after training,8-10 but
little is known about the quality of clinical performance on patients. Aufderheide
et al11 recently observed short periods with
inappropriately high ventilation rates during advanced cardiac life support
(ACLS), and van Alem et al12 found long pauses
in CPR when first responders used automated external defibrillators.
We therefore studied the performance of paramedics and nurse anesthetists
during out-of-hospital ACLS by continuously monitoring all chest compressions
and ventilations during resuscitation episodes using online defibrillators
modified to collect such data.
The study was approved by the regional ethics committees for Akershus,
Norway, Stockholm, Sweden, and London, England. Informed consent for inclusion
in the study was waived as decided by these committees in accordance with
paragraph 26 in the Declaration of Helsinki.13 The
study was a case series involving patients older than 18 years with out-of-hospital
cardiac arrest of all rhythms. Noncardiac causes of cardiac arrest were included.
Patients with cardiac arrest occurring between March 2002 and October 2003
were included in the study.
Prototype defibrillators based on Heartstart 4000 (Philips Medical Systems,
Andover, Mass) were deployed in 6 ambulances in each of the 3 regions. These
ambulances were chosen based on historically high rates of cardiac arrest
at their sites. The defibrillators were fitted with an extra chest pad to
be mounted on the lower part of the sternum with double adhesive tape. This
chest pad was fitted with an accelerometer (ADXL202e, Analog Devices, Norwood,
Mass) and a pressure sensor (22PCCFBG6, Honeywell International Inc, Morristown,
NJ). The heel of the rescuer’s hand was placed on top of the chest pad
and movement of the chest pad was considered equal to that of sternal movement
during chest compressions. To avoid registering movements of the entire patient
as chest compressions, only movements of the sternal chest pad with a parallel
compression force greater than 2 kg were used in the automated analysis. A
second accelerometer of the same kind was fitted within the defibrillator.
Signals from this accelerometer were subtracted from signals from the chest
pad accelerometer prior to depth calculation to compensate for possible vertical
motion of the entire supporting surface. This technology has previously been
reported to measure chest compression depth with an accuracy of ± 1.6
All ambulances were staffed by paramedics; in Stockholm, the second
rescue vehicle at the scene also included a nurse anesthetist. Immediately
prior to the study period, all involved personnel underwent a refresher course
in ACLS according to international CPR guidelines2,3 and
in use of the modified defibrillator. In Akershus, a modification required
that patients with ventricular fibrillation or pulseless ventricular tachycardia
received 3 minutes of CPR before the first direct current shock and between
unsuccessful series of 3 direct current shocks.15 Resuscitation
was otherwise attempted in accordance with the guidelines.2,3 The
defibrillators were used in manual mode in Akershus and in semiautomatic mode
in the 2 other regions. The personnel were aware that we intended to study
CPR performance and that the sternal pad recorded chest compressions. They
were not informed that a primary focus was duration of time CPR was performed.
Data from each resuscitation episode were collected in 2 data cards;
1 standard card collected electrocardiographic signals, time, and events,
and a second card fitted specially for this study recorded signals from the
extra chest pad and thoracic impedance between the defibrillator pads as measured
by applying a nearly constant sinusoidal current. After each CPR episode,
all data were extracted and collected and the memory of the cards was cleared.
One person at each site was responsible for this.
The raw data consisted of timeline and events, electrocardiographic
signals, thoracic impedance, and values from the extra chest pad, all sampled
at 500 Hz. For each episode, a copy of the ambulance record and other written
documentation, including the Utstein format for out-of-hospital cardiac arrest,16 were collected. All data were collected on a designated
server at the facilities of Laerdal Medical Corp, Stavanger, Norway, and Laerdal
personnel preprocessed the data by filtering and down-sampling to 50 Hz to
facilitate display of the data for annotation and review. A custom-made computer
program designed for the study (Sister Studio, Laerdal Medical) was used to
view and annotate each cardiac arrest case. A second standard computer program
(CodeRunner Web Express, Philips Medical, Andover, Mass) was used in parallel
to provide further details about electrocardiography. For each episode, the
initial rhythm and each subsequent change in rhythm were annotated. Pulseless
electrical activity was defined as QRS complexes without blood flow, indicated
either by a clinically detected pulse or blood flow–induced changes
in thoracic impedance. Impedance changes coincident with cardiac contractions
and arterial pressure pulses have been validated with echocardiography and
blood pressure measurements in pigs.17 In a
pilot study, we found these changes to be in the range of 87 to 477 mΩ
in 21 healthy volunteers, and an impedance amplitude of greater than 50 mΩ
was used to indicate blood flow in the present study.
Spontaneous circulation was defined as QRS complexes with blood flow
as indicated by the same factors. Time markers were set at the start of the
first chest compression, 5 minutes thereafter, and at the end of the resuscitation
episode, defined as discontinued monitoring or the end of treatment as judged
from recordings and written information. The term time is
used for time intervals in this article and time point for
a specific point in time. The annotations were made by an experienced anesthesiologist
with training and clinical practice in ACLS together with a research engineer
with working knowledge of the Sister Studio program and the measurement systems.
Compressions were calculated by integrating the difference between the
2 accelerometers over a time window defined by the 2-kg threshold from the
force transducer. Compression depth was characterized as appropriate for 38
to 51 mm (1.5-2 in),2,3 too deep,
or too shallow. Incomplete compression release was annotated if the chest
pad pressure did not fall below 4 kg at any time during the compression-decompression
cycle. Duty cycle was defined as the percentage of time with downward movement
of the chest pad divided by the total cycle time. For each time period, the
actual number of compressions per minute as well as the rate during compression
periods (defined as a period with <1.5 seconds between 2 compressions)
were determined. No-flow time (NFT) was defined as total time minus the time
with chest compressions or spontaneous circulation (NFT = timetotal − timecompressions − timespontaneous circulation), and the ratio between NFT and the total time
without spontaneous circulation was defined as the no-flow ratio (NFR) [NFR = NFT/(timetotal − timespontaneous circulation)]. The
NFT and NFR represent the total time during the resuscitation episode without
cerebral and myocardial circulation.
According to the guidelines,2,3 chest
compressions should not be given during rhythm analysis, defibrillator charging,
shock delivery, and pulse checks. Adjusting the NFT by subtracting the time
required for these procedures (NFTadj = NFT − timedefibrillator) thus indicates time without blood flow due to performance
of the rescuer team without interfering with rhythm analysis, defibrillation
attempts, or pulse checks. Timedefibrillator was determined for
each episode. With the defibrillator in semiautomatic mode, actual recorded
times from the defibrillator for automatic analysis, charging, and shock delivery
were used. In manual mode, a maximum of 5 seconds was allowed for rhythm analysis.
If an organized rhythm was present, palpation of pulse was allowed for a maximum
of 10 seconds and included in timedefibrillator [NFRadj =
NFTadj/(timetotal − timespontaneous
The NFTadj and NFRadj represent the potential
for reducing time without circulation without interfering with guidelines
recommendations2,3 and are less
than the unadjusted values, which include NFT, as recommended in the guidelines.
Ventilations were automatically detected by changes in thoracic impedance,
filtered and corrected for compression and blood flow–related signals.
Ventilation measurement by impedance has been reported in many studies since
194418 and was recently validated for the use
of defibrillator electrodes during cardiac arrest in pigs.17 A
recent study using the present defibrillator setup in volunteers showed strong
correlation between impedance and spirometer waveforms.19
The primary outcome measure was adherence to international guidelines
for CPR. Target values for compression rate were 100/min to 120/min; for depth,
38 to 52 mm; and for ventilation rate, 2 ventilations for every 15 compressions
before intubation and 10/min to 12/min after intubation.
All data from each resuscitation episode were collected and described
using a spreadsheet program (Excel 2002, Microsoft Corp, Redmond, Wash) and
a statistical analysis program (SPSS 12.0.1, SPSS Inc, Chicago, Ill). All
statistical analyses were performed by J.K.-J. at the University of Oslo,
Oslo, Norway. All numbers are given as mean (standard deviation) for the first
5 minutes after the start of recorded CPR and for the entire resuscitation
episode. When variables had very skewed distributions, medians were used as
the mid-point estimate and interquartile ranges as the variability measure.
The results for the first 5 minutes of the resuscitation episode were analyzed
vs the rest of the episode by a paired 2-sided t test,
and 95% confidence intervals (CIs) are presented for these variables.
The annual statistics and demographic data from the 3 emergency medical
service systems are shown in Table 1.
The outcomes according to initial rhythm for patients in this study are shown
in Table 2.
Of the total 243 episodes correctly included, 67 were excluded because
of incompleteness of data. The main reasons for exclusion were failure to
apply the additional chest pad (35/67) and technical problems with the 2 data
cards or the defibrillator pads (26/67). In 13 episodes, signal quality made
ventilation count impossible; thus, ventilation data are reported for 163
Compression data are summarized in Table
3. For the first 5 minutes and for the entire resuscitation episode,
the mean (SD) fractions of the time without CPR (NFR) were 49% (21%) and 48%
(18%), respectively, and when subtracting the time necessary for analysis
and defibrillation, the NFRsadj were 42% (19%) and 38% (17%), respectively.
There was no difference in the mean NFR in the first 5 minutes vs during the
rest of the episode (49%; 95% CI, 46%-52% vs 50%; 95% CI, 47%-54%; P = .58), but there was a significant difference in NFRadj (42%; 95% CI, 39%-45% vs 38%; 95% CI, 35%-41%; P = .004).
For the first 5 minutes and for the entire resuscitation episode, mean
(SD) compressions were 60/min (25/min) and 64/min (23/min), respectively,
significantly lower during the first 5 minutes than during the rest of the
episode (60/min; 95% CI, 57-64/min vs 65/min; 95% CI, 61-69/min; P = .02). There were no significant differences with time
for any other variables. For the first 5 minutes and for the entire resuscitation
episode, mean (SD) chest compression rates were 120/min (20/min) and 121/min
(18/min); mean (SD) compression depth was 35 mm (10 mm) and 34 mm (9 mm);
the mean (SD) percentages of compressions with a depth between 38 and 51 mm
were 27% (30%) and 28% (25%); and the mean (SD) percentages of inappropriately
shallow compressions were 59% (37%) and 62% (33%). The compression parts of
the duty cycle were 41% (5%) and 42% (4%). Incomplete release occurred after
a median (interquartile range) of 0% (0%-1%) and 0% (0%-2%) of the compressions.
During the first 5 minutes, there was no occurrence of incomplete release
of compressions in 101 of 173 episodes (58%), and in only 16 episodes, more
than 10% of the compressions had incomplete release. Mean (SD) ventilations
were 8/min (4.6/min) and 11/min (4.7/min) for the first 5 minutes and for
the entire episode, respectively (Table 3).
A total of 61 patients (35%) achieved return of spontaneous circulation,
34 (19%) were admitted to the hospital, and 6 (3%) were discharged from the
hospital. Five of 6 patients who survived to hospital discharge had nearly
normal neurological function (Table 2).
Survival according to CPR quality indicators for patients with ventricular
fibrillation as initial rhythm are presented in Table 4.
In this study of 176 adults with out-of-hospital cardiac arrest, chest
compressions were given only half of the available time during these resuscitation
events. Van Alem et al12 reported that police
and firefighters performed CPR a mean (SD) of only 45% (15%) of the duration
during a median of 5 minutes of resuscitation before ambulance personnel took
over. In that study, two thirds of the time without CPR could be explained
by programmed interruptions from automated defibrillators. In our present
study, CPR was performed by paramedics and nurse anesthetists, and only 15%
to 20% of the time without CPR could be attributed to defibrillator use and
required pulse checks. The periods without chest compressions and the relatively
shallow compressions are not easily explained by focus on other tasks such
as intubation or placement of an intravenous cannula. These interventions
should occur during the initial minutes of ACLS, and there were only small
differences in the results for the first 5 minutes and the rest of the episodes.
Only good-quality CPR improved the chance of survival in 3 studies of
cardiac arrest patients.5-7 Chest
compressions appear to be the most important factor, both in human6 and animal studies,20,21 and
even short 4- to 5-second interruptions in chest compressions decrease coronary
perfusion pressure.22 In addition to periods
without chest compressions, more than half of chest compressions given in
the present study were too shallow, indicating less-than-optimal circulatory
effect of the CPR given. Arterial blood pressure increases with increasing
compression force in humans,23 and coronary
blood flow increases with increasing compression depth from 38 mm to 64 mm
in large pigs.24 Most compressions in the present
study were less than the recommended depth. This is in contrast with mannequin
studies of professional rescuers, in which 30% to 50% of the compressions
were too deep.25,26
In addition to compression depth, blood flow is dependent on compression
rate, compression/decompression ratio, and low intrathoracic pressure in the
decompression phase, avoiding “leaning” on the chest by the rescuer.
In canine and swine models, highest blood flows are reported with chest compression
rates of 90/min to 120/min,27-29 leading
to the guidelines recommendation of 100/min.2,3 Mean
compression rate tended to be too high in the present study, which might decrease
cardiac output because of insufficient time for venous return to the heart
during the decompression periods. “Leaning” on the chest wall
during compressions was not a serious problem, although we cannot exclude
that pressures lower than the 4 kg used to define leaning in the present study
could have an unwanted effect. The compression/decompression ratio was satisfactory,
with 41% to 42% compression time. The main problems were the long periods
without any chest compressions and the shallow compression depth.
We did not find abnormally high ventilation rates, although we recorded
the rate average over a minimum of 5 minutes. In contrast, Aufderheide et
al11 recently reported average ventilation
rates of 30/min (3/min) with maximal rates during any 16-second period.11 In animal models, ventilatory rates of 30/min vs
12/min decreased coronary perfusion pressure and also appears to decrease
survival if sustained for 4 minutes.11
Training programs for CPR have been implemented worldwide during the
last 4 decades following guidelines from the American Heart Association2 and the European Resuscitation Council.3 These
programs specify criteria for correct performance of CPR, but neither the
effects of such training programs on clinical CPR nor the effects of specific
criteria or overall quality of ACLS on patient survival have been clinically
documented. The present study was not powered to evaluate the effects of quality
of CPR in a proper multivariate analysis with other factors known to influence
survival, such as initial rhythm. A crude comparison between survivors and
nonsurvivors with ventricular fibrillation as initial rhythm showed a tendency
toward relatively less time without chest compressions among survivors, with
no difference in compression depth or ventilation rate (Table 4).
All paramedics and nurse anesthetists in the present study had previous
ACLS training with regular retraining, and all underwent a refresher course
immediately prior to study initiation. Some of the deviations from the international
2000 guidelines2,3 could be due
to lack of knowledge retention, as most studies have reported deterioration
in the performance of CPR within a few months after a course.8,10,30 The
failure to perform chest compressions half the available time has not been
reported in such studies, but they are all in mannequins,8,10,30 not
in patients. It is possible that the highly complex physical and mental situation
of treating a patient with cardiac arrest is too different from the training
situation on mannequins, making the performance dramatically different and
possibly less efficient. Based on this, the extrapolation from mannequin performance
can be questioned, and as a recent international consensus document states,
there is an urgent need to promote better CPR and improve the way CPR is taught.31
Whatever the reason, the resuscitation performance we measured was dramatically
different from that recommended in the ACLS guidelines. It is tempting to
question the focus on and the importance of details such as ventilation/compression
ratios of 1:5 or 2:15 or biphasic vs monophasic defibrillators in our efforts
to adjust evidence-based CPR guidelines, if the performance of vital skills
is so far from the guidelines recommendations.
Whether some of these deficiencies can be improved by specific focus
during training needs attention. Through better understanding of the mistakes
made in a real-life cardiac arrest situation, training courses might be designed
to focus on these aspects. Another approach would be to develop online tools
that prompt the rescuer to improved performance. Audiotapes giving instructions
on chest compression rate have been reported to improve the compression rate
during cardiac arrest in patients.16 In mannequin
studies, audio feedback based on continuous online automated evaluation dramatically
improved CPR performance within the first 3 minutes.32,33 According
to the international consensus, the ideal would be to have identically configured
aids during both training and resuscitation attempts.31
If our study represents how CPR is delivered during resuscitation from
out-of-hospital cardiac arrest in other communities, there is a great opportunity
to improve CPR quality and, hopefully, patient survival by focusing on delivery
of chest compressions of correct depth and rate, with minimal “hands-off”
Corresponding Author: Lars Wik, MD, PhD,
NAKOS, Institute for Experimental Medical Research, Ulleval University Hospital,
N-0407 Oslo, Norway.
Financial Disclosure: Mr Myklebust is an employee
of Laerdal Medical Corp, which developed the monitor/defibrillator.
Author Contributions: Drs Wik, Kramer-Johansen,
and Steen had full access to all of the data in the study and take responsibility
for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Wik, Kramer-Johansen,
Acquisition of data: Kramer-Johansen, Myklebust,
Sørebø, Svensson, Fellows.
Analysis and interpretation of data: Wik, Kramer-Johansen,
Myklebust, Svensson, Steen.
Drafting of the manuscript: Wik, Kramer-Johansen,
Critical revision of the manuscript for important
intellectual content: .Wik, Kramer-Johansen, Myklebust, Sørebø,
Svensson, Fellows, Steen.
Statistical expertise: Kramer-Johansen.
Obtained funding: Myklebust, Svensson, Steen.
Administrative, technical, or material support:
Wik, Myklebust, Svensson, Fellows, Steen.
Study supervision: Wik, Myklebust, Sørebø,
Svensson, Fellows, Steen.
Funding/Support and Role of Sponsors: Laerdal
Medical Corp (Stavanger, Norway) supplied defibrillators, the custom-made
computer program used for viewing and annotating the data, and the server
used. Laerdal paid the salaries for Mr Myklebust and other of their personnel
who preprocessed the data by filtering and down-sampling to 50 Hz. Laerdal
paid for 40 hours of instructor time for refresher ACLS courses in Stockholm,
Sweden, for overtime required for data handling at study sites, and for all
travel to study sites and investigator meetings. All other funding was obtained
from the following independent foundations: Norwegian Air Ambulance Foundation,
Laerdal Foundation for Acute Medicine, and Jahre Foundation. As stated in
the protocol, Laerdal Medical could not influence manuscript submission (their
employee, Mr Myklebust, could have withdrawn as an author). Except as stated
herein, the sponsors played no role in the design and conduct of the study,
in the collection, analysis, and interpretation of the data, or in the preparation,
review, or approval of the manuscript.
Acknowledgment: We thank all of the paramedics
and nurses who performed CPR for their contribution to this study. In addition,
the following CPR instructors were of exceptional value: Jan Ottem, Lars Didrik
Flingtorp, Helena Borovszky, RN, Lars Safsten, RN, Andrew Nord, and Allan
Bromley. We also thank Ståle Freyer, Mette Stavland, Linn Somme, and
Geir Inge Tellnes for their important technical help. Finally, we thank our
US collaborators, Lance Becker, MD, and Ben Abella, MD, MPhil, for their input
during the planning and performance of our study.