Injury (1992)
23,
(7), 489-493
Printed in Greut Britain
489
Correlation between injury severity scores and subjective ratings of injury severity: a basis for trauma audit B. T. Collopy, B. R. Tulloh, G. C. Rennie, R. L. W. Fink, J. H. Rush and G. W. Trinca University
Department
of Surgery, St Vincent’s Hospital, Melbourne,
A retrospective review of 1900 road accident victims attending the emergency departments of two Melbourne hospitals was undertaken fo identify lnjuy Severi@ Score levels which could distinguish between nlinor, moderate, severe and critical inju y. Injuries scoring 1SS 6 or below were designated ‘minor’ because they were associated with a low risk of requiring admission to hospital. Case notes of patients scoring above ISS 6 were then reviewed by a panel of clinicians, who independently rafed each pafienb’s overall injury severi/y as moderate, severe or c&al according to what was recorded in the notes and their ‘clinical’judgement. I.% values were compared with clinicians’ ratings. Measures of each clinician’s individual rating conk&my, and correlation between pairs of clinicians wifh respect to inter-rater consistency, were made. By combining datafrom both hospituls it emerged that ‘moderate’ injury corresponded to ISS 8-13, ‘severe’ to IS!? 74-20 and ‘critical’ to 15521 and above. These 1% breakpoints will be useful in selecting groups of injured patients for future trauma audit studies.
Introduction The Injury Severity Score (ES) was developed in 1974 from the Abbreviated Injury Scale (AIS) to quantify multisystem injury and has gained worldwide acceptance as the foremost anatomic severity scale available (Baker et al., 1974; Committee on Injury Scaling, 1990). Derived by adding together the squares of the three highest AIS scores from different regions of the body, it is of value in both clinical and research settings. It has considerable potential in the evaluation of quality of care, for which several applications are available. First, it aids in comparisons between different trauma populations by controlling for injury severity case mix. Second, it can be used to monitor diagnostic accuracy in audits of early trauma management. Differences have already been noted between ISS values calculated on the basis of information available in the emergency department and those calculated on the basis of confirmed final diagnoses (Mackenzie et al., 1985; Civil and Schwab, 1989; Smejkal et al., 1989). These differences have been utilized to assess diagnostic accuracy in at least one study (Tulloh, 1988). Third, cut-off values or breakpoints may be used to identify particular groups for study. Several researchers have used ISS > 16 when selecting patients with ‘severe’ injury for study (Dearden and Rutherford, 1985; Civil et al., 1987; (Q 1992 Butterworth-Heinemann 0020-1383/92/070489-04
Ltd
Victoria, Australia
Civil and Schwab, 1989), but this discriminatory value has been selected arbitrarily. In this study we aimed to develop a method for categorizing patients’ injuries as minor, moderate, severe or critical according to ISS values. The discriminatory values (breakpoints) of the ISS thus derived would then be available for trauma audit and peer review studies, identifying groups of patients with ‘clinically’ similar injury severity.
Methods A retrospective review was conducted of road accident victims attending the emergency department of an innercity teaching hospital and a busy suburban general hospital in Melbourne over a 12-month period. At each institution the patients’ histories were examined and injuries were recorded. An ISS was calculated in each case. It was also noted if the patient had been admitted to hospital. The first step was to separate the minor injuries, which formed the vast majority of cases. Table1 shows the distribution of ISS in the study population and indicates the cases which were admitted to hospital. The probability of being admitted rose sharply above SO per cent with ISS > 6. Thus the first cut-off was identified: scores of 6 or below were classified as ‘minor’ as they were unlikely to require admission to hospital. Note that there were very few patients with ISS 3, for which a patient must sustain injuries scoring AIS 1 in three separate regions of the body - an unusual occurrence. Note also that ISS 7 does not appear in the table. Such a score is impossible to obtain by the sum of squares as required for ISS calculation: similarly, a score of 15 does not exist. The second part of the study was to select the records of patients with ISS > 6 for review by three clinicians at the city hospital and by two clinicians at the suburban hospital. In the former instance the clinicians comprised an orthopaedic surgeon (OS), a general surgeon (GS,) and a resident medical officer (RMO) in his third postgraduate year, while in the case of the suburban hospital the two clinicians comprised another general surgeon (GS,) and the same RMO. These doctors, each unaware of the ISS for the patients involved, were asked to rate independently the overall injury severity for each patient as ‘moderate’, ‘severe’ or
Injury: the British Journal of Accident Surgery (1992)Vol. 23/No. 7
490
TableI.
The vast majority of patients suffered minor injury (ISS < 6). The likelihood that a patient was admitted to hospital rose sharply as the ISS became > 6
ISS
Total number of patients
Number admitted
13 1065 1 13 283 2 3 3 22 4 106 38 157 5 30 95 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .._....................................... 11 8 27 ;: 9 37 46 10 36 40 11-14 52 52 >14 Total
1900
279
Table II. Maximum indices of consistency for each clinician,and the consensus figure calculated by combining data from both hospitals. The optimum breakpoints between moderate, severe and critical categories are also shown
Percentage admitted
Optimum breakpoints between
1 5 -
Hospital
21 24 32 73 71 80 90 100
Clinician
City
Both
Moderate and severe
Severe and critical
0.65 0.72 0.71
14 14 14
29 21 20
0.74 0.57
14 10
26 29
0.67
14
21
Consistency
Consensus
15
‘critical’ on the basis of the information available in the records and their clinical judgemenf. They were not given any further instruction as to how their assessments should be made, nor given any opportunity to confer with one another. The clinicians’ ratings were then compared with each patient’s corresponding ISS. The ES values of all patients were listed in rank order and the clinicians’ ratings were recorded alongside these. It fhen was possible to determine two breakpoints in the listings for each clinician - one between ‘moderate’ and ‘severe’ and one between ‘severe’ and ‘critical’ - positioned to give the maximum consistency of rating for each clinician, i.e. the maximum number of equal ratings above and below each breakpoint. A numerical expression of consistency could then be determined. This was defined as the number of consistent (equal) ratings between each breakpoint divided by the total number of cases rated. It was also possible to determine the extent to which a particular pair of clinicians agreed in their rating of a given patient. This was assessed using Pearson’s correlation coeficient, r. Finally, the discriminatory power of each clinician - i.e. the ability of each clinician to distinguish between moderate, severe and critical injuries in relation to the more objective ISS - was assessed using the Kruskal-Wallis statistic (Kruskal and Wallis, 1952). This is maximized when the numbers of patients in each category of severity are equal and the ratings are 100 per cent consistent; it is zero when the ratings are each uniformly spread throughout the frequency distribution of the scores. The Kruskal-Wallis statistic, H, for each clinician was interpreted as a measure of the relative discriminatory power of each clinician. Because they involved different populations of patients, rated by dtierent clinicians, the results from each hospital are reported separately. However, data from both centres were combined to provide consensus breakpoints for use at any hospital.
Results A total of 1900 road accident victims presented to the two hospitals in the study period. Of these, 820 were at the city hospital and 1080 at the suburban hospital. The majority of patients had suffered only minor injury, as only 119 at the city hospital (15 per cent) and 84 at the surburban hospital (8 per cent) had an ISS greater than 6.Owing fo variation in
Table III. Rating agreement (expressed as Pearson’s correlation coefficients) between pairs of clinicians at each hospital. Agreement was greatest between the two specialists at the city hospital Correlation coefficient Hospital
Comparison
0)
City
RMO vs. GS, RMO vs. OS OS vs. GS,
0.62 0.64 0.78
Suburban
RMO
0.48
vs. GS.
availability of both individual records and clinicians throughout the study period, only 98 out of the 119 cases were suitably scored at the city hospital and 61 out of the 84 cases af the suburban hospital. This left a total of 159 patients to have their injuries categorized as moderate, severe or critical. Their ISS values ranged from 6 to 50. The individual rating consistencies for the RMO, general surgeon and orthopaedic surgeon at the city hospital and for the RMO and general surgeon at the suburban hospital are set out in Tub/eIZ,together with the corresponding breakpoints for each clinician and consensus breakpoints for both hospitals combined. At the city hospital, both of the two specialists were more consistent than the RMO. Although this was as expected on the basis of experience, the difference was not significant (exact test binomial distribution). The converse was found at the suburban hospital, where the RMO was more consistent than the GB,. Here the difference was more marked but it also failed to reach significance at the 5 per cent level. The levels of agreement between pairs of the three clinicians at the city hospital and between the two clinicians at the suburban hospital, expressed as Pearson’s correlation coefficients, are set out in Ttzble111.Agreement was greafesf between the two specialists at the city hospital (Pearson’s correlation coefficient r = 0.78).The RMO and the GS, showed poor concordance (r = 0.48), which is not surprising as they differed in their individual consistencies. The optimum breakpoints between moderate/severe and severe/critical for each clinician, positioned to provide maximum individual consistency, are shown in TubleII. Also shown are the consensus breakpoints for both hospitals combined. ISS 14 emerged repeatedly as the breakpoint between moderate and severe injury for all clinicians apart from the GS, for whom a breakpoint at ISS 10 gave
Collopy et al.: IS and ratings of injury severity
marginally higher index of consistency (0.57) than one at ISS 14 (0.56). There was less agreement about the breakpoint between severe and critical injury. At the city hospital, the RMO’s breakpoint (ISS 29) was substantially higher than that ‘for either specialist (ISS 20 and 21) which suggests that the less experienced RMO required a patient to have sustained more serious injury before rating it as ‘critical’. However, at the suburban hospital, the breakpoint of the RMO (ISS 26) was lower than that of the GS, (ISS 29). This disparity may have been partly due to the lower number of patients, and fewer with major injury, at the suburban hospital than at the city hospital. When data from both hospitals were combined, the consensus severe/critical breakpoint fell between scores of 20 and 21. On the basis of data from both hospitals, it was concluded that injuries scoring between 6 and 14 should be classified as ‘moderate’, those between 14 and 20 as ‘severe’ and those scoring 21 and above should be classified as ‘critical’.
Discussion It is useful to have breakpoints between levels of injury severity to aid in the selection of groups of patients for study. Dearden and Rutherford (1985) and Civil et al. (198i’), Civil and Schwab (1989) used ISS 16 as a cut-off to select patients with ‘severe’ injuries for audits of medical care, but this value was selected arbitrarily. The findings of the present study suggest that patients with ISS 14 or more be used when selecting those with ‘severe’ injury. One difficulty with trying to quantify accurately such inaccurate and subjective terms as ‘moderate’, ‘severe’ and ‘critical’ was that there were many patients in the middle range of injury severity (ISS 9-17). These patients were classified diversely between the three categories depending on physiological criteria such as circulatory status or level of consciousness, or according to complications or injury outcomes to which the reviewing clinicians had access in each patient’s medical record. Examples of outcomes, real or anticipated, which might affect the clinicians’ ratings include the possible development of pneumonia following rib fractures, or the onset of paraplegia following fractures to the spine. None of these physiological criteria or clinical outcomes could be measured by the ‘anatomical’ ISS. The moderate-severe breakpoint fell in the midst of a group of patients with ISS 14, but it was decided to designate all injuries scoring ISS 14 as ‘severe’ rather than ‘moderate’ because of the presence of possible outcomes or complications as mentioned above. In other words, we elected to err on the side of overrating the injury severity. The sever-ritical breakpoint was easier to deal with, however, as it lay between scores of 20 and 21. Injuries scoring 21 and above were classified as ‘critical’ while injuries scoring from 14 to 20 were classified as ‘severe’. Overall injury severity and mortality were very low in this study population but comparable to those from other Australasian series (Civil et al., 1987; Deane et al., 1988, 1990). In Melbourne at the time in which this review was set, there was no stratification of hospitals according to the trauma service they provided. At that time, the only triage guidelines for prehospital personnel were to transport injured patients to the nearest public hospital (Ambulance Service Melbourne, 1982). It is therefore likely that the sample of patients reviewed in the present paper was indeed representative of the injury spectrum of the local road accident population. It has been the experience of the
491
clinicians involved in this study that patients with ISS > 20 do not present frequently, and that a great number of minor injuries is seen for every case of major trauma. This contrasts with the experience of dedicated trauma centres in the USA and Europe where elaborate triage systems endeavour to ensure that all patients sent to such a centre have significant injury. The ISS breakpoints identified in this study, which related to the local experience and local resources available for management, are therefore unlikely to be applicable in these overseas centres. It was noted at the city hospital that the junior doctor tended to underrate injury severity in patients with more serious injuries and had the lowest individual consistency of rating. However, the converse was found at the suburban hospital where the general surgeon was significantly less consistent than the RMO and appeared to be less able to discriminate between different categories of severity with the crude ratings used. This disparity in clinicians’ consistencies reflects the subjectivity inherent in injury assessment and the inability of the ISS to account for physiological changes in the injured patient.
Conclusions This study has demonstrated that particular ISS levels can be established to classify the severity of injury of patients admitted to hospital into moderate, severe and critical groups which correspond well to subjective ‘clinical’ ratings. The particular ISS levels in the population studied were scores of 6-13,14-20 and 21 and above, respectively. Thus, the ISS has been confirmed as an appropriate tool for peer review in injury management and these ISS ranges can be used to select groups of patients for future trauma audits.
Acknowledgements The authors are grateful to Misses M. Dalton, N. Burford, L. Dix and W. Anderson for their assistance in the retrieval of information from case records.
References Ambulance Service Melbourne (1982) Gnr& of i%+roPe/i~~rrr Hes@rls. Staff Notice 8215, June 1982. Baker S. P., O’Neil B., Haddon W. et al. (1974) The ISS: A method for describing patients with multiple injuries and evaluating emergency care 1. Traitrun 14, 187. Civil 1. D., Ross S. E. and Schwab C. W. (1987) Major trauma in an urban New Zealand setting: resource requirements. Au.d AJ.Z. J Surg. 57, 543. Civil I. D. and Schwab C. W. (1989) Clinical prospective injury scoring: When is it accurate? J Tmrlrtinu29, 613. Committee on Injury Scaling (1990) The Abbreviated Injury Scale, 1990 Revision. Association for the Advancement of Automotive Medicine, Des Plaines, IL 60018, USA. Deane S. A., Gaudry P. L., Woods P. et al. (1988) The management of injuries -a review of deaths in hospital. Awt N. Z. J Strrg.58, 463. Deane S. A., Gaudry P. L., Woods W. P. D. et al. (1990) Interhospital transfer in the management of acute trauma. Aus/ N. Z. J Sug. 60, 441. Dearden C. H. and Rutherford W. H. (1985) The resuscitation of the injured in the accident and emergency department - a medical audit. Iniun/ 16, 249.
Injury: the British Journal of Accident Surgery (1992) Vol. 23/N0. 7
492
Kruskal W. H. and Walks W. A. (1952) Use of ranks in one-criterion variance analysis. J. Amer. Stdist. Assoc. 4, 583. Mackenzie E. J., Shapiro S. and Eastham J. N. (1985) Rating AIS severity using emergency department sheet vs. inpatient charts. I. Truuma 25, 984. Smejkal R., Civil I., Unkle D. et al. (1989) Injury severity scoring: a comparison of early clinical versus discharge diagnosis. Ad. Anal. Prev. 21, 386.
Tulloh B. R. (1988) The use of injury severity scales in the early management of road trauma. M.S. Thesis, University of Melbourne, Australia. Paper accepted
26 February
1992.
Requests for reprintsshouti be aa%ssed fo: Mr B. R. Tulloh, Echuca District Hospital, Francis Street, Echuca, Victoria 3564, Australia.
Book review Blunt Multiple Trauma: Comprehensive Pathophysiology and Care. J. R. Border, M. Allgiiwer, S. T. Hansen and T. P. Ruedi. Marcel Dekker, Inc, New York. 82478172 4 $234.00,
1990, 1024 pp.
This large and challenging book has come out of Swiss-American Co-operation fostered by the A0 Group. It is designed to consider tissue response to trauma in general rather than to act as a specific ‘cook book in the fashion of the A0 manual. There is, accordingly, heavy emphasis on the problems of devitalised tissue, sepsis, oxygen transport, and pulmonary and gut function. The book is divided into nine sections commencing with an excellent historical review detailing the evolution of trauma management and the role played by the control of sepsis and resuscitation. The diminishing treatment-free interval is documented along with the diminishing use of ARDS and fat embolism syndrome. The second section details basic cellular pathology and the response of the different body systems to injury. The third section deals with the epidemiology of trauma and injury severity scoring. There then follows a fourth section on resuscitation and a
fifth section on evaluation. Section 6 is devoted to emergency surgery and this is perhaps the least appropriate part of this volume. There are chapters within this that are highly irrelevant such as multiple team operating and intra-operative monitoring while other sections such as re-plantation, brachial plexus injuries and free tissue transfers are all far too isolated topics to be valuable here, particularly when they are covered only in outline and without indications as to overall results. Indications and end results need to feature more prominently if these chapters are to be justified. Overall, however this is an excellent and authoritative volume. The authors are to be commended on pulling together the essence of basic science and the practical clinical applications derived from it. References are voluminous and available by chapter. It is essential reading for those involved in the treatment of major trauma and will undoubtedly help inspire new thought and better clinical standards in the treatment of patients with major injuries.
J. L. Plewes, FRCS
23,
(7), 489-493
Printed in Greut Britain
489
Correlation between injury severity scores and subjective ratings of injury severity: a basis for trauma audit B. T. Collopy, B. R. Tulloh, G. C. Rennie, R. L. W. Fink, J. H. Rush and G. W. Trinca University
Department
of Surgery, St Vincent’s Hospital, Melbourne,
A retrospective review of 1900 road accident victims attending the emergency departments of two Melbourne hospitals was undertaken fo identify lnjuy Severi@ Score levels which could distinguish between nlinor, moderate, severe and critical inju y. Injuries scoring 1SS 6 or below were designated ‘minor’ because they were associated with a low risk of requiring admission to hospital. Case notes of patients scoring above ISS 6 were then reviewed by a panel of clinicians, who independently rafed each pafienb’s overall injury severi/y as moderate, severe or c&al according to what was recorded in the notes and their ‘clinical’judgement. I.% values were compared with clinicians’ ratings. Measures of each clinician’s individual rating conk&my, and correlation between pairs of clinicians wifh respect to inter-rater consistency, were made. By combining datafrom both hospituls it emerged that ‘moderate’ injury corresponded to ISS 8-13, ‘severe’ to IS!? 74-20 and ‘critical’ to 15521 and above. These 1% breakpoints will be useful in selecting groups of injured patients for future trauma audit studies.
Introduction The Injury Severity Score (ES) was developed in 1974 from the Abbreviated Injury Scale (AIS) to quantify multisystem injury and has gained worldwide acceptance as the foremost anatomic severity scale available (Baker et al., 1974; Committee on Injury Scaling, 1990). Derived by adding together the squares of the three highest AIS scores from different regions of the body, it is of value in both clinical and research settings. It has considerable potential in the evaluation of quality of care, for which several applications are available. First, it aids in comparisons between different trauma populations by controlling for injury severity case mix. Second, it can be used to monitor diagnostic accuracy in audits of early trauma management. Differences have already been noted between ISS values calculated on the basis of information available in the emergency department and those calculated on the basis of confirmed final diagnoses (Mackenzie et al., 1985; Civil and Schwab, 1989; Smejkal et al., 1989). These differences have been utilized to assess diagnostic accuracy in at least one study (Tulloh, 1988). Third, cut-off values or breakpoints may be used to identify particular groups for study. Several researchers have used ISS > 16 when selecting patients with ‘severe’ injury for study (Dearden and Rutherford, 1985; Civil et al., 1987; (Q 1992 Butterworth-Heinemann 0020-1383/92/070489-04
Ltd
Victoria, Australia
Civil and Schwab, 1989), but this discriminatory value has been selected arbitrarily. In this study we aimed to develop a method for categorizing patients’ injuries as minor, moderate, severe or critical according to ISS values. The discriminatory values (breakpoints) of the ISS thus derived would then be available for trauma audit and peer review studies, identifying groups of patients with ‘clinically’ similar injury severity.
Methods A retrospective review was conducted of road accident victims attending the emergency department of an innercity teaching hospital and a busy suburban general hospital in Melbourne over a 12-month period. At each institution the patients’ histories were examined and injuries were recorded. An ISS was calculated in each case. It was also noted if the patient had been admitted to hospital. The first step was to separate the minor injuries, which formed the vast majority of cases. Table1 shows the distribution of ISS in the study population and indicates the cases which were admitted to hospital. The probability of being admitted rose sharply above SO per cent with ISS > 6. Thus the first cut-off was identified: scores of 6 or below were classified as ‘minor’ as they were unlikely to require admission to hospital. Note that there were very few patients with ISS 3, for which a patient must sustain injuries scoring AIS 1 in three separate regions of the body - an unusual occurrence. Note also that ISS 7 does not appear in the table. Such a score is impossible to obtain by the sum of squares as required for ISS calculation: similarly, a score of 15 does not exist. The second part of the study was to select the records of patients with ISS > 6 for review by three clinicians at the city hospital and by two clinicians at the suburban hospital. In the former instance the clinicians comprised an orthopaedic surgeon (OS), a general surgeon (GS,) and a resident medical officer (RMO) in his third postgraduate year, while in the case of the suburban hospital the two clinicians comprised another general surgeon (GS,) and the same RMO. These doctors, each unaware of the ISS for the patients involved, were asked to rate independently the overall injury severity for each patient as ‘moderate’, ‘severe’ or
Injury: the British Journal of Accident Surgery (1992)Vol. 23/No. 7
490
TableI.
The vast majority of patients suffered minor injury (ISS < 6). The likelihood that a patient was admitted to hospital rose sharply as the ISS became > 6
ISS
Total number of patients
Number admitted
13 1065 1 13 283 2 3 3 22 4 106 38 157 5 30 95 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . .._....................................... 11 8 27 ;: 9 37 46 10 36 40 11-14 52 52 >14 Total
1900
279
Table II. Maximum indices of consistency for each clinician,and the consensus figure calculated by combining data from both hospitals. The optimum breakpoints between moderate, severe and critical categories are also shown
Percentage admitted
Optimum breakpoints between
1 5 -
Hospital
21 24 32 73 71 80 90 100
Clinician
City
Both
Moderate and severe
Severe and critical
0.65 0.72 0.71
14 14 14
29 21 20
0.74 0.57
14 10
26 29
0.67
14
21
Consistency
Consensus
15
‘critical’ on the basis of the information available in the records and their clinical judgemenf. They were not given any further instruction as to how their assessments should be made, nor given any opportunity to confer with one another. The clinicians’ ratings were then compared with each patient’s corresponding ISS. The ES values of all patients were listed in rank order and the clinicians’ ratings were recorded alongside these. It fhen was possible to determine two breakpoints in the listings for each clinician - one between ‘moderate’ and ‘severe’ and one between ‘severe’ and ‘critical’ - positioned to give the maximum consistency of rating for each clinician, i.e. the maximum number of equal ratings above and below each breakpoint. A numerical expression of consistency could then be determined. This was defined as the number of consistent (equal) ratings between each breakpoint divided by the total number of cases rated. It was also possible to determine the extent to which a particular pair of clinicians agreed in their rating of a given patient. This was assessed using Pearson’s correlation coeficient, r. Finally, the discriminatory power of each clinician - i.e. the ability of each clinician to distinguish between moderate, severe and critical injuries in relation to the more objective ISS - was assessed using the Kruskal-Wallis statistic (Kruskal and Wallis, 1952). This is maximized when the numbers of patients in each category of severity are equal and the ratings are 100 per cent consistent; it is zero when the ratings are each uniformly spread throughout the frequency distribution of the scores. The Kruskal-Wallis statistic, H, for each clinician was interpreted as a measure of the relative discriminatory power of each clinician. Because they involved different populations of patients, rated by dtierent clinicians, the results from each hospital are reported separately. However, data from both centres were combined to provide consensus breakpoints for use at any hospital.
Results A total of 1900 road accident victims presented to the two hospitals in the study period. Of these, 820 were at the city hospital and 1080 at the suburban hospital. The majority of patients had suffered only minor injury, as only 119 at the city hospital (15 per cent) and 84 at the surburban hospital (8 per cent) had an ISS greater than 6.Owing fo variation in
Table III. Rating agreement (expressed as Pearson’s correlation coefficients) between pairs of clinicians at each hospital. Agreement was greatest between the two specialists at the city hospital Correlation coefficient Hospital
Comparison
0)
City
RMO vs. GS, RMO vs. OS OS vs. GS,
0.62 0.64 0.78
Suburban
RMO
0.48
vs. GS.
availability of both individual records and clinicians throughout the study period, only 98 out of the 119 cases were suitably scored at the city hospital and 61 out of the 84 cases af the suburban hospital. This left a total of 159 patients to have their injuries categorized as moderate, severe or critical. Their ISS values ranged from 6 to 50. The individual rating consistencies for the RMO, general surgeon and orthopaedic surgeon at the city hospital and for the RMO and general surgeon at the suburban hospital are set out in Tub/eIZ,together with the corresponding breakpoints for each clinician and consensus breakpoints for both hospitals combined. At the city hospital, both of the two specialists were more consistent than the RMO. Although this was as expected on the basis of experience, the difference was not significant (exact test binomial distribution). The converse was found at the suburban hospital, where the RMO was more consistent than the GB,. Here the difference was more marked but it also failed to reach significance at the 5 per cent level. The levels of agreement between pairs of the three clinicians at the city hospital and between the two clinicians at the suburban hospital, expressed as Pearson’s correlation coefficients, are set out in Ttzble111.Agreement was greafesf between the two specialists at the city hospital (Pearson’s correlation coefficient r = 0.78).The RMO and the GS, showed poor concordance (r = 0.48), which is not surprising as they differed in their individual consistencies. The optimum breakpoints between moderate/severe and severe/critical for each clinician, positioned to provide maximum individual consistency, are shown in TubleII. Also shown are the consensus breakpoints for both hospitals combined. ISS 14 emerged repeatedly as the breakpoint between moderate and severe injury for all clinicians apart from the GS, for whom a breakpoint at ISS 10 gave
Collopy et al.: IS and ratings of injury severity
marginally higher index of consistency (0.57) than one at ISS 14 (0.56). There was less agreement about the breakpoint between severe and critical injury. At the city hospital, the RMO’s breakpoint (ISS 29) was substantially higher than that ‘for either specialist (ISS 20 and 21) which suggests that the less experienced RMO required a patient to have sustained more serious injury before rating it as ‘critical’. However, at the suburban hospital, the breakpoint of the RMO (ISS 26) was lower than that of the GS, (ISS 29). This disparity may have been partly due to the lower number of patients, and fewer with major injury, at the suburban hospital than at the city hospital. When data from both hospitals were combined, the consensus severe/critical breakpoint fell between scores of 20 and 21. On the basis of data from both hospitals, it was concluded that injuries scoring between 6 and 14 should be classified as ‘moderate’, those between 14 and 20 as ‘severe’ and those scoring 21 and above should be classified as ‘critical’.
Discussion It is useful to have breakpoints between levels of injury severity to aid in the selection of groups of patients for study. Dearden and Rutherford (1985) and Civil et al. (198i’), Civil and Schwab (1989) used ISS 16 as a cut-off to select patients with ‘severe’ injuries for audits of medical care, but this value was selected arbitrarily. The findings of the present study suggest that patients with ISS 14 or more be used when selecting those with ‘severe’ injury. One difficulty with trying to quantify accurately such inaccurate and subjective terms as ‘moderate’, ‘severe’ and ‘critical’ was that there were many patients in the middle range of injury severity (ISS 9-17). These patients were classified diversely between the three categories depending on physiological criteria such as circulatory status or level of consciousness, or according to complications or injury outcomes to which the reviewing clinicians had access in each patient’s medical record. Examples of outcomes, real or anticipated, which might affect the clinicians’ ratings include the possible development of pneumonia following rib fractures, or the onset of paraplegia following fractures to the spine. None of these physiological criteria or clinical outcomes could be measured by the ‘anatomical’ ISS. The moderate-severe breakpoint fell in the midst of a group of patients with ISS 14, but it was decided to designate all injuries scoring ISS 14 as ‘severe’ rather than ‘moderate’ because of the presence of possible outcomes or complications as mentioned above. In other words, we elected to err on the side of overrating the injury severity. The sever-ritical breakpoint was easier to deal with, however, as it lay between scores of 20 and 21. Injuries scoring 21 and above were classified as ‘critical’ while injuries scoring from 14 to 20 were classified as ‘severe’. Overall injury severity and mortality were very low in this study population but comparable to those from other Australasian series (Civil et al., 1987; Deane et al., 1988, 1990). In Melbourne at the time in which this review was set, there was no stratification of hospitals according to the trauma service they provided. At that time, the only triage guidelines for prehospital personnel were to transport injured patients to the nearest public hospital (Ambulance Service Melbourne, 1982). It is therefore likely that the sample of patients reviewed in the present paper was indeed representative of the injury spectrum of the local road accident population. It has been the experience of the
491
clinicians involved in this study that patients with ISS > 20 do not present frequently, and that a great number of minor injuries is seen for every case of major trauma. This contrasts with the experience of dedicated trauma centres in the USA and Europe where elaborate triage systems endeavour to ensure that all patients sent to such a centre have significant injury. The ISS breakpoints identified in this study, which related to the local experience and local resources available for management, are therefore unlikely to be applicable in these overseas centres. It was noted at the city hospital that the junior doctor tended to underrate injury severity in patients with more serious injuries and had the lowest individual consistency of rating. However, the converse was found at the suburban hospital where the general surgeon was significantly less consistent than the RMO and appeared to be less able to discriminate between different categories of severity with the crude ratings used. This disparity in clinicians’ consistencies reflects the subjectivity inherent in injury assessment and the inability of the ISS to account for physiological changes in the injured patient.
Conclusions This study has demonstrated that particular ISS levels can be established to classify the severity of injury of patients admitted to hospital into moderate, severe and critical groups which correspond well to subjective ‘clinical’ ratings. The particular ISS levels in the population studied were scores of 6-13,14-20 and 21 and above, respectively. Thus, the ISS has been confirmed as an appropriate tool for peer review in injury management and these ISS ranges can be used to select groups of patients for future trauma audits.
Acknowledgements The authors are grateful to Misses M. Dalton, N. Burford, L. Dix and W. Anderson for their assistance in the retrieval of information from case records.
References Ambulance Service Melbourne (1982) Gnr& of i%+roPe/i~~rrr Hes@rls. Staff Notice 8215, June 1982. Baker S. P., O’Neil B., Haddon W. et al. (1974) The ISS: A method for describing patients with multiple injuries and evaluating emergency care 1. Traitrun 14, 187. Civil 1. D., Ross S. E. and Schwab C. W. (1987) Major trauma in an urban New Zealand setting: resource requirements. Au.d AJ.Z. J Surg. 57, 543. Civil I. D. and Schwab C. W. (1989) Clinical prospective injury scoring: When is it accurate? J Tmrlrtinu29, 613. Committee on Injury Scaling (1990) The Abbreviated Injury Scale, 1990 Revision. Association for the Advancement of Automotive Medicine, Des Plaines, IL 60018, USA. Deane S. A., Gaudry P. L., Woods P. et al. (1988) The management of injuries -a review of deaths in hospital. Awt N. Z. J Strrg.58, 463. Deane S. A., Gaudry P. L., Woods W. P. D. et al. (1990) Interhospital transfer in the management of acute trauma. Aus/ N. Z. J Sug. 60, 441. Dearden C. H. and Rutherford W. H. (1985) The resuscitation of the injured in the accident and emergency department - a medical audit. Iniun/ 16, 249.
Injury: the British Journal of Accident Surgery (1992) Vol. 23/N0. 7
492
Kruskal W. H. and Walks W. A. (1952) Use of ranks in one-criterion variance analysis. J. Amer. Stdist. Assoc. 4, 583. Mackenzie E. J., Shapiro S. and Eastham J. N. (1985) Rating AIS severity using emergency department sheet vs. inpatient charts. I. Truuma 25, 984. Smejkal R., Civil I., Unkle D. et al. (1989) Injury severity scoring: a comparison of early clinical versus discharge diagnosis. Ad. Anal. Prev. 21, 386.
Tulloh B. R. (1988) The use of injury severity scales in the early management of road trauma. M.S. Thesis, University of Melbourne, Australia. Paper accepted
26 February
1992.
Requests for reprintsshouti be aa%ssed fo: Mr B. R. Tulloh, Echuca District Hospital, Francis Street, Echuca, Victoria 3564, Australia.
Book review Blunt Multiple Trauma: Comprehensive Pathophysiology and Care. J. R. Border, M. Allgiiwer, S. T. Hansen and T. P. Ruedi. Marcel Dekker, Inc, New York. 82478172 4 $234.00,
1990, 1024 pp.
This large and challenging book has come out of Swiss-American Co-operation fostered by the A0 Group. It is designed to consider tissue response to trauma in general rather than to act as a specific ‘cook book in the fashion of the A0 manual. There is, accordingly, heavy emphasis on the problems of devitalised tissue, sepsis, oxygen transport, and pulmonary and gut function. The book is divided into nine sections commencing with an excellent historical review detailing the evolution of trauma management and the role played by the control of sepsis and resuscitation. The diminishing treatment-free interval is documented along with the diminishing use of ARDS and fat embolism syndrome. The second section details basic cellular pathology and the response of the different body systems to injury. The third section deals with the epidemiology of trauma and injury severity scoring. There then follows a fourth section on resuscitation and a
fifth section on evaluation. Section 6 is devoted to emergency surgery and this is perhaps the least appropriate part of this volume. There are chapters within this that are highly irrelevant such as multiple team operating and intra-operative monitoring while other sections such as re-plantation, brachial plexus injuries and free tissue transfers are all far too isolated topics to be valuable here, particularly when they are covered only in outline and without indications as to overall results. Indications and end results need to feature more prominently if these chapters are to be justified. Overall, however this is an excellent and authoritative volume. The authors are to be commended on pulling together the essence of basic science and the practical clinical applications derived from it. References are voluminous and available by chapter. It is essential reading for those involved in the treatment of major trauma and will undoubtedly help inspire new thought and better clinical standards in the treatment of patients with major injuries.
J. L. Plewes, FRCS
