Abella BS, Alvarado JP, Myklebust H, Edelson DP, Barry A, O’Hearn N, Vanden Hoek TL, Becker LB. Quality of Cardiopulmonary Resuscitation During In-Hospital Cardiac Arrest. JAMA. 2005;293(3):305-310. doi:10.1001/jama.293.3.305
Author Affiliations: Sections of Emergency
Medicine (Drs Abella, Edelson, Vanden Hoek, and Becker, and Mr Alvarado and
Ms Barry) and Critical Care (Mr O’Hearn), University of Chicago Hospitals,
Chicago, Ill; and Laerdal Medical Corporation, Stavanger, Norway (Mr Myklebust).
Context The survival benefit of well-performed cardiopulmonary resuscitation
(CPR) is well-documented, but little objective data exist regarding actual
CPR quality during cardiac arrest. Recent studies have challenged the notion
that CPR is uniformly performed according to established international guidelines.
Objectives To measure multiple parameters of in-hospital CPR quality and to determine
compliance with published American Heart Association and international guidelines.
Design and Setting A prospective observational study of 67 patients who experienced in-hospital
cardiac arrest at the University of Chicago Hospitals, Chicago, Ill, between
December 11, 2002, and April 5, 2004. Using a monitor/defibrillator with novel
additional sensing capabilities, the parameters of CPR quality including chest
compression rate, compression depth, ventilation rate, and the fraction of
arrest time without chest compressions (no-flow fraction) were recorded.
Main Outcome Measure Adherence to American Heart Association and international CPR guidelines.
Results Analysis of the first 5 minutes of each resuscitation by 30-second segments
revealed that chest compression rates were less than 90/min in 28.1% of segments.
Compression depth was too shallow (defined as <38 mm) for 37.4% of compressions.
Ventilation rates were high, with 60.9% of segments containing a rate of more
than 20/min. Additionally, the mean (SD) no-flow fraction was 0.24 (0.18).
A 10-second pause each minute of arrest would yield a no-flow fraction of
0.17. A total of 27 patients (40.3%) achieved return of spontaneous circulation
and 7 (10.4%) were discharged from the hospital.
Conclusions In this study of in-hospital cardiac arrest, the quality of multiple
parameters of CPR was inconsistent and often did not meet published guideline
recommendations, even when performed by well-trained hospital staff. The importance
of high-quality CPR suggests the need for rescuer feedback and monitoring
of CPR quality during resuscitation efforts.
Survival from cardiac arrest remains low despite the introduction of
cardiopulmonary resuscitation (CPR) more than 50 years ago.1- 3 The
delivery of CPR, with correctly performed chest compressions and ventilations,
exerts a significant survival benefit in both animal and human studies.4- 8 Conversely,
interruptions in CPR or failure to provide compressions during cardiac arrest
(“no-flow time”) have been noted to have a negative impact on
survival in animal studies.7 Consensus guidelines
clearly define how CPR is to be performed,9 but
the parameters of CPR in actual practice are not routinely measured, nor is
the quality known.
There are multiple reasons for concern regarding the quality of CPR.
Even though CPR training programs are ubiquitous, a number of studies demonstrate
that these learned resuscitation skills deteriorate over time.10,11 Furthermore,
issues such as translation of skills from training environments to actual
cardiac arrest settings, as well as rescuer fatigue during resuscitation,12 may limit CPR quality. Recent investigations have
revealed that patients may be hyperventilated during out-of-hospital arrest,13 and that low chest compression rates are present
during in-hospital arrest.14,15
Given the proven survival benefit of high-quality CPR and the lack of
data on actual performance, we sought to determine whether well-trained hospital
staff perform CPR compressions and ventilations according to guideline recommendations.
The in-hospital environment was selected because it offers the added advantage
of thorough pre-arrest documentation as well as resuscitation by ample numbers
of highly trained personnel.
The study protocol and consent materials were approved by the institutional
review board at the University of Chicago Hospitals, Chicago, Ill. Data collection
was carefully structured to comply with all relevant Health Insurance Portability
and Accountability Act of 1996 regulations. Consent was obtained from all
members of the resuscitation teams via an oral consent process.
Resuscitation events were studied among inpatients at the University
of Chicago Hospitals who experienced cardiac arrest, defined by the documented
loss of a pulse and respirations as well as the delivery of chest compressions.
Patients were excluded for analysis if they experienced arrest in the operating
room or emergency department, were younger than 18 years, or if the CPR-sensing
defibrillator was used without its chest compression–detecting mechanism.
During in-hospital cardiac arrests, an investigational monitor/defibrillator
(IDE G020121) was used. This device is based on a commercially available monitor/defibrillator
(Heartstart 4000SP, Laerdal Medical Corporation, Stavanger, Norway) with the
additional investigational capabilities for capturing and recording rate and
depth of chest compressions, rate and volume of ventilations, presence or
absence of a pulse, as well as standard electrocardiogram and defibrillator
shock event data. In addition, customized software for data analysis collected
these parameters and calculated the no-flow time and no-flow fraction (NFF,
fraction of cardiac arrest time without compressions being performed). These
additional device features and analysis software were developed by engineers
at Laerdal Medical Corporation.
Chest compression data were captured via a special chest compression
pad outfitted with an accelerometer sensor (ADXL202e Analog Devices, Norwood,
Mass) and a pressure sensor (22PCCFBG6, Honeywell, Morristown, NJ). The pad
was placed on the mid-sternum of the patient under the hands of the rescuer
performing compressions. This method has been previously validated in the
laboratory setting, with compression depth data accurate to within 1.6 mm.16,17 Components of the sensing and recording
software have also been tested, validated, and published elsewhere.18,19 Additional testing has demonstrated
the use of impedance measurement for ventilation monitoring, in both swine20 and healthy human volunteers (P. A. Steen, oral communication,
2003). This latter human study was performed as a validation pilot study to
our current study and demonstrated a strong correlation between impedance
and spirometry waveforms.
Ventilation and pulse data were obtained using impedance measurements
captured from the defibrillation pads. All data collected by the device were
stored on data cards for subsequent analysis using additional custom software
that allowed for calculation of rates and other parameters. Per hospital regulation,
all users of the device and CPR performers were originally certified in either
basic life support (medical students and nurses), advanced cardiovascular
life support (all physicians), or both. The study device was utilized by the
hospital team that responds to all cardiac arrests. The study design was purely
observational with no alteration in therapy or suggested change from standard
resuscitation practice. Resuscitation teams were blinded to the results of
defibrillator measurements during the arrest. The patients studied represented
a convenience sample of all cardiac arrests during the study period, in that
during some other cardiac arrests another defibrillator was used instead of
the study device.
To determine CPR parameters, chest compression rate, depth, ventilation
rate, no-flow time, and NFF were calculated by Sister Studio software (Laerdal
Medical Corporation). Correct chest compression depth was defined as between
38 and 51 mm (1.5-2.0 in). (Current CPR guidelines do not take adult patient
characteristics into account in recommendations for CPR parameters; therefore,
we did not perform adjustments for any of these variables.) Pauses in chest
compressions of more than 1.5 seconds (for pulse checks and intubation) were
excluded from rate calculations so as to not artifactually lower chest compression
rate. Mean (SD) values were calculated for CPR parameters. No-flow time (time
periods of cardiac arrest without compressions being performed) was mathematically
defined as total time minus the time with chest compressions or spontaneous
circulation, and NFF was defined as the no-flow time divided by cardiac arrest
time (ie, total time minus time periods with spontaneous circulation). This
measure of NFF represents the fraction of time during the resuscitation episode
without cerebral or myocardial circulation.
All data were sent to the study investigator (H.M.) at Laerdal Medical
Corporation, where data were processed by filtering and down sampling to 50
Hz to prepare files for annotation and review. Proprietary software designed
for the study (Sister Studio) was used for processing each cardiac arrest
file. Raw data from each patient were collected as 2 separate data files.
One file contained impedance and chest compression data, while the second
file contained elements collected by the recording defibrillator (electrocardiogram
and shock times). These 2 data files were then conditioned, filtered, and
merged into a single data set for each patient by the study sponsor. At this
time the study sponsor did not analyze the data or perform interpretation
of waveforms. The merged conditioned files were then sent back to the study
site, where all data annotation, analysis, and interpretation were conducted.
This analysis involved a full annotation of the file to determine when a pulse
was present vs when cardiac arrest was present; the software would then read
compressions and ventilations, which were confirmed by a study investigator,
before a final data file was prepared that contained the parameters of interest
(compression rate, compression depth, ventilation rate, no-flow time). The
study sponsor did not perform interpretation or access the data during this
analysis phase. Secondary data analysis was performed using a spreadsheet
application (Excel, Microsoft Corp, Redmond, Wash).
For our outcome measures of CPR quality, we analyzed the first 5 minutes
of CPR, which was presumed to be both the best rescuer effort based on study
of rescuer fatigue12 and the most clinically
important. Each 5-minute resuscitation episode was divided into 30-second
segments, and both compression and ventilation rates were calculated. Segments
in which either chest compression or ventilation signals were obscured by
signal noise were excluded from analysis. Segments without compressions or
ventilations were excluded from calculations of mean compression or ventilation
rates, respectively. All files were manually evaluated by a physician investigator
to ensure appropriate software marking of events such as compressions, ventilations,
and rhythms. Similar analysis was also performed for entire cardiac arrest
episodes to provide comparison with the initial 5-minute data. No-flow fraction
was only calculated for the first 5-minute period.
Our study was not designed or powered to find CPR quality differences
between survivors and nonsurvivors; however, we undertook this evaluation
as a secondary analysis. Of the 67 arrest episodes, 60 had complete data sets
for comparison of all parameters. Cardiopulmonary resuscitation parameters
were compared between the cohort of patients that achieved return of spontaneous
circulation (ROSC) vs those who died during resuscitation. This analysis was
only conducted on data from the first 5 minutes of resuscitation efforts.
All means (SDs) were calculated using a spreadsheet application (Excel).
Differences in CPR parameters for outcome evaluation were assessed using a
2-tailed t test. Statistical evaluation of data was
performed independent of the study sponsor in consultation with a biostatistician
at our institution. P <.05 was considered statistically significant.
A total of 67 patients with cardiac arrest were treated using the study
defibrillator with data collection from December 11, 2002, to April 5, 2004.
Data analyzed from this cohort included 1073 segments (536.5 minutes) with
chest compression and ventilation data. Patient demographic and cardiac arrest
data are shown in Table 1. Mean (SD)
patient age was 62.2 (17.4) years, and 34.3% of patients were women. Patient
race included black (65.7%), white (23.9%), and other/unknown (10.5%) individuals.
Cardiac arrest events took place in intensive care settings (52.2%), general
wards (44.8%), or other locations (3.0%, radiology [n=1] and cardiac catheterization
laboratory [n=1]). Frequencies of the presenting rhythm were 14.9% ventricular
fibrillation/ventricular tachycardia, 59.7% pulseless electrical activity,
10.4% asystole, and 14.9% other (indeterminate). Return of spontaneous circulation
was achieved in 40.3% of patients. Baseline characteristics and rate of ROSC
are similar to data reported in other studies of in-hospital cardiac arrest.21
Cardiopulmonary resuscitation characteristics for the entire patient
cohort are shown in Table 2. During
the first 5 minutes of resuscitation, mean chest compression rate was less
than 90/min 28.1% of the time and less than 80/min 12.8% of the time. Chest
compression depth data revealed that chest compressions were too shallow (<38
mm depth) 37.4% of the time. Ventilation rates were calculated in a similar
fashion to chest compression rates. In contrast with compressions, ventilation
rates tended to be high; during 60.9% of segments, ventilations were performed
at a rate of more than 20/min. Ventilation volumes did not appear to deviate
greatly from physiological ranges and are not reported herein. Analysis of
the time with cardiac arrest but without compressions (NFF) yielded a mean
(SD) of 0.24 (0.18) with 40.3% of the segments having an NFF of more than
Although the intent of this investigation was only to objectively describe
multiple parameters of CPR during cardiac arrest, we considered whether ROSC
was associated with better CPR quality. We did not find any statistically
significant differences in chest compression rate, depth, ventilation rate,
or NFF between patients who achieved ROSC vs those who did not (Table 3). A trend toward lower NFF was observed for patients with
ROSC compared with nonsurvivors. We did not expect to find clinical outcome
differences given our small patient cohort and the nonrandomized nature of
the study; therefore, we cannot draw any conclusions regarding the direct
clinical impact of the quality of CPR on survival.
Our study represents, to our knowledge, the first multiparameter, quantitative
recordings of actual CPR during in-hospital cardiac arrest. Using impedance
measurement techniques, we found that quality of CPR was often deficient from
guideline recommendations9 in several specific
parameters, including chest compression rate, compression depth, ventilation
rate, and NFF. Specifically, chest compression rates were often less than
the recommended 100/min, compression depth was often more shallow than the
minimum 38 mm, ventilation rate was higher than the recommended 12 to 16/min,
and NFF was longer than adherence to recommendations might allow (although
not clearly specified in the guidelines, a 10-second pulse check every minute
of CPR would yield an NFF of 0.17).
These data confirm other recent investigations13- 15 suggesting
that CPR quality may be highly variable in actual practice. Just as we observed
frequent overventilation, Aufderheide et al13 recently
showed that paramedics hyperventilate patients during out-of-hospital cardiac
arrest, and parallel animal experiments confirmed that this degree of hyperventilation
led to decreased survival. We recently documented low chest compression rates
during in-hospital cardiac arrest in a multicenter study when recorded by
observers equipped with a handheld device to record compression rate.15 A smaller observer-based study found low chest compression
rates during in-hospital arrest.14
Cardiopulmonary resuscitation performance in our study may have been
affected by the knowledge that rescuers were being studied. This “Hawthorne
effect”22 would likely have led to improved
CPR quality and would minimize our findings of significant deviations from
recommended practice. In addition, due to institutional review board requirements,
we did not link individuals performing CPR with CPR-quality data. However,
resuscitation teams change each month (with resident rotations), with completely
new rescuers. Therefore, it is unlikely that an individual rescuer performed
CPR in more than approximately 4 to 5 cardiac arrests.
The paramount importance of CPR has been confirmed in both animal and
human studies. In 2 clinical studies, survival from ventricular fibrillation
arrest was improved if CPR was performed before defibrillation attempts.23,24 In animal studies, coronary perfusion
pressure, hemodynamic function, and survival were adversely affected by even
short pauses in chest compressions.25,26 Moreover,
pauses in chest compression just before defibrillation worsened outcomes in
a swine model.27 Additionally, laboratory study
has shown that physiological and survival outcomes are sensitive to CPR quality.28,29 Mechanical devices that provide chest
compressions at consistent rate and depth have shown promise toward improving
There are several limitations to our study. A primary limitation is
that the precise contribution to survival of the specific parameters that
were measured is unknown. Although an isolated compression rate of less than
100/min can be considered a failure to adhere to a published recommendation
of the American Heart Association, we cannot determine whether this “deficiency”
is directly linked to worsened survival. Support for objective CPR quality
monitoring lies in the fact that this technology will allow future studies
to carefully examine the effects of CPR parameters on survival.
Additional limitations are that filtered electrocardiogram and ventilation
signals were occasionally overcome by artifact, which caused us to exclude
some segments. Chest compression depth as studied was calibrated for presence
of a backboard and therefore depth may be overestimated if a backboard was
not used during the resuscitation. For this reason, we describe in our analysis
only compressions that are too shallow. Although our study is limited by use
of a single site for data collection, we believe these results are likely
generalizable to other hospitals, just as our prior results demonstrated chest
compression rate deficiencies when studied at 3 hospitals.15 Performance
difficulties during stressful and disorganized cardiac arrest settings, the
lack of reliable internal timing to pace chest compressions, rescuer fatigue,12 and infrequent recertification in CPR31 may
all contribute to the observed deficiencies. It is therefore likely that our
findings are representative of a more general dilemma in resuscitation. Human
factors in CPR performance are important and at this point underinvestigated
areas of research.32
Our study has implications for the conduct and design of future clinical
CPR studies. Cardiopulmonary resuscitation quality is currently an unmeasured
but potentially important confounder in most published clinical studies involving
cardiac arrest outcomes. The importance of this variable given the current
ability to measure these parameters should be considered by researchers attempting
to study methods for improving survival from cardiac arrest.
There are several potential practical solutions for helping to improve
poor CPR quality. The first involves mechanical devices that can provide chest
compressions reliably at a set rate and depth.33 These
devices may generate better hemodynamic characteristics than manual chest
compressions.34,35 Another solution
is to improve monitoring and feedback to reduce human error during manual
CPR, by using devices such as end-tidal CO2 monitors36 and “smart defibrillators,” which can
measure CPR characteristics and provide audio feedback to alert the rescuers
to errors such as incorrect chest compression or ventilation rate.18,19
Corresponding Author: Lance B. Becker, MD,
University of Chicago Hospitals, Section of Emergency Medicine, 5841 S Maryland
Ave, MC 5068, Chicago, IL 60637 (firstname.lastname@example.org).
Financial Disclosures: Mr Myklebust is an employee
of Laerdal Medical Corporation, which developed the monitor/defibrillator.
Dr Becker has received grant/research support from Philips Medical Systems,
Laerdal Medical Corp, and Alsius Corp, and has served as a consultant for
Abbott Laboratories and Philips Medical Systems.
Author Contributions: Dr Becker had full access
to all of the data in the study and takes responsibility for the integrity
of the data and the accuracy of the data analysis.
Study concept and design: Abella, Myklebust,
Vanden Hoek, Becker.
Acquisition of data: Abella, Alvarado, Myklebust,
Barry, O’Hearn, Becker.
Analysis and interpretation of data: Abella,
Myklebust, Edelson, Barry, O’Hearn, Vanden Hoek, Becker.
Drafting of the manuscript: Abella, Barry,
O’Hearn, Vanden Hoek, Becker.
Critical revision of the manuscript for important
intellectual content: Abella, Alvarado, Myklebust, Edelson, Vanden
Obtained funding: Abella, Vanden Hoek, Becker.
Administrative, technical, or material support:
Alvarado, Myklebust, Edelson, Barry, O’Hearn, Vanden Hoek, Becker.
Study supervision: Abella, Vanden Hoek, Becker.
Funding/Support: This study was supported by
a grant from the Laerdal Medical Corporation, Stavanger, Norway.
Role of the Sponsor: One of the authors, Mr
Myklebust, is employed at Laerdal Medical Corporation and was involved in
study conception and design; however, Laerdal Medical Corporation had no role
in data collection, interpretation of results, or drafting of the manuscript.
Acknowledgment: We thank Raina Merchant, MD,
David Beiser, MD, and Kuang-Ning Huang, for constructive discussions during
our study; Lynne Harnish, for expert administrative assistance; and Ted Karrison,
PhD, for assistance with statistical analysis. We also thank our European
collaborators, Petter Andreas Steen, MD, PhD, Lars Wik, MD, PhD, Fritz Sterz,
MD, and Jo Kramer-Johansen, MD, for important input during the planning and
execution of our study.