Page 100 - Journal of Special Operations Medicine - Summer 2015

Table 5 Summary Statistics for Injury Severity Scores by Reviewer Group* major blood loss). No investiga-
tion of the ability of surgeons or
ICC
Score Reviewer Mean Median SD Weighted κ (95% CI) TNCs to estimate and attribute
traumatic blood loss was identi-
Surgeon 3.53 4 1.39 0.89
MaxAIS(M) 0.88 (p < .001) fied from the literature.
TNC 3.53 4 1.39 (0.78–0.94)
Surgeon 22.4 17 18.5 0.94 When comparing anatomic re-
ISS 0.63 (p < .001)
TNC 22.5 20 18.2 (0.88–0.97) gions, the study investigators
tended to miss injuries to the
Surgeon 29.2 25.5 21.7 0.41 0.84
NISS head and face. This may relate
TNC 30.6 27.0 20.9 (p = .020) (0.70–0.92) to surgeons scoring shortly af-
Note: *Interrater agreement and reliability for each score given by weighted κ and ICC, respec- ter initial damage control op-
tively. CI, confidence interval; ICC, intraclass correlation coefficient; ISS, Injury Severity Score; erations, where attention had
MaxAIS(M), maximum Abbreviated Injury Scale (Military version); NISS, New Injury Severity
Score; SD, standard deviation; TNC, trauma nurse coordinator. been focused on torso and limb
injuries with only a provisional
Figure 4 Comparison of Injury Severity Score (ISS) between radiographic diagnosis of head and facial injuries avail-
Joint Theatre Trauma Registry (JTTR) and investigators. able. In contrast, the definitive radiology report was
available for JTTR coding. Surgeons tended to record
more torso, limb, and external injuries than TNCs. This
may be an effect of direct involvement with casualties
(and would be consistent with previous reports that
surgeon involvement improves data quality), whereas
27
conventional coding relies on the level of detail recorded
in written records and the ability to match that detail to
the specific descriptors within the relevant AIS diction-
ary. Such discrepancies between individuals when calcu-
lating injury severity scores are not unique; variation in
calculation of ISS of up to 80% from the mean has been
28
noted. In a large study from Queensland, Australia, six
raters independently coding 120 sets of notes achieved
17
almost perfect agreement for ISS. However, that vali-
Figure 5 Comparison of New Injury Severity Score (NISS) dation exercise was based on repeated examination of
between Joint Theatre Trauma Registry (JTTR) and standardized data rather than the comparison of visual-
investigators. ized injuries against written records.

Despite the differences in this study between investiga-
tor and JTTR AIS(M) scores, there were no statistically
significant differences between the groups in the derived
ISS and NISS for these casualties. This may suggest that
discrepancies in AIS(M) do not affect the predictive and
prognostic function of the JTTR data. This is consistent
with a previous report of “almost perfect” interrater
reliability despite only 36% agreement regarding AIS.
17
However, this current study is small; statistically signifi-
cant differences might be identified if a larger cohort of
patients were examined as part of a study design that
removed the potential for confounding discussed in the
limitations section.
comment on the relative abilities of TNCs and surgeons
to score injuries accurately. Rather, our results likely rep- Our suggestion that the differences in assigned scores
resent a difference in the perception of an injury visualized arise from different exposures to the casualty is consis-
at the time of surgery compared with its written record tent with previous findings that involving trauma sur-
in medical documentation. This may lead to differences geons in coding at a Level I trauma center resulted in
in both the description of an injury and the interpreta- coding amendments in 5.2% of cases scored by non-
tion of secondary features required for accurate coding surgeon coders alone, with 71% of these amendments
27
(e.g., the identification of specific injuries responsible for being an upward revision of ISS. If accurate scoring

90 Journal of Special Operations Medicine Volume 15, Edition 2/Summer 2015