Content uploaded by Amir Khorram-Manesh
Author content
All content in this area was uploaded by Amir Khorram-Manesh
Content may be subject to copyright.
Www.jedm.org JEDM2013.3 | Page 12
Journal of Emergency & Disaster Medicine 2, 1 (2013) | Article
Validation of Swedish Emergency Medical Index in trauma pa-
tients
Schagerlind L
1, 2,*
, Örtenwall P
1
, Widgren BR
2
, Taube M
3
, Asplén B
1
, Örninge P
1
, Khorram-Manesh
A
1,*
1. Prehospital and Disaster Medicine Centre, Gothenburg, Sweden
2. Research and development unit (FFOU), Halmstad, Sweden
3. Sahlgrenska University Hospital, Gothenburg, Sweden
*
Corresponding author: Lars Schagerlind, MD, Regionens Hus, SE-405 44, Gothenburg, Sweden (Email:
lars.schagerlind@vgregion.se; Tel: +46 706 779 000)
Amir Khorram-Manesh MD, PhD, Regionens Hus, SE-405 44, Gothenburg, Sweden (Email: amir.khorram-
manesh@surgery.gu.se; Tel: +46 707 22741)
Abstract: Introduction: All incoming medical calls to a Swedish dispatch center are triaged using a medical index
(MI). In a recent study, we presented a discrepancy between priority set by the dispatchers and the ambulance crews
using a clinical, knowledge based, triage. In this study, we aim at investigating the accuracy of this medical index, by com-
paring it to a validated prehospital/hospital triage; RETTS (Rapid Emergency Triage and Treatment System), in a group of
trauma patients during a known period of time, using available medical records, including outcome. Material and
Method: All ambulance transports in Gothenburg Sweden, registered during 2010 and all trauma patients triaged with
both systems; MI by dispatchers and RETTS by ambulance crews were reviewed. Outcome of each patient could be re-
trieved by using the national trauma registry “KVITTRA.” In dubious cases, the conversation between the caller and emer-
gency dispatcher could also be reviewed. Results: We found over-triage of 35% and under-triage of 15.5% among 3079
patients, triaged with both MI and RETTS. Of 27 under-triaged patients, 3 died within 30 days after trauma (all over 80
years of age). Analyses of conversation tapes revealed only one preventable death. Conclusion: We conclude that,
when compared to a validated triage method, Swedish MI seems to have an acceptable level of accuracy. It might be
possible to improve its accuracy, to some extent, by adding some specific questions into its algorithm. However, to prove
this hypothesis; we would need a larger study with a more unselected study population.
Received: May 11
th
, 2012, Accepted: September 1
st
, 2012, Published: February 10
th
, 2013
D
uring the last decades there has been an increas-
ing focus on the effectiveness and optimization of the hospital
and prehospital systems in Sweden due to a growing problem
with overcrowded emergency departments (ED). One reason
for this could be an increasing number of ambulance trans-
ports to the hospitals (1). Triage of incoming calls for a proper
provision of ambulances is essential for resource optimiza-
tion. All ambulances in Sweden are dispatched from an emer-
gency dispatch center.
Www.jedm.org JEDM2013.3 | Page 13
Until January 2012, only one medical index (MI) has been
used by dispatchers to triage all incoming calls. This MI has
been used since the 80´s in different countries, but has never
been critically and scientifically validated in Sweden or to the
best of our knowledge anywhere else. MI is an American de-
veloped 4-grade triage algorithm, adapted to Norwegian and
Swedish conditions. It is an algorithm developed to serve
dispatchers to decide a priority level based upon the informa-
tion received by a phone call (2).
Recently, we presented a discrepancy between priority set
by the emergency dispatcher using MI and that of ambulance
crews using a clinical, knowledge based, triage (1). There
were both over- and under-triage. While over-triage is an eco-
nomic loss, under-triage is considered a hazard for the pa-
tient.
In this study, we aim at investigating the accuracy of the
MI by comparing it to a validated prehospital/hospital triage –
RETTS (Rapid Emergency Triage and Treatment System),
earlier called METTS, in a group of trauma patients during a
known period of time, using available registry with data in-
cluding input variables and outcome (3).
RETTS is a triage algorithm combining vital signs, chief
complaints, symptoms, and signs to give the priority on five
different levels. The highest priority is labeled Red. The fol-
lowing priorities are in order of declining priority: Orange, Yel-
low, Green and Blue. The RETTS protocol is reasonably vali-
dated (1, 3). It has also been used in the prehospital setting
during several years in Gothenburg. The prehospital protocol,
which was used by ambulance crews in this study, has a low
over-triage and a negligible under-triage in a trauma popula-
tion (4).
Method
All ambulance transports in greater Gothenburg, Sweden
(with around 1 million inhabitants and an area of 3694.86 km
2
) are registered in a database called AmbuLink (1). This reg-
istry contains information regarding the patient, priorities as-
signed by the medical dispatchers using the MI and the am-
bulance crews (RETTS), reason for transport, origin, destina-
tion, etc.
In order to have a group of patients with known input data
and accessible outcome, we chose trauma patients. This
group of patients is registered in AmbuLink and also in the
National trauma registry “KVITTRA” (5). We used data ex-
tracts from the AmbuLink registry saved as Excel files
(Microsoft Corp, USA). Out of these files, all patients, coded
as S (Injuries), K18 (Concussion), K19 (Severe head injury)
and TOL (Traffic) were extracted.
Among these ambulance transports, those coded as ur-
gent/emergency runs were selected. All patients who had
priority given in both MI, and RETTS were recognized. The
outcome of each case could be obtained as 30- day mortality
as well as the reason of death by using KVITTRA. We chose
to compare P1 in MI with combined Red and Orange in
RETTS. Priority 2 was compared to Yellow priority in RETTS
and priority 3 in MI with that of Green and Blue in RETTS.
All calls to the dispatch center are recorded and stored.
Selected cases with discrepancies in triage between the
emergency dispatch center and the ambulance crews were
evaluated by one of the authors and a very experienced dis-
patcher with knowledge in both index and priority setting.
Statistic
In order to evaluate the discordance/concordance of the
obtained results, Cohen´s kappa with Fleiss-Cohen weights
was used as a measure of concordance. Bowker´s test of
symmetry was performed to investigate the systematic differ-
ence between the two ratings. The difference in mean ratings
between RETTS and Index priority was used as rough meas-
ures to quantify the magnitude of the systematic difference
between ratings.
The SAS statistical package was used for the analysis of
data (6).
Results
Missions and priorities
There were 40,893 ambulance missions recorded in Am-
buLink between 1
st
January and 31 December 2010 in Goth-
enburg. Out of these, 4753 missions corresponded to diag-
nose codes; S, K18, K19, and TOL. Among these missions,
4137 were identified as Urgent. From this data 3079 patients
with triage in both MI and RETTS could be extracted. Table 1
shows the results obtained by the methodological review of
data in numbers and percent.
Of those prioritized as priority 1 according to MI, 954
(65%) were assessed as Red or Orange according to
RETTS. In patients prioritized as priority 2 by MI, 493 (34%)
were assessed as Red or Orange by RETTS and of those
with priority 3 using MI, there were 26 Red or Orange priori-
ties (15.5%). Thus, in this material, comparing MI with
RETTS, we found an over-triage of 35% (n=519) using MI
priority 1. We also found an under-triage of 15.5% (n=26) i.e.
26 patients were triaged as priority 3 (P3) according to MI,
but were found to be Red or Orange based on RETTS. There
was also one patient who was assessed as Priority 4 accord-
ing to MI, but Orange according to RETTS. To identify the
Lars Schagerlind, et al.| Validation of Swedish Emergency Medical Index in trauma patients
Article
Www.jedm.org JEDM2013.3 | Page 14
reason for under-triage, this group of patients (n= 26 + 1) was
further analyzed by a review of their medical records.
Analysis of 27 under-triaged patients
One patient had a rapid atrial fibrillation and was thus ex-
cluded. Of the remaining 26 patients, in five patients no pre-
hospital or hospital emergency record could be found to clar-
ify the reason for their priority-setting. Another patient had an
emergency record from ED but no prehospital records. How-
ever, in the ED record, it was stated that the patient was pri-
oritized Red due to a fall from more than 3 meters height. Of
the remaining 20 patients, one was Red according to RETTS
due to unconsciousness at arrival. Of 19 patients, prioritized
as Orange, 10 were prioritized RETTS Orange based on ESS
(Emergency Symptoms and Signs) algorithms; four patients
due to anti-thrombosis treatment and the remaining six due to
intoxication, suspected face fractures, severe pain or swelling
and impaired sensibility in extremities. Five patients had an
Orange priority due to low oxygen saturation (POX <90 with-
out O2 supply). Two patients had a respiratory rate > 25/
minute. One patient had a heart rate > 110/minute. The P4
patient who turned Orange based on ESS had an anti-
thrombosis treatment (Figure 1).
Analysis of the conversation between callers and dis-
patchers
All 27 under-triaged patients were further examined by
listening to their calls to the emergency dispatch centre. Of
three cases with Red priority, two were found to be correct
triaged by the dispatcher based on the conversation, and the
information dispatcher received from the caller. In one case,
there was a possibility that more detailed questioning based
on the algorithm could have revealed the patient’s ongoing
SDH (Subdural Hematoma) and consequently changed the
priority he was given from P3 to P1. Of remaining 24 cases
with Orange priority, there were no audio files in four cases.
In remaining 20 cases all were found to be correctly priori-
tized based on the conversations and information given to
dispatchers by callers.
30- Day mortality
In this study, three patients died within 30 days after the
trauma. They were all over 80 years of age (81, 85, and 90
years old, mean + SD = 85.3 + 4.5). The death reasons were
SDH in two cases and unknown in the third patient, who was
discharged from the hospital and died at home (Table 2).
Statistical analysis
The statistical analysis with Cohen´s kappa and Fleiss-
Cohen’s weights showed a weighted kappa of 0.3035 with an
ASE (average sum of error) of 0.0173 (95% CI; 0.2697-
0.3373). The Bowker´s test of symmetry resulted in a p-value
of <0.0001, which indicates a systematic difference between
RETTS and Index priorities.
A t-test of differences in mean points between RETTS and
Index priorities showed a value of 0.06 (95% CI: 0.033-0.087)
and a p-value of 0.00001. It is thus clear that, although small,
there is a difference between RETTS and Index priorities.
There is also a tendency of low RETTS priorities giving
higher priorities in Index, while higher RETTS priorities give
lower value in index.
Discussion
The most common causes of the ambulance requests are
respiratory and circulatory disorders. However, in this study,
we chose to focus on triage of trauma patients since this
group of patients were easy to define and follow up. Our main
objective was to compare two different triage methods; one to
be used to prioritize ambulance provision based on an incom-
ing call and symptom`s description and the other based on
physiological parameters.
RED Orange Yellow Green Blue Total
Priority 1 348 (23.7%) 606 (41.3%) 387 (26.4%) 122 (8.3%) 3 (0.3%) 1466 (100%)
Priority 2 35 (2.4%) 458 (31.7%) 766 (53.1%) 178 (12.3%) 7 (0.5%) 1444 (100%)
Priority 3 3 (1.8%) 23 (13.7%) 92 (54.8%) 48 (28.5%) 2 (1.2%) 168 (100%)
Table 1: Shows the number (percent) of patients triaged by MI in each priorities and their corresponding distribution in all levels of priority in RETTS
Lars Schagerlind, et al.| Validation of Swedish Emergency Medical Index in trauma patients
Article
Www.jedm.org JEDM2013.3 | Page 15
Study limitations
There are several limitations in this study. Firstly, we com-
pare a method with four grade scale priority (MI) with a five
grade scale (RETTS) (1-4). We chose to compare P1 in MI
with combined Red and Orange in RETTS. Priority 2 was com-
pared to Yellow priority in RETTS and priority 3 in MI with that
of Green and Blue in RETTS. There can be some objections
on how we matched these two grades; however, this grading
was suggested by investigators and was approved by a panel
of experts in both triage and specialties (Internal and emer-
gency medicine, trauma and anesthesiology), based on the
existing definition of each priority (1, 3). We have chosen to
use RETTS Red and Orange as the equivalent to P1 because
Red and Orange priority results in a trauma alert at receiving
hospital. It might have been more correct to translate P3 to
only RETTS Green instead of Green and Blue and use P4 as
the equivalent to Blue. However, we have chosen not to do so
because the very scarce number of blue patients. This has not
changed the result obtained from this study, but it may be a
more correct definition in a larger cohort of patients.
Another limitation is the time difference between the first
triage performed by medical dispatchers using MI, and the
second triage performed by ambulance crews using RETTS.
However, as the goal for ambulance crews in Gothenburg is to
cover 90% of all priority one calls within 10 minutes and antici-
pating that all trauma calls are emergencies and with high pri-
orities, there is a reason to believe that the time difference
between these two triage occasions is limited. We have not
analyzed time tables for the patients.
Discrepancies between two methods
Looking at the reasons for discrepancies between two
methods in our material, we could identify factors such as
blood pressure, oxygen saturation, pulse and respiratory rate
and medication with anti-thrombosis medication (1-4). In one
patient, unconsciousness was not prioritized higher than P3;
this could possibly be a clinical error. Another patient priori-
Figure 1: The outcome of 27 under-triaged patients in this study.
Lars Schagerlind, et al.| Validation of Swedish Emergency Medical Index in trauma patients
Article
Www.jedm.org JEDM2013.3 | Page 16
tized P3 by MI, but red according to RETTS, was a patient on
a ferry who had fallen more than 3 meters. Since there was no
possibility to assess the patient before the ferry reached the
port, he was prioritized as P3.
Of the total number of 27 under-triaged patients, three died
within the 30 days after the trauma. All patients were older
than 80 years. It is well known that elderly suffers from higher
mortality when exposed to trauma (7).
Evaluation of audio files of the 27 under-triaged patients
revealed that in only one patient prioritized by MI, this priority
could have been up-graded to a higher priority (Red as in
RETTS) if the dispatcher asked more detailed questions re-
garding the mechanism of the injury. Typical fluctuation of the
status and the mechanism of injury could have revealed the
possibility of SDH. Remaining cases all were found to be cor-
rectly prioritized based on the conversations and information
given to dispatchers by callers.
The statistical method used showed that there is a concor-
dance between the results, meaning that the discrepancies
found in this study are significant.
The over-triage of 35% is higher than the internationally
accepted level of around 25% (1, 8). Considering that all over-
triaged are just an economic and administrative problem, we
focused on all under-triaged who might have suffered a nega-
tive medical impact. Our material has an under-triage of 15.5%
(patients prioritized as P3 in MI but as Red or Orange in
RETTS). There are fundamental differences between the com-
pared triage methods as the first one (MI), only relies on the
symptom description given by a caller and maybe also the
clinical background of the dispatcher who receives the call,
while RETTS is a protocol, including triage algorithm combin-
ing vital signs, chief complaints, symptoms, and signs to give
the priority level. RETTS can never replace MI as a tool for
triage in incoming emergency calls to the dispatch center
since the operator needs an algorithm based purely on the
history told by the caller. The MI algorithm might be improved
by adding optional information and questions on the question
list. Possible questions could, for example, be “known or esti-
mated age of the patient?” alternatively, “the use of anti-
thrombosis medication?” However, such completion of MI
needs further and larger studies with statistical analysis.
Conclusion
We conclude that when compared to a validated triage
method, Swedish MI, has a higher over-triage than the interna-
tionally accepted figures, but more seriously an under-triage of
over 15% in the P3 group. However, in this study, the only
preventable under triage was just one patient. Thus, MI seems
to have an acceptable level of accuracy. It might be possible
to improve its accuracy, to some extent, by adding some spe-
cific questions into its algorithm, but to prove this hypothesis;
we would need a larger study with a more unselected study
population. We suggest that similar work be done for more
complicated ambulance calls such as cardiac emergencies.
Acknowledgement
The authors would like to thank Mr Kjell Pettersson, School
of Business, Economics and Law at University of Gothenburg
for his contribution to the statistic analysis and Laerdal Foun-
dation for financial support.
Age 30 day mortality Triage priority Death reason
69 Alive Red
27 Alive Red
85 Death Red Subdural hema-
toma
87 Alive Orange
28 Alive Orange
55 Alive Orange
85 Alive Orange
67 Alive Orange
81 Alive Orange
75 Alive Orange
81 Death Orange Subdural hema-
toma
48 Alive Orange
90 Death Orange Not available
90 Alive Orange
75 Alive Orange
56 Alive Orange
86 Alive Orange
68 Alive Orange
44 Alive Orange
48 Alive Orange
39 Alive Orange
85 Alive Orange
96 Alive Orange
83 Alive Orange
20 Alive Orange
87 Alive Orange
Table 2: Shows the 30-day mortality in the under triaged group of pa-
tients.
Lars Schagerlind, et al.| Validation of Swedish Emergency Medical Index in trauma patients
Article
Www.jedm.org JEDM2013.3 | Page 17
References
1. Khorram-Manesh A, Lennquist Montan K, Hedelin A,
Kihlgren M, Ortenwall P. Prehospital triage, discrepancy in
priority-setting between emergency medical dispatch cen-
tre and ambulance crews. Eur J Trauma Emerg Surg.
2011; 37(1): 73-78
2. Robertson-Steel. Evolution of triage system. Emerg Med J.
2006; 23: 154-155
3. Widgren BR, Jourak M. Medical Emergency Triage and
Treatment System (METTS): a new protocol in primary
triage and secondary priority decision in emergency medi-
cine. J Emerg Med. 2011; 40(6): 623-628
4. Widgren BR, Nilsson G, Ortenwall P. Prehospital triage
enligt METTS-T ger mer effektiv traumasjukvård. Läkar-
tidningen. 2009; 106 (11): 746-749
5. Ortenwall P. Start of a national trauma registry. Läkartidnin-
gen 2000; 97 (42): 4760-4762
6. Bland JM. Measurement in health and disease; Cohen´s
kappa. http://martinbland.co.uk/
7. Sharma OP, Oswanski MF, Sharma V, Stringfellow K, Raj
SS. An appraisal of trauma in the elderly. Am Surg 2007;
73: 354-358
8. Lehmann RK, Arthurs ZM, Cuadrado DG, Casey LE, Beek-
ley AC, Martin MJ. Trauma team activation: simplified crite-
ria safely reduces overtriage. Am J Surg 2007; 193: 630-
635
Lars Schagerlind, et al.| Validation of Swedish Emergency Medical Index in trauma patients
Article
