World J Surg (2015) 39:356–362

DOI 10.1007/s00268-014-2830-6

ORIGINAL SCIENTIFIC REPORT

The Utility of the Kampala Trauma Score as a Triage Tool

in a Sub-Saharan African Trauma Cohort
Bryce Haac • Carlos Varela • Andrew Geyer •

Bruce Cairns • Anthony Charles

Published online: 15 October 2014

Ó Société Internationale de Chirurgie 2014

Abstract
Background Trauma scoring systems have been developed to assess injury severity and may have triage potential.
We sought to evaluate the ability of the Kampala trauma score (KTS) to assess injury severity and its potential as an
outcome predictive tool in Malawi.
Methods This is a prospective cohort study of trauma patients presenting to Kamuzu Central Hospital in 2012. We
recorded admission KTS and Revised trauma score (RTS), emergency department disposition, and hospital length of
stay (LOS) and survival. Logistic regression and ROC curve analyses were used to compare the KTS to the widely
accepted RTS.
Results 15,617 patients presented with trauma. 2,884 (18 %) were admitted, of which 2,509 (95 %) survived. The
mean admission KTS was 14.5 ± 0.6, and RTS was 11.9 ± 0.3. For KTS and RTS, the odds of admission with each
increment increase in score was 0.44 and 0.3, respectively. Similarly, odds of mortality is 0.48 and 0.36. Neither KTS
(p = 0.96, ROC area 0.5) nor RTS (p = 0.25, ROC area 0.5) correlated significantly with hospital LOS. KTS and
RTS performed equally well as predictors of mortality, but KTS was a better predictor of need for admission (KTS
ROC area 0.62, RTS ROC area 0.55, p \ 0.001).
Conclusions Both the KTS and RTS were significantly associated with need for admission and final outcome on
logistic regression analysis; however, they may not be strong enough predictors to merit their use as a screening tool
in our setting.

Introduction
B. Haac
B. Cairns
A. Charles (&)
Department of Surgery,UNC School of Medicine, University of
North Carolina at Chapel Hill, 4008 Burnett Womack Building, Traumatic injury is a significant cause of morbidity and
Chapel Hill 7228, USA mortality globally. The WHO estimates that injuries con-
e-mail: anthchar@med.unc.edu stitute 16 % of the global burden of disease [1]. This
translates into 5.8 million injury-related deaths at a rate of
C. Varela
Department of Surgery, Kamuzu Central Hospital, Lilongwe, 97.9 per 100,000 worldwide. Injuries further account for
Malawi between 10 and 30 % of all hospital admissions and render
at least 78 million people disabled each year [2, 3]. The
A. Geyer burden resulting from injuries tends to be higher in low- to
Air Force Institute of Technology (AFIT/ENC),
Wright-Patterson AFB, OH, USA middle-income countries (LMICs) and communities.

123
World J Surg (2015) 39:356–362
Approximately 87.9 % of all road traffic deaths and 88.3 %
of lost disability-adjusted life years (DALYs) were from
LMICs [4]. Despite the heavy burden from the human
immunodeficiency virus and the acquired immune defi-
ciency syndrome (HIV/AIDS), malaria, and other infec-
tious diseases in LMIC’s, injuries were still responsible for
19.93 % of all deaths of those between the ages of 15 and
59 years, and for nearly one in every four deaths of those
between 15 and 29 years. Most of these deaths resulted
from road traffic injuries, wars, and interpersonal violence.
Sub-Saharan Africa has some of the highest mortality
rates and the greatest number of disability-adjusted life
years lost from injury [5]. At the same time, this region
suffers from a dearth of resources with only 3 % of the
world’s health care workers and less than 1 % of health
care resources [6]. Malawi is one of the poorest countries in
Sub-Saharan Africa, and, despite offering universal free
health care, the country’s total health care expenditure is
only $23 per capita [7]. At the same time, injury is the
second most common non-communicable cause of mor-
tality in Malawi and accounts for 9 % of deaths [8].
Inadequate pre-hospital and emergency healthcare systems
contribute to the devastation of injury in this resource-
limited setting.
With such a great trauma disease burden in a resource-
poor setting, it is imperative that protocols be developed
to strategically allocate resources in a manner that max-
imizes patient survival and minimizes injury-related
morbidity. Many predictive scoring systems have been
developed and used in high-income countries to aid with
injury description, triage, and outcome prediction since
the 1970s including the revised trauma score (RTS),
injury severity score (ISS), and the trauma and ISS
(TRISS) [9–11]. These scoring tools perform well in
industrialized nations, but their calculation can be com-
plex and can require diagnostic tools not available in
resource-poor settings [12]. The Kampala trauma score
(KTS) was developed in 1996 in Uganda to address this
difficulty [13]. The scoring system combines the RTS and
the ISS using both anatomic and physiologic data to
create a score similar to TRISS but simplified to be useful
in sub-Saharan Africa. In Uganda, the KTS has been
shown to be a good predictor of need for hospital
admission, hospital length of stay (LOS) greater than
2 weeks, and death [13, 14].
We sought to evaluate the utility of the KTS at Kamuzu
Central Hospital (KCH), a tertiary care referral hospital in
Lilongwe, Malawi and compare the performance of the
KTS to the RTS in predicting the outcomes of hospital
admission, LOS greater than 2 weeks, and mortality in our
setting. Finally, we examined the feasibility of training lay
people to calculate the score; a system that could aid in
triage in settings where clinicians are scarce.
357
Methods
Study setting, design and population
KCH has five surgical wards with 235 adult surgical beds,
over 5,000 surgical admissions a year, and an average of
15,000 trauma patients seen each year. The hospital serves
a catchment of 800,000 from the capital city, Lilongwe,
and referrals from the country’s entire central region (pop.
5 million). The vast majority of trauma patients present to
the casualty department for evaluation, but a few patients
are evaluated in the orthopedic outpatient clinic and never
pass through the casualty department. The General Surgical
Staff available at KCH during the study period included
four full-time general surgery consultants, one urologist,
and sixteen full-time clinical officers. Also operating dur-
ing the study period was eleven Malawian general and
orthopedic surgical residents and two American general
surgery residents. The main operating theater at KCH has
four fully functional operating rooms with one anesthesi-
ologist and six clinical officer anesthetists.
We conducted a prospective cohort study of patients
presenting to KCH with trauma from January 1 through
December 31, 2012. All adult patients presenting with
trauma to the KCH casualty department were eligible.
Trauma patients who were admitted directly through the
orthopedic outpatient clinic were also eligible once found
on the surgical wards. Patients admitted through outpatient
orthopedics were included in the trauma registry, but only
had an admission KTS calculated if all the necessary vital
signs were documented in the chart by the admitting cli-
nician. Due to lack of pediatric blood pressure cuffs,
trauma scores could not be calculated for pediatric patients.
Data collection
Trauma registry clerks collected all data. Clerks were lay
people who had completed secondary school or higher and
were able to read and write in both English and Chichewa,
the local language. The former trauma registry form as
described in Samuel et al. was revised to include a section
for KTS and RTS scoring [15].
We held an initial 3-hour training to orient the clerks to
the updated trauma registry form and the KTS and RTS
scoring systems. In addition, we taught the clerks to check
a manual blood pressure, heart rate, and respiratory rate.
They were also educated on how to calculate the AVPU
neurologic score. The clerks were given a chart with nor-
mal vital sign ranges and guidelines for contacting a nurse
or clinician if an abnormal vital sign is found. Registry
clerks completed a take home test with example patient
problems following the training. We reviewed any calcu-
lation errors individually with each clerk.
123
358 World J Surg (2015) 39:356–362

Table 1 Coding values and calculation of the Kampala trauma score (KTS) and revised trauma score (RTS) KTS coded value
Age (years) A. __________
B5 1
6–55 2
[ 55 1
Number of serious Injuries B. ___________
None 3
One 2
Two or more 1
Systolic blood pressure (mmHg) C. __________
[ 89 4
50–89 3
1–49 2
Undetectable 1
Respiratory rate (breath/min) D. __________
10–29 3
C 30 2
B9 1
Neurological status (AVPU system) E. ___________
Alert 4 KTS Total:
Responds to verbal stimuli 3 A ? B?C ? D?E = ______________
Responds to painful stimuli 2
Unresponsive 1
Glasgow coma scale Systolic blood pressure Respiratory rate RTS Coded value

13–15 [89 10–29 4

9–12 76–89 [29 3
6–8 50–75 6–9 2
4–5 1–49 1–5 1
3 0 0 0

After orientation but prior to the study period, a US-
trained medical student or surgery resident rounded with
the trauma clerks daily on the wards and verified all
manually taken vital signs for 2 weeks at which point
measurements taken by the clerks corresponded to those
taken by the resident or student. The medical student or
resident continued to spot check vital signs including
rechecking any abnormal vital signs for 2 additional weeks.
During the study, clerks were provided with and trained to
use automatic OmronBP760 7 Series Upper Arm Blood
Pressure Monitor machines to measure vital signs. If the
machines failed, clerks were instructed to revert to manual
vital sign measurements. The clerks collected respiratory
rate manually.
Data were collected in two stages. First, on presentation
to the casualty department patient demographic informa-
tion, injury description, initial vital signs, KTS, RTS, and
patient disposition were recorded on the trauma registry
form. Trauma clerk coverage was provided 24 h a day,
7 days a week in the casualty department. General and
orthopedic surgeons made all decisions regarding need for
admission, discharge, and surgical interventions. For
patients admitted directly through outpatient orthopedics,
the clerks completed the registry form when the patient was
identified on the hospital wards, usually on admission day
2. KTS and RTS were only calculated for these patients if
the admitting physician recorded admission vital signs. All
scores were retrospectively re-calculated by a US-trained
physician to confirm correct calculations for the purposes
of analysis. (Table 1).
The second stage of data collection occurred on the
hospital wards. Trauma clerks rounded daily on the surgi-
cal wards to collect outcome data including final disposi-
tion and LOS on all registry patients and to identify any
trauma patients missed in the casualty department.
Data analysis
Trauma registry forms were pre-numbered to give each
patient admission a unique identifier and prevent

123
World J Surg (2015) 39:356–362 359

Table 2 Basic demographic characteristics of admitted trauma Table 3 Mechanisms of Injury for Presenting Trauma Patients
patients
Mechanism of injury Number of Percentage of total
Characteristic Number and percentage injuries injuries (%)

Age (years) Road Traffic accidents: 3,704 24

Mean, range 24 (0–96) Pedestrian hit by:
Missing values 9 Car 777 5
Gender Bicycle 181 1
Male 2,096 (73 %) Motorcycle 30 \1
Female 786 (27 %) Minibus/pickup 145 1
Missing values 2 Bus/lorry 37 \1
Head injuries (includes head, skull, eyes, ears, nose, jaw/mandible) Ox Cart 14 \1
No 1,870 (65 %) Other 1 \1
Yes 1,002 (35 %) Driver/passenger in a:
Missing values 12 Car 598 4
Chest injuries Cab of lorry 62 \1
Yes 2,662 (93 %) Minibus 253 2
No 209 (7 %) Bus 36 \1
Missing values 13 Open bed of lorry 335 2
Spinal injuries (includes neck, cervical, lumbar and thoracic spine Motorcycle 119 \1
injuries) Bicycle 1,060 7
Yes 347 (12 %) Ox Cart 56 \1
No 2,524 (88 %) Animal bite 493 3
Missing values 13 Human bite 67 \1
Abdominal Injuries Gun shot wound 36 \1
Yes 282 (10 %) Foreign body 527 3
No 2,589 (90 %) Fall 4,740 31
Missing values 13 Electric shock/ 22 \1
Pelvic Injuries lightening
Yes 65 (2 %) Occupational injury/ 130 \1
No 2,806 (98 %) machine
Missing values 13 Burn 681 4
Extremity Injuries Assault 3,681 24
Yes 1,818 (63 %) Collapsed structure 829 5
No 1,053 (37 %) Drowning 70 \1
Missing values 13 Laceration (non- 507 3
assault)
Final Outcome
Discharged 2,423 (86 %)
Absconded 84 (3 %) Validation of KTS was done using ROC (receiver
Died 127 (5 %) operating characteristics) curves to assess the overall per-
Transferred 2 (0.1 %) formance of the score assuming equal importance of sen-
Missing values 71 sitivity and specificity. Separate ROC curves were
Surgical Procedure performed constructed to compare the performance of KTS to the
Yes 1,149 (42 %) more widely used RTS in prediction of each outcome
No 1,595 (58 %) variable: mortality and LOS greater than 2 weeks. Areas
Missing values 110 under the curve and 95 % confidence intervals were com-
pared. Treating each score as a continuous variable, odds of
death and hospitalization at 2 weeks was calculated using
logistic regression. Two weeks were used as a marker for
duplication. Data were entered into Access 2007 and ana- consistency with other published papers that have used that
lyzed using Stata 11. Duplicates were removed and outlier time frame to determine score association with LOS [14].
variables were verified manually. Two-dimensional scatter plots were generated for each

123
360 World J Surg (2015) 39:356–362

Table 4 Injury setting for presenting trauma patients in 2012

1.00
Place where injury Number of Percentage of patients
occurred patients (%)

0.75
Sensitivity
Home 7,066 46
Work 1,239 8

0.50
Road/street 5,184 33
School 489 3

0.25
Farm 136 \1
Sport/recreation 498 3

0.00
Public space 740 5
0.00 0.25 0.50 0.75 1.00
Lake/river 58 \1
1-Specificity
Field/forest 39 \1
Other 10 \1 KTS ROC area: 0.6904 RTS ROC area: 0.6703
Reference

Fig. 1 Comparison of ROC curves of KTS and RTS as predictors

Table 5 KTS and RTS as predictors of outcome. odds of death for of death for all presenting traumas
every one-point increase in the trauma score
OR 95 % CI p value

All traumas
*KTS 0.26 0.20–0.33 \0.001 1.00
*RTS 0.18 0.13–0.25 \0.001
0.75

Admitted traumas only

Sensitivity

*KTS 0.48 0.36–0.64 \0.001

0.50

*RTS 0.35 0.26–0.49 \0.001

0.25

trauma score component. Patient demographic data point

0.00

was color-coded based on patient mortality outcome. This

was a way to separate patients who lived or died based on 0.00 0.25 0.50 0.75 1.00

any single or combination of trauma score components or 1-Specificity

patient demographic information. If the two populations of
patients were separated in any dimension on these plots, it
would be considered a good indicator that a statistical
classifier such as logistic regression can be used effectively
to find the relationship between trauma score and patient
mortality.
Ethics
Both the Malawi National Health Science Research
Council and the University of North Carolina IRB com-
mittees approved this study.
Results
From January 1st 2012 to December 31st 2012, a total of
15,617 patients presented with trauma, 2,884 (18 %) of
which were admitted to the hospital wards. The basic
demographic description of admitted trauma patients is
described in Table 2. Of these traumas, the most common
KTS ROC area: 0.5929 RTS ROC area: 0.6371
Reference
Fig. 2 Comparison of ROC curves of KTS and RTS as predictors
of death for all admitted traumas
causes of injury were fall (31 %, n = 4,740), road traffic
accident (24 %, n = 3,704), and assault (24 %, n = 3,681)
(Table 1). The most common locations where injury
occurred were at the patient’s home (46 %, n = 7,066) and
on the road (34 %, n = 5,184) (Table 3 and 4).
The mean age of all presenting trauma patients was
23 ± 16 years, (range 0–97). The majority of trauma
patients (72 %, n = 11,129) were male. Alcohol was a
contributing factor in 5 % (n = 731) of traumas. The mean
delay in presentation from time of injury was
1.3 ± 7.5 days, with the majority of patients presenting on
the same day of injury (median and mode 0 days, range
0–366 days). The mean KTS was 14.5 ± 0.6 (range 9–16,
n = 9180), and the mean RTS was 11.9 ± 0.3, (range
5–12, n = 8,921) for all presenting traumas.

123
World J Surg (2015) 39:356–362
For admitted patients, the mean KTS was 14.2 ± 0.7
(range 10–16, n = 1,350), and the mean RTS was
11.8 ± 0.5 (range 8–12, n = 1,305) on presentation. Of
admitted patients with a recorded final outcome, 2,509
(95 %) lived (2,423 were discharged, 84 absconded and 2
were transferred). 177 patients were recorded as lost to
follow-up and were excluded from the final outcome ana-
lysis. The mean LOS was 12 ± 16 days (range 0-16,
median 5 days). 761 (27 %) of patients had a LOS greater
than 2 weeks.
Both the KTS and RTS trauma score were significantly
associated with final outcome (Table 5). In admitted
patients, for every one-point increase in the admission
KTS, the odds of death was halved (OR 0.48, 95 %
CI = 0.36–0.64, p \ 0.001). Similarly, for every one-point
increase in the admission RTS, odds of death was reduced
by approximately one third (OR 0.36, 95 % CI 0.26-0.49,
p \ 0.001). There was no significant correlation between
LOS greater than two weeks and KTS (p = 0.96, area
under the ROC curve 0.5) or RTS (p = 0.25, area under the
ROC curve 0.5).
KTS and RTS performed equally well as predictors of
death. There was no significant difference in ROC curves
for KTS and RTS as predictors of death in all trauma
patients (p = 0.56, Fig. 1) and in admitted traumas only
(p = 0.23, Fig. 2). Some patients with low trauma scores
survived and multiple patients with high trauma scores
died; however, resulting in ROC curves that do not vary
greatly from the no-discrimination line. Consistent with
these findings, preliminary scatter plot analysis indicated
that the population of patients who died was not separated
from the population who lived for any individual or com-
bination of trauma score components or patient demo-
graphic information. The two populations of patients were
intermixed in all graphs, indicating that there may not be a
strong statistical relationship between the trauma score and
patient mortality in the sample data. Further statistical
analysis using the corrected Akaike Information Criterion
(AICc) was performed on both RTS and KTS. The AICc
for RTC and KTS were is 612.981 and 635.136, respec-
tively. AICc showed no difference in model performance.
Discussion
The performance of KTS was similar to the universally
accepted RTS as a predictor of mortality based on the ROC
curve analysis in this study, but KTS was a statistically
significantly better predictor of need for admission. The
superior performance of KTS in predicting need for
admission may be attributable to the score components.
KTS includes data on the number of injuries unlike the
RTS, which only includes physiologic data points. The
361
anatomic variables (age and number of injuries) are likely
to factor into a clinician’s decision to admit a patient.
Patients with a larger number of injuries and normal vital
signs might be admitted for observation due to concern for
missed injury or more serious trauma mechanism. Older
and younger patients with less severe injuries may be
admitted out of concern for decreased physiologic reserve.
Neither the KTS nor the RTS was good indicators of
prolonged hospitalization in our population. Multiple fac-
tors affect LOS in a resource-poor setting. Poor patient
follow-up, minimal infrastructure, high cost of transporta-
tion, and increasing distances between home and the hos-
pital, compounded by a dearth of health care workers and
other unknown cultural norms can all increase LOS. Sim-
ilarly, social factors including a patient’s support network
at home or a patient’s need to return to a work may reduce
LOS. For example, 84 patients in our study absconded
suggesting they left the hospital before a physician deemed
them medically appropriate for discharge. As a result, LOS
is often influenced by subjective data that cannot be
accounted for in a scoring tool making the predictive
probability for LOS of the KTS or the RTS in our envi-
ronment unreliable.
Incremental increases in both scores were significantly
associated with decreased need for admission and
decreased mortality according to the logistic regression
models, consistent with previous studies [13, 14, 16].
However, the ROC curves for these outcomes did not vary
significantly from the line of no discrimination, suggesting
that they may not be strong enough predictors to merit their
use as a screening tool in our setting. Multiple factors not
accounted for by the trauma scores likely affect survival in
our setting. The aforementioned obstacles to care, poor
nutritional reserve, poor patient monitoring, and reliance
on family members to alert health workers on critical
changes in clinical status, increasing clinical complica-
tions, and ultimately failure to rescue can all affect sur-
vival. Unlike the validation studies in Uganda, we followed
patients until discharge or death, which was often longer
than their two-week follow-up. With longer follow-up
comes increased potential for complications and deaths. In
addition, potentially lethal injuries can be missed on
admission or in the initial stages of treatment. Neither of
these scores account for the anatomy of injury, medical
comorbidities, or potential injuries that could be caused by
blunt mechanisms; all of which factor into a patient’s
overall prognosis and are important for triage decisions.
These factors could account for some of the deaths we saw
in patients with less severe injury as defined by their
admission trauma score.
The mean KTS and RTS of our population are relatively
high. The lack of a structured emergency response system
in Malawi can make it difficult for trauma patients to get to
123
362
the hospital. As a result, patients with more severe injuries
often do not make it to the hospital or die in transit like the
132 patients in our study sample who were brought in dead
and excluded from analysis. If these patients were able to
make it to the hospital alive, we may see greater correlation
between the trauma scores and mortality because we would
have a larger number of patients with non-survivable
injuries whose outcomes are not affected by disparities in
post-trauma care.
Despite the low percentage of severe injuries and deaths,
we believe our trauma population is representative of the
population of patients that would present to any similar
tertiary care center in a resource-poor setting with similar
pre-hospital emergency support. Preliminary scatter plot
analysis failed to identify any individual or combination of
trauma score components that separated the population of
patients who died from those who lived in our study. The
lack of a strong enough statistical relationship between the
trauma scores and patient mortality suggests that the
functionality of KTS and RTS as triage tools is limited in
our setting. While the scores may serve as good retro-
spective classifiers of injury severity, care must be taken
before using these scores to aid in clinical decision making
with a rudimentary trauma system.
Acknowledgments The authors wish to thank Jared Tomlinson MD
for his help in facilitating data collection. This research was supported
by the Doris Duke Charitable Foundation; The UNC Institute of
Global Health and Infectious Diseases; the UNC Center for AIDS
Research.
References
1. World Health Organization (WHO) (2002) World report on
violence and health. WHO, Geneva
123
World J Surg (2015) 39:356–362
2. Berger L, Mohan D (1996) Injury Control: A Global View.
Oxford University Press, New York
3. World Health Organization (WHO) (2010) Injuries and violence:
the facts. WHO, Switzerland
4. Mathers CD, Sadana R, Salomon JA et al (2001) Healthy Life
Expectancy in 191 Countries, 1999. Lancet 357:1685–1691
5. World Health Organization (WHO) (2002) The Injury Chart
Book. Geneva, Switzerland. Available at whqlibdoc.who.int/
publications/924156220x.pdf. Accessed 5 April 2014
6. World Health Organization (WHO) (2006) The World Health
Report 2006: working together for health. Geneva. Available at
http://www.who.int/whr/2006/whr06_en.pdf?ua=1. Accessed 5
April 2014
7. World Health Organization (WHO) WHO Africa region: Malawi
statistics summary (2002 to present). Available at http://apps.
who.int/gho/data/?theme=country&vid=12800. Accessed 5 April
2014
8. World Health Organization (WHO) Noncommunible diseases
country profiles 2011. Available at http://www.who.int/nmh/
countries/mwi_en.pdf?ua=1. Accessed 5 April 2014
9. Baker S, O’Neill B, Haddon W et al (1974) The injury severity
score: a method for describing patients with multiple injuries and
evaluating emergency care. J Trauma 14:187–196
10. Boyd C, Tolson M, Copes W (1987) Evaluating trauma care: the
TRISS method. J Trauma 27:370–378
11. Champion H, Sacco W (1989) A revision of the trauma score.
J Trauma 29:623–629
12. Kingston R, O’Flanagan SJ (2000) Scoring systems in trauma. Ir
J Med Sci 169(3):168–172
13. Kobusingye OC, Lett RR (2000) Hospital-based trauma registries
in Uganda. J Trauma 48(3):498–502
14. MacLeod JBA, Kobusingye O, Frost C et al (2003) A comparison
of the kampala trauma score (KTS) with the revised trauma score
(RTS), injury severity score (ISS) and the TRISS method in a
Ugandan trauma registry: is equal performance achieved with
fewer resources? Eur J Trauma 29(6):392–398
15. Samuel JC, Akinkuotu A, Baloyi P et al (2010) Hospital-based
injury data in Malawi: strategies for data collection and feasibility
of trauma scoring tools. Trop Doctor 40(2):98–99
16. Chalya PL, Mabula JB, Dass RM et al (2012) Injury character-
istics and outcome of road traffic crash victims at Bugando
Medical Center in Northwestern Tanzania. J Trauma Manag
Outcomes 6(1):1
Reproduced with permission of the copyright owner. Further reproduction prohibited without
permission.