Article

The motor component does not convey all the mortality prediction capacity of the Glasgow Coma Scale in trauma patients

Unlabelled imageAmerican Journal of Emergency Medicine (2012) 30, 1032-1041

Original Contribution

The motor component does not convey all the Mortality prediction capacity of the Glasgow Coma Scale in trauma patients?,??

Benoit Vivien MD, PhD a,1, Jean-Michel Yeguiayan MDb,1, Yannick Le Manach MD c,1, Claire Bonithon-Kopp MD, PhD d,1, Sebastien Mirek MDb,1, Delphine Garrigue MDe,1,

Marc Freysz MD, PhD b,1, Bruno Riou MD, PhD f,?,1

aUniversity Paris Descartes-Paris 5, Service d’Aide Medicale Urgente (SAMU) 75 and Department of Anesthesiology and Critical Care, Centre Hospitalo-Universitaire (CHU) Necker-Enfants Malades, Paris, France

bUniversite de Bourgogne, SAMU 21 and Trauma Critical Care Unit, CHU de Dijon, Dijon, France

cDepartment of Anesthesiology and Critical Care, CHU Pitie-Salpetriere, Assistance Publique-Hopitaux de Paris (APHP),

Paris, France

dINSERM CIE 01, Center of Clinical Investigation and Department of Clinical Epidemiology, CHU de Dijon, Dijon, France

eDepartment of Anesthesiology and Critical Care, CHU de Lille, Lille, France

fUniversite Pierre et Marie Curie-Paris 6, Department of Emergency Medicine and Surgery, CHU Pitie-Salpetriere, APHP, Paris France

Received 24 May 2011; accepted 25 June 2011

Abstract

Purpose: We tested the hypothesis that the motor component of the Glasgow Coma Scale (GCS) conveys most of the predictive information of triage scores (Triage Revised Trauma Score [T-RTS] and the Mechanism, GCS, Age, arterial Pressure score [MGAP]) in trauma patients.

Method: We conducted a Multicenter prospective observational study and evaluated 1690 trauma patients in 14 centers. We compared the GCS, T-RTS, MGAP, and Trauma Related Injury Severity Score (reference standard) using the full GCS or its motor component only using logistic regression model, area under the receiver operating characteristic curve, and reclassification technique.

Results: Although some changes were noted for the GCS itself and the Trauma Related Injury Severity Score, no significant change was observed using the motor component only for T-RTS and MGAP when considering

(1) the odds ratio of variables included in the logistic model as well as their discrimination and calibration characteristics, (2) the area under the receiver operating characteristic curve (0.827 +- 0.014 vs 0.831 +- 0.014, P = .31 and 0.863 +- 0.011 vs 0.859 +- 0.012, P = .23, respectively), and (3) the reclassification technique. Although the mortality rate remained less than the predetermined threshold of 5% in the low-risk stratum, it slightly increased for MGAP (from 1.9% to 3.9%, P = .048).

? This study was supported by the Programme Hospitalier de Recherche Clinique 2003 of the French Ministry of Health (National PHRC), the Societe Francaise d’Anesthesie et de Reanimation (SFAR), and the Centre Hospitalier Universitaire de Dijon.

?? The authors have declared no conflict of interest.

* Corresponding author. Service d’Accueil des Urgences, CHU Pitie-Salpetriere, 47 Boulevard de l’Hopital, 75651 Paris Cedex 13, France. Tel.: +33 1 42 17 72 49; fax: +33 1 42 17 72 69.

E-mail address: [email protected] (B. Riou).

1 FIRST Study Investigators. See Appendix for the complete listing.

0735-6757/$ – see front matter (C) 2012 doi:10.1016/j.ajem.2011.06.022

Conclusion: The use of the motor component only of the GCS did not change the global performance of triage scores in trauma patients. However, because a subtle increase in mortality rate was observed in the low-risk stratum for MGAP, replacing the GCS by its motor component may not be recommended in every situation.

(C) 2012

Introduction

In most developed countries, a regional Trauma system has been developed that categorizes hospitals according to the resources required to provide various Levels of care for traumatic injuries. It has recently been demonstrated that the overall risk of death was 25% lower when care was provided at a specialized trauma center [1], emphasizing the importance of appropriate prehospital triage of trauma patients who need to be referred to the most appropriate facility. Among the trauma scores developed to facilitate prehospital triage, the Revised Trauma Score remains the most widely cited [2], particularly its triage version (T-RTS) [3]. Recently, we have designed the Mechanism, Glasgow Coma Scale (GCS), Age, and arterial Pressure (MGAP) score [4] that predicts mortality better than the T-RTS and is more specific, approaching the specificity of the reference standard, the Trauma Related Injury Severity Score (TRISS) [5].

In most, if not all, trauma scores, the GCS score is considered as one of the key variables [2-5]; and it has also been introduced in various outcome scores in the intensive care unit (ICU), such as the Acute Physiology And Chronic Health Evaluation score [6]. The GCS was initially developed to assess consciousness level in patients with brain injury using a 3-component (eye, motor, and verbal) scoring system [7] and has been widely accepted as a numeric GCS score (from 3 to 15) for both traumatic and nontraumatic patients to assess the degree of coma and to predict outcome. However, several important issues have been emphasized concerning the GCS score: (1) only moderate cross-assessment reliability, even in skilled hands; [8] (2) different eye/motor/verbal profiles may provide the same GCS score, whereas they are associated with significantly different outcome [9,10]; (3) relatively poor statistical performance [11,12], including redundancy between components; [10] (4) inability to calculate GCS score in some patients [13,14] because they are inebriated, intubated, or sedated. As a result of these shortcomings, several previous studies have suggested that the motor component of the GCS could replace the full GCS [10,15,16].

The purpose of this study was to test the hypothesis that the motor component of the GCS conveys most of the predictive information of triage scores in trauma patients. We compared the diagnostic performance of various scores available in the prehospital phase (GCS, T-RTS, and MGAP) using either the motor component only or the full GCS. We also tested the TRISS as a reference standard, although TRISS can be determined only later after complete assessment of all trauma lesions. We tried to assess why

the verbal and eye components of the GCS convey few additional predictive information.

Methods

This prospective Epidemiological study was an ancillary study of the French Intensive Care Recorded in Severe Trauma (FIRST) study that involved ICU and emergency departments from 14 university hospitals and corresponding to Level 1 trauma centers. Between December 2004 and March 2007, study centers were asked to record data regarding consecutive patients with severe blunt trauma in a computerized and anonymous database. Inclusion criteria were age (18 years or over) and severe blunt trauma, defined as trauma requiring admission into an ICU within 72 hours after injury or, in the case of early death before ICU admission, trauma managed by a mobile emergency care unit (MECU) units [17]. Exclusion criteria were penetrating trauma and death occurring before any life-sustaining treatment. A total of 3205 patients were eligible for inclusion in the FIRST study. After checking for Data quality, patients with either incomplete (n = 97) or poor-quality data (aberrant or illogical data, n = 281) regarding prehospital management, injury severity score (ISS), and vital status were subsequently excluded. Patients who died before any evaluation of lesions (n = 124) were also excluded. Thus, the database was restricted to 2703 patients suffering from severe trauma, alive upon arrival at the hospital, for whom complete and high- quality data were available in the major variables of interest. According to French law, this noninterventional study did not require approval by an ethics committee. The study was declared to, and approved by, the National Commission for Data Processing and Civil Liberties (CNIL, Paris, France). All patients or their families received information about the study. Our study followed the Standards for Reporting of Diagnostic Accuraccy (STARD) recommendations concern- ing the report of studies of diagnostic accuracy [18].

Data collection

Intensive care unit physicians collected data from the medical records of MECU, emergency divisions, and ICU regardless of the hospital of first admission. In each center, ICU physicians entered data into the FIRST database with the help of local research assistants. The eligibility criteria were checked online by the research assistants of the Coordination Center in Dijon, France. Every month, the Coordination

Center extracted data for quality control. In cases of missing, aberrant, or illogical mandatory data, queries were sent to local research assistants. At the end of inclusion, data monitoring was performed by the Coordination Center to validate data quality on a 7% random sample of patients. Unreliable variables were discarded from analysis.

The ICU physicians collected (1) patient characteristics;

(2) data about accident circumstances, condition of the victim following a traffic-related accident, rescue services mobilized for Patient transport (MECU or first aid pro- viders); (3) hospital units involved in early care of the patient before admission in the ICU; 4) clinical and biological data on the prehospital phase, at first hospital admission, and at 24 and 72 hours after trauma; and (5) summary of clinical variables at patient discharge or death. During the prehospital phase, the following data were recorded: initial physiological variables (arterial pressure, respiratory rate, pulse oximetry), pupil status, GCS, and life-sustaining treatments. The first available measurement, either at the prehospital phase or upon hospital admission, was used to describe the initial physiological status of the patient. At patient discharge from the ICU or death (within 30 days), anatomical injury diagnoses with corresponding Abbreviated Injury Scale codes [19] and the ISS [20] were recorded from medical records. The Abbreviated Injury Scale was coded, according to the 1998 updated classification [21], by local research assistants using medical, radiological, and surgical reports. Local ICU physicians reviewed all problematic cases. The Probability of survival was calculated using the TRISS [5]. When respiratory rate measurement was lacking, the value was coded as normal (20 beats per minute), as previously described [22]. Survival was defined as survival within 30 days after trauma.

End points

We compared the GCS score itself, the triage scores (T- RTS and MGAP), and the TRISS (the reference standard) when considering the full GCS or its motor component only. To assess the diagnostic performance, we performed the following comparisons: (1) the multivariate logistic model using the variables of each score [2,4,5]; (2) the area under the receiver operating characteristic (ROC) curve; and (3) the reclassification techniques [27,28], with the population being divided into 3 risk categories: low (b5%), intermediate, and high (N50%) risk for death, as previously described [4].

Statistical analyses

Data are presented as mean +- SD or median [25th-75th interquartile] for non-Gaussian variables (D’Agostino- Pearson omnibus test). Comparison of 2 proportions was performed using the comparison of confidence interval. Comparison of 2 areas under the ROC curve was performed using a nonparametric technique [23].

The observed survival was compared with the expected survival obtained by summing the individual TRISS values [24,25]. To compare our population with that of the Major Trauma Outcome Study (MTOS) [26], we calculated the M, W, Z, Ws, and Zs scores. A value of M less than 0.88 indicates a disparity in the severity match between the study group and the MTOS. The W score is the percentage of survivors more or less than would be expected. A Z score of between -1.96 and +1.96 indicates no significant difference (P N .05) between the actual number of survivors and that expected. The standardized W score (Ws) and Z score (Zs) represent the scores that would have been observed if the case mix of severity was identical to that of the MTOS [24]. A multiple logistic regression was performed to assess variables associated with death for each score (R-RTS, MGAP, and TRISS). We entered only variables that have been previously incorporated in these scores [2,4,5]. We compared the logistic models obtained when using either the full GCS or its motor component only. Discrimination of the logistic models was assessed by the C-statistic (area under the ROC

curve); and calibration, by the Hosmer-Lemeshow statistic.

As the comparison of ROC curves is recognized to be potentially insensitive, 2 complementary reclassification methods (net reclassification improvement [NRI] and integrated reclassification improvement [IDI]) were used

[27]; NRI requires predefined strata of risk, whereas IDI does not and can be seen as a continuous version of NRI with a probability of end point differences used instead of predefined strata [28]. This approach allowed us to assess the changes of predictive abilities given by the use of the motor component of the GCS only. Because NRI and IDI are powerful statistical tools, significant results might only have a moderate clinical impact. To illustrate the changes provided by the use of the motor component only, we provide reclassification tables that enable us to quantify the change in terms of number of patients correctly reclassified by using the motor component only. Furthermore, the reclassification tables of patients with or without the end point offer a practical representation of both the relationship between false positive and false negative and the magnitude of the gain of predictability in quantitative terms (number of patients) [29]. All P values were 2-sided, and P b .05 was considered significant. Statistical analyses were conducted using NCSS (Statistical Solutions Ltd, Cork, Ireland) and R software with

specific packages (http://www.R-project.org).

Results

Among the 2703 blunt trauma patients who fulfilled the criteria for inclusion in the database, the GCS was measured in 2628 (97%) patients; and the different components of the GCS were available in 1690 (63%) patients who were retained for the analysis.

The main characteristics of the trauma populations were as follows: age, 42 +- 18 years; 1289 (76%) men; initial systolic

Fig. 1 Bimodal distribution of the GCS score in the blunt trauma population (n = 1690). Patients with a GCS of 3 and 15 represent 19% and 57%, respectively.

arterial pressure, 115 +- 37 mm Hg; respiratory rate, 18 +- 8 beats per minute, median GCS, 15 [6-15]; and median duration of prehospital care and transportation, 1.75 [1.25- 2.4] hours. The cause of trauma was road accident in 1061 (63%) patients. The median T-RTS was 12 [10-12]; median

RTS, 7.84 [5.97-7.84]; median MGAP, 22 [15-24]; median

ISS, 25 [17-34]; and median TRISS, 0.95 [0.71-0.99]. The

observed number of deaths was 324 (19.2%), whereas the expected number of deaths was 345 (20.4%), providing a W score of 1.2% (P b .05). However, the M score (0.66) indicated a significant disparity in the severity match between the studied groups and the MTOS group, mainly due to a lower proportion of patients with a very high probability of survival (TRISS >=0.96, 49% vs 83%) and a higher proportion of patients with a low probability of survival (TRISS <=0.25, 11% vs 4%). The Ws score was 0.0%, indicating that it was not significantly different from that of the MTOS.

Fig. 1 shows the bimodal distribution of GCS among the studied population. Fig. 2 shows the relationship between GCS and its 3 components with mortality rate. The motor components had the highest correlation coefficient.

There was a high and significant correlation between the motor component and the verbal (? = 0.93, P b .001), eye opening (? = 0.92, P b .001), and verbal + eye opening (? = 0.93, P b .001) components. Fig. 3 shows the high correlation between the motor component and the verbal + eye opening components of the GCS and the corresponding distribution of paired values. Several profiles did not exist or were poorly represented.

Table 1 shows the multivariate analysis of variables associated with mortality in the different prehospital scores (T-RTS, MGAP) and TRISS using the full GCS or its motor component only. The use of the motor component only was not associated with significant modifications of the odds ratio. Conversely, there were no noticeable changes in the calibration and discrimination of the logistic models.

When considering the GCS and its ability to predict mortality, the use of its motor component only was associated with a significant decrease in the area under the ROC curve (0.807 +- 0.014 vs 0.823 +- 0.013, P b .001) (Fig. 4). This

significant decrease was also observed for the TRISS (0.876 +- 0.011 vs 0.882 +- 0.010, P = .01) (Fig. 4). However, when considering the scores available during the prehospital phase, T-RTS (0.827 +- 0.014 vs 0.831 +- 0.014, P = .31) and MGAP

(0.859 +- 0.012 vs 0.863 +- 0.011, P = .23), no significant changes in the area under the ROC curve were noted (Fig. 4). When considering the GCS, the lower Predictive ability of the motor component only was confirmed by the NRI (-0.510 +- 0.053, P b .001) but not IDI (-0.015 +- 0.006, P =

.06). When considering the TRISS score, the NRI and IDI indicated a reduced predictive ability when using the motor component only (-1.11 +- 0.062, P b .001 and -0.0089 +- 0.0021, P b .001, respectively). When considering the T-RTS, the NRI and IDI also indicated significant changes (-0.740 +- 0.062, P b .001 and -0.116 +- 0.0023, P b .001, respectively), indicating a slightly reduced predictive ability for events when using the motor component only, illustrated by the reclassi- fication table (Table 2). When considering the MGAP score, the NRI (-0.010 +- 0.062, P = .87) as well as the IDI (0.0015 +-

0.0034, P = .66) indicated no significant modification of the global predictive ability when using the motor component only. In fact, as depicted by the reclassification table (Table 3), the decrease of predictive ability for events was entirely balanced by the increase of predictive ability for nonevents. Although the global risk classification was not modified, dead patients were more frequently classified in a stratum of low risk by the triage scores (T-RTS and MGAP) using the motor component only when compared with the full GCS. As a consequence and although the mortality rate remained less than the predetermined threshold of 5% in the low-risk stratum, it slightly increased for MGAP (from 1.9% to 3.9%, P = .048) (Fig. 5). For T-RTS, the increase was not significant (from 5.5% to 6.8%, P = .26); but both values were greater than the 5% threshold.

Discussion

We demonstrated that the motor component conveys most of the GCS’s capacity to predict mortality in blunt trauma patients. When considering the logistic regression models, the areas under the ROC curve, and the reclassification technique, the triage scores (T-RTS and MGAP) were not significantly affected when the motor component was used in place of the full GCS. This was not completely the case for the reference

Fig. 2 Percentage of deaths observed according to the GCS score (A) and its verbal (B), eye opening (C), and motor (D) components. R2 is the coefficient of correlation and indicates the proportion of variation explained by the linear model (dotted line) (n = 1690).

Fig. 3 Frequency distribution of the motor and eye + verbal components of the GCS score (n = 1690). Some profiles do not exist or are poorly represented.

standard, the TRISS, which cannot be used for triage, or when considering the GCS itself. However, a subtle increase in mortality was observed in the low-risk stratum for MGAP, which has better predictive characteristics than T-RTS [4].

An ideal prehospital triage score for trauma patients should accurately predict death, be easily applied in the field, and be clear enough for a person on the telephone to understand, transmit the case, and justify sending the patient to an appropriate trauma center. Although the T-RTS has been used widely, we recently demonstrated that the MGAP score predicted mortality better than the T-RTS and, when an objective of 5% undertriage was fixed, was more specific than T-RTS, approaching the specificity of the reference standard, TRISS, which incorporates information concerning all trauma lesions not available in the prehospital phase [4]. The MGAP score was also able to clearly delineate patients with low (b5%), intermediate, and high (N50%) risk of mortality [4] and was a good predictor of ISS greater than 15, prolonged ICU stay, and massive hemorrhage [30]. Emphasis is usually given on the reliability and the ease of use of prehospital scores, while minimizing overtriage of minor trauma and undertriage

T-RTS

SAP

1.41

[1.14-1.73]

.001

1.43 [1.17-0.76]

b.001

GCS

2.00

[1.84-2.18]

b.001

0.69

0.83

1.93 [1.77-2.10]

b.001

0.21

0.83

RR

1.10

[0.90-1.36]

.35

1.12 [0.91-1.37]

.29

MGAP

SAP b60 mm Hg

3.22

[1.79-5.78]

b.001

3.00 [1.65-5.43]

b.001

SAP 60-120 mm Hg

1.54

[1.13-2.11]

b.001

0.30

0.86

1.50 [1.09-2.05]

b.001

0.83

0.86

GCS

1.29

[1.25-1.33]

b.001

1.82 [1.69-1.95]

b.001

Age N65 y

4.82

[3.35-6.93]

.007

4.80 [3.34-6.88]

.01

TRISS

RTS

1.90

[1.75-2.05]

b.001

1.85 [1.72-2.00]

b.001

ISS

0.96

[0.95-0.97]

b.001

2.52

0.88

0.96 [0.95-0.97]

b.001

2.16

0.88

Age N55 y

3.92

[2.82-5.46]

b.001

3.93 [2.83-5.47]

b.001

OR indicates odds ratio; CI, confidence interval: HL, Hosmer-Lemeshow ?2.

of major trauma or death. Our study indicates that the simplification of using only the motor component for prehospital triage scores does not decrease their global prediction capacity (Fig. 4). Nevertheless, a slight but

Table 1 Comparison of the logistic regression models of mortality prediction from various scores, using the full GCS or its motor component only (n = 1690)

Score Variables

With full GCS

OR [95% CI]

With motor component of the GCS only

P

HL

C-statistic

OR [95% CI]

P

HL

C-statistic

significant increase in mortality rate in the low-risk stratum was observed for MGAP when only the motor component was considered, suggesting that undertriage may deteriorate and that the motor component cannot replace the full GCS in every

Fig. 4 Receiving operating characteristics curves of the GCS score (A), T-RTS (B), MGAP (C), and TRISS (D) using the full GCS or its motor component only (GCSm, T-RTSm, MGAPm, TRISSm) (n = 1690). P values refer to the comparison of the area under the ROC curve between the 2 scores (eg, GCS vs GCSm ). All areas under the ROC curves were significantly different from 0.50 (ie, no discrimination). The dotted line corresponds to the nondiscrimination curve.

Death

Low risk

51 (100%)

0 (0%)

0 (0%)

51

intermediate risk

18 (30%)

40 (67%)

2 (3%)

60

High risk

0 (0%)

12 (6%)

201 (94%)

213

Total

69

52

203

324

Survival

Low risk

919 (100%)

0 (0%)

0 (0%)

919

Intermediate risk

89 (29%)

211 (70%)

3 (1%)

303

High risk

0 (0%)

11 (8%)

133 (92%)

144

Total

1008

222

136

1366

Patients were stratified according to T-RTS or T-RTSm among low (12 and 12, respectively), moderate (10-11 and 9-11, respectively), and high risk (0-

9 and 0-8, respectively).

situation or all algorithm schemes. Because an insufficient interrater reliability of the GCS has been previously reported [8], it is possible that this simplification may increase the interrater reliability of prehospital triage scores. Further studies are needed to test this hypothesis.

Table 2 Reclassification table for the T-RTS when considering the full GCS score or its motor component only (T-RTSm) (n = 1690)

T-RTS T-RTSm

Low risk Intermediate High risk Total

risk

Although the comparison of areas under the ROC curves remains the most popular metric to capture discrimination, it appears that, for models containing clinical risk and possessing reasonably good discrimination, significant associations between the test and the end point are required to provide significantly different areas under the ROC curves [27,28]. In other words, comparisons of the areas under the

Table 3 Reclassification table for the MGAP score when considering the full GCS score or its motor component only (MGAPm) (n = 1690)

MGAP MGAPm

Low risk Intermediate High risk Total risk

Death

Low risk 8 (100%)

0

(0%)

0 (0%)

8

Intermediate risk 28 (33%)

55

(65%)

2 (2%)

85

High risk 0 (0%)

18

(8%)

213 (92%)

231

Total 36

73

215

324

Survival

Low risk 425 (100%)

0

(0%)

0 (0%)

425

Intermediate risk 468 (62%)

287

(38%)

3 (1%)

758

High risk 0 (0%)

37

(20%)

146 (80%)

183

Total 893

324

149

1366

Patients were stratified according to MGAP or MGAPm among low (24-26 and 20-22, respectively), moderate (15-23 and 13-18, respec- tively), and high risk (3-14 and 2-12, respectively). Because of rounding, the sum may not provide 100%.

Fig. 5 Comparison of mortality according to risk stratification (low, intermediate, high) obtained with the T-RTS and MGAP score when considering the full GCS score or its motor component only (T-RTSm, MGAPm).

ROC curves might be considered as being powerless in identifying changes of interest in such situations. To address this problem, new ways of evaluating the usefulness of markers have been described based on the quantification of the reassignment of subjects into adapted risk categories [27-29]. In our study, the areas under the ROC curves were not significantly modified for T-RTS and MGAP (Fig. 4); and the lack of significant changes in NRI and IDI should be considered as further arguments suggesting that the GCS can be replaced by its motor component for MGAP and probably for RTS. Nevertheless, the reclassification tables (Tables 2 and 3) underlined more subtle changes that were not captured by more global assessment of the diagnostic performance. Therefore, our study emphasizes the impor- tance of reclassification methods in assessing performance of diagnostic tests [27-29].

Several mechanisms may explain why the motor component conveys most of the GCS’s capacity to predict mortality. The redundancy between components has already been noted and is confirmed by the high Correlation coefficients between components we observed (all greater than 0.92). This high correlation has been used to predict the

verbal component from the eye and motor components in intubated patients [14,30]. This characteristic has been identified as one of the elements of the relatively poor statistical performance of the GCS [11,12]. A highly skewed frequency distribution has been previously reported and is probably enhanced by the naturally bimodal frequency distribution of the Probability of death in trauma patients [25]. It should be noted that Raux et al [31] recently demonstrated that the duration of stay in the ICU, a variable linked to mortality, also had a bimodal distribution. As shown in Fig. 1, most trauma patients had a GCS of 3 or 15, implying an obligatory high correlation between the 3 components of the GCS. In our study, the proportion of patients with a GCS of 3 or 15 was 78% (Fig. 1) but has been reported to be as high as 86% in North American data banks [11], this difference being explained by a severity selection bias [4,22,25]. Moreover, many eye/motor/verbal profiles did not exist in our trauma population or were poorly represented (Fig. 3). In the T-RTS and TRISS, the GCS is used as an ordered categorical variable, whereas in the MGAP, score is used as a continuous variable and a nonlinear relation to mortality complicating modeling of the GCS [10]. Healey et al [11] proposed using fractional polynomials to improve the model fit, whereas Moore et al

[12] reported that spline regression performed even better. However, these complex transformations cannot be applied for simple prehospital triage purpose. We also observed that the linear relation with mortality was not very good for the GCS and was greater for the motor component as compared with the GCS or its verbal and eye components (Fig. 2). The fact that different eye/motor/verbal profiles may also provide the same GCS score but significantly different outcomes has been previously pointed out [9,10] and may aggravate the above mechanisms.

It is not possible to calculate GCS in some patients [13,14], particularly because they are inebriated, intubated, or sedated. The simplification provided by considering only the motor component of the GCS may decrease the level of initial missing data. In our study, missing values of GCS were rare (3%), although the detailed components of the GCS were more frequently lacking (47%). Raux et al [22] also observed that missing GCS values were rare (1%) in a physician-staffed prehospital system, which is in marked contrast with the situation sometimes observed in paramedic- staffed prehospital system, where GCS is often not recorded in patients with normal consciousness [12]. In our study, GCS was measured before intubation and sedation; but the assessment of GCS may be also important to detect clinical deterioration either during transportation to the hospital or in the hospital. Because the motor component is thought to be less altered by intubation and sedation than the eye or verbal components, we think that it may be more useful than the GCS to monitor trauma patients, although some authors have proposed to predict the verbal components from the eye and motor component [14,31]. Further studies are required to test these hypotheses.

Some authors have even proposed a more drastic simplification of the GCS, describing a very simple 3-point score [32,33]. They demonstrated test performance similar to the GCS on 4 clinically relevant traumatic brain injury outcomes, including mortality. However, the performance of GCS is markedly lower than T-RTS and MGAP (Fig. 2) in trauma patients, mainly because the prognostic value of low arterial pressure is high in trauma patients, and these results cannot be simply extrapolated. Moreover, in those studies, they only considered ROC curves that are now recognized as relatively powerless tools [32,33]. Our study provides some evidence that a complete analysis including reclassification methods is necessary before accepting simplification of scores. Some limitations in our study deserve consideration. First, our study was conducted in an adult population and thus may not apply to pediatric patients [34]. Second, the GCS score and its components were measured in a physician-staffed prehospital system, and the result might be different in a paramedic-staffed prehospital system, although comparable results have also been reported in these conditions [11,15]. Third, it should be pointed out that prehospital triage should probably not be limited to only one score whatever its accuracy. Algorithms or decision schemes have been proposed by expert panels [35,36], and it is possible that an algorithm incorporating a prehospital triage score could be a valuable tool. Finally, we studied only trauma patients; and our results probably do not apply to other medical conditions such as stroke because the verbal and the eye components are likely to add valuable prognostic informa-

tion in these clinical conditions [37].

In conclusion, the motor component conveys most of the GCS’s capacity to predict mortality in blunt trauma patients; and its use does not significantly modify the global performance of prehospital triage scores. Several mecha- nisms may explain that result: redundancy between compo- nents, bimodal distribution with nonexistent eye/motor/ verbal profiles or poorly represented profiles, nonlinear relation to mortality, and different profiles providing the same GCS but different outcomes. Nevertheless, because a subtle increase in mortality rate was observed in the low-risk stratum for MGAP, replacing the full GCS by its motor component cannot be recommended unless another clear advantage, such as increase in reliability, is demonstrated [38].

Acknowledgments

We thank the physicians of SAMU/SMUR, emergency, and intensive care units who participated for their cooperation with the FIRST study, and all the research assistants of the INSERM CIE 01, Centre d’Investigation clinique-Epidemiologique clinique, CHU de Dijon, Dijon, France. We also thank Dr DJ Baker (Department of Anesthesiology, CHU Necker-Enfants Malades, Assistance Publique-Hopitaux de Paris, Paris, France) for reviewing the manuscript.

Appendix A. Investigators of the FIRST study

Steering committee : Pr Claire Bonithon-Kopp, Pr Jacques Duranteau, Pr Claude Martin, Pr Bruno Riou, Dr Jean- Michel Yeguiayan, Pr M Freysz (study coordinator).

The following investigators participated in the FIRST study group:

Besancon: Pr Annie Boillot, Dr Gilles Blasco, Pr Gilles Capelier, Pr Emmanuel Samain, Dr Thibault Desmettre, Dr Gabriel Hamadi.

Dijon : Pr Marc Freysz, Dr Jean-Michel Yeguiayan, Dr Christophe Avena, Dr Sebastien Andre, Dr Philippe Reviron, Dr Dalila Serradj.

Grenoble : Dr Claude Jacquot, Dr Celine Gourle, Dr Julien Brun, Dr Frederic Mongenot, Dr Elisabeth Rancurel, Dr Benedicte Bourgeois, Dr Isabelle Favier, Dr Francois Coppo. Lille : Dr Patrick Goldstein, Dr Herve Coadou, Dr Vincent Marel, Dr Delphine Garrigue, Dr Sandrine Rosen-

berg, Dr Philippe Poidevin, Dr Bernard Leroy.

Limoges : Dr Dominique Cailloce, Dr Stephanie Sebban. Lyon : Dr Francois Artru, Dr Frederic Dailler, Dr Thomas Lieutaud, Dr Carole Bodonian, Dr Jacqueline Convert, Dr Sarah Lorge, Dr Philippe Rague, Dr Marie-Christine Laplace, Dr Carine Delaleu-Rague, Pr Jean-Stephane

David, Dr Laure Besson, Pr Pierre Yves Gueugniaud. Marseille : Dr Francois Antonini, Pr Claude Martin. Nantes : Dr Antoine Andre, Dr Jean-Pierre Gouraud, Pr

Michel Pinaud, Dr Philippe Champin, Dr Dominique Demeure, Dr Pierre Joachim Mahe.

Nimes : Pr Jean Yves Lefrant, Dr Sophie Louvard, Pr Jean- Emmanuel de La Coussaye, Dr Pierre Geraud Claret, Dr Aurelie Dardalhon.

Le Kremlin-Bicetre : Pr Jacques Duranteau, Dr Christian Laplace, Dr Gaelle Cheisson, Dr Bernard Vigue, Dr Pierre- Etienne Leblanc, Dr Olivier Huet, Dr Catherine Ract

Paris : Pr Bruno Riou, Dr Danielle Sartorius, Dr Yan Zhao, Pr Olivier Langeron, Dr Frederic Marmion, Dr Sabine Roche, Dr Julien Amour, Dr Armelle Nicolas-Robin, Dr Caroline Telion, Dr Jean-Sebastien Marx, Dr Yael Ichay, Dr Kim An, Dr Benoit Vivien, Pr Pierre Carli.

Poitiers : Dr Jean Yves Lardeur, Dr Etienne Quoirin, Dr Fatima Rayeh, Pr Olivier Mimoz,

Coordination center for data monitoring and statistical analysis-Centre d’Investigation clinique-Epidemiologique clinique du CHU de Dijon (INSERM CIE 01), Dijon, France: Pr Claire Bonithon-Kopp (coordinator), Dr Christine Bin- quet (head statistician), Elodie Gautier and Sandrine Vinault (statisticians), Alexandra Felin (study monitor).

Local research assistants : Nathalie Berger (Nantes, Poitiers), Brigitte Lafond and Francoise Casano (Lyon, Marseille, Nimes), Carine Piatek (Le Kremlin-Bicetre, Lille, Paris), Alexandra Felin (Grenoble, Besancon, Dijon).

References

  1. MacKenzie EJ, Rivara FP, Jurkovich GJ, et al. A national evaluation of the effect of trauma-center care on mortality. N Engl J Med 2006;354:366-78.
  2. Champion HR, Sacco WJ, Copes WS, et al. A revision of the trauma score. J Trauma 1989;29:623-9.
  3. Moore L, Lavoie A, Abdous B, Le Sage N, Liberman M, Bergeron E, et al. Unification of the Revised Trauma Score. J Trauma 2006;61: 718-22.
  4. Sartorius D, Le Manach Y, David JS, et al. Mechanism, Glasgow Coma Scale, Age, Arterial Pressure (MGAP): a new simple prehospital triage score to predict mortality in trauma patients. Crit Care Med 2010;38:831-7.
  5. Boyd CR, Tolson MA, Copes WS. Evaluating trauma care: the TRISS method. J Trauma 1987;27:370-8.
  6. Cho DY, Wan YC. Comparison of the APACHE III, APACHE II and Glasgow Coma Scale in acute head injury for prediction of mortality and functional outcome. Intensive Care Med 1997;23:77-84.
  7. Teasdale G, Jennett B. Assessment of coma and impaired conscious- ness: a practical scale. Lancet 1974;II:81-3.
  8. Gill M, Reiley DG, Green SM. Interrater reliability of Glasgow Coma Scale scores in the emergency department. Ann Emerg Med 2004;43: 215-23.
  9. Hirai S, Ono J, Yamaura A. Clinical grading and outcome after early surgery in Aneurysmal subarachnoid hemorrhage. Neurosurgery 1996;39:441-6.
  10. Teoh LSG, Gowardman JR, Larsen PD, Green R, Galletly DC. Glasgow Coma Scale: variation in mortality among permutations of specific total scores. Intensive Care Med 2000;26:157-61.
  11. Healey C, Osler TM, Rogers FB, et al. Improving the Glasgow Coma Scale score: motor score alone is a better predictor. J Trauma 2003;54: 671-80.
  12. Moore L, Lavoie A, Camden S, et al. Statistical evaluation of the Glasgow Coma Score. J Trauma 2006;60:1238-44.
  13. Marshall LF, Becker DP, Bowers SA, et al. The national traumatic coma data bank: part I–design, purpose, goals, and results. J Neursurg 1983;59:548-51.
  14. Rutledge R, Lentz CW, Fakhry S, Hunt J. Appropriate use of the Glasgow Coma Scale in intubated patients: a linear regression prediction of the Glasgow verbal score from the Glasgow eye and motor scores. J Trauma 1996;41:514-22.
  15. Ross SE, Leipold C, Terregino C, O’Malley KF. Efficacy of the motor component of the Glasgow Coma Scale in trauma triage. J Trauma 1998;45:42-4.
  16. Al-Salameh MA, McDowell I, Stiell IG, et al. Initial emergency department trauma scores from the OPALS study: the case for the motor score in blunt trauma. Acad Emerg Med 2004;11:834-42.
  17. Yeguiayan JM, Garrigue D, Binquet C, Jacquot C, Duranteau J, Martin C, et al, The FIRST (French Intensive Care Recorded In Severe Trauma) Study Group. Medical pre-hospital management reduces mortality in severe blunt trauma: a prospective epidemiological study. Crit Care 2011;15:R34.
  18. Bossuyt PM, Reitsma JR, Bruns DE, et al. The STARD statement for reporting studies of diagnostic accuracy: explanation and elaboration. Ann Intern Med 2003;138W1-12.
  19. Champion HR, Copes WS, Sacco WJ, et al. A new characterization of injury severity. J Trauma 1990;30:539-45.
  20. Baker SP, O’Neill B, Haddon W, Long WB. The Injury Severity Score: a method for describing patients with multiple injuries and evaluating emergency care. J Trauma 1974;14:187-96.
  21. Association for the Advancement of Automotive Medicine. The Abbreviated Injury Scale. 1998 revision. USA: Des Plaines, IL; 1998.
  22. Raux M, Thicoipe M, Wiel E, et al. Comparison of respiratory rate and peripheral oxygen saturation to assess severity in trauma patients. Intensive Care Med 2006;32:405-12.
  23. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988;44:837-45.
  24. Champion HR, Copes WS, Sacco WJ, et al. The Major Trauma Outcome Study: establishing national norms for trauma care. J Trauma 1990;30:1356-65.
  25. Hollis S, Yates DW, Woodford M, Foster P. Standardized comparison of performance indicators in trauma : a new approach to case-mix variation. J Trauma 1995;38:763-6.
  26. Riou B, Landais P, Vivien B, et al. The distribution of the probability of survival is a strategic issue in randomized trial in trauma. Anesthesiology 2001;95:56-63.
  27. Pencina MJ, D’Agostino RB, D’Agostino RB, Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008;27:157-72.
  28. Ray P, Le Manach Y, Riou B, Houle TT. Statistical evaluation of a biomarker. Anesthesiology 2010;112:1023-40.
  29. Fellahi JL, Le Manach Y, Daccache G, Riou B, Gerard JL, Hanouz JL. The combination of EuroScore and postoperative cardiac troponin I improves the prediction of fatal and nonfatal adverse outcome following cardiac surgery. Anesthesiology 2011;114:330-9.
  30. Meredith W, Rutlefge R, Fakhry SM, Emery S, Kromhout-Shiro S. The conundrum of the Glasgow Coma Scale in intubated patients: a linear regression prediction of the Glasgow verbal score from the Glasgow eye and motor scores. J Trauma 1998;44:839-44.
  31. Raux M, Sartorius D, Le Manach Y, et al. What do prehospital trauma scores predict besides mortality? J Trauma 2011;70 [Epub ahead of print].
  32. Gill M, Windemuth R, Steele R, Green SM. A comparison of the Glasgow Coma Scale score to simplified alternatives scores for the prediction of traumatic brain injury outcomes. Ann Emerg Med 2005;45:37-42.
  33. Haukoos JS, Gill MR, Rabon RE, Gravitz CS, Green M. Validation of the simplified motor score for the prediction of brain injury outcomes after trauma. Ann Emerg Med 2007;50:18-24.
  34. Fortune PM, Shann F. The motor response to stimulation predicts outcome as well as the full Glasgow Coma Scale in children with severe head injury. Pediatr Crit Care Med 2010;11:339-42.
  35. Sasser SM, Hunt RC, Sullivent EE, et al. Guidelines for field triage of injured patients recommendations of the National Expert Panel on Field Triage. MMWR 2009;58:1-35.
  36. Riou B, Thicoipe M, Atain-Kouadio P, Carli P. Comment evaluer la gravite ? SAMU de France ed. Actualites en reanimation prehospitaliere: le traumatise grave. SFEM Editions: Paris; 2002.

p. 115-28.

  1. Weir CJ, Bradford APJ, Lees KR. The prognostic value of the component of the Glasgow Coma Scale following acute stroke. Q J Med 2003;96:67-74.
  2. Edmonds ZV, Mower WR, Lovato LM, Lomeli R. The reliability of vital signs measurements. Ann Emerg Med 2002;39:233-7.

Leave a Reply

Your email address will not be published. Required fields are marked *