Scoring of anatomic injury after trauma: AIS 98 versus AIS 90—do the changes affect overall severity assessment?

Introduction 

To improve quality of care for severely injured trauma patients, continuous analysis of performance data over time and versus other institutions or a recognised standard is essential. Standardised tools are required to enable such comparisons.

Appropriate classification of injuries by type and severity is fundamental for quality assessment analyses. The worldwide recognised tool for anatomic injury scoring following trauma is the Abbreviated Injury Scale (AIS). The most common edition in use is AIS 98.5 The coding dictionary is divided into nine anatomic body regions; in each region every anatomic injury is given a seven-digit injury description indicating body region, type and specification of anatomic structure, level and severity.

AIS classifies individual injuries by body region on a six-point ordinal severity scale ranging from AIS 1 (minor) to AIS 6 (lethal) and is the foundation for Injury Severity Score (ISS),8 a recognised tool for assessment of overall injury severity. The ISS is the sum of the squares of the highest AIS code in each of the three most severely injured ISS body regions.5 The New Injury Severity Score (NISS)16 is the sum of the squares of the three highest AIS codes disregarding body region localisation of the injuries. The probability of survival (Ps) of a trauma patient is most frequently estimated with the Trauma Score—Injury Severity Score (TRISS) method, which since 1989 has been based on classification of overall anatomic severity assessment according to ISS, physiological derangement according to the Revised Trauma Score (RTS), age and mechanism of injury (blunt versus penetrating).10, 11, 13

The 1998 AIS edition, AIS 98,5 is implemented for instance in the Trauma Audit & Research Network (TARN)1 in UK, in South Western Sydney Regional Trauma Registry4 in Australia and in our own registry at Ulleval University Hospital (UUH) in Oslo, while National Trauma Data Bank (NTDB)3 in the US still uses the 1990 version, AIS 90.6 To our knowledge no existing registry has yet implemented the new revised and expanded AIS 2005 edition.7

The use of different versions of the coding dictionary could possibly invalidate comparisons between systems. In AIS 985 numerous injury descriptions were changed, and a major change was introduced in the algorithm for calculating ISS for patients with skin injuries. While it is stated in the preface to the coding manual5 that these changes should not affect overall severity assessment, no quantitative evaluation of the consequences for calculating ISS, NISS or probability of survival according to TRISS has to our knowledge been published.

The aim of the present study was to analyse whether the changes in injury descriptions and/or ISS calculations introduced in AIS 985 caused significant changes in ISS, NISS, Ps or body region specific ISS in our trauma population.

Materials and methods 

Changes in coding dictionary and guidelines 

The number of injury descriptions increased from 1312 in AIS 90 to 1339 in AIS 98. The changes comprised 51 new codes, 26 codes were changed and 24 removed (Table 1). Among the removed codes were injury descriptions for accidental hypothermia, inexpedient for coding of trauma in our country where hypothermia on admission is a serious concern. In addition to these changes in injury descriptions, changes were introduced in the algorithm for calculation of ISS in cases of skin injury. In AIS 90, skin injuries of minor severity (AIS 1 and AIS 2) were marked with an asterisk. The guidelines for ISS calculation were (AIS 90 coding manual,6 page 11):

•If the asterisked injury is the only injury in a body region, locate it under the body region in which it occurs, but assign it to ISS body region, external.

•If the asterisked injury accompanies an injury to a deeper structure, code the asterisked injury under the body region in which it occurs. The injury to the deeper structure in that body region will take precedence over the external injury for ISS.In AIS 98, all asterisks were removed from the text, and the algorithm was changed as follows (AIS 98 coding manual,5 page xix):

•If a skin injury occurs in combination with an underlying injury, both are coded under the specific body region. The skin injury is assigned to external body region for ISS.

Table 1. Modifications from AIS 90 to AIS 98

		N	%
	Injury descriptions AIS 90	1312
	Injury descriptions AIS 98	1339	100.0
	New	51	3.8
	Removed	24	1.8
	Changed	26	1.9
	Six-digit injury identifier	15	1.1
	AIS severity score	11	0.8
	Reduced	10	0.7
	Increased	1	0.1

Exceptions to this rule are open fractures and penetrating injuries where the skin injury is implicit in the code of the deeper injury and should not be coded separately. This change in algorithm for coding of skin injuries in AIS 98 is a simplification with respect to injury coding.

Results 

Of the 1681 patients included in the trauma registry during the study period, 27 were excluded due to lack of anatomic injury suitable for AIS scoring. This included patients with hypothermia as single injury, drowning, strangulation or syncope, all registered in our trauma database due to trauma team activation. Of the remaining 1654 patients constituting the study population, 147 (8.9%) were coded differently using AIS 98 versus AIS 90 due to different injury descriptions. Seventy-one (4.3%) of those patients were hypothermic. 369 patients (22.3%) had a change in ISS value in one or more regions due to the change in the scoring algorithm for skin injuries. Overall ISS score changed in 378 patients (22.9%).

Descriptive statistics for ISS, NISS and ISS body region external are presented in Table 2. The number of patients that underwent coding changes was very small (<10) in all ISS body regions except external, and hence statistical analyses have not been performed for these subgroups (head or neck, face, chest, abdomen or pelvic contents, extremities or pelvic girdle). For ISS body region external the mean difference in ISS score for the total study population was 0.19 (95% CI 0.14–0.24) with an increase for 348 patients (21% of total study population), a decrease for 21 patients (1.2%) and unchanged for 498 patients (30.1%). Mean ISS score difference from AIS 90 to AIS 98 for the 867 patients with ISS region external injuries was 0.37, 95% CI 0.28–0.45, cf. Table 2. The intraclass correlation coefficient (ICC) for the ISS scores in ISS region external (Table 3) suggests substantial agreement between the AIS 90 and AIS 98 scoring systems.

Table 2. Difference between AIS 98 and AIS 90 for individual patients

	Change in ISS value	ISS region “External”		Total ISS		Total NISS
		N	%	N	%	N	%
	−16	1	0.1	1	0.1	1	0.1
	−15	1	0.1	1	0.1
	−12	1	0.1
	−9	2	0.2
	−8	1	0.1	2	0.1	2	0.1
	−7
	−5			8	0.5	9	0.5
	−4	3	0.3	2	0.1	4	0.2
	−3	6	0.7	8	0.5	3	0.2
	−1	6	0.7	5	0.3	6	0.4
	0	498	57.4	1276	77.1	1613	97.5
	1	322	37.1	321	19.4	2	0.1
	3	4	0.5	4	0.2
	4	22	2.5	23	1.4	2	0.1
	5			1	0.1	8	0.5
	7			2	0.1	4	0.2
	Number of patients	867	100.0	1654	100.0	1654	100.0
	Patients with changes	369	42.6	378	22.9	41	2.5

Table 3. Group differences between AIS 98 and AIS 90 for ISS region “External”, total ISS, total NISS and Ps

	Variable	Mean score AIS 90	Mean score AIS 98	Mean difference AIS 98–AIS 90 (with 95% CI)	Intraclass correlation coefficient (ICC) (with 95% CI)
	ISS value in ISS region “External” for patients with injuries in that region (n=867)	1.06	1.42	0.37 (0.28–0.45)	0.820 (0.797–0.841)
	ISS value in ISS region “External” for total study population (n=1654)	0.55	0.74	0.19 (0.14–0.24)	0.842 (0.827–0.855)
	Total ISS	14.54	14.74	0.19 (0.14–0.24)	0.997 (0.9968–0.9974)
	Total NISS	19.07	19.05	−0.02 (−0.06–0.02)	0.999 (0.9988–0.9990)
	Ps	0.8985	0.8983	−0.0003 (−0.0008–0.0002)	0.999 (0.9989–0.9991)

Total ISS changed in 378 patients. This was mainly due to the changes in ISS region external (Table 2). Thus, the mean differences for the total study population were identical for overall ISS and ISS region external (both 0.19, 95% CI 0.14–0.24; Table 3). The ICC value for ISS of 0.997 (95% CI 0.9968–0.9974) indicates excellent agreement between the two editions of the scoring system (Table 3).

NISS changed for only 41 patients (2.5%), with no significant difference in mean NISS (Table 3). The ICC value of 0.999 (Table 3) indicates excellent agreement. Neither was there a significant difference for the mean probability of survival, with ICC 0.999 (Table 3).

Discussion 

For 1654 trauma patients admitted to Ulleval University Hospital during a 2-year period in 2000–2002 there were no important differences in mean Injury Severity Score, New Injury Severity Score or probability of survival with the 1990 versus the 1998 edition of the Abbreviated Injury Scale.

This is important, as it potentially enables comparison of survival for registries which use AIS 98, like Ulleval, the Trauma Audit & Research Network (TARN)1 in the UK and the South Western Sydney Regional Trauma Registry4 in Australia, with data from registries which still employ AIS 90, e.g. the US National Trauma Data Bank.

The intraclass correlation coefficients calculated also indicate excellent agreement between the two editions of the coding dictionary, as all ICC values were above 0.99 except for ISS region external where it was 0.82. This was to be expected, as the most dramatic changes in the AIS coding system from the 1990 to 1998 edition were the removal of hypothermia as an injury description and the changes for the coding of skin injuries in the external region. These changes might therefore be important for studies and registries with focus on hypothermia, burn and other skin injuries.

The removal of injury descriptions for accidental hypothermia is potentially problematic for evaluating systems or hospitals with a large percentage of hypothermic patients. If the occurrence of accidental hypothermia is very different in study and reference populations, this might markedly affect the results using AIS 98. Although the overall effect was small in our material, 26 of the 71 patients in our study with significant accidental hypothermia on admission (<35°C) had lower ISS with AIS 98 than AIS 90. Hypothermia is a risk factor after injury, and used as a prognostication tool for an individual patient, this becomes particularly important. Six of the patients with accidental hypothermia can be identified in Table 2, region ISS external, displaying a reduction in ISS of between 8 and 16 with AIS 98. Their corresponding increase in Ps was between 5% and 27%. We therefore strongly support the re-implementation of descriptions for accidental hypothermia in the recently published 2005 edition of the coding dictionary. However, the actual consequences of this change should be evaluated for registries currently using AIS 98.

Due to the changes in coding of ISS external, total ISS changed in 42.6% of those with an injury in the external region—corresponding to 22.3% of the total patient population. One consequence of the AIS 98 revision, probably not foreseen, is that patients with an isolated open fracture may actually end up with a lower ISS than with a similar closed fracture with minimal skin contusion: In AIS 98, a patient with an open fracture of the femoral shaft is assigned an AIS severity code of 3 and consequently ISS 9, as the skin injury is part of the code for the deeper structure (coding rule 8, page xvi, AIS 985). However, a patient with a closed fracture and smaller skin contusion receives ISS 10; 9 for the femoral shaft fracture and 1 for skin contusion/laceration (ISS body region external).

For burn injuries, it is important to recognise that a new code for minor third degree burn injuries (<100cm2, face <25cm2) was added in AIS 98, reducing the region specific ISS from four to one for this injury.

The mean difference in Ps between AIS 90 and AIS 98 in the study population was virtually zero (Table 3) due to ISS being the only element that had changed in the Ps calculation and, in fact, its change was minimal between the two coding schemes.

Osler et al.16 have launched NISS as an improved way to calculate overall anatomic injury based on AIS coding. Their study indicated improved survival prediction over ISS. Recently, Lavoie et al.14 have confirmed and extended this finding. They conclude that NISS has its particular strength versus ISS in populations with high injury severity, such as patients with head and neck injuries. This study is of specific interest for populations like ours, with a large portion of patients with head injuries. Others have been skeptic of Osler's conclusion; Tay et al.19 and Meredith et al.15 reported similar predictive values and misclassification rates for ISS and NISS. However, in the latter study duplicate coding (e.g., bilateral femoral fracture), which is one of the surmised strengths of NISS,15 was not allowed.

In our present study, the change in AIS protocol affected NISS in only 41 patients (2.5%), with virtually identical mean values and ICC of 0.999. This robustness to changes in AIS protocol might be a further argument for implementing NISS in survival prediction models.

Conclusion 

We found excellent agreement between the two different editions of the AIS coding dictionaries, AIS 90 and AIS 98. There were no differences in mean NISS or probability of survival (Ps) estimates. The differences in mean total ISS and external body region specific ISS between the two editions are so minimal that we do not consider them to be of clinical relevance when comparing trauma systems that use different AIS versions. A possible exception might be systems with an exceptionally high incidence of accidental hypothermia or skin/burn injuries. We strongly support the re-implementation of injury descriptions for accidental hypothermia in the recently published AIS 2005.