Injury, Int. J.

Care Injured 44 (2013) 2328

Contents lists available at SciVerse ScienceDirect

Injury
journal homepage: www.elsevier.com/locate/injury

UK triage An improved tool for an evolving threat

Simon Horne a,*, James Vassallo a, James Read b, Susan Ball c
a

Ministry of Defence Hospital Unit Derriford, Derriford Hospital, Plymouth Hospitals NHS Trust, Plymouth, Devon PL6 8DH, United Kingdom
Derriford Hospital, Plymouth Hospitals NHS Trust, Plymouth, Devon PL6 8DH, United Kingdom
c
Centre for Health and Environmental Statistics, Plymouth University, Devon PL4 8AA, United Kingdom
b

A R T I C L E I N F O

A B S T R A C T

Article history:
Accepted 8 October 2011

Introduction: A key challenge at a major incident is to quickly identify those casualties most urgently
needing treatment in order to survive triage. The UK Triage Sieve (TS) advocated by the Major Incident
Medical Management (MIMMS) Course categorises casualties by ability to walk, respiratory rate (RR) and
heart rate (HR) or capillary rell time. The military version (MS) includes assessment of consciousness.
We tested whether the MS better predicts need for life-saving intervention in a military trauma
population. Ideal HR, RR and Glasgow Coma Score (GCS) thresholds were calculated.
Methods: A gold standard Priority 1 casualty was dened using resource-based criteria. Pre-hospital data
from a military trauma database allowed calculation of triage category, which was compared with this
standard, and presented as 2  2 tables. Sensitivity and specicity of each physiological parameter was
calculated over a range of values to identify the ideal cut-offs.
Results: A gold standard could be ascribed in 1657 cases. In 1213 both the MS and TS could ascribe a
category. MS was signicantly more sensitive than TS (59% vs 53%, p < 0.001) with similar specicity (89
vs 88%). Varying the limits for each parameter allowed some improvements in sensitivity (7080%) but
specicity dropped rapidly.
Discussion: Previous studies support the inclusion of GCS assessment for blunt as well as penetrating
trauma. Optimising the physiological cut-offs increased sensitivity in this sample to only 71% a Sieve
based purely on physiological parameters may not be capable of an acceptable level of sensitivity.
Conclusions: The MS is more sensitive than the TS. Major incident planners utilising the Sieve should
consider adopting the military version as their rst line triage tool. If validated, altering the HR and RR
thresholds may further improve the tool.
2011 Elsevier Ltd. All rights reserved.

Keywords:
Major incident
Mass-casualty
Penetrating trauma
Triage
Sieve
Sensitivity
Military

Introduction
One of the main challenges at the scene of major incident is to
quickly identify those casualties who most urgently need
treatment in order to survive a process called triage. Most
triage systems aim to identify those who need immediate lifesaving intervention (Priority 1) and separate them from those who
need intervention but who can safely wait a short while (Priority
2), and those likely to survive even if treatment is delayed (Priority
3). The UK system advocated by the Major Incident Medical
Management (MIMMS) Course assesses the casualties on the basis
of their ability to walk, their respiratory rate and their heart rate or
capillary rell time.1 These parameters feed into an algorithm that

* Corresponding author. Tel.: +44 01752 792516; fax: +44 07092 023445;
mobile: +44 07882 432220.
E-mail addresses: psihorne@doctors.org.uk, simon.horne@phnt.swest.nhs.uk
(S. Horne).
00201383/$ see front matter 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.injury.2011.10.005

then categorises each casualty as Priority 1 (P1), Priority 2 (P2) or

Priority 3 (P3) as seen in Fig. 1.
Triage must be a dynamic process as it is well accepted that no
tool predicts the clinical course perfectly patients may
deteriorate as time passes, or may improve as a result of
interventions. The MIMMS Triage Sieve (TS) is designed as a rst
look which separates P1s from P2s in around 30 s each on the
basis of respiratory rate (RR) (if breathing present) and heart rate
(HR). In a conventional major incident the TS would normally be
performed in situ. A more in depth assessment, the Triage SORT,
would be undertaken a little further away (if a casualty clearing
station has been established) or at receiving hospitals.
The UK military have adapted the TS. Their version of the Triage
Sieve (MS) includes an estimate of consciousness as part of the nal
step of the algorithm.2 There is good evidence from literature
looking at eld triage (the process of determining whether single
casualties from normal civilian trauma need to go to Major
Trauma Centres) that the GCS is the physiological parameter most
strongly predictive of serious injury.3,4 One study found it to have
an odds ratio (OR) for need for signicant intervention (as opposed

S. Horne et al. / Injury, Int. J. Care Injured 44 (2013) 2328

24

Fig. 1. The Triage Sieve. In the nal step, prolonged vs normal Capillary Rell Time
may be used instead of heart rate over or below 120.
Fig. 2. The Military Sieve, with assessment of consciousness alongside heart rate as
the nal step.

to no need) of 75.5 Of particular value is the motor component

(Glasgow Motor Score, GMS).
The next most sensitive parameter is the systolic blood pressure
(OR 32), with RR and HR being far less predictive (ORs 2.53.5).
The current military version does not specify an exact GCS for a P1
merely whether or not the patient appears unconscious (Fig. 2:
the military version of the Triage Sieve). There is no clearly dened
best cut-off for the GCS. A Glasgow Motor Score (GMS) of 6/6
(obeying commands, or the Hey Bub, touch your nose test) has
been shown to be effective, but does not seem to meet the current
description of unconscious in the MS. Whatever assessment is
settled upon, it should not add to the time taken to perform the Sieve.
The military triage context may differ somewhat from that
encountered by most civilian major incident responders. In
particular the model of a casualty clearing station may not be
applicable. Casualties may be evacuated direct from scene by
helicopter and the transport response times may well be very short.
As a result, patients may be transferred to hospital on the basis of a
single triage assessment, with no time for secondary triage (SORT).
Transport is likely to be in cohorts of patients rather than as a steady
stream of ambulances. As the rst transport will not be able to take
all the casualties at once, the sensitivity of the rst tool must be as
high as possible. This will reduce under-triage (false negatives) and
the risk of the sickest patients being missed from the rst transport
and having to wait much longer for their immediate intervention.
Patients are also more likely to suffer penetrating trauma (ballistic or
blast injury) in military incidents than in civilian ones, although with
the increase in global terrorism, civilian mass casualty events
featuring penetrating trauma are becoming more common. There
has been very limited research into the behaviour of triage systems
in penetrating trauma.6
There is only one assessment of the validity of the TS in a clinical
context. Garner et al.5 retrospectively used resource-based criteria
to examine a database of consecutive patients arriving at two
Major Trauma Centres. The TS was calculated from pre-hospital

physiological data, and compared with a gold standard denition

of a P1 based on whether the patient needed a life-saving
intervention within 6 h. They found that the TS had a sensitivity of
46%, with a specicity of 88%. Other systems tested (START from
the US and Careight from Australia) performed considerably
better sensitivities of 85% and 82%, respectively. A comparison of
the various triage systems and the parameters used is shown at
Table 1.
Experience after bombings in Israel has shown that experienced
clinicians using no particular system at the door to the hospital can
have a sensitivity of 50% for casualties with ISS > 15.7 This serves
as a benchmark that any triage tool should exceed.
Aims of this study
1. To compare the ability of the civilian and military versions of the
Triage Sieve (TS and MS) to identify patients who needed
immediate, life-saving intervention in Afghanistan and Iraq.
2. To identify the HR, RR and GCS thresholds that best predict P1
patients.
Methods
A resource-based denition of a Priority 1 (P1) casualty has been
described previously.5,8 We undertook a modied Delphi process
involving all of the consultants involved in trauma management at a
British Military Hospital (Camp Bastion, Helmand province,
Afghanistan, March 2010) to ensure that it reected the most
current military trauma practice. The starting point was the set of
criteria described by Garner et al. Eight people offered changes on
the rst round, and one on the second. The third round provoked no
alterations and the interventions listed in Box 1 became our gold
standard. If a patient received one or more of the listed interventions

Table 1
Comparison of criteria used in UK, US and Australian triage systems.
Method

Ist assessment

2nd assessment

3rd assessment

Sieve

Walking?

Heart rate >120

Military Sieve

Walking?

START

Walking?

Careight

Walking?

Breathing?
10 < rate > 30?
Breathing?
10 < rate > 30?
Breathing?
Rate > 29?
Obeying commands?

4th assessment

Heart rate >120

Unconscious?

Palpable pulse?

Obeying commands?

Breathing?

S. Horne et al. / Injury, Int. J. Care Injured 44 (2013) 2328

Box 1. Resource based definition of a P1 casualty (modified

from Garner5)
A patient was designated a P1 casualty if they required one or
more of the following interventions, or died in the department:
Airway:
 Intubation for low GCS or airway obstruction (actual or
impending) or Surgical airway.
 Oral or nasal airway for impaired ventilation with GCS < 13.
Breathing:
 Any kind of thoracostomy (needle, finger, tube).
 Positive pressure ventilation for ventilatory inadequacy.
Circulation:
 Tourniquet or haemostatic agents applied to control bleeding.
 Central line or IO access for resuscitation.
 >4 U blood products, >4l crystalloids or inotropes given.
 Proximal amputations.
 Fasciotomies for actual/suspected compartment syndrome.
 Laparotomy or thoracotomy/pericardial window.
 Ex-fix to pelvis or open femur fracture for haemorrhage
control.
 Surgical proximal vascular control.
 Peri-arrest rhythm or cardiac arrest requiring A(C)LS.
Disability:
 Immediate neurosurgery.
 Spinal nursing for proven unstable c-spine fracture.
Environmental:
 Active re-warming for initial temperature less than 32 8C.
 Chemical antidotes (OPs, CO, HCN).

then they were classied as a P1 patient. The key differences from

previously published criteria were to do with uid administration.
Blood has now become the resuscitation uid of choice for trauma in
Afghanistan, so the uid requirement was adjusted to reect this. As
it was not uncommon for patients with less severe injuries to receive
crystalloid pre-hospital (soldiers are often presumed to be
dehydrated at the point of wounding) the threshold for crystalloid
resuscitation was increased to 4 L. Major orthopaedic injuries were
also added, along with catastrophic haemorrhage control measures
in the ED such as tourniquet application or topical haemostatics.
Finally environmental factors such as hypothermia or poisoning
were included. We also felt that the 6 h timeframe previously used
was too long, and recommended 2 h. Unfortunately the timing of
interventions is not captured by the database. We instead accepted
emergency interventions that had taken place in the ED or theatres
of the receiving facility, excluding those that were done at denitive
care hospitals.
We interrogated the Joint Theatre Trauma Registry (JTTR), held
by the Academic Department of Military Emergency Medicine
(Royal Centre for Defence Medicine, Birmingham, UK). This registry
holds data on the mechanism of injury, physiology, interventions
carried out and outcomes for all trauma patients managed in UK

25

military facilities in Iraq and Afghanistan. Data was extracted on all

patients (military or civilian) for whom there was pre-hospital
information from 2005 to 2010. The JTTR data is collected as
routine and the patients are not identiable. The chair of the
Ministry of Defence Research and Ethics Committee conrmed that
ethical approval was not required for this study.
A patient was designated P1 if they met any of the criteria in Box
1 or P2/3 otherwise. This was considered the gold standard. From
the physiological data they were then triaged using the civilian and
military Triage Sieves. For the MS criteria of unconscious a cut off
of GCS < 13 was used.
If there was a complete set of physiological data then the
appropriate category by either the TS or MS would be ascribed.
When the data was incomplete, but a single parameter met the P1
cut-off then they were classied P1, as they would have still met
that criterion if the dataset had been complete. If the dataset was
incomplete and the data present did not automatically classify the
patient as P1, then that patient was excluded from the analysis.
A 2  2 table was then generated to compare each version of the
Sieve with the gold standard and sensitivity and specicity
calculated. From this table, the numbers of disagreements between
the two tests could be identied and a McNemar test performed to
establish whether the tests were signicantly different.
The individual physiological variables were also correlated with
outcome to model the most useful cut-offs. The impact of altering
the cut-offs for RR to below 10 or 12, or above 24, 27 or 30 was
assessed. For HR the impact of introducing lower limits was
examined (0, 40, 60) as well as variation of the upper limit (100 vs
120). Finally, for the GCS the impact of scoring above 3 was
assessed, along with cut-offs at every point 8 and above.
Incomplete datasets were used for this part of the analysis as
all the datasets included outcome then any that contained one or
more physiological parameter could be assessed. This means that
the sensitivity and specicity predicted for the TS and MS by the
model differ somewhat from the comparative analysis of the two
triage tools where data was only be included if the set was
complete enough to determine both TS and MS categories.
The performance of a selection of cut-offs is shown in Table 3.
Results
The dataset comprised 1657 cases. Of these, 1213 contained
adequate data to ascribe a category for both TS and MS. Data for the
surgical outcomes and interventions necessary to determine gold
standard status was available for all patients. The vast majority of
patients met multiple criteria. In no cases were the patients
categorised T1 solely on the basis of hypothermia. There were no
chemical antidotes administered.
The data for the TS are presented in Table 1, and for the MS in
Table 2.
Addition of the consciousness assessment to the Triage Sieve
(dened here as a GCS < 13) resulted in an absolute increase in
sensitivity of 5.2% (relative increase 9.8%) without reducing the
specicity. McNemars test indicated that the difference in
sensitivities was signicant (p < 0.001).
The second stage of the study was to examine the impact of
altering the cut-offs for each physiological parameter to explore

Table 2
Number of patients by category P1 or P2/3, gold standard against TS.

Triage Sieve (TS) P1

Triage Sieve (TS) P2

Gold standard P1

Gold standard P2/3

370
326
Total P1: 696
Sensitivity = 53.2% (95% CI 49.456.8)

63
454
Total P2/3: 517
Specicity = 87.8% (95% CI 84.790.4)

433
780
Total: 1213

S. Horne et al. / Injury, Int. J. Care Injured 44 (2013) 2328

26

Table 3
Number of patients by category P1 or P2/3, Gold standard against MS.

Military Sieve (MS) P1

Military Sieve (MS) P2

Gold standard P1

Gold standard P2/3

407
289
Total P1: 696
Sensitivity = 58.5% (95% CI 58.462.1)

56
461
Total P2/3: 517
Specicity = 89.2% (95% CI 84.790.4)

whether the performance of the MS can be improved further. All

1657 datasets were used for this analysis as all contained at least
one physiological measurement and the gold standard category.
Reconstructing the tools in this model gave a predicted sensitivity
and specicity of 41.8% (95% CI 38.245.4) for the TS and 65.2%
(95% CI 6268.4) for the MS. Specicity for the TS was 91.7% (95% CI
89.594) and for the MS, 89.2% (95% CI 86.691.9). The sensitivity
and specicity predicted for a variety of combinations of different
RR, HR and GCS limits are shown in Table 3.
Changing the upper and lower limits for HR and RR produced
marked improvements in the potential sensitivity of the MS, with a
consequent loss of specicity (Table 4).
Introduction of a lower limit for HR (40 bpm or 60 bpm)
produced modest improvements in sensitivity (+0.3% and +1.6%,
respectively), but the decrease in specicity was not marked
( 0.1% for 40 and 3.1% for 60). Reducing the upper limit of HR (to
100 bpm) provoked a large drop in specicity ( 21%).
Altering the upper threshold for RR from 30 to 27 produced
modest changes in sensitivity in addition to the above (+3.2%) with
a similar drop in specicity ( 3.9%). Decreasing the threshold
further (to 24) again gave a small improvement in sensitivity of
around +2% with a loss of sensitivity of nearly 5%. Increasing the
lower threshold from 10 to 12 breaths per minute produces a more
modest increase in sensitivity (+0.4%) but also only a small
reduction in specicity ( 0.5%). Overall, changing HR and RR
thresholds can allow sensitivities to be increased into the 70%
range whilst maintaining specicities close to 80%.
Fig. 3 shows the relationship between GCS and sensitivity/
specicity of the MS. For a given HR and RR range, a change in the
GCS cut-off has a marked impact on the sensitivity at the lower
range, but has only a small impact after a GCS of 8. Equally the
specicity is steady to a GCS of about 10 and then drops off steadily
at a rate of 0.5% per point increase.

463
750
Total: 1213

Discussion
Two methods have commonly been used to evaluate triage.
Some studies have reported on the outcomes of patients involved
in genuine MCIs.9,10 Others have looked at cohorts of patients
arriving at hospitals from single-casualty incidents, assuming that
their needs would be unchanged if they had instead came from one
MCI. They then asked whether their physiology accurately
predicted their outcomes and need for interventions.5,11 Both
have signicant limitations. The key advantage of the latter
method is that the physiological ndings on which the MCI triage is
based should not be affected by the fact that the patient was not
actually involved in a MCI and so large numbers can be analysed
without the chaos of a major incident affecting data quality. There
is also one prospective study but this only looked at paediatric
injury.12 Interestingly, none of the triage tools examined
performed well, with the highest sensitivity being 46% which is
no better (and may be worse) than an end of the bed assessment
by a senior doctor.
Whatever the method chosen, the same fundamental problem
has to be addressed what does in immediate need of a lifesaving intervention mean? Some authors have used an Injury
Severity Score (ISS) of >15 as a denition of severe trauma as this
has been used in many other aspects of healthcare research and
planning,13 but there is evidence that the correlation between ISS
and the need for intervention may not be strong.8,12 An alternative
to the ISS is a resource-based denition of triage criteria. Here the
resources used to treat the patient determine whether they were a
true P1 casualty i.e. did they actually receive immediate lifesaving intervention. This method has problems too; the decision
whether or not to operate may be very clinician-dependent, or in a
military context may be inuenced by the number and types of
other casualties presenting at the same time. Also, whereas there

Table 4
Sensitivity/specicity for TS and MS if physiological parameter cut-offs are altered from the current recommendations, calculated using all available data (n = 1657).

TS
MS
MS1
MS2
MS3
MS4
MS5
MS6
MS7
MS8
MS9
MS10
MS11
MS12
MS13
MS14
MS15
MS16
MS17

HR (low)

HR (high)

RR (low)

RR (high)

GCS (low)

Sen %

Spec %

0
0
40
60
40
60
40
60
40
60
40
60
40
60
40
60
40
40
40

120
120
120
120
100
100
120
120
120
120
120
120
120
120
120
120
120
120
120

10
10
10
10
10
10
10
10
10
10
12
12
12
12
12
12
12
12
12

30
30
30
30
30
30
24
24
27
27
24
24
24
24
24
24
24
24
24

N/A
13
13
13
13
13
13
13
13
13
13
13
3
3
8
8
9
10
11

41.8
65.2
65.4
66.8
77.9
79.3
70.8
71.9
68.7
70
71.2
72.3
56.8
59.5
68.9
70.6
69.5
69.8
70.5

91.7
89.2
89.1
86.1
67.5
65
79.9
77.7
84.7
82.2
79.3
77.1
81
78
80.8
77.8
80.8
80.8
79.6

S. Horne et al. / Injury, Int. J. Care Injured 44 (2013) 2328

85
80

Percent %

75
70
Sensitivity

65

Specificity

60
55
50

9 10 11 12 13 14

Glasgow Coma Score

Fig. 3. The impact of changing GCS cut-off on sensitivity and specicity (for given
HR and RR cut-offs of 40120 and 1224, respectively).

has been broad agreement as to what constitutes a life-saving

intervention, there is debate as to what constitutes immediate.
Baxt and Upenieks originally suggested 48 h.8 This was later
reduced to 6 h.5 Our Delphi concluded that 2 h is more
representative and is in line with UK MIMMS recommendations.
A key limitation of our study is that we were generally unable to
conrm whether the Gold standard P1s actually received their
treatment immediately. We were unable to analyse time to
intervention as this is not routinely captured by the database. It is
conceivable that the patients were indeed less sick and that some
of the interventions (e.g. laparotomies and chest drains) were
being carried out after 2 h which could reduce the sensitivity of the
tools in our study. Personal experience (SH) at the UK facility in
Camp Bastion, Afghanistan refutes this, but there is still considerable scope for bias here. However there is no reason to believe that
one group should have been affected more than the other.
The literature for eld triage species an acceptable accuracy
for tools for identifying which patients should go to Major Trauma
Centres. The Center for Disease Control states that under-triage is
unacceptable and that the sensitivity of the tools should be high at
95100%. It suggests that an over-triage rate of 5060% is
acceptable.14 There is no such consensus for MCI triage, but as
the desired outcomes are similar (i.e. that eld triage should
identify those most in need of high level trauma care and that MCI
triage should identify those in most need of lifesaving intervention) it might be a reasonable starting point. MCI triage may
require a higher specicity to prevent resources being swamped as
there is some evidence that over-triage increases the death rates
after major incidents,15,16 although this may not always be the
case.17
Previous estimates of the sensitivity and specicity of the MCI
triage tools vary. Kahn et al.9 showed that START detected all of the
P1 casualties after a train crash, but as there were only 2 dened by
the gold standard, the condence intervals ranged from 16 to 100%.
The challenges of a retrospective study of a naturally chaotic event
means that there may be problems with the data used in this study
in particular it is not possible to separate errors made by the tool
from errors made by the people using it.
Garner et al.5 used 1144 consecutive civilian patients as the test
group. START outperformed the TS (sensitivity 85% vs 46%). In this
study both START and the TS used capillary rell time as their
cardiovascular assessment. The physiological parameters were the
same in both tools except that an assessment of consciousness
(ability to follow commands) appeared in START. This criterion also
appears in Careight, which had a sensitivity of 82%. Our data
suggests that in military trauma the sensitivity increase from the
addition of the GCS (MS) is more modest. The specicities of the MS
(89%) and TS (88%) in this study mirror those found by Garner

27

(START 86%, TS 88% and Careight 95%). The striking concordance

between this work and that of Garner (despite the very different
patient populations and mechanisms involved) suggests that the
advantage of MS over TS will translate to civilian blunt trauma.
Adding a GCS cut-off of <13 clearly enhances the triage. There
may be an element of bias in that some key life-saving
interventions relate to GCS (for example the need for intubation)
and so of course the assessment of GCS will predict this need. The
fact remains that these patients still require intervention. Still
unclear is the ideal test for consciousness, or cut-off for GCS. The
ability to obey commands (GMS of 6/6, or the Hey Bub test)
might be the quickest, simplest means of making a meaningful
assessment. Our analysis shows that most of the advantage gained
from the addition of a consciousness component comes from
identifying those patients whose GCS is 8 or below. The benets
from incremental scores above this are much smaller. Any test that
can discriminate between conscious patients and unconscious
ones will likely be of value. Actual calculation of the GCS is unlikely
to be useful because of the time required to do it properly.
The evidence suggests that current RR thresholds are too high18
and our analysis shows that modest gains in sensitivity can be
made by reducing the RR range to 1224 instead of 1030. There is
a slight drop in specicity associated with this but it still remains
around 80%.
The choice of best cardiovascular assessment is still hotly
debated. Systolic blood pressure is the gold standard and
hypotension has long been associated with increased mortality
in trauma19,20 but its measurement is not practical in the early
stages of a MCI. The correlation between presence of pulse and
actual blood pressure has been shown to be highly variable
whilst Advanced Trauma Life Support suggests that a radial pulse is
usually lost as the systolic blood pressure drops to 80 mmHg,21
several studies have suggested that 7022,23 or even 6024 may be
more accurate. Assumptions about the presence of pulses based on
blood pressures have weakened previous studies in this eld.
Weak pulse vs normal pulse has been strongly predictive of
mortality in one study25 but is surely too observer dependent to be
of use in this context. HR might seem a strong candidate, but there
is evidence that a marked tachycardic response to hypovolaemia is
absent in a large proportion of patients.2628 Our data suggests that
if HR is used, then a lower limit (40 or 60 bpm) can safely be
introduced allowing modest increases in sensitivity whilst
maintaining acceptable specicity. Reduction in the upper limit
from 120 to 100 causes a larger drop in specicity, potentially
making it unworkable.
Prospective research is still needed to determine whether
meaningful improvement is achieved by manipulation of the
physiological cut-offs. Our study suggests there may be a benet in
reducing the RR range to 1224, and the introduction of a lower HR
range of 40 or 60. According to our model these changes would
maintain a specicity of 79.3% (95% CI 75.982.7) whilst increasing
the sensitivity to 71.2% (95% CI 68.274.1). Even if these benets
are shown to be genuine, it seems likely that the MS has only
limited potential for improvement and that its sensitivity will
probably never reach 80% as long as it remains a purely
physiological assessment. It is likely that this limitation will be
shown to extend to START and Careight too when they are tested
prospectively. There is mounting evidence that some form of
hybrid tool that assesses a combination of mechanism, anatomical
injury and physiology may be the only practical solution. Whatever
further adjustments may be implemented, the system needs to be
kept as simple as possible.29
Our data clearly shows that in a population suffering mostly
penetrating trauma (ballistic and blast) the MS outperforms the TS.
This is in line with the literature on GCS as a predictor of serious
injury. In view of the agreement between the Garner data and ours,

28

S. Horne et al. / Injury, Int. J. Care Injured 44 (2013) 2328

it would be reasonable to conclude that the MS would perform

better than the TS in a classical civilian MCI as well as in the
penetrating trauma context described here.
Conclusions
The military version of the Triage Sieve adds an assessment of
consciousness into its nal step. This change increases the
sensitivity for patients in need of immediate life saving intervention by at least 5% (relative increase 9.8%) (p < 0.001).
In the absence of a denitive, prospective comparison of the
various methods, countries using the Sieve can easily improve their
triage by moving to the military version of the Triage Sieve.
Conict of interest statement
The authors declare that they have no conict of interest. SH
and JV are both serving members of the UK Armed Forces.
Acknowledgments
This article reects the personal opinions of the authors only
and in no way represents the position of the Ministry of Defence.
The authors would like to acknowledge the help and support
from the Academic Department of Emergency Medicine, Royal
College of Defence Medicine, Birmingham.
References
1. Hodgetts TJ, Mackaway-Jones K, editors. Major incident medical management
and
.
support, the practical approach. Plymouth, MA: BMJ Publishing Group; 1995
2. Clinical Guidelines for Operations. Joint doctrine publication 4-03.1. 1st edition.
2008. Available from publications@dcdc.org.uk.
3. Meredith W, Rutledge R, Hansen A, Oller D, Thomason M, Cunningham P, et al.
Field triage of trauma patients based upon the ability to follow commands: a
study in 29,573 injured patients. J Trauma 1995;38(1):12935.
4. Ross SE, Leipold C, Terregino C, OMalley KF. Efcacy of the Glasgow Coma Score
in trauma triage. J Trauma 1998;45:424.
5. Garner A, Lee A, Harrison K, Schultz CH. Comparative analysis of multiplecasualty incident triage algorithms. Ann Emerg Med 2001;38(5):5418.
6. Sacco WJ, Navin M, Waddell R, Fiedler KE, Long WB, Buckman R. A new
resource-constrained triage method applied to victims of penetrating injury.
J Trauma 2007;63:31625.
7. Ashkenazi I, Kessel B, Khashan T, Haspel J, Oren M, Olsha O, et al. Precision of inhospital triage in mass-casualty incidents after terror attacks. Prehospital
Disaster Med 2006;21(1):203.

8. Baxt WG, Upenieks V. The lack of full correlation between the Injury Severity
Score and the resource needs of injured patients. Ann Emerg Med 1990;19:1396
400.
9. Kahn CA, Schultz C, Miller K, Anderson C. Does START triage work? An outcomes
assessment after a disaster. Ann Emerg Med 2009;54:42430.
10. Malik Z, Pervez M, Safdar A, Masood T, Tariq M. Triage and management of mass
casualties in a train accident. J Coll Physicians Surg Pak 2004;14(2):10811.
11. Gebhart M, Pence R. START triage: does it work? Disaster Manage Response
2007;5:6873.
12. Wallis L, Carley S. Comparison of paediatric major incident primary triage tools.
Emerg Med J 2006;23:4758.
13. Trauma who cares? A report of the national condential enquiry into patient outcome
and death.2007. Available from http://www.ncepod.org.uk/2007report2/Downloads/SIP_summary.pdf [cited 19.01.10].
14. Sasser SM, Hunt RC, Sullivent EE, Wald MM, Mitchko J, Jurkovich G, et al.
Guidelines for eld triage of injured patients: recommendations of the National
Expert Panel on Field Triage. MMWR Recomm Rep 2009;58(RR-1):135.
15. Hirschberg A, Scott B, Granchi T. How does casualty load affect trauma care in
urban bombing incidents? A quantitative analysis. J Trauma 2005;58:68695.
16. Frykberg ER. Triage principles and practice. Scand J Surg 2005;94:2728.
17. Aylwin CJ, Konig TC, Brennan NW, Shirley P, Gareth D, Walsh M, et al. Reduction
in critical mortality in urban mass casualty incidents: analysis of triage, surge,
and resource use after the London bombings on July 7, 2005. Lancet
2006;368:221925.
18. Husum H, Gilbert M, Wisborg T, Van Heng Y, Murad M. Respiratory rate as a
prehospital triage tool in rural trauma. J Trauma 2003;55:46670.
19. Franklin G, Boaz P, Spain D, Lukan J, Carrilo E, Richardson J. Prehospital
hypotension as a valid indicator of trauma team activation. J Trauma
2000;48(6):10349.
20. Heckbert S, Vedder N, Hoffman. Wilma RN, Winn R, Hudson L, et al. Outcome
after hemorrhagic shock in trauma patients. J Trauma 1998;45(3):5459.
21. Committee on Trauma, American College of Surgeons. ATLS: advanced trauma
life support program for doctors. 8th edition. Chicago: American College of
Surgeons; 2008. ISBN: 9781880696316. OCLC OL22228190M.
22. Deakin CD, Low JL. Accuracy of the advanced trauma life support guidelines for
predicting systolic blood pressure using carotid, femoral, and radial pulses:
observational study. BMJ 2000;321:6734.
23. Poulton TJ. ATLS paradigm fails. Ann Emerg Med 1988;17(January (1)):107.
24. Poloujadoff M-P, Lapostolle F, Lockey D, Amathieu R, Merouani M, Galinski M,
et al. Survival of severely shocked patients who present with absent radial pulse
and unrecordable blood pressure in the pre-hospital phase. Resuscitation
2006;69:1859.
25. McManus J, Yershov A, Ludwig D, Holcomb J, Salinas J, Dubick M, et al. Radial
pulse character relationships to systolic blood pressure and trauma outcomes.
Prehosp Emerg Care 2005;9(4):4238.
26. Victorino GP, Battistella F, Wisner DH. Does tachycardia correlate with hypotension after trauma? J Am Coll Surg 2003;196:67984.
27. Demetriaded D, Chan LS, Bhasin P, Berne T, Ramicone E, Huicochea F, et al.
Relative bradycardia in patients with traumatic hypotension. J Trauma
1998;45:5349.
28. Little RA, Kirkman E, Driscoll P, Hanson J, Mackway-Jones K. Preventable deaths
after injury: why are the traditional vital signs poor indicators of blood loss? J
Accid Emerg Med 1995;12:114.
29. Batchelor JS. Adult prehospital scoring systems: a critical review. Trauma
2000;2:25360.