The quality of chest compressions by trained personnel: The effect of feedback, via the CPREzy, in a randomized controlled trial using a manikin model

Abstract

Even after training, the ability to perform effective cardiac compressions has been found to be poor and to decrease rapidly. We assessed this ability with and without a non-invasive feedback device, the CPREzy, during a 270s CPR session in an unannounced, single-blinded manikin study using 224 hospital employees and staff chosen at random and using a non-cross over design. The two groups self-assessed their knowledge and skills as adequate. However, the control group (N=111) had significantly more difficulty in delivering chest compressions deeper than 4 cm (25 versus 1 candidate in the CPREzy group), P=0.0001. The control group compressed ineffectively in 36% (+/-41%) of all compressions as opposed to 6+/-13% in the CPREzy group (N=112, P=0.0001). If compressions were effective initially, the time until >50% of compressions were less than 4 cm deep was 75+/-81s in the control group versus 194+/-87 s in the CPREzy group (P=0.0001 [-180 to -57.5]). After a few seconds of training in its use, our candidates used the CPREzy effectively. Against the background knowledge that estimation of compression depth by the rescuer or other team members is difficult, and that performing effective compressions is the cornerstone of any resuscitation attempt, our data suggests that a feedback device such as the CPREzy should be used consistently during resuscitation.

Full text

Resuscitation (2006) 69, 241—252
TRAINING AND EDUCATION
The quality of chest compressions by trained
personnel: The effect of feedback, via the
CPREzy, in a randomized controlled trial using a
manikin model夽
Gerrit J. Noordergraafa,∗, Bianca W.P.M. Drinkwaardb,
Paul F.J. van Berkomc, Hans P. van Hemert c, Alyssa Venemac,
Gert J. Schefferd, Abraham Noordergraaf e
aDepartment of Anaesthesiology, St. Elisabeth Hospital, Tilburg, Hilvarenbeekseweg 60, 5022 GC
Tilburg, The Netherlands
bDepartment of Emergency Medicine (during the study: Resident, Department of Anaesthesiology),
St. Elisabeth Hospital, Tilburg, The Netherlands
cInstructor CPRLab, Sector of the Department’s for Education & Research and Anaesthesiology,
St. Elisabeth Hospital, Tilburg, The Netherlands
dDepartment of Anaesthesiology, University Medical Centre Nijmegen Geert Grooteplein 10,
Nijmegen, NL, USA
eUniversity of Pennsylvania, Philadelphia, PA, USA
Received 11 May 2005; received in revised form 27 July 2005; accepted 3 August 2005
KEYWORDS
Basic life support (BLS);
Cardiac massage;
Cardiopulmonary
resuscitation (CPR);
Impedance;
Manikin;
Chest compliance;
CPREzy
Summary Even after training, the ability to perform effective cardiac compres-
sions has been found to be poor and to decrease rapidly. We assessed this ability with
and without a non-invasive feedback device, the CPREzyTM, during a 270 s CPR ses-
sion in an unannounced, single-blinded manikin study using 224 hospital employees
and staff chosen at random and using a non-cross over design.
The two groups self-assessed their knowledge and skills as adequate. However, the
control group (N= 111) had significantly more difficulty in delivering chest compres-
sions deeper than 4 cm (25 versus 1 candidate in the CPREzy group), P= 0.0001. The
control group compressed ineffectively in 36% (±41%) of all compressions as opposed
to 6 ±13% in the CPREzyTM group (N= 112, P= 0.0001). If compressions were effec-
tive initially, the time until >50% of compressions were less than 4 cm deep was
75 ±81 s in the control group versus 194 ±87 s in the CPREzyTM group (P= 0.0001
[−180 to −57.5]). After a few seconds of training in its use, our candidates used the
CPREzyTM effectively.
夽A Spanish translated version of the summary and keywords of this article appears as Appendix in the online version
at 10.1016/j.resuscitation.2005.08.008.
∗Corresponding author. Tel.: +31 13 539 8017; fax: +31 13 50 44 926.
E-mail address: gj.noordergraaf@wxs.nl (G.J. Noordergraaf).
0300-9572/$ — see front matter © 2005 Published by Elsevier Ireland Ltd.
doi:10.1016/j.resuscitation.2005.08.008

242 G.J. Noordergraaf et al.
Against the background knowledge that estimation of compression depth by the res-
cuer or other team members is difficult, and that performing effective compressions
is the cornerstone of any resuscitation attempt, our data suggests that a feedback
device such as the CPREzyTM should be used consistently during resuscitation.
© 2005 Published by Elsevier Ireland Ltd.
Introduction
Hospitals are settings to which the ‘chain of sur-
vival’ applies. Interest has recently shifted from
improvements in advanced cardiac life support
towards improvements in basic life support (BLS).
As in any other emergency medical system, in our
hospital first responders provide basic life support
until the rapid response team (RRT) arrives. The
first responders continue to assist as needed, for
example in performing the chest compressions. The
importance of consistent and adequate chest com-
pressions has been stressed,1,2 with extrapolation
suggesting that chest compressions may be the cru-
cial factor in improving outcome.3,4
All personnel in our hospital, not only nursing
staff, participate in structured, intensive training
and certification, but the frequency of training may
be too low to guarantee continuing adequate skills.
Loss of skills has been observed as early as a few
days after training.5,6 A further complicating factor
is the poor ability of rescuers to assess their own
resuscitation skills,7,8 and those of others, espe-
cially in actual cases of resuscitation.9Feedback
has now been introduced10 to reduce subjective
evaluation of rescuers during external chest com-
pressions. Initial tests have been encouraging.11,12
For the purpose of expanding these tests and to
gain further understanding of the mechanical engi-
neering involved, we performed two types of eval-
uation with the CPREzyTM (Health Affairs, London,
UK). These were: (1) assessment of compression
skills of hospital personnel with and without the
use of the non-invasive feedback device, and (2)
description of the indicators on the CPREzyTM as a
mechanical model on a stiff surface with and with-
out an underlying manikin or human chest.
Materials and methods
All medical, nursing and support staff members in
our hospital were eligible for participation in the
study. Over a period of 10 days, during which no
training was being given, and using a roster to insure
representation of different departments, the inves-
tigators toured the hospital, recruiting staff as they
were seen. By referring to management support if
needed, staff could not ‘‘bow out’’ using work load,
other tasking, or even coffee breaks as excuses.
Candidates were ‘‘asked’’ individually to come
to a quiet, private, location where the assess-
ment was to be performed. The first questionnaire
(Appendix A) was filled out, followed by randomiza-
tion to the ‘control’ or ‘Ezy’ group, and a standard-
ized briefing given. Assessment followed using one
of four SkillReporter Resusci Annes (Laerdal, Sta-
vanger, Norway). The practical session, with ven-
tilation being performed by an investigator who
offered neither suggestions nor comments, contin-
ued for at least 240 s, but no longer than 270 s. This
period reflects the maximal time a caregiver might
be expected to perform chest compressions over
one uninterrupted period in our hospital, as set out
in the hospital resuscitation standing orders. It was
deemed to be long enough to allow for fatigue, as
suggested by Baubin.13 After the practical session,
a second questionnaire (Appendix B) was filled out.
The candidate was also asked to sign a form giv-
ing permission to use the data, and to access their
training database. The two questionnaires and the
print out from the SkillReporter were labelled with
a unique randomization number. Two investigators
experienced in the use of the standardized scoring
system, blinded for the randomization, scored the
results based on a fixed list of variables (Table 1),
and entered directly into the SPSS database. The
candidate was not told how he/she had performed.
The candidates randomized to the ‘Ezy’ group
used the CPREzyTM (Health Affairs, London, UK).
This device was not in use in our hospital and was
unknown to all the candidates. It is a small, light
(260 g) device, with a resting height of 5 cm, a
base of 5.5 cm ×18 cm, and is powered by a 9 V
battery (Figure 1). The device is placed on the
sternum, and features an illustration to assist with
the correct location. It has a non-slip pad on the
surface applied to the sternum. When turned on
(green) indicator lights show compression force cor-
related to the weight of the patient. The indications
start with one light for a ‘child’ (40 kg/90 lbs); two
lights for a ‘small adult’ (55 kg/120 lbs), three for
an ‘average adult’ (75 kg/165 lbs), four lights for a
‘large adult’ (90 kg/200 lbs) and a fifth (red) light

The quality of chest compressions by trained personnel 243
Table 1 Listing of points scored for each candidate
Overall Per compression set
(15:2)
Randomization (blinded) Block number
Manikin code Time block started (s)
Total # compressions Number leaning
Total time (s) Actual depth/leaning
compression
Total number 15:2 Direction leaning
Mean compression
frequency
Actual
depth/compression
Mean compression depth Mean depth/15
compressions
Compression depth lost
(mm)
Depth loss/15
compressions
Total # compressions Number ‘out of range’/15
compressions
Total time (s) Compression frequency
Compression:relaxation
ratio
Scoring was done by one of two investigators blinded for the
randomization and entered into a database directly.
for ‘extra large adult’. The force needed to activate
the indicator lights for the different sizes is 23, 32,
41, 50, and 54 kg, respectively.11,12 A metronome
flashes an orange light and emits a 60 dB beep at
100 min−1in accordance with International Liaison
Committee on Resuscitation (ILCOR) guidelines.14
Candidates randomized to the ‘control’ group per-
formed standard chest compressions. Candidates
were requested to maintain secrecy if asked the
Figure 1 The CPREzy as used in the trial. The large
surface for compression (A) is labelled, as is the LED indi-
cation for pressure applied (B). The on/off button (C) is
next to the metronome light (D). The arrows indicate the
moving planes.
reason for their temporary absence. Approval for
the study was given by the Medical Ethics Commit-
tee.
For the registration of skills the Resusci AnneTM
was connected to its external monitor, the
‘‘SkillReporter’’, which uses lights to give feedback
and which also produces a written record although
not in ‘real time’. During the assessment period,
the lights on the SkillReporter were not used and
the device turned so that the record and print out
were not visible to the candidate. On the writ-
ten record, the chest compression is recoded as
a stroke and movement of the chest, and venti-
lation volumes are recorded as curves. Compres-
sion:relaxation ratios must be calculated. The Skill-
Reporter also has an internal data log system which
can be printed at the end of a compression stroke.
The usefulness of this data in our experimental
setup was limited as we analyzed 270 s and worked
on a compression to compression basis.
Each compression stroke was inspected care-
fully, using a standard technique first developed by
Berden et al. as a standardizing evaluation tool.15
This system is in regular use for the courses given in
our hospital, and has been adapted to current ILCOR
guidelines. A measuring ‘ruler’ is used by us to
improve accuracy further and limit inter-observer
scoring variations.
The SkillReporter Resusci Annes were tested
beforehand by the hospital medical instrumenta-
tion service to validate the methods. In the first
series, a standardized force was applied to the
manikin and correlated with the SkillReporter writ-
ten record for compression depth. No relevant
(<5%, ≤2 mm inter-manikin variation) differences
could be found in the range expected to be rele-
vant during the study of 20, 30, 40, 50 and 60 kg. A
second test evaluated these forces applied to the
CPREzyTM in relation to the indicator lights and the
actual depth in the manikin while on a firm surface.
The force to activate the CPREzy indicator lights
confirmed the manufacturer’s specifications,12 but
also demonstrated a larger spread during dynamic
testing, as the interpretation of the indicator light
scale is approximate with steps between lights of
4—9 kg. This reflected itself in a maximum 5 mm
spread, both within and between manikins. The
difficulty lay in applying the exact force needed
to just light the indicator light during dynamic
testing. During static testing, we confirmed the
data from Boyle and Perkins,12,13 in that at least
three lights, and optimally, four lights should (just)
be illuminated to achieve adequate depths in our
model (Appendix B). The results of these tests
were also used to check for inter-observer variabil-
ity, which was negligible. This pilot used a tech-

244 G.J. Noordergraaf et al.
nique described elsewhere.16 The validation tests
were performed using the DPM3 Universal Biome-
ter (DPM3, BIO-Tek Instruments Inc. Winooski, VT
USA), which can measure pressure applied to 1% of
accuracy.
The principal endpoint was the number of cor-
rect compressions a rescuer could perform during
the test session. Numeric assessment of compres-
sion depth, incomplete relaxation, compression to
relaxation ratio and compression frequency were
scored as defined by the 2000 ILCOR guidelines.
As each candidate had been trained, these scores
were related to their skills during the course exit
test. More specifically, hand position was scored as
incorrect if a lateral force was registered by the
manikin, and incomplete relaxation if the chest
did not return to the ‘resting’ position (≥1 mm)
between compressions. Each variable was scored
independently of the other. Secondary endpoints
were: the ability of the candidates to assess their
compressions and the work required, loss of com-
pression depth during the course of the session, fre-
quency with which the cardiac compressions were
performed, and whether individual characteristics,
such as age, weight, body-mass-index (BMI) influ-
enced their capabilities.
For each candidate, the results were entered
into an SPSS (v12) database for statistical analysis.
P< 0.05 was taken to be significant, with data being
presented as mean (±standard deviation), with 95%
confidence intervals presented in brackets. After
testing for normality using the Kurtosis test and the
Levene’s test for equality in variance, the (two-
tailed) independent samples T test was used for
parametric data, with correlations sought using
the (two-tailed) Pearson’s coefficient, with miss-
ing cases excluded pair wise, for parametric val-
ues. Using the general linear model, repeated mea-
sures analysis as performed with time and depth
as variables. The Huynh-Feldt epsilon was used
if spherecity conditions were not demonstrated.
The Mann—Whitney Uand Chi-squared tests were
used for independent sample analysis of non-
parametric data. The (two-tailed) Spearman’s rho,
was then also used, with missing cases excluded
pair wise.
Results
Two-hundred and twenty-four candidates, includ-
ing physicians, nursing and non-nursing staff, were
included (Table 2). One candidate was not random-
ized due to recent neck complaints and was not
replaced. The 223 remaining were randomized to
the ‘control’ and the ‘Ezy’ groups (N= 111 and 112
candidates, respectively). All candidates were able
to produce usable records that were entered into
the database for analysis. Two were unable to com-
plete the nominal 240 s of chest compressions (pro-
tocol deviation) due to technical difficulties with
the manikin. In four cases, the candidate stopped
early due to shortness of breath (n= 1), or other
physical discomfort (n= 3). Their data remained in
Table 2 Descriptive statistics for the candidates
Control mean (±S.D.) CPREzy mean (±S.D.) P[95% CI]
N111 112
Gender Female: 72 Female: 82 N.S. (0.521)
Male: 39 Male: 30
Age (years) 36.6 (±10.5) 35.4 (±10.2) N.S. (0.401)
Weight (kg) 72.8 (±14.4) 71.6 (±11.9) N.S. (0.518)
Height (cm) 173.2 (±9.5) 172.2 (±8.8) N.S. (0.414)
BMI (kg/m2) range 23.8 (23.5—24.8) 23.9 (23.5—24.9) N.S. (0.822)
Ability at end of training (number and
% of ineffective compressions)
155 (±50) 159 (±48) N.S. (0.626)
4.5 (±9) 4.8 (±10) N.S. (0.880)
Function Nurse: 62 Nurse: 71 N.S. (0.159)
Non-nurse: 31 Non-nurse: 31 Between groups
Physician: 18 Physician: 10
Time since training (months) 16.4 ±15 14.5 ±14 N.S. (0.09)
Actual (recent) BLS experience (n) 26 23 N.S. (0.603)
Data is presented as mean ±standard deviation and the 95% confidence interval in square brackets, except when non-parametric
(median + 95% confidence interval). n= number of cases. This is reported as the actual number for that individual variable if data
was missing. No significant differences were found between the two groups, for any category even though there was a tendency
to more physicians in the control group. The candidate’s abilities are reported on the basis of the post-test skill measurements.

The quality of chest compressions by trained personnel 245
Table 3 Pre- and post-assessment questionnaire
Control mean (±S.D.) CPREzy mean (±S.D.) P[95% CI]
VAS score N= 111 N= 112
Pre-assessment
Knowledge: at the end of last training
course?
7.0 (±1.4) (N= 109) 7.0 (±1.3) (N= 109) N.S. (0.891)
Knowledge: this minute? (±1.7) 5.8 (±1.8) N.S. (0.538)
Post-assessment
Opinion: (0—10) how good were
compression?
5.9 (±1.6) 5.9 (±1.8) (N= 110) N.S. (0.591)
Opinion: depth (5 = correct) 5.2 (±1.8) 4.7 (±1.8) 0.039 [0.025—0.98]
Opinion: effective compressions (%) 53 (±21) 53 (±20) N.S. (0.911)
Opinion: how tiring was CPR? 4.8 (±2.2) (N= 110) 4.7 (±2.2) N.S. (0.714)
Correct depth is? (knowledge) 32 39 N.S. (0.338)
Opinion (0—10) how good was frequency? (±1.6) (N= 109) 6.2 (±1.7) N.S. (0.407)
Correct frequency is? (knowledge) 46 40 N.S. (0.381)
Opinion of own confidence 5.9 (±1.8) 6.0 (±1.9) N.S. (0.688)
The VAS score is used with a non-calibrated 10 cm line on which the candidate marked their answer. In the two questions labelled
‘knowledge’, the candidate was asked to answer with the ILCOR guidelines. This was scored as correct or as not correct. N.S.:
not significant. Data is presented as mean (±standard deviation), except when non-parametric. If significant the 95% confidence
interval [95% CI] is reported. N: number of cases. This is reported in the individual variable if data was missing. No significant
differences were found between the two groups, except for in the difference between estimation of compression depth (P= 0.039).
This difference does not seem clinically relevant.
the database for analysis. Failures of the CPREzyTM
did not occur.
The two groups were well balanced in their phys-
ical characteristics and potential skills. In particu-
lar, similarity in the ‘time since last training’ (16.4
versus 14.5 months) and the ‘compressions skills at
the end of training’ (95 versus 95% effective com-
pressions) in the control and the CPREzyTM groups,
respectively, suggest that randomization was ade-
quate (Table 2). In the pre-assessment question-
naire, all the candidates reported that they were
capable of performing BLS-CPR {6.9 (±1.4) versus
7(±1.3); P= 0.536}at the end of their last course
(Table 3). They estimated that their current skills
were 5.7 (±1.7) out of 10, versus 5.8 (±1.8), (con-
trol and CPREzy groups, respectively; P= 0.538),
with their knowledge estimations essentially the
same. In the post-assessment questionnaire, these
estimations became 5.9 (±1.6) and 5.9 (±1.8),
P= 0.591. Both groups reported that they felt that
they had performed about 53% of the compressions
correctly, while the CPREzy group was less confi-
dent in actually having achieved and maintained
adequate depth {4.7 (±1.8) versus 5.2 (±1.8) for
the control group, respectively, with P= 0.039 [95%
CI = 0.025—0.98]). Similar results were reported for
the physical work needed to perform chest com-
pressions. These were 4.8 (±2.2) versus 4.7 (±2.2),
respectively (P= 0.714), on a 10-point scale with 0
being ‘extremely tiring’ and 10 being ‘no effort at
all’ (Table 3).
Practical skills differed markedly between the
two groups (Table 4). The control group had signifi-
cantly more difficulty with achieving and maintain-
ing effective compressions over time even though
this was not reflected in their opinion during the
post-assessment questionnaire (P= 0.591). This lack
of performance could be expressed in the number
of candidates unable to perform compression depth
of more than 4 cm (25 in the control group versus
1 in the CPREzy group: P= 0.0001). Of the remain-
ing candidates, 48 members of the control group
started with adequate compressions but lost com-
pression depth progressively, reaching<4cm by 75
(±81) s, without recognizing or correcting this inad-
equacy for the remainder of the trial period. In
the CPREzy group, the minimum effective depth
threshold was maintained by all but 11 members up
to 194 (±87) s (P= 0.0001 between groups for time
as well as number of candidates reaching thresh-
old, see also Figure 2 {CI −180 to −57.5}). The
number of ineffective compressions for the con-
trol group, within each block of 15 compressions,
ranged from a mean of 4 (±6) initially to 7 (±7) at
240 s, as opposed to 1 (±4) consistently throughout
the trial for the CPREzy group (see also Figure 3,
P= 0.0001 CI = 59—100). The overall percentage of
candidates not reaching a mean 4 cm compression
depth during each block of 15 compressions ranged
from 21% at 15 s to 38% at 270 s. The incidence of
incomplete relaxation was limited in both groups
(13 and 14 cases, respectively), while 50 candidates

246 G.J. Noordergraaf et al.
Table 4 Practical skills assessment (compression and frequency aspects)
Control mean (±S.D.) CPREzy mean
(±S.D.)
P[95% CI]
N111 112
Actual time (s) 260 (±26) 258 (±23) N.S. (0.589)
Actual number of 15:2 cycles 21 (±4) 21 (±3) N.S. (0.784)
Mean depth (mm) 40 (±9) 45 (±4) 0.0001 [−0.32 to −2.15]
Depth loss (mm) 4 (±5) (N= 106) 3 (±4)
(N= 109)
N.S. (0.143)
Incidence of ineffective chest
compressions (N)
25 (never correct) 1 (never
correct)
0.0001
No. of candidates compressing <4 cm
consistently (after an adequate start)
48 11 0.001
Time until compressions become and
remained <4 cm (s)
75 (±81) 194 (±87) 0.0001 [−180 to −57.5]
Total ineffective compression (% of all
compressions)
36 (±41) 6 (±13) 0.0001 [22.7—38.7]
Ineffective compressions (n)94(±104) 15 (±33) 0.0001 [59—100]
Leaning (incidence) 13 14 N.S.
Incorrect hand position (incidence
15:2−1)
136 88 0.001
Actual compression frequency (cpm) 106 (±21) 102 (±10) N.S. (0.056)
This table summarizes the practical skills. mm: millimetres of impression; s: seconds after initiation of CPR; N: number of cases;
cpm: compressions/min. Ineffective compressions are those outside the ILCOR range. If a subgroup is reported, the actual number
of cases is listed. Leaning was scored as present if the registration showed >1 mm non-return to resting position. Hand position
was scored if an ‘incorrect hand position’ exclamation mark was listed on the written record (not specific for the four potential
sites of the error: see text for explanation).
in the control group and 98 in the CPREzy group
demonstrated consistently correct hand positions
(P= 0.001). Compression frequency did not differ
significantly between groups (Table 4).
Correlations between the use of the CPREzy, the
adequacy of their compressions and potential con-
founders such as caregiver physiognomy, time since
last training, weight, and the self-opinion of their
abilities, could not be demonstrated. Notably, a
rescuers opinion could not be correlated to their
actual ability when the skill was expressed as ade-
quate compressions (P= 0.38, r=−0.084).
Discussion
This unannounced study assessed the compression
skills of 224 trained rescuers with and without the
use of an unfamiliar non-invasive feedback device,
the CPREzyTM, in a manikin setting. Two investi-
Figure 2 Compression depth over time (data presented as mean ±S.D. for each time segment). Compression depth
measured from 0 (resting). Measurements of a compression to compression basis, with the series of compressions being
scored closest to the time interval.

The quality of chest compressions by trained personnel 247
Figure 3 Number of ineffective compressions by group over time (data presented as mean ±S.D.). Measured on a
compression to compression basis. Ineffective defined as compressions (measurably) less than 4 cm depth.
gators, blinded for the randomization, and using a
standardized scoring system, evaluated the effec-
tiveness on a compression-to-compression basis and
scored pre and post assessment questionnaires. The
use of a feedback device in a large, non cross-
over, manikin study with rescuers at known levels
of training and time since training, and a question-
naire, had not been performed.
We found, as expected, a generally moderate
ability to perform adequate chest compressions at
an average of 15 months after the most recent
training, with a mean of 6 of each 15 compres-
sion series being inadequate over time. While the
assessment lasted up to 270 s, it demonstrates
(Figure 3) that the period during which one res-
cuer may compress the chest effectively, may need
to be limited further to about 120 s. When the
CPREzyTM is used, the overall percentage of ade-
quate compression during the assessment period
was increased significantly, from an average of 9
(±8) compressions of sufficient depth in the control
group to at least 14 (±2) per 15 compression series
in the Ezy group. This improvement in effectiveness
is both for consistency of depth as well as for ade-
quacy of depth. Use of the CPREzyTM was not cor-
related with improvements in frequency, nor did it
increase incomplete relaxation between compres-
sions. However, the low incidence of these errors
may have contributed to this outcome.
Our study demonstrates that better definition
and consistency of force during chest compressions
is feasible in a population of rescuers not specif-
ically trained in the use of the CPREzy. This is a
significant prerequisite for application in patient
resuscitation. Compression force and depth assess-
able during chest compressions has received lit-
tle attention. This may have been caused by early
emphasis on getting the lay public to perform
compressions to the exclusion of aspects difficult
to teach. Outcome studies have been careful to
score ‘early access’ while avoiding estimation of
adequacy.14,17
Early work, performed by Thomas et al., reports
on the use of a force indicating gauge to improve
depth estimation. They report improvement of
compression efficacy from 33 to 96% in a manikin
study of trained flight nurses using a cross-over
design.10 However, they drew no general conclu-
sions. Their impressive results could not be repro-
duced entirely by Elding et al., although this
study also demonstrated statistically significant
improvement in compression efficacy and general
technique.18
The CPREzyTM, first described by Boyle et al.
in 2002 in a limited group assessed without and
with the CPREzyTM on consecutive days, found a
3-fold improvement (13—42%) in effective compres-
sions if the CPREzyTM was used.11 The suggestion
that this improvement may not translate to clini-
cal improvement seems justified in their study, as
the rate of effective compressions remains low even
after instruction and introduction of the feedback
device. Perkins et al. validated Boyle’s results using
a small group of medical students when resusci-
tating on a bed.12 In this later study, the students
received instructions, practiced, and were told how
many indicator lights were optimal, with some of
the test candidates having been active in the vali-
dation series.
Our data confirms but also expands on this expe-
rience. While our improvement, expressed as a per-
centage is smaller, it brings the percentage of ade-
quate compressions in line with what seem to be
realistic clinical demands. In addition, we used an
interrupted compression model in order to simulate
the discontinuity this brings with it, incorporating
a maximum time that one rescuer may need to
perform BLS before being relieved. We avoided a

248 G.J. Noordergraaf et al.
cross-over design to exclude a learning curve as
may have occurred in Boyle’s and Perkins’s studies.
A potential individual learning effect of uncertain
magnitude cannot be ruled out. We increased the
number of participants to limit any bias caused by
skills (Table 2). Perhaps most importantly, we did
not train our rescuers in the use of the CPREzyTM:
they were confronted with it as randomized, and
were instructed to ‘‘use the lights as indicated and
begin immediately’’, so as to standardize unfamil-
iarity most likely to occur in the clinical setting. We
also demonstrate that improvement is independent
of time since training, rescuer weight and function
within the hospital. The mean time since training in
our assessment was 15 months (range 0—37 months
with a normal distribution). A sub-group analysis
of those with training less than 6 months earlier
and those with training between 12 and 18 months
demonstrated that the CPREzy maintains skills, as
suggested by others. These results demonstrated
that the use of the CPREzy might increase the use-
ful interval extensively, giving it both a teaching as
well as practical role.
The need for quality in basic life support, both
early and during the advanced stages of a resusci-
tation effort need not be expanded upon.1,2,12,19
Without a feedback device, such as the CPREzyTM,
the rescuer as well as the physician have to depend
on their ‘experience’ and ‘memory’ to evaluate the
effectiveness of the compressions, while evidence
suggests that this is neither taught in courses nor
is it clinically possible for instructors or rescuers.9
Their experience relates to manikins in the train-
ing situation, and negates understanding of the
variables such as the loss of rescuer capabilities
over time, individual chest wall stiffness and chest
diameter.2,4 The characteristics of a device, such
as the CPREzy, with its effect on the force applied
to the individual patient (Appendix B) is, perhaps
regrettably, unclear to many. The CPREzyTM pro-
vides one of these variables and empirically a sec-
ond, thereby reducing the number of unknowns in
a user-friendly manner. The feedback device also
gives the rescuer a benchmark for compressions,
allowing physician delegation and monitoring of this
basic life support task. Its use can be considered an
adjuvant to CPR as is ETCO2monitoring, and expen-
sive methods such as impedance evaluation built
into defibrillators.
In our hospital the CPREzy is brought to the scene
by the advanced life support team, and imple-
mented as the first step in their protocol: although
the first minutes of chest compressions may be sub-
optimal, it allows improved maintenance of chest
compressions during advanced life-support proce-
dures, which may continue for up to an hour.
The importance of chest compressions has been
rediscovered.2,4,19 The exact amount of force
required to create an optimal artificial circulation
in humans is still a matter of discussion. Force
needed for compressions labelled as adequate vary
from manikin to manikin, and 20—70 kg for human
adults.4Suggestions that force and depth should be
individualized have been expressed,2but may need
to remain in the realm of advanced skills. Perkins
et al. described the force needed when working
with the CPREzy as a range from child to extra-
large adult as 23—54 kg. Timerman et al, working
with a novel chest compression device, reported
that they used 51 ±20 kg of peak force in their
population.12,20 Experimental determination of the
force to compression relationship with the CPREzy
on a rigid surface demonstrated agreement with the
values listed above. Doubts have been expressed in
the past about whether such force indicators retain
their meaning when the CPREzy is used on a flexi-
ble or compressible support,12,14 such as the human
chest. Our experience confirms this accuracy, while
recognizing that the amount of physical work on a
multiple layer support exceeds that of compression
on a rigid surface. Clinically, this allows the res-
cuer performing chest compressions to recognize
that, should the patient be lying in a hospital bed,
the total distance their hands move may be more
than 4—5 cm, and may initially require extra work
to achieve correct depth, even on modern hospital
beds, as this position often is not used for training.
The Resusci Anne uses its SkillReporter lights
to specify incorrect hand position. Incorrect posi-
tion of the hands, and the application of force in
a non vertical direction while the hands are cor-
rectly placed, will be scored as incorrect. During
our study, we used the written record which does
not specify the location/direction of the error. Ear-
lier investigation found a propensity for too low a
position of the CPREzy12; our study design, which
relied on the written record, does not allow us to
confirm or repudiate this aspect in chest compres-
sions.
Our study has a number of limitations. We did
not use a cross-over design as discussed above; by
including large groups of candidates at random, we
allowed for variables such as motivation, physiog-
nomy, and skills to correct themselves in the sam-
pling. As the candidates were unaware of which task
they would perform next, any bias should be lim-
ited. Human evaluation of the written record was
also used: while the blinded investigators, dedi-
cated resuscitation officers, have extensive skills
and practices in evaluating the records, and the
benchmarking did not demonstrate relevant dif-
ferences, more checks during scoring may have

The quality of chest compressions by trained personnel 249
increased security as to the value of the scores.
However, a post hoc analysis of the database, using
the investigator as a variable, could not demon-
strate any systematic differences.
While the device increases the possibility for the
rescuer to choose a force and judge the consistency
of their compressions, it does not allow insight into
the force which might be optimal for that patient.
In manikin studies this effect is difficult to simu-
late and the manikin may not be related to actual
clinical conditions.21 Our study also used manikins,
and relied on the written record as an accurate rep-
resentation of reality.15 However, even the CPREzy
does not allow for more accuracy than increments
of 3—9 kg of force. The actual force required, or
even the optimal impression depths in humans are
still subject to debate4and may be greater than
currently thought (i.e. 70 kg or more needed).
In order to evaluate compression skills optimally,
we de-emphasized ventilation. We used the com-
pression ventilation ratios for one-rescuer CPR, but
supplied a non-intervening rescuer to perform these
ventilations. While introducing interruptions for the
ventilations, our protocol did not require the move-
ment of hands to open the airway and the additional
fatigue caused by mouth-to-mouth ventilation was
not evaluated. This did allow for a large number of
compressions to be evaluated in a brief period of
time.
Conclusions
Our study demonstrates a marked improvement in
achieving and maintaining adequate depth during
chest compressions when a feedback device, the
CPREzyTM, is used. This difference was found even
though the rescuers were not specifically trained in
its use. Although it requires more work, the vari-
ation in depth is significantly smaller than without
the device, regardless of physiognomy of the care-
giver. It also shows that this device de-emphasises
the interval after training without compromising
quality. The improvements in efficacy should be an
important factor in optimization of the ‘Chain of
Survival’.
Acknowledgements
Preliminary data was presented as a poster at the
3rd Conference on Emergency Medicine (Leuven,
B). No outside support was given for the study.
None of the authors has a financial relationship or
interest with or in the development and sale of the
CPREzyTM. The authors would like to thank F. Jansen
and B. Tax for their technical assistance, and Dr. J.
Giele for her help with the statistics.
Appendix A
A.1. The pre-assessment questionnaire
This was filled in by the candidates before they had
been briefed on the explicit purpose of the assess-
ment. The visual analogue scale (VAS) was mea-
sured to be exactly 10 cm. The candidates placed
an ‘x’ at a position of their self evaluation. Time
to last course was validated using the hospital’s
training database. BLS-CPR = basic life support car-
diopulmonary resuscitation.

250 G.J. Noordergraaf et al.
A.2. The post-assessment questionnaire
This was filled in immediately after completion of
the practical session.
Appendix B. The effects of combining
springs with different properties as
applied to CPR
During CPR courses, caregivers are taught to com-
press the chest and to strive for 4—5 cm compres-
sion depth. Little is said about the role of force
needed for this. Applying sufficient force to com-
press the chest deep enough can be a challenge.
The CPREzy technology suggests that a force of
50—54 kg on the CPREzy may be adequate for com-
pressions in an adult. Most adults can produce such
a force for limited periods of time and we suggest
that this feedback may help in controlling and main-
taining this effort. But does the use of the feedback
system also influence the work the caregiver must
produce?
This appendix approaches that question from a
simplified, mathematical, point of view.
In Figure 4, an object such as a manikin or the
chest wall of a human being, represented as a weak
spring is compressed while lying on a firm surface
[22]. The stiffness of the chest wall, d, and the
spring compression x1are related to the loading
force F1via the equation:
F1=dx1(1)
In Figure 5, a spring-loaded device (e.g. the CPREzy)
is put on a firm surface. The stiffness of the spring
Dand the compression of the spring x2are related
to a loading force F2via the equation:
F2=Dx2(2)
If the above two springs are placed one on top of the
other (Figure 6), i.e. in series, such as the CPREzy
on the sternum of the chest, with the patient lying
on a firm surface, continuity of force requires that:
F1=F2(3)
Figure 4 See Appendix B for explanation of the symbols.
Schematic diagram of a weak spring (e.g., the human
chest) on a firm surface.

The quality of chest compressions by trained personnel 251
Figure 5 See Appendix B for explanation of the symbols.
Schematic diagram of a stiff spring on a firm surface.
Hence
x1=d
Dx2(4)
The work done on the spring in the first example
(Figure 4) equals:
W1=
Fx1
2(5)
The work done on the combined springs (Figure 6)
is equal to:
W2=
Fx2
2+
Fx1
2=F
2(x1+x2)
=Fx1
21+d
D (6)
Figure 6 See Appendix B for an explanation of sym-
bols. Schematic diagram of the CPREzy on the chest of a
manikin lying on a firm surface demonstrating the accu-
mulation of work.
Hence, for the same compression of the lighter
spring (i.e. the patient or the manikin, x1) the ratio
of the amounts of work equals:
W2
W1
=1+ d
D (7)
For the particular combination of the CPREzy with
our manikin, measurements show that d/D= 1.07.
For example using Eq. (7) compressing the combi-
nation of springs will require close to twice as much
work as compressing the CPREzy or the manikin
alone, without affecting the force needed. The
results of the study demonstrate that this should
not be a clinical concern.
It does clarify why caregivers may feel that their
hands are uncomfortable and why they are tiring:
the amount of work being performed has increased.
This increase is most likely both in terms of the
force being applied (due to the feedback) as well
as to the hard, a relatively small surface on which
it is applied. In effect, it has also become possible
to individualize force and depth both as a teaching
as a clinical tool.
B.1. Conclusions
In these models, force indicators are not signif-
icantly disturbed by multiple layers, when their
masses can be ignored. However, levels of com-
pression may be entirely different, depending of
the stiffnesses of the springs. The physical work
performed by the caregiver also depends on spring
stiffnesses and is increased in our example.
References
1. Steen S, Liao Q, Pierre L, Paskevicius A, Sjorberg T. The crit-
ical importance of minimal delay between chest compres-
sions and subsequent defibrillation: a hemodynamic expla-
nation. Resuscitation 2003;58:249—58.
2. Abella BS, Alvarado JP, Myklebust H, Edelson DP, Barry A,
O’Hearn N, et al. Quality of cardiopulmonary resuscitation
during in-hospital cardiac arrest. JAMA 2005;293(3):305—10.
3. Eisenberger P, Safar P. Life supporting first aid training
of the public—–review and recommendations. Resuscitation
1999;41:3—18.
4. Noordergraaf GJ, Tilborg GFAJB van, Schoonen JAP, Otte-
sen J, Noordergraaf A. Thoracic CT-scans and cardiovascular
models: the effect of external force in CPR. Int J Cardiovasc
Med Sci 2005;5(1):1—7.
5. Berden HJJM, Bierens JJLM, Willems FF, Hendrick JMA, Pijls
NHJ, Knape JTA. Resuscitation skills of lay public after recent
training. Ann Emerg Med 1994;23(5):1003—8.
6. Swor R, Compton S, Vining F, Osoky Farr L, Kokko S, Pascual
R, et al. A randomized controlled trial of chest compres-
sion only CPR for older adults—–a pilot study. Resuscitation
2003;58:177—85.

252 G.J. Noordergraaf et al.
7. Crunden EJ. An investigation into why qualified nurses inap-
propriately describe their own cardiopulmonary resuscita-
tion skills. J Adv Nursing 1991;16:597—605.
8. Hightower D, Thomas SH, Stone CK, Dunn K, March JA. Decay
in quality of closed-chest compressions over time. Ann Emerg
Med 1995;26(3):300—3.
9. Brennan RT, Braslow A. Skill mastery in cardiopul-
monary resuscitation training classes. Am J Emerg Med
1994;13(5):505—8.
10. Thomas SH, Stone C, Austin PE, March JA, Brinkley S. Utiliza-
tion of a pressure-sensing monitor to improve in-flight chest
compressions. Am J Emerg Med 1995;13(2):155—7.
11. Boyle AJ, Wilson AM, Connelly K, McGuigan L, Wilson J, Whit-
bourn R. Improvement in timing and effectiveness of exter-
nal cardiac compressions with a new non-invasive device:
the CPR-Ezy. Resuscitation 2002;54:63—7.
12. Perkins GD, Augre C, Rogers H, Allan M, Thickett DR.
CPREzyTM: an evaluation during simulated cardiac arrest on
a hospital bed. Resuscitation 2005;64(1):103—8.
13. Baubin MA, Schirmer M, Nogkler M, Semenitz B, Falk
M, Kroessen G, et al. Rescuer’s work capacity and
duration of cardiopulmonary resuscitation. Resuscitation
1996;33:135—9.
14. Guidelines 2000 for cardiopulmonary resuscitation and emer-
gency cardiovascular care. Circulation 1994;102(8):I-1—384.
15. Berden HJJM, Pijls NHJ, Willems FF, Hendrick JMA, Crul JF. A
scoring system for basic cardiac life support skills in training
situations. Resuscitation 1992;23:21—31.
16. Kopka A, Crawford J. Cricoid pressure: a simple yet effective
biofeedback trainer. Eur J Anaesthesiol 2004;21:443—7.
17. Halperin H, Chandra NC, Levin HR, Rayburn BK, Tsitlik JE.
Compression techniques and blood flow during cardiopul-
monary resuscitation. Respir Care 1995;40(4):380—92.
18. Elding C, Baskett P, Hughes A. The study of effectiveness
of chest compressions using the CPR-plus. Resuscitation
1998;36:169—73.
19. Wik L. Rediscovering the importance of chest compres-
sions to improve outcome from cardiac arrest. Resuscitation
2003;58:267—9.
20. Timerman S, Cardoso LF, RamiresF J.A.F., Halperin H.
Improved hemodynamic performance with a novel chest
compression device during treatment of in-hospital cardiac
arrest. Resuscitation 2004;61:273—80.
21. Baubin MA, Gilly H, Posch A, Schinnerl A, Kroesen GA.
Compression characteristics of CPR manikins. Resuscitation
1995;30:117—26.
22. Tsitlik JE, Weisfeldt ML, Chandra N, Effron MB, Halperin
HR, Levin HR. Elastic properties of the human chest
during cardiopulmonary resuscitation. Crit Care Med
1983;11(9):685—91.