Journal of Critical Care (2006) 21, 271 – 279

Clinical Research—Pediatric

The pediatric early warning system score:

A severity of illness score to predict urgent
medical need in hospitalized childrenB,BB
Heather Duncana, James Hutchisonb,c,d, Christopher S. Parshuramb,c,d,*
a
Paediatric Intensive Care Unit, Diana, Princess of Wales Children’s Hospital, Steelhouse Lane, B4 6NH Birmingham, UK
b
Department of Critical Care Medicine, University of Toronto, Toronto Ontario, Canada M5G 1X8
c
The Research Institute, Hospital for Sick Children, University of Toronto, Toronto Ontario, Canada M5G 1X8
d
Department of Pediatrics, University of Toronto, Toronto Ontario, Canada M5G 1X8

Keywords: Abstract
Early detection;
Purpose: We developed and performed the initial retrospective validation of a pediatric severity of
Prevention;
illness score. The score is to preemptively identify hospitalized children who are likely to require
Code blue;
resuscitation to treat cardiopulmonary arrest.
Severity of illness
Materials and Methods: The Pediatric Early Warning System (PEWS) score was developed using expert
opinion. The score generated contained 20 items, 16 of which were able to be retrospectively abstracted.
Validation used a case-control study design in a Canadian university–affiliated pediatric hospital. Eligible
patients were younger than 18 years, were admitted to a hospital ward, and had no level-of-care
restrictions. Case patients had a code blue called to obtain immediate assistance for treatment of impending
or actual cardiopulmonary arrest. Control patients had no code blue event and were not urgently admitted
to the intensive care unit within 48 hours of study. A total of 128 controls and 87 cases were compared.
Results: The PEWS score area under the receiver operating characteristic curve was 0.90. The sensitivity
was 78% and the specificity was 95% at a score of 5.
Conclusions: Application of the score may have identified more than 3 quarters of code blue calls in our
hospital with at least an hour’s warning. After further refinement and validation, the PEWS score has great
potential to increase the efficiency of care delivery and to improve the outcomes of care provided to
hospitalized children.
D 2006 Elsevier Inc. All rights reserved.

B
This work was supported by internal funding from the Department of Critical Care Medicine and the Research Institute at the Hospital for Sick Children,
Toronto, Ontario, Canada. The investigators functioned independent of these funding sources, and have no conflict of interest to declare with respect to this
publication. Drs C Parshuram and J Hutchison have received peer-reviewed funding from the Heart and Stroke Foundation of Canada to support ongoing
research in the prevention and treatment of pediatric cardiopulmonary arrest. This funding was in part related to the work described in the submitted manuscript.
BB
H Duncan contributed to the concept, design, interpretation, and manuscript revisions. J Hutchison contributed to the design, interpretation,
and manuscript revisions. CS Parshuram contributed to the concept, design, interpretation, and drafting of the manuscript and assisted with its revisions. All
authors approved the final submitted manuscript.
* Corresponding author. Department of Critical Care Medicine, Research Institute, Hospital for Sick Children, Department of Pediatrics, University of
Toronto, Toronto, Ontario, Canada M5G 1X8.
E-mail address: christopher.parshuram@sickkids.ca (C.S. Parshuram).

0883-9441/$ – see front matter D 2006 Elsevier Inc. All rights reserved.
doi:10.1016/j.jcrc.2006.06.007
272
1. Introduction
Up to 3% of children admitted to hospital wards require
immediate medical assistance for treatment of actual or
impending cardiopulmonary arrest [1,2]. These children
have increased morbidity and mortality associated with
cardiopulmonary arrest [3,4]. Currently, this catastrophic
clinical deterioration is treated by bcode blueQ teams called
to provide immediate resuscitation.
In adults, preemptive management may prevent cardio-
pulmonary arrest and improve mortality [5]. The provision
of critical care boutreachQ services to deliver preemptive
care is recommended in Britain [6]. However, the provision
of preemptive care relies on the timely identification of
patients at risk and referral to the responding medical
emergency team. The incomplete identification of patients
may have contributed to the negative findings of the MERIT
study, a cluster randomized trial of medical emergency
teams in 23 Australian hospitals [7].
Two approaches to the timely identification of patients at
risk may be used. First is the use of calling criteria, where
patients meeting one or more specific btriggeringQ criteria
are referred. Alternatively, b early warningQ scores may be
used. These severity of illness scores combine clinical
parameters into a single score. Patients with scores greater
than a threshold are identified and referred.
Early warning scores compliment clinical decision
making and can identify trends independent of practitioner
experience or clinical workload. Our systematic review
found no scores for use in children and found one partially
validated adult score [8]. We decided to develop a simple
bedside score to preemptively identify children who require
resuscitation to treat actual or impending cardiopulmonary
arrest (Fig. 1).
2. Methods
A score to identify children with increasing severity of
illness was developed using expert opinion synthesized by
a modified Delphi method [9]. The performance of the
score was evaluated with a frequency-matched case-control
design. Hospitalized children without code blue or urgent
intensive care unit (ICU) admission were compared with the
sicker children for whom a code blue was called.
2.1. Score development
Items were generated by 2 focus groups of experienced
acute care nurses, facilitated by one of the investigators
(CP). Items were divided into 3 domains: patient-related
physiologic measurements, patient-related demographic,
and care-related factors. These were reclassified into
different domains: dynamic, where measurements were
anticipated to change frequently (eg, heart rate, blood
pressure, respiratory rate); static, where measurements
change infrequently (days-months, eg, previous ICU admis-
H. Duncan et al.
Fig. 1 The PEWS concept. Trajectories of illness in 3
hospitalized children. Child A is admitted to hospital, underwent
initial stabilization, and has an uneventful hospital course before
discharge. Child B and Child C continue to deteriorate once
admitted to hospital. Child B is identified by the PEWS score,
receives preemptive therapy, and improves. The clinical deteriora-
tion in child C was not recognized and culminated in a
cardiopulmonary arrest.
sion, postoperative); and staffing-related items (eg, nurse-
patient ratio). We excluded staffing-related items from the
score, as staffing variables are planned as part of the graded
responses to Pediatric Early Warning System (PEWS) score
within the pediatric early warning system.
One focus group met again to reduce the dynamic items
to those that were thought to be feasibly measured and
documented as part of routine care. A modified Delphi
method was used [9]. Items required a 75% majority to be
either excluded or included, or were included after 3 rounds
of inconclusive voting. The focus group then categorized
selected items into age-based and age-independent catego-
ries, and assigned weights. The investigators reviewed
the static items. Items with unclear definitions and those
that would be cumbersome to use in a bedside score
were excluded.
2.2. Clinical data
Eligible children were admitted to a hospital ward at the
Hospital for Sick Children, were younger than 18 years at
hospital admission, and had no prespecified care limitations.
Case patients were defined as children who had code blue
calls made as part of their care. In our hospital, a code blue
is called for children who are assessed to require additional
and immediate medical assistance for the treatment of actual
or impending cardiopulmonary arrest. Case patients were
retrospectively identified from the resuscitation committee
database for the 28-month period ending March 2003.
Control patients were defined as children who had no
code blue event and were not admitted to the pediatric
The pediatric early warning system score 273

Table 1 Items generated by nurse focus groups

ICU in the 48 hours after the period studied. The control
patients were retrospectively identified from a list of child-
Dynamic Static Staffing related
ren selected by matching the admission ward and age
Heart rate History Nurse-patient category of the code blue patients with other patients ad-
ratio mitted to the hospital during the study period. The controls
Respiratory rate Diagnosisa Physician were selected from the first medical records available for
(primary, secondary) availability
review until a ratio of 1 control patient to 1 case patient
Respiratory effort Airway abnormality Skill mix
Temperature Aspiration Years of
was exceeded.
pneumoniaa experience Clinical data were abstracted as directed by the focus
Systolic blood Severe Ward groups. Documented vital signs, clinical findings, and their
pressure developmental delay assessment times were recorded. In case patients, data
Diastolic blood Allergies Off service ward collection began 25 hours before the code blue call. In
pressure control patients, data were collected for the 24 hours
O2 saturation Past cardiac arresta beginning at the first 1:00 am of either hospitalization or
after ICU discharge.
Level of Intercurrent factors
consciousness
2.3. Initial analysis and score development
Pupils Recent transfer
from ICU The proportions of children who were cases and controls
Central venous Postoperative for the static domain items were compared. Static items
pressure within 24 h
Urine output Previous intubationa
Perfusion Paced
Color Change
primary service
Pulses Change of ward
within 24 h

Behavior Psychosocial factors

Laboratory Patient anxiety
values
Therapy Parental concern

Fluid bolus Illness severity

Oxygen therapy Home oxygen
Gastrostomy tube
Central venous line
Transplantb
Tracheostomy
Previous ICU
admission

Patient complexity
Comorbiditiesa
No. of medications

18 Total, 22 Total, 6 Total, none

7 evaluated 13 evaluated evaluated
a
Data for these items were not collected because it was assumed to
be unreliable due to either potential elective admission in other hospitals
(previous intubation) or high probability of subsequent admission to
ICU or death (cardiac arrest). Aspiration pneumonia was not included
because this diagnosis lacked clear definition and has potential overlap Fig. 2 PEWS score development and validation. Representation
with severe (dependent) cerebral palsy. Data on bpatient complexityQ of the sequence of processes comprising the development of the
and bother morbiditiesQ were not evaluated to avoid the assessment of a PEWS score. Initial item generation by 2 expert focus groups
long list of specific diagnoses and the creation of a complex score. The was followed by item reduction by 3 methods and resulted in
location of the patients was not analyzed because it was modified as part removal of 22 items. Four items were unable to be abstracted
of the experimental design (frequency matching).
b from medical records, and after exclusion of a further 4 items
Transplant recipients included children with bone marrow and
solid organ transplants.
based on initial evaluation of the clinical data, 16 items remained
in the score.
274 H. Duncan et al.

Table 2 Evaluation of static domain items

studied. The area under the receiver operating characteristic
(ROC) curve was calculated.
Factor Case N Control N P (*)
The performance of the evolving score was assessed in a
All patients 88 128 stepwise fashion. First, the dynamic items were evaluated.
Included The static items were then formed into 2 groups (medi-
Abnormal airway 9 1 .0016 cations and other items). These groups were sequentially
Home oxygen 6 1 .03
added to the dynamic items and the evolving score was
Previous ICU admission 43 5 b.0001
Central line in situ 21 5 b.0001
reevaluated. Finally, the investigators experimentally mod-
Transplant (any) 11 1 .0003 ified the weighting of the score.
Severe cerebral palsy 16 1 b.0001
G tube 26 1 b.0001 2.5. Evaluation of final score
z3 Medical specialties 30 9 b.0001
Excluded The resulting PEWS score was evaluated by 2 methods.
Tracheostomy 3 2 .40 First, the ability of the score to discriminate between case
Postoperative 7 11 .87 and control patients was tested in the entire sample and
in last 24 h within each age category. Second, changes in the score over
Transfers time were evaluated. In case patients, scores from 24 to
Ward in last 24 h 10 13 .77 12 hours, 12 to 6 hours, 6 to 2 hours, and in 1 hour ending
From ICU in last 24 h 3 2 .40 60 minutes before the code blue event were compared. In
Primary specialty 1 2 1.0 control patients, PEWS scores were compared in similar
in last 24 h periods (24-12, 12-6, and 6-2 hours and in the last hour)
Comparisons were made by v 2 test unless any cell count was less than ending at the completion of the 24 hours of clinical data
5, when Fisher exact test was used.
collected (1:00 am).
Institutional research ethics board approval was obtained
were included if they occurred in more than 5% of cases and before commencing the study. Funding was provided by
if comparisons were significant at the P = .10 level. the Department of Critical Care Medicine and the Re-
Because the information about the dynamic items from search Institute at the Hospital for Sick Children, Toronto,
each child was anticipated to contain missing values, it was Ontario, Canada.
decided a priori that missing values were to be assumed
normal. To reduce the impact of missing information, 3. Results
measurements were bchunkedQ into 1-hour blocks and the
most extreme values within each block were used for score 3.1. Focus groups
calculation. Next, the maximum score was determined for
each child from the 24 hours ending 1 hour before the code Ten nurses (with more than 140 years of combined
blue call in case patients and for the specified 24-hour nursing experience) formed the 2 focus groups (ICU and
period for control patients. The number of 1-hour blocks ward based). The focus groups decided that the PEWS
with information was compared for each group. should be able to be scored repeatedly throughout the child’s
hospital admission. They decided that PEWS score would
2.4. Assessment of potential scores be calculated by summation of subscores from changing
clinical data and more static background factors.
The ability of scores to discriminate between case The focus groups generated 47 items (Table 1). After
patients and control patients was assessed by logistic exclusion of 6 staffing-related and 7 dynamic items by
regression using the maximum score for the 24-hour period the second meeting of the ICU focus group and a further

Table 3 PEWS score performance assessed by logistic regression

Score components ca P % Discordant
Dynamic 0.823 b.0001 11.5
Dynamic + medications 0.855 b.0001 9.9
Dynamic + medications + demographic 0.904 b.0001 7.0
2  Dynamic + medications + demographic 0.892 b.0001 8.9
5  Dynamic + medications + demographic 0.865 b.0001 12.1
The dynamic items are heart rate, respiratory rate, systolic blood pressure, transcutaneous oxygen saturation, temperature, oxygen therapy, and fluid
therapy. Medications is the categorization of the number of medications taken per day, and demographic items are the presence of abnormal airway, home
oxygen, previous ICU admission, central line, transplant (any), severe cerebral palsy, G tube, 3 or more medical specialties. Experimental modification of
weighting by the developers is included (2VS = 2  vital signs, 5VS = 5  vital signs).
a
The c statistic from logistic regression was used to measure the area under the ROC curve.
The pediatric early warning system score 275

Table 4 The PEWS score

Item subscores
2 1 0 1 2
Age-specific items
b3 mo
HR b90 90-109 110-150 151-180 N180
RR b20 20-29 30-60 61-80 N80
SBP b50 50-59 60-80 81-100 N100
3-12 mo
HR b80 80-99 100-150 151-170 N170
RR b20 20-24 25-50 51-70 N70
SBP b70 70-79 80-100 99-120 N120
1-4 y
HR b70 70-89 90-120 121-150 N150
RR b15 15-19 20-40 41-60 N60
SBP b75 75-89 90-110 111-125 N125
4-12 y
HR b60 60-69 70-110 111-130 N130
RR b12 12-19 20-30 31-40 N40
SBP b80 80-90 90-120 120-130 N130
N12 y
HR b50 50-59 60-100 101-120 N120
RR b8 8-11 12-16 15-24 N24
SBP b86 85-101 100-130 131-150 N150
General items
Pulses Absent Doppler Present Bounding
O2 saturation (%) b85 85-95 N95
Capillary refill CRT N3 2-3 CRT b2
LOC b7 7-11 12-15
Oxygen therapy N50% or N4l/min Any b50% or b4l/min None
Bolus fluid Any None
Temperature b35 35-b36 36 N38.5-b40 N40
Score is calculated by the summation of the demographic and medication sub-scores. One point is scored for each item present from the following:
abnormal airway (not tracheostomy), home oxygen, any previous admission to an ICU, central venous line in situ, transplant recipient, severe cerebral
palsy, gastrostomy tube, and greater than 3 medical specialties involved in care. The medication subscore is from the number of medication administered in
24 hours. V3 = 0, 4-6 = 1, 7-9 = 2, 9-12 = 3, 12-15 = 4, z16 = 5. The maximum score from the full score is 34; however, 4 items were not tested, making
the maximum total 26 points.
HR indicates heart rate (beats/min); RR, respiratory rate (breaths/min); SBP, systolic blood pressure (mm Hg); LOC, level of consciousness measured
with the Glasgow Coma Scale.

Table 5 Maximum PEWS scores

Cases Controls Pa Area under ROC curveb
N Mean N Mean
Overall 87 7.9 128 3.2 b.0014 0.895
b3 mo 13 6.9 30 3.2 .0020 0.896
3-12 mo 19 9.7 15 2.4 b.0001c 1.0
1-4 y 15 7.3 32 3.1 .0004 0.875
4-12 y 21 7 21 3.2 .0023 0.833
N12 y 19 8.2 30 3.6 .0017 0.896
Maximum PEWS scores from the 24 hours ending 1 hour before code blue events in cases, and for 24 hours beginning on the second day of hospital
admission or after ICU discharge in controls. Scores were calculated from the vital signs aggregated using the most extreme vital signs for 1-hour periods.
Absent values were assumed as normal. The maximum score for each child was determined and mean scores were calculated for the specified groups.
a
Probability of type 1 error from logistic regression.
b
The c statistic from logistic regression was used to measure the area under the ROC curve.
c
Logistic estimates unstable as complete separation of cases and controls, therefore t test comparison reported.
276 H. Duncan et al.

Fourteen static domain items were assessed in 88 case

patients (one case had no vital signs in the medical
record) and 128 control patients; 9 (64%) met inclusion
criteria (Table 2) and were included in the score. The
number of medications in 24 hours was categorized and
scored a priori and was included as a separate item. Case
patients had more hour-length blocks of data in the 24-hour
interval than control patients (mean, 17.5 vs 11.9; P b .0001).

3.3. Assessment of potential scores

Sequential assessment of the scores showed that the
addition of medications and then the bbackgroundQ items
improved the performance of the score from the area
under the ROC curve (from 0.82 to 0.90). Manipulation
of the weighting by the investigators did not improve the
score (Table 3).

3.4. Final score evaluation

Fig. 3 The ROC curve. The comparison of sensitivity vs
specificity for PEWS scores for 87 patients with code blue and
128 well control patients. The area under the ROC curve is 0.90.
9 static items by the investigators, 25 items remained.
These remaining items comprised 11 dynamic and 14 static
items (Fig. 2).
Five age categories were defined. Three items were
classified by age. The other items were classified without
reference to age. The static items (excepting the number of
medications per day) were assigned 1 point when present
and none when absent.
3.2. Clinical data and initial analysis
During the study period, 99 (0.31%) of the 32,233
children admitted to our hospital had code blue events.
Complete clinical data were obtained from 87 (87%) of
these sick children and from 128 (16%) of the 782 children
that were identified as potential controls. Four of the
dynamic items (urine output, perfusion, pulses, and level of
consciousness) could not be reliably abstracted and were
not analyzed.
The final PEWS score evaluated included 16 items
(9 static and 7 dynamic) and could range from 0 to 26
(Table 4). The score was greater in case patients than
control patients overall (mean maximum score, 7.9 vs 3.2;
P b .0001) and within each age category (Table 5). The
score could discriminate between cases and controls, both
overall and within each age category (area under the ROC
curve, 0.83-1.0). At a threshold score of 5, the sensitivity
was 78% and specificity was 95% (Fig. 3, Table 6). The
subscore from the static items (medication plus demo-
graphics) were greater than 5 in 12 (14%) of case patients
and in none of the control patients. Despite this observa-
tion, the PEWS scores of children with code blue events
tended to increase as the code blue event neared ( P = .08).
A time-related change in PEWS scores was not found in
controls ( P = .27).
4. Discussion
The PEWS score is the first severity of illness score for
children admitted to hospital wards. In the initial evaluation,
the PEWS score could differentiate between bwell Q children

Table 6 PEWS: sensitivity, specificity, and positive predictive value

Threshold Cases correctly Controls incorrectly Sensitivity Specificity Positive predictive No. of false
score identified identified (%) (%) valuea (%) positivesa
0 87 126 100 2 0.31 31 729
1 87 114 100 11 0.34 28 708
2 83 77 95 40 0.49 19 390
3 79 52 91 59 0.68 13 095
4 72 25 83 80 1.3 6 296
5 68 7 78 95 4.2 1 763
6 59 4 68 97 6.2 1 007
7 47 2 54 98 9.6 504
8 39 0 45 100 100 0
a
Assuming an incidence of code blue calls of 0.31% of admissions.
The pediatric early warning system score
requiring routine ward-based care and sicker children who
subsequently had an immediate need for medical interven-
tion. At a threshold score of 5, the sensitivity and specificity
were 78% and 95%, respectively. The PEWS score
identifies patients with at least a 1-hour warning before
the code blue event. This should be sufficient time to initiate
ward-based management and to arrange transfer to higher
acuity patient-care area as required.
4.1. Strengths and weaknesses
This is the most detailed development of any early
warning score. The score was developed for nurses to use at
the bedside and was based on the knowledge of experienced
acute care nurses, with minor modifications by the
physician-developers. It incorporates adjustments for age-
related physiologic differences and for more constant
background risk factors. In this validation study, we
calculated PEWS scores by using an electronic PEWS-
scoring algorithm, rather than the frontline nurses who are
anticipated to calculate the score in real time.
The largest component (dynamic items) of this expert-
derived score was validated in an independent data set
(Table 3), and alone has an area under the ROC curve of
0.83. However, the clinical data were used to assess and
remove 5 static items from the score; thus, the validation of
the PEWS score is not completely independent of the
development data set. In addition, the contribution of 4
dynamic items could not be assessed because of incomplete
or inconsistent documentation in the medical record.
However, despite the absence of these items, the score
performs well. The ability of the dynamic items to
discriminate between case and control patients is good.
Discrimination is improved with the addition of the static
items, although 12 children who went on to have a code
blue event were above threshold on the basis of these
background factors alone.
Three potential issues remain. First, the impact of biased
measurement endorsement, second the use of extreme
groups, and third the use of the bmost availableQ medical
record to select control patients. These factors may have
inflated the differences between groups, and artificially
enhanced score performance. Finally, the bscoreQ was
developed to be used in all bpediatricQ ages; as such, it
could be argued that 5 scores were developed and tested in
5 samples. Fortunately, the score performs well, both overall
and within each subgroup.
4.2. Relationship to other studies
This is the first pediatric early warning score. We found,
as in adults, that clinical deterioration in patients with
evolving critical illness is detectable well before immediate
resuscitation is initiated [10]. In adult hospitals the
application of unvalidated bcalling criteriaQ in combination
with medical emergency response teams has been associated
with reductions in hospital cardiac arrests, a 50% reduction
277
in mortality and reduced ICU admission [5,11]; however, a
recent cluster randomized trial showed that the introduction
of medical emergency teams in adult hospitals did not
reduce rates of cardiac arrest, unplanned ICU admission, or
unexpected death in bmedical emergency teamsQ vs control
hospitals. Only 30% of patients at risk were identified to the
medical emergency team [7]. Other studies have shown that
early intervention in populations of critically ill adults
improves survival [12] and late admission to ICU is
associated with increased mortality and nonoptimal care
[13]. If these observations are valid in children then the
application of the PEWS score to identify children with
evolving critical illness has significant potential to decrease
hospitalization-associated mortality and morbidity.
4.3. Implications of findings
We found that code blue events are infrequent. This is
fortunate, but poses challenges to the development of
screening mechanisms to identify children who are at risk
for this event. Application of the PEWS score would have
identified 58 (78%) of the children who were resuscitated by
the code blue team. With intervention after identification,
these children may have avoided the morbidity associated
with urgent resuscitation [14], may have had less disease-
associated organ dysfunction, and consequently avoided or
had a shortened ICU admission. Conversely, 22% of case
patients would not have been identified, and 5% (1763
children during the study period) of the hospitalized
population would have been bfalselyQ identified as needing
review (Table 6).
There may be advantages to reviewing children with high
PEWS scores. These children are likely to be the sickest pa-
tients on the hospital wards. The immediate medical attention
of the code blue team may not be required; nevertheless,
review by early-response personnel may be beneficial.
The PEWS score identifies patients at risk at least 1 hour
before imminent cardiopulmonary arrest. Earlier and graded
interventions in response to rising PEWS scores may offer
opportunity to reduce ICU admissions. These interventions
may include altered monitoring, higher nurse-patient ratios,
more frequent physician review, in addition to response-
team notification, and or ICU admission. In turn, this may
promote efficient use of available resources.
Although the PEWS score has great potential in this initial
assessment, issues with the application of the PEWS score as
a routine bedside tool should be addressed before its wide-
spread use; specifically the usability and score-associated
workload. The investigators believe that independent pro-
spective validation and refinement is required before allo-
cating resources to the application of PEWS at the bedside.
5. Conclusions
A severity of illness score for hospitalized children was
developed by using nursing-expert focus groups. The PEWS
278
score can discriminate between children who had a code
blue event and those who did not. At a threshold score of 5,
the sensitivity and specificity were 78% and 95%, respec-
tively. Prospective evaluation and refinement are required
before the score can be used routinely to detect children at
risk for this uncommon but highly significant event. Despite
these reservations, the PEWS score has significant potential
to identify children who will benefit from preemptive
therapy and to thus improve the delivery and outcome of
hospital-based care.
Acknowledgments
The authors thank the following individuals: BP Kava-
nagh for review of the manuscript and suggestions about the
figures; P Rochon and D Scales for their reviews and
thoughtful comments about the manuscript; C Hyslop, K
Dryden-Palmer, B Bruinsie, L Mak, R Gateiro, L McCarthy,
K LeGrow, J Cvar, L Liske, and E Speed who participated
in the focus groups.
Dr Christopher S Parshuram is the corresponding author
and acts as guarantor of the manuscript and the underpin-
ning data.
References
[1] Reis AG, Nadkarni V, Perondi MB, Grisi S, Berg RA. A prospective
investigation into the epidemiology of in-hospital pediatric cardio-
pulmonary resuscitation using the international Utstein reporting style.
Pediatrics 2002;109:200 - 9.
[2] Suominen P, Olkkola KT, Voipio V, Korpela R, Palo R, Rasanen J.
Utstein style reporting of in-hospital paediatric cardiopulmonary
resuscitation. Resuscitation 2000;45:17 - 25.
[3] Gillis J, Dickson D, Rieder M, Steward D, Edmonds J. Results of
inpatient pediatric resuscitation. Crit Care Med 1986;14:469 - 71.
[4] Young KD, Seidel JS. Pediatric cardiopulmonary resuscitation:
a collective review. Ann Emerg Med 1999;33:195 - 205.
[5] Buist MD, Moore GE, Bernard SA, Waxman BP, Anderson JN,
Nguyen TV. Effects of a medical emergency team on reduction of
incidence of and mortality from unexpected cardiac arrests in hospital:
preliminary study. BMJ 2002;324:387 - 90.
[6] Day V, The Comprehensive Critical Care Expert Group. Comprehen-
sive critical care: a review of adult critical care services. http://www.
dh.gov.uk/assetRoot/04/08/28/72/04082872.pdf [accessed 20 August
2004] 2000.
[7] Hillman K, Chen J, Cretikos M, Bellomo R, Brown D, Doig G, et al.
Introduction of the medical emergency team (MET) system: a cluster-
randomised controlled trial. Lancet 2005;365:2091 - 7.
[8] Hodgetts TJ, Kenward G, Vlachonikolis IG, Payne S, Castle N. The
identification of risk factors for cardiac arrest and formulation of
activation criteria to alert a medical emergency team. Resuscitation
2002;54:125 - 31.
[9] Jones J, Hunter D. Qualitative research: consensus methods for
medical and health services research. BMJ 1995;311:376 - 80.
[10] Hillman K, Bristow PJ, Chey T, Daffurn K, Jacques T, Norman SL,
et al. Duration of life-threatening antecedents prior to intensive care
admission. Intensive Care Med 2002;28:1629 - 34.
[11] Bristow PJ, Hillman K, Chey T, Daffurn K, Jacques TC, Norman
SL. Rates of in-hospital arrests, deaths and intensive care admis-
H. Duncan et al.
sions: the effect of a medical emergency team. Med J Aust 2000;
173:236 - 40.
[12] Rivers E, Nguyen B, Havstad S, Ressler J, Muzzin A, Knoblich B,
et al. Early goal directed therapy in the treatment of severe sepsis
and septic shock. N Engl J Med 2001;345:1368 - 77.
[13] McQuillan P, Pilkington S, Allan A, Taylor B, Short A, Morgan G,
et al. Confidential inquiry into quality of care before admission to
intensive care. BMJ 1998;316:1853 - 8.
[14] Kozer E, Seto W, Verjee Z, Parshuram C, Khattak S, Koren G, et al.
Prospective observational study on the incidence of medication errors
during simulated resuscitation in a paediatric emergency department.
BMJ 2004;329:1321.
Commentary
The PEWS score: Potential calling criteria for critical
care response teams in children’s hospitals
The article by Duncan et al in this issue describes the first
ward-based severity-of-illness score specifically for children,
the pediatric early warning system (PEWS) score. The
authors used a retrospective case-control study design at a
large tertiary children’s hospital to validate the newly
developed PEWS scoring system. Importantly, the PEWS
scoring system appropriately incorporates adjustments for
age-related physiologic differences. The authors conclude
that the application of the PEWS score may have identified
more than three quarters of ward code-blue calls in their
children’s hospital with at least an hour’s warning. The article
is timely given the recent interest in the potential of critical
care response teams (CCRTs) to reduce morbidity and
mortality associated with in-hospital cardiopulmonary arrest.
Although in-hospital ward cardiopulmonary arrests are
relatively rare in children, estimated at 0.7% to 3% of
admissions [1,2], most pediatric hospitals have established
code-blue teams to respond to impending or actual
cardiopulmonary arrests. Management of cardiopulmonary
arrests in children in hospital wards is problematic for a
variety of reasons. Early recognition of children at risk for
cardiopulmonary arrest is often difficult because of the
infrequent nature of such events and the fact that nurse-
patient ratios in many pediatric facilities may be increasing.
Furthermore, children at greatest risk for in-hospital
cardiopulmonary arrest tend to be complex patients with
preexisting comorbidity [2], making successful resuscitation
challenging. Finally, given the rarity of these cardiopulmo-
nary arrests in children’s hospitals, it is difficult to maintain
personnel optimally skilled for code-blue coverage. These
problems associated with cardiopulmonary arrests in
children highlight the importance of identifying children
at risk for cardiopulmonary arrest early, with aggressive
treatment interventions overseen by trained personnel with
critical care expertise.
We believe that the article by Duncan et al is timely for
the pediatric critical care audience, particularly given the