Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

C
URRENT
O
PINION
Trauma scoring systems
Rolf Lefering
Purpose of review
Trauma scoring systems are used by researchers, registries, or individuals to describe injury severity or to
estimate the prognosis of trauma patients. Triage scores also may influence the treatment of a trauma case.
Recent findings
The first trauma scores appeared about 40 years ago. Since then, multiple attempts to improve their
performance have been published, or new scores were introduced. However, only few scoring systems
manage to survive. Actual research focuses on further optimization of the available scores as well as the
consequences of coding the injuries.
Summary
Cross-national comparisons evaluating different scores will further help to identify the optimal scoring
system, based on the available information. The inclusion of patients with partially missing data is also an
important task for the future.
Keywords
scoring systems, trauma registries, wounds and injuries
INTRODUCTION
Trauma score systems now have a history of about
40 years. One of the very first systems, the Injury
Severity Score (ISS) was published in the early 1970s.
Most interestingly, this score is still the most
frequently used scoring system in trauma research
today.
Trauma scores are developed to describe the
severity of injuries or the prognosis of a patient
(which correlates with severity) with a single
numerical value. This one-dimensional description
necessarily will have to disregard details, and some-
times very different types of injuries are given the
same score value. The purpose of scores is to improve
and simplify communication about trauma cases.
For example, if you report about a group of patients
you might give a comprehensive list of all injuries,
or you simply give the average injury severity by
presenting a score value. In this sense, score values
become a common language for those involved in
trauma research, care, and management.
Furthermore, severity of injuries is an important
factor, which has to be considered when trauma
outcome is analysed. Survival rates could only be
compared if the severity of injuries were similar.
However, if there are differences regarding injury
severity, then direct comparison of, for example,
mortality rates, is noninformative and misleading.
In these situations it is recommended to consider
risk-adjusted outcome measures, like standardized
mortality ratios, or to compare observed with
expected outcome. This is what trauma registries
all over the world do when they benchmark trauma
outcome across different hospitals.
A third reason for introducing trauma score
systems is to support decision making in the indi-
vidual patient. Such systems, called triage scores, are
simple combinations of findings, as there is usually
no time for sophisticated formulas. Such triage
scores could also be used to define guidelines and
rules for the process of care.
The present review describes, besides some
general remarks about scores, some actual develop-
ments from the literature as well as an overview of
actual use of trauma scores in large registries.
QUALITY CRITERIA FOR SCORES
What is a good trauma score? A trauma severity
score should be able to discriminate between minor
Institute for Research in Operative Medicine (IFOM), University of Witten/
Herdecke, Cologne, Germany
Correspondence to Rolf Lefering, PhD, Institute for Research in Oper-
ative Medicine (IFOM), University of Witten/Herdecke, Ostmerheimer
Str. 200, Cologne 51109, Germany. Tel: +49 221 98957 16; fax: +49
221 98957 30; e-mail: [email protected]
Curr Opin Crit Care 2012, 18:637640
DOI:10.1097/MCC.0b013e3283585356
1070-5295 2012 Wolters Kluwer Health | Lippincott Williams & Wilkins www.co-criticalcare.com
REVI EW
Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
and severe injuries. Severity, however, is usually
associated with mortality and other unfavourable
outcomes. Thus, mortality or survival rates are
frequently used to evaluate trauma scores. The
following three areas have to be considered when
the quality of a score is measured:
(1) discrimination,
(2) precision, and
(3) calibration.
Discrimination is the extent to which a score
gives different values to survivors and nonsurvivors.
The most frequently usedmeasure for discrimination
is the area under the receiver operating characteristic
curve (AUC ROC), a summary measure of using
all possible score values for prediction of survival
(or death). Precision is the extent to which a prog-
nostic score (i.e. a score whichprovides a risk of death
estimate for each case) is able to closely predict the
observed mortality rate. Finally, calibration is the
extent to which the above-mentioned precision
is equally valid for low-risk and high-risk patients.
Calibration is usually measured by the Hosmer
Lemeshow statistic, which evaluates the precision
in different subgroups.
An important point should be mentioned here.
AUC ROC values are very helpful in describing the
discriminative power of different scoring systems
when applied to the same dataset. However, results
from different publications could not directly be
compared because the AUC strongly depends on
the portion of easy to predict cases. In datasets
where 95% of cases survived much better results
are expected than in more severely injured patient
groups.
AVAILABE SCORES
A rough impression about what trauma scores
actually are used in medical literature could be seen
from a Medline search. A quick (but maybe not
complete) search including the last 5 years and
the text word trauma revealed the following
results.
The winner is, not surprisingly, again the ISS
(902 hits). This dinosaur among the available
trauma scores has gained such a large publicity that
it could be considered as the de facto standard,
despite its multiple limitations. The New ISS, or
NISS, a simplification that is easier to determine
and also excludes some of the limitations of the
ISS, managed to get the second place (132 hits).
The Revised Trauma Score (RTS; 97 hits) which
considers the physiological response to trauma,
and the Trauma and Injury Severity Score (TRISS;
91 hits) follow closely. All other scores are men-
tioned rarely, for example the Revised Injury
Severity Classification (RISC, 11 hits) which was
developed and used by the Trauma Register of the
German Trauma Society (TR-DGU) [1].
It should be mentioned that the Glasgow Out-
come Scale (GCS) which is not a classical trauma
score but only regards consciousness has a similar
world-wide attention as the ISS (780 hits).
WHATS NEW?
Several articles actually deal with the basis of the ISS,
the Abbreviated Injury Scale (AIS). The most recent
major revision in 2005 (and a further minor revision
in 2008) has again fired the discussion about the
reliability of severity scoring. The AIS is a codebook
of about 2000 different diagnoses, each given a
severity level of 1 (minor) to 6 (actually untreatable).
Of course, changes in the severity level of injury
codes will influence the derived score values. Inves-
tigations from the Victorian State Trauma Registry
in Australia suggest that using the new AIS 2005/
2008 leads to a lower number of ISS more than
16 cases [2,3
&
,4
&
]. This could in part be caused by
classifying some injuries as less critical for outcome
than before, when compared to decades ago. Each
update of a coding system causes problems, and
these problems need a close investigation, but with-
out any updates we would still use the 74 injuries the
AIS initially was based on. Future updates of the AIS
should better care for the ability to map existing
data into the new codes.
An interesting and extensive analysis aiming to
optimize the TRISS was published by Schluter [5
&
,6
&
]
from New Zealand. The idea was to take the indi-
vidual component variables of the TRISS (age, blood
pressure, respiratory rate, GCS, and mechanism of
injury) and to find a superior categorization than
used in the original scores (RTS, TRISS). Analyses
were performed on two large samples from the
National Trauma Data Base (NTDB) and a data col-
lection from New Zealand. Most interestingly, one
of the provided models is a classical TRISS model
with actual coefficients for each variable. The
KEY POINTS
Despite their limitations, the well known classical
trauma score systems are still most frequently used in
medical literature.
Trauma registries use updated trauma score systems
for benchmarking.
Trauma
638 www.co-criticalcare.com Volume 18 Number 6 December 2012
Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
respiratory rate, which is a kind of a problem
variable with lots of missing values, received a
coefficient close to zero for blunt cases, and only
a very small one for penetrating cases. This confirms
the clinical impression and suggests to disregard the
variables in future predictive models. The RISCscore
developed and used by the TR-DGU has already
eliminated this variable from its prediction model.
Regarding the optimal representation of the
variables in the model the suggested best five
categories representation seems to be a good com-
promise; these categories were superior to the linear
representation for all five variables considered.
Experts from the leading European trauma
registries recently reached a consensus about a core
dataset, the UtsteinTrauma Template, whichshould
be the basis of future cross-country comparisons of
trauma care and a European trauma registry [7]. This
core dataset has now been validated regarding
feasibility and completeness in an international
study. Participants from 24 different institutions
(hospitals and registries) provided sample data. Of
the 36 Utstein variables, 28 had a completeness rate
of more than 80% [8
&
]. Results of this validation
study will be used to refine the definition of the core
dataset.
An interesting article with prediction of compli-
cations rather than mortality was presented by de
Jongh et al. [9
&
]. In a large database, institution-
related and diagnosis-related complications were
identified, and prediction models were derived for
both types of complications. Institution-related
complications were found to be related only to
injury severity whereas diagnosis-related ones also
depend on age and GCS. This approach should be
further investigated, in larger samples and with
further predictors, because mortality alone only
covers an important but small area of trauma care.
Regarding triage scores, a recently published
score for prediction of mass transfusion has now
been validated with TR-DGU data. The Trauma-
Associated Severe Hemorrhage score uses infor-
mation available shortly after a patient has been
admitted to hospital. It predicts the risk that this
patient will need a mass transfusion of 10 or more
units of packed red blood cells. The validation has
shown that the score itself is stable, that is, there
is no need for updating the point values [10
&
].
However, the decreasing incidence of blood trans-
fusion in general, and specifically mass transfusion,
required an adjustment of the formula for deriving
the estimated risk. A comparison with other score
systems for prediction of mass transfusion will soon
be published.
An interesting alternative to the GCS, the
EppendorfCologne Score, is published by
Hoffmann et al. [11
&
,12
&
]. The motor component
of the GCS is already known to be the most import-
ant predictive part. This component is reduced to
four categories (instead of six: normal/specific/non-
specific/none) and was combined with pupil size
(normal/anisocoric/bilateral dilated) and pupil reac-
tivity (brisk/sluggish/fixed). This new score has
superior quality results and seems to be easier to
collect.
WHAT TRAUMA SCORE SYSTEMS DO
LARGE TRAUMA REGISTRIES USE?
Most trauma registries use the TRISS, or a system
similar to the TRISS, for outcome adjustment.
Because the original coefficients of the TRISS are
based on the Major Trauma Outcome Study data,
they represent the expected outcome from trauma
patients treated in the 1980s in the United States.
Therefore, calculating new coefficients from actual
data is the method of choice. This has been done, for
example, by the NTDB in the United States.
The British trauma registry Trauma Audit and
Research Network has developed its own prediction
model, called PS06, PS07, and so on, in which
updated coefficients are calculated every year [13].
The actual model is available fromthe TARNwebsite
(http://www.tarn.ac.uk). This model considers in
principle the same data as the TRISS, however, the
ISS is included as a transformation, and interaction
terms (age sex) are added as well.
The trauma registry of the German Society for
Trauma (DGU, Deutsche Gesellschaft fur Unfall-
chirurgie), TR-DGU, has developed an own predic-
tion system based on data from 1993 to 2000, the
RISC [1]. Ten different variables including first
laboratory results after admissions were considered
to predict outcome. The superior predictive power
of the RISC, which is based on the inclusion of
coagulopathy and acidosis causes, on the contrary,
an increased problemof patients with partially miss-
ing data.
CONCLUSION
Repeatedly, newtrauma score systems are suggested,
or improvements for existing ones. Only few of
these scores/suggestions manage to survive. The
adoption of a system by large trauma registries as
well as the wide-spread publicity of the old scores
will determine the future of trauma scores. It is an
important future task to combine highly predictive
quality with adequate management of missing data.
Acknowledgements
None.
Trauma scoring systems Lefering
1070-5295 2012 Wolters Kluwer Health | Lippincott Williams & Wilkins www.co-criticalcare.com 639
Copyright Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.
Conflicts of interest
The author is statistical consultant for the TR-DGU.
There are no further conflicts of interest.
REFERENCES AND RECOMMENDED
READING
Papers of particular interest, published within the annual period of review, have
been highlighted as:
&
of special interest
&&
of outstanding interest
Additional references related to this topic can also be found in the Current
World Literature section in this issue (p. 723).
1. Lefering R. Development and validation of the Revised Injury Severity Classi-
cation (RISC) score for severely injured patients. Europ J Trauma Emerg
Surg 2009; 35:437447.
2. Cameron PA, Palmer C. Developing consensus on injury coding (editorial).
Injury 2011; 42:1011.
3.
&
Palmer CS, Franklyn M. Assessment of the effects and limitations of the 1998
to 2008 Abbreviated Injury Scale map using a large population-based
dataset. Scand J Trauma Res Emerg Med 2011; 19:1.
This article shows the difculties in using the ofcial mapping of AIS98 codes into
the new AIS05 codes, using a large trauma registry database.
4.
&
Palmer CS, Niggemeyer LN, Charman D. Double coding and mapping using
Abbreviated Injury Scale 1998 and 2005: identifying issues for trauma data.
Injury 2010; 41:948954.
Manual double coding was performed for selected patients, showing the critical
points of the new AIS version. Specically, there seem to be fewer patients with
major trauma (ISS 16) if the AIS05 version was used.
5.
&
Schluter PJ. Trauma and Injury Severity Score (TRISS): is it time for variable
re-categorization and re-characterization? Injury 2011; 42: 8389.
Using a large database, different representations of key predictor variables
(categorical, linear, transforms) were compared, and suggestions for an optimal
representation were made. This could be important for future developments of
trauma score systems.
6.
&
Schluter PJ. The Trauma and Injury Severity Score (TRISS) revised. Injury
2011; 42:9096.
This article makes an attempt to include cases with partially missing data into the
prediction model based on the TRISS score.
7. Ringdahl KG, Coats TJ, Lefering R, et al. The Utstein template for
uniform reporting of data following major trauma: a joint revision by
SCANTEM, TARN, DGU-TR and RITG. Scand J Trauma Res Emerg
Med 2008; 16:7.
8.
&
Ringdal KG, Lossius HM, Jones JM, et al. Collecting core data in severely
injured patients using a consensus trauma template: an international multi-
centre study. Crit Care 2011; 15:R237.
This validation study is a mix of directly collected data and subsets from trauma
registries. It is a pilot study focussing on data quality for a future European trauma
registry.
9.
&
de Jongh MAC, Bosma E, Verhofstad MHJ, Leenen LPH. Prediction
models for complications in trauma patients. Br J Surg 2011; 98:790
796.
This article considers complications rather than mortality as dependent variable,
and thus opens a door to considering alternative outcome measures and the
analysis of factors associated to them.
10.
&
Maegele M, Lefering R, Wafaisade A, et al. Revalidation and update of the
TASH score: a scoring system to predict the probability of massive transfu-
sion as a surrogate for life-threatening haemorrhage after severe injury. Vox
Sanguinis 2011; 100:231238.
The risk of needing mass transfusion could be predicted with considerable validity,
and thus offers its application not only for research projects but also for routine
guidelines for treating coagulopathy.
11.
&
Hoffmann M, Lehmann W, Rueger JM, Lefering R. Introduction of a novel
trauma scale. J Trauma (in press).
Based on the analysis of pupil reactivity and pupil size, an easy to determine score
is presented which is a real challenge for the established GCS.
12.
&
Hoffmann M, Lefering R, Rueger JM, et al. Pupil evaluation in addition to the
Glasgow Coma Scale (GCS) components in traumatic brain injury. Br J Surg
2012; 99 (Suppl.1):122130.
The GCS gives important advice regarding early symptoms of head injury; the
present article now proves that there are additional symptoms with prognostic
relevance which are easy to determine.
13. Bouamra O, Wrotchford A, Hollis S, et al. A new approach to outcome
prediction in trauma: a comparison with the TRISS model. J Trauma 2006;
61:701710.
Trauma
640 www.co-criticalcare.com Volume 18 Number 6 December 2012