American Journal of Emergency Medicine 33 (2015) 1687–1691

Contents lists available at ScienceDirect

American Journal of Emergency Medicine

journal homepage: www.elsevier.com/locate/ajem

Brief Report

Simple Triage Algorithm and Rapid Treatment and Sort, Assess,

Lifesaving, Interventions, Treatment, and Transportation mass
casualty triage methods for sensitivity, specificity, and
predictive values☆,☆☆
Mary Colleen Bhalla, MD a,b,⁎, Jennifer Frey, PhD a, Cody Rider, DO a,
Michael Nord, DO a, Mitch Hegerhorst, DO a,c,1
a
Summa Akron City Hospital, Akron, OH
b
Northeast Ohio Medical University, Rootstown, OH 44272
c
Kadlec Medical Center, Richland, WA 99352

a r t i c l e i n f o a b s t r a c t

Article history: Objective: Two common mass casualty triage algorithms are Simple Triage Algorithm and Rapid Treatment
Received 16 June 2015 (START) and Sort, Assess, Lifesaving, Interventions, Treatment, and Transportation (SALT). We sought to deter-
Received in revised form 28 July 2015 mine the START and SALT efficacy in predicting clinical outcome by appropriate triage.
Accepted 11 August 2015 Methods: We performed a retrospective chart review of trauma registry of patients from our emergency depart-
ment (ED). We applied the triage algorithms to 100 patient charts.
The end points categories were defined by patient outcomes and the need for intervention: minor/green,
discharged without intervention other than minor ED procedure; delayed/yellow, patients get an intervention
more than 12 hours after arrival to the ED; immediate/red, patients get an intervention less than 12 hours
after arrival; dead/expectant/black, patients die within 48 hours after arrival.
Results: The mean age was 47 years (range, 17-92 years), and 72% were male. The mechanism of injury was 41%
motor vehicle collision, 32% fall, and 16% penetrating trauma. Hospital outcome was 60% minor/green, 5% de-
layed/yellow, 29% immediate/red, and 6% dead/black. The SALT method resulted in 5 patients overtriaged (95%
confidence interval [CI], 1.6-11.2), 30 undertriaged (95% CI, 21.2-40), and 65 met triage level (95% CI, 54.8-
74.3). The START method resulted in 12 overtriage (95% CI, 6.4-20), 33 undertriaged (95% CI, 23.9-43.1), and
55 at triage level (95% CI, 44.7-65). Within triage levels, sensitivity ranged from 0% to 92%, specificity from 55%
to 100%, positive predictive values from 10% to 100%, and negative predictive value from 65% to 97%.
Conclusion: Overall, neither SALT nor START was sensitive or specific for predicting clinical outcome.
© 2015 Elsevier Inc. All rights reserved.

1. Introduction needed to be applied, but how do we organize this system of priorities

Triage is defined as “the sorting of and allocation of treatment to pa-
tients and especially battle and disaster victims according to a system of
priorities designed to maximize the number of survivors” [1]. Although
the word is clearly defined, the process of how to carry out triage is less
well defined. As the definition states, “a system of priorities” will be
☆ No outside funding was provided.
☆☆ Presented as a poster at the Society of Academic Emergency Medicine annual meeting,
May 2015.
⁎ Corresponding author at: Department of Emergency Medicine, Summa Akron City
Hospital, 525 East Market St, Akron, OH 44304-1619. Tel.: +1 330 375 7530; fax: +1
330 375 7564.
E-mail addresses: bhallam@summahealth.org (M.C. Bhalla), freyja@summahealth.org
(J. Frey), cjrider84@gmail.com (C. Rider), mtn417@gmail.com (M. Nord),
mitchhegerhorst@gmail.com (M. Hegerhorst).
1
Current location, formally with Summa Akron City Hospital.
and how do we know it is effective? One such system, known as Simple
Triage and Rapid Treatment (START) (Fig. 1), has been the standard tri-
age algorithm since the 1980s when it was developed [2]. The START al-
gorithm is appealing because it applies the same approach of evaluating
airway, breathing, and circulation as taught in advanced trauma life
support certification. However, there is not a single way to approach
triage, and retrospective studies have shown that START is not nearly
as sensitive and specific as it claims to be and also has significant
overtriage [3]. A second algorithm for triage has more recently been
developed, the Sort, Assess, Lifesaving Interventions, Treatment/
Transport algorithm, commonly known as SALT (Fig. 2A and B) [4].
This model has been endorsed by the American College of Emergency
Physicians, American Trauma Society, and American College of Sur-
geons Committee on Trauma [5]. The SALT model involves initial global
sorting as well as basic lifesaving interventions such as controlling hem-
orrhage, opening airway/rescue breaths, autoinjector antidotes, and
chest compressions.

http://dx.doi.org/10.1016/j.ajem.2015.08.021
0735-6757/© 2015 Elsevier Inc. All rights reserved.
1688 M.C. Bhalla et al. / American Journal of Emergency Medicine 33 (2015) 1687–1691

All Walking Wounded Respirations

MINOR NO YES

Position Airway Under 30/min Over 30/min

NO Respirations Respirations
IMMEDIATE
DECEASED IMMEDIATE
Radial Pulse
Perfusion Present
Radial Pulse Absent
Over 2 Under 2
seconds
or
seconds
Mental Status
Capillary Refill
Control CAN NOT CAN
Bleeding Follow Follow
Commands Commands
IMMEDIATE
IMMEDIATE DELAYED
Fig. 1. START triage method.

The nature of mass causality incidents (MCIs) does not allow for ran- 2. Methods
domized studies to compare the START and SALT triage algorithms.
There is limited literature directly comparing the 2 triage algorithms 2.1. Study design
when applying them to the same patient. One study evaluated the effi-
cacy of START triage to predict mortality but did not assess the correla- We applied the START and SALT triage algorithms to patient data from
tion with other outcomes [6]. The goal for this pilot study is to our trauma registry to assess the sensitivity and specificity of the algo-
retrospectively apply the START and SALT triage methods to patients rithms. This study was approved by the local institutional review board.
presenting to our level I trauma center as surgical or trauma activations
and evaluate the accuracy based on patient outcomes and interventions 2.2. Study setting and population
required. Our hypothesis is that START and SALT triage methods are
sensitive and specific and predict clinical outcome. We performed a retrospective chart review of trauma patients, in
2013, who presented for evaluation at our level I Midwestern trauma
center, starting on January 1, 2013. The emergency department (ED)
has more than 80 000 adult patient visits per year from a large trauma
A catchment area. We collected data from the first 100 charts with com-
Walk
plete data available from the first health care encounter either emergen-
Assess 3rd
cy medical service (EMS) or ED, if self-transported. Patients transferred
Step 1: Wave/Purposeful Movement from other hospitals or freestanding EDs were excluded from the study.
SORT Assess 2nd Our trauma registry is collected as part of our membership in the North-
eastern Ohio Regional Trauma Network.
Still / Obvious Life Threat
Assess1st
2.3. Study protocol
B
We collected demographic data on age, sex, and trauma mechanism.
Step 2: Assess Mechanism categories were defined as motor vehicle collision, fall, pen-
No
Lifesaving Interventions: Breathing? Dead etrating trauma, pedestrian/bicycle struck, and industrial accident.
• Control majorhemorrhage The START triage (Fig. 1) review included determining from review
• Open airway (if child, Minimal
consider 2 rescue breaths) if patient could walk, which would triage them as green. If unable to
Yes Yes
• Chest decompression walk, respirations reported less than 30 breaths per minute and systolic
• Auto injector antidotes • Obeys commands or makes All
purposeful movements? Minor
blood pressure (BP) greater than 80 mm Hg (correlating with radial
Yes
• Hasperipheral pulse? injuries pulse or normal cap refill), and patient was following commands, pa-
• Not in respiratory distress? only?
• Major hemorrhage is controlled?
tient was triaged as yellow. If there is any abnormality in the aforemen-
tioned group with respirations, BP, and ability to follow commands,
Any No No patient was triaged as red. If patient was apneic, they were triaged as
Delayed black/expectant.
Likely to The SALT triage (Fig. 2) breaks triage into 2 steps. Sorting is first, in
Expectant
survive given
Immediate which patients who can walk are assessed last, those who cannot
current
No
resources?
Yes walk but can wave/purposefully respond are assess second, and those
patients who are still/unresponsive are immediately seen. Assess and
lifesaving treatment is next. Patients are triaged as green if they can
obey commands or make purposeful movements, have a peripheral
Fig. 2. A and B, SALT triage method. pulse, are not in respiratory distress, do not have a hemorrhage, and
 M.C. Bhalla et al. / American Journal of Emergency Medicine 33 (2015) 1687–1691 1689

Table 1 Other
Triage outcomes 10%

Triage category Clinical features

Minor/green tag Discharged from the ED or hospital without

Delayed/yellow tag
Immediate/red tag
Dead/expectant/black tag
intervention other than minor ED procedure
(splint/sling, observation, suture) Motor Vehicle
Fall
Patients get an intervention (group together: 32% Collision
surgery, blood product transfusion, chest tube, 41%
angio procedure) sometime after the first 12 h
after arrival to the ED
Patients get an intervention (group together:
surgery, blood product transfusion, chest tube,
angio procedure) sometime within the first 12 h
after arrival to the ED
Patients die within 48 h after arrival to the ED Pedestrian or Bike
or have a Cerebral Performance Category Scale of 1%
Penetrating Trauma
4 or 5 upon discharge 16%

Fig. 3. Mechanisms of injury.

only present with minor injuries. This obviously required more inter-
pretation from the chart. Patients are yellow if they meet all of the
green criteria, but injuries are not considered minor. If the answer is
no to any of the green criteria and the patient is likely to survive given
resources (hemorrhage control, chest compressions, autoinjector anti-
dotes, and opening airway), then the patient is triaged as red.
When reviewing charts, we considered a systolic BP less than 80 mm
Hg to be an equivalent finding to absent radial pulse/delayed capillary
refill [7]. We considered “minor injuries” to mean minor abrasions or
lacerations, contusions, sprains, or strains. We considered documenta-
tion of ambulation at scene by EMS as able to walk. The START triage di-
vides patients by a respiratory rate less than 30 or greater than 30
breaths per minute. We changed this to less than 30 or 30 or higher
breaths per minute. Using information obtained from EMS reports and
ED triage notes, patients were triaged using SALT and START criteria, re-
spectively. This was often with limited information, as would be expect-
ed if one was triaging during a disaster scenario. Two different
researchers reviewed the prehospital and ED triage notes, and if a dis-
crepancy in the interpretation occurred, it went to a third researcher.
The EMS records were used primarily for the triage determination,
and the ED triage notes were used if there was missing data or if the pa-
tient came in by private vehicle.
2.4. Measurements
Our initial triage category was compared with the patients' eventual
hospital outcomes. We designated different levels of treatment received
while in the hospital to correspond with different triage categories so
there could be a shared variable (Table 1).
2.6. Data analysis
We entered data into REDcap (Nashville, TN) and use STATA statisti-
cal software (StataCorp LP, College Station, TX) for analysis. We provide
descriptive statistics for demographic data. We present sensitivity, spec-
ificity, negative predictive values, and positive predictive values with
95% CIs.
3. Results
3.1. Patient characteristics
The mean age was 47 years (range, 17-92 years), and 72% were male.
The most common mechanism of injury was motor vehicle collision
(41%) (Table 2; Fig. 3). Both SALT and START triaged more than half of
the patients as minor (Fig. 4). Hospital outcome was 60% minor/green,
5% delayed/yellow, 29% immediate/red, and 6% dead/black (Fig. 4).
3.2. Main results
The START method resulted in 12 overtriage (95% CI, 6.4-20), 33
undertriaged (95% CI, 23.9-43.1), and 55 at triage level (95% CI, 44.7-
65). The overall sensitivity was 55% (44.7-65), and specificity was 85%
(80.4-88.9). The SALT method resulted in 5 patients overtriaged (95%
CI, 1.6-11.2), 30 undertriaged (95% CI, 21.2-40), and 65 met triage
level (95% CI, 54.8-74.3). The overall sensitivity was 65% (95% CI, 54.8-
74.2), and specificity was 88.3% (95% CI, 84.1-91.7). Within all triage
levels, sensitivity ranged from 0% to 92%; and specificity, from 55% to
100%. The positive predictive values ranged from 10% to 100% and neg-
ative predictive value from 65% to 97% (Tables 3 and 4).

2.5. Sample size calculation

Dead or Expectant/Black
We collected data on 100 patient encounters for a 95% confidence in- 6%
terval (CI) of 10% for SALT and START sensitivity and specificity.

Minor/Green
60%
Immediate/Red
29%
Table 2
Results—triage START vs SALT

START SALT

Minor/green 66% 76% Delayed/Yellow

Delayed/yellow 22% 10% 5%
Immediate/red 9% 11%
Dead or expectant/black 3% 3%
Fig. 4. Hospital outcomes.
1690 M.C. Bhalla et al. / American Journal of Emergency Medicine 33 (2015) 1687–1691

Table 3
Results START

Sensitivity (n, 95% CI) Specificity (n, 95% CI) Positive predictive value (n, 95% CI) Negative predictive value (n, 95% CI)

Minor/green 80% (48/60, 67.7-89.2) 55% (22/40, 38.5-70.7) 72.4% (48/66, 60.4-83) 79.2% (22/34, 46.5-80.3)
Delayed/yellow 0% (0/5, 0-52.2) 76.8% (73/95, 67.1-84.9) 0% (0/22, 0-15.4) 93.6% (73/78, 85.7-97.9)
Immediate/red 13.8% (4/29, 3.9-31.7) 93% (66/71, 84.3-97.7) 44.4% (4/9, 13.7-78.8) 72.5% (66/91, 62.2-81.4)
Dead or Expectant/black 50% (3/6, 11.8-88.2) 100% (94/94, 96.2-100) 100% (3/3, 29.2-100) 96.9% (94/97, 91.2-99.4)
Overall 55% (55/100, 44.7-65) 85% (255/300, 80.4-88.9) NA NA

4. Discussion
Both triage algorithms were compared with each patient's hospital
outcome, to approximate the algorithms' ability to appropriately deter-
mine each patient's need for intervention. Both triage systems did triage
the majority of patients correctly: START 55% and SALT 65%. Although
they did have notable specificity, START 85% and SALT 88%, they lacked
sensitivity. Both the START and SALT algorithms were noted to have
poor sensitivity for appropriately identifying the severity of injury in pa-
tients and subsequent need for intervention. The largest inaccuracy of
the triage systems' flaws was in the “red” or critically ill category. Mean-
ing, the most severely injured patients who did receive treatment with-
in 12 hours of arrival were often not triaged appropriately, of which
most were undertriaged. The SALT system was less likely to overtriage,
but both would frequently undertriage. The positive and negative pre-
dictive values for death were good for START and SALT. Although
these triage algorithms did not seem to give promising results, it may
be that other systems would do no better in this type of evaluation.
In the ED, we have many tools to guide us in our decision making
and ample resources at most times to deal with trauma patients who
come in one at a time. Patients are seen immediately upon arrival
with the blood bank, radiology, and laboratory already alerted. We can
use ultrasound and computed tomographic scan to rapidly identify
dangerous conditions. We have serial vital signs and constant nursing
supervision for critically ill patients so we can make changes immedi-
ately as the need arises.
Of specific concern was the undertriage of both methods. In an MCI,
undertriage may mean an underestimation of resources needed and
may delay care for some. As MCI are rare events, overtriaging and
expecting patients to be more ill and require more resources than
they eventually use would not necessarily be excessively taxing on the
system. In a system that uses the incident command model, having
extra staff or physicians on hand than was absolutely needed can be eas-
ily remedied. It may be that neither of these triage methods should be
used and instead an alternative triage method should be developed
based on further research. A recent study using a trauma database to
identify patients who would be triaged with START as green/minor
who died found that the accuracy improved by simply making elderly
patients yellow [8].
The largest limitation is that this is a retrospective study and infor-
mation had to be interpreted from EMS run reports and initial triage
notes. First-hand visualization of the patient could not be considered
in retrospectively assigning patients through triage. It is possible in
the future to design a study in which patients are enrolled as they
enter the ED for trauma evaluation or are enrolled by EMS, but it
would require extensive training of providers for accurate data collec-
tion. With the advent of prehospital electronic records, it may be possi-
ble to build in a research study to their documentation that is triggered
by a trauma diagnosis. Enrolling them in this method may still not test
the triage methods fully as asking someone who is part of a mass casu-
alty event to walk is very different from asking a patient from an isolated
traumatic incidence to walk.
The study is also limited by its sample size and demographic. The
large majority of the patients (41%) were blunt trauma (motor vehicle
crash) mechanisms, which may not accurately represent patients need-
ing triaged after a mass casualty event such as a large explosion with
primary to quaternary blast type injuries. A future study would be
more reliable if it were to remove retrospective limitations. Such
would be the case if patients in a disaster occurred at some sort of
major event where medical professionals are volunteering and there is
an established protocol of which triage criteria are to be used (ie, as
commonly arranged at concerts, marathons, sporting events, large pub-
lic gatherings, etc) and then patients could be followed through defini-
tive treatment as done in our study and more formally compared. A
study applying START, Manchester Sieve, and CareFlight triage systems
to patient data from the transport bombings in London in 2005 found
that missing data hindered their study [9]. They also found that triage
tags from more than 1 manufacturer came in on patients and concluded
that respondents were using their own supplies as well as official equip-
ment [9]. The nature of disaster makes studying the response difficult.
A future study that included more penetrating trauma, more yellow
and black category patients, and more elderly or included children could
allow for subgroup analysis of the accuracy of the triage methods on dif-
ferent patient populations. It may be that 1 triage system is better than
another for specific types of mass casualty events. It also may be that 1
triage system is better than another for different size mass casualty
events. A mass casualty event such as a shooting may overwhelm local
resources for a few hours, whereas a natural disaster with loss of infra-
structure may overwhelm resources for days. Only the SALT method
takes into account likelihood to survive given current resources and
the use of lifesaving interventions.
5. Conclusion
Neither SALT nor START algorithm was appropriately sensitive for
determining a victim's level of triage, especially in the critically injured
who would require immediate intervention. Both START and SALT tri-
age algorithms did have high specificity for predicting death.

Table 4
Results SALT

Sensitivity (n, 95% CI) Specificity (n, 95% CI) Positive predictive value (n, 95% CI) Negative predictive value (n, 95% CI)

Minor/green 91.7% (55/60, 81.6-97.2) 47.5% (19/40, 31.5-63.8) 72.4% (55/76, 60.9-82) 79.2% (19/24, 57.9- 92.9)
Delayed/yellow 20% (1/5, 0.1-71.6) 90.5% (86/95, 82.8-95.6) 10% (1/10, 0.3-44.5) 95.6% (86/90, 89-98.8)
Immediate/red 20.7% (6/29, 8-39.7) 93% (66/71, 84-97.7) 54.5% (6/11, 23.3-83.3) 74.2% (66/89, 63.8-82.9)
Dead or expectant/black 50% (3/6, 11.8-88.2) 100% (94/94, 96.2-100) 100% (3/3, 29.2-100) 96.9% (94/97, 91.2-99.4)
Overall 65% (65/100, 54.8-74.2) 88.3% (265/300, 84.1-91.7) NA NA
 M.C. Bhalla et al. / American Journal of Emergency Medicine 33 (2015) 1687–1691
Acknowledgments
We would like to thank the Summa Akron City Hospital Department
of Surgery, Trauma Division for their assistance in data collection and
project development.
References
[1] Triage—Definition and More from the Free Merriam-Webster Dictionary. [Internet]. [Cited
2013 Nov 15]. Available from: http://www.merriam-webster.com/dictionary/triage.
[2] START adult triage algorithm—CHEMM. [Internet]. [Cited 2013 Oct 4]. Available from:
http://chemm.nlm.nih.gov/startadult.htm.
[3] Kahn CA, Schultz CH, Miller KT, Anderson CL. Does START triage work? An outcomes
assessment after a disaster. Ann Emerg Med 2009;54(3):424–430.e1.
1691
[4] SALT mass casualty triage algorithm—CHEMM. [Internet]. [cited 2013 Oct 4]. Available
from: http://chemm.nlm.nih.gov/salttriage.htm.
[5] SALT mass casualty triage: concept endorsed by the American College of Emergency
Physicians, American College of Surgeons Committee on Trauma, American Trauma
Society, National Association of EMS Physicians, National Disaster Life Support Educa-
tion Consortium, and State and Territorial Injury Prevention Directors Association. Di-
saster Med Public Health Prep 2008;2(4):245–6.
[6] Gebhart ME, Pence R. START triage: does it work? Disaster Manag Response 2007;
5(3):68–73.
[7] Roth R, Idris A, Fowler R. Hypotension and shock. Emerg Med Serv Clin
Asp Prehospital Med. Dubuque, IA: Kendall Hunt Publishing Company;
2009 57.
[8] Cross KP, Petry MJ, Cicero MX. A better START for low-acuity victims: data-driven re-
finement of mass casualty triage. Prehosp Emerg Care 2015;19(2):272–8.
[9] Challen K, Walter D. Major incident triage: comparative validation using data from
7th July bombings. Injury 2013;44(5):629–33.