External Validation of the Emergency Trauma Score for Early Prediction of Mortality in Trauma Patients*

Pieter Joosse, MD1; Willem-Jan J. de Jong, MSc2; Johannes B. Reitsma, PhD3; Klaus W. Wendt, PhD2; Niels W. Schep, PhD1; J. Carel Goslings, PhD1

Objectives: The Emergency Trauma Score has been developed for early estimation of mortality risk in adult trauma patients with an Injury Severity Score of 16 or higher. Emergency Trauma Score combines four early predictors available at the trauma resuscitation room: age, Glasgow Coma Scale, base excess, and prothrombin time. Our goal was to validate the Emergency Trauma Score in two large external cohorts. As the Injury Severity Score is not accurately known at the time patients present at the resuscitation room, we evaluated the performance of Emergency Trauma Score in all trauma patients. Design: External validation study using data from two prospectively collected trauma registries. Setting: Two academic level 1 trauma centers. Patients: Adult patients admitted to the hospital after treatment at the trauma resuscitation room. Intervention: Calibration and discrimination of the original Emergency Trauma Score were assessed within each cohort separately. Measurement and Main Results: A total of 4,418 consecutive patients were evaluated. Discrimination was good in both validation cohorts, with areas under the receiver-operating curve curves that were even higher (0.94 and 0.92, respectively) than that in the original cohort (0.83). Predicted mortality was systematically too high compared with actual mortality in patients with low-tomedium expected risk (< 25%). Calibration improved in the lower expected risk range after exclusion of patients with Injury Severity Score less than 16. Conclusions: The Emergency Trauma Score model performs well in discriminating between trauma patients who will survive and who
*See also p. 207. 1 Trauma Unit, Department of Surgery, Academic Medical Center, Univer sity of Amsterdam, Amsterdam, The Netherlands. 2 Department of Surgery, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands. 3 Department of Epidemiology, Julius Center for Health Sciences and Primary Care, University Medical Center Utrecht, University Utrecht, Utrecht, The Netherlands. The authors have disclosed that they do not have any potential conicts of interest. For information regarding this article, E-mail: pieterjoosse@hotmail.com Copyright 2013 by the Society of Critical Care Medicine and Lippincott Williams & Wilkins DOI: 10.1097/CCM.0b013e31829e53f5

will not. If applied to all trauma patients, predicted mortality risks are too high in the low-risk category. (Crit Care Med 2014; 42:8389) Key Words: emergencies; Injury Severity Score; prognosis; trauma severity indices; validation; wounds and injuries/mortality

score to quantify the probability of survival of trauma patients shortly after admission to the trauma resuscitation room enables caregivers to assess the patients injury severity to guide clinical decisions and allows for efcient allocation of resources. In 2009, Raum et al (1, 2) introduced the Emergency Trauma Score (EMTRAS) for early estimation of mortality risk in adult trauma patients. EMTRAS combines four early predictors from the emergency resuscitation room and demonstrated favorable discrimination compared with more complex scores. The early predictors used by EMTRAS are age (yr), Glasgow Coma Scale (GCS), base excess (mmol/L), and prothrombin time (%). For each predictor, a subscore of 0, 1, 2, or 3 points is assigned, based on the actual value of the predictor. EMTRAS is dened as the sum of these subscores, that is, the lowest (best) EMTRAS is zero and the highest (worst) is 12. The EMTRAS was developed in a large cohort (n = 4,808) of trauma patients with an Injury Severity Score (ISS) (3) of 16 or higher, derived from the German Trauma Registry (http:// www.traumaregister.de). The EMTRAS has been internally validated in a second cohort (n = 1,292) from a later period within the same database. The EMTRAS is intended to be an easy-to-compute scoring system for the emergency resuscitation room based on a limited number of clinical predictors that are commonly and early available. Despite its apparent advantages, the EMTRAS has limitations that have contributed to its relative obscurity. First, generalizability of the EMTRAS has not been established as no validation in an external database has been undertaken yet. Second, EMTRAS excludes patients with an ISS below 16. Anatomic injury scales, like the ISS, can only be scored reliably further along the diagnostic or therapeutic process and are usually not available during the initial phase at the trauma resuscitation room. Therefore, the ISS criterion forms a limitation in
www.ccmjournal.org

Critical Care Medicine

83

Joosse et al

practice to identify patients who are eligible for early outcome prediction using the EMTRAS. Ideally, all potentially severely injured trauma patients should benet from EMTRAS, including those patients who turn out to have an ISS below 16. The objective of this study was to validate the EMTRAS in two external cohorts. In particular, we examined whether the applicability of the EMTRAS can be extended to all trauma patients regardless of their ISS. Therefore, all adult trauma patients that were admitted to the trauma resuscitation room were evaluated using the EMTRAS.

TaBle 1. Elements of the Emergency Trauma Score and Their Respective Weights
Predictor Category Weight

Age (yr)

< 40 4060 6175 > 75

0 1 2 3 0 1 2 3 0 1 2 3 0 1 2 3

Glasgow Coma Scale

1315 1012 69 35

MATERIALS AND METHODS

The EMTRAS was externally validated in two separate cohorts of trauma patients treated at the trauma resuscitation room in an urban level 1 trauma center (A) and a rural level 1 trauma center (B) in The Netherlands. Data were derived from the prospectively collected trauma registries of both trauma centers. The rst cohort consisted of consecutive patients who were admitted during the period from 2004 to 2010 (center A). The second cohort consisted of consecutive patients who were admitted during the period from 2006 to 2011 (center B). In center A, all patients treated at the trauma resuscitation room were included. In center B, patients triaged as code red were judged potentially severe trauma patients and eligible for inclusion in the cohort. The decision for a patient to be transported to a level 1 trauma center is based on the triage and expertise of the prehospital emergency medical services personnel, following national and international protocols (4). Triage criteria include vital variables and mechanism of injury. The initial evaluation and resuscitation of trauma patients was performed following Advanced Trauma Life Support guidelines (5). Trauma patients admitted to the hospital, patients who died in the trauma resuscitation room, or patients referred immediately after trauma were included. Patients declared dead on arrival, patients discharged home directly from the emergency department, and children (< 16 yr) were excluded. This is a retrospective observational study. The Dutch Medical Research Involving Human Subjects Act (Wet medisch-wetenschappelijk onderzoek met mense) does not apply to this type of study, and therefore, it was exempt from approval by the ethics committee. EMTRAS Calculation Four predictors for mortality are included in the EMTRAS: age in years, GCS, base excess (mmol/L), and prothrombin time (%). Each predictor value is scored into four categories ranging from 0 to 3. EMTRAS is dened as the sum of these subscores, that is, the lowest (best) EMTRAS is zero and the highest (worst) is 12. Weights for the predictors (Table 1) and coefcients for calculation of predicted mortality were taken from the original publication (1). The GCS was recorded immediately at admission. In intubated or pharmacologically paralyzed patients, the most recent GCS before intubation was recorded. Laboratory tests were obtained directly at admission at the trauma resuscitation room. Base excess was obtained by arterial blood gas analysis, and the test results
84
www.ccmjournal.org

Base excess (mmol/L)

> 1 1 to 5 5.1 to 10 < 10

Prothrombin time (%)

> 80 8050 4920 < 20

were available within 15 minutes. Prothrombin time was expressed as percent of the reference value, with lower percentages denoting a prolonged prothrombin time (1), and was available within 3045 minutes. Outcome in the two validation cohorts was in-hospital mortality, as it was in the original EMTRAS study. Statistical Analysis Multiple imputation methods (10 rounds) were used to impute missing values on the EMTRAS predictors for both validation cohorts (6, 7). All EMTRAS variables plus the outcome variable were included in the imputation model (8). Other available variables deemed valuable for imputation of EMTRAS predictors were sex, ISS, heart rate, systolic blood pressure, SO2, respiratory rate, serum lactate, hemoglobin, platelet count, and activated partial thromboplastin time. Imputation was performed on the original continuous measurement scale of the predictors based on multivariate normal distributions. Imputation was performed separately for both validation cohorts. Values for age (center A and B) and GCS (center B) were complete, so no imputation was needed. The highest percentage of values that was imputed was 11% for GCS in cohort A and 17% for base excess in cohort B. For analysis involving imputed data, the results of 10 imputed datasets were combined to obtain nal estimates. Performance of the EMTRAS was studied by determining discrimination and calibration within both cohorts after imputation. Subgroup analyses were performed for patients with an ISS more than or equal to 16 and for patients with an ISS less than 16. Discrimination can be explained as the
January 2014 Volume 42 Number 1

Clinical Investigations

TaBle 2. Characteristics of Potentially Severe Trauma Patients Who Were Admitted to Trauma Centers A and B
Validation Cohorts Center A Predictor Original Cohort Available Percentage Value Available Percentage Center B Value

Number of patients General Male (%)  Injury Severity Score Blunt (%) Emergency Trauma Score predictors Age (yr)  Glasgow Coma Score  Base excess (mmol/L)  Prothrombin time (%) Complete (n) Physiologic systolic blood pressure (mm Hg)   Heart rate (per min) SO2 (%)  Respiratory rate (per min) Intubation (%) Laboratory test Hemoglobin (g/dL)  Platelet count (10E9/L)  Activated partial thromboplastin time (s) Lactate (mmol/L)  International normalized ratio Outcome Mortality (%)

11,533 73.4 30.1 ( 12.5) 96.3 42.3 ( 19.2) 10 ( 4.9) 3.3 ( 5.4) 73.9 ( 23.6) 4,808 119 ( 32) 91 ( 24) 97 ( 6) 14.9 ( 5.4) 11.0 ( 3.2) 190 ( 82) 38.8 ( 24.1) 5.3 ( 8.6) 21.9 100 100 100 100 89 92 92 78 92 91 85 34 93 98 95 92 58 11 100

3,001 71 12.8 ( 11.4) 91.3 42.2 ( 18.7) 13 ( 4) 2.5 ( 4.6) 97 ( 15) 2,353 138 ( 28) 87 ( 20) 98 ( 7) 17 ( 6) 16.8 8.1 ( 1.3) 230 ( 71) 31.7 ( 20.8) 3.0 ( 3.0) 1.5 ( 1.0) 6.7 100 100 100 100 100 83 90 77 0 0 0 0 99 99 98 91 88 89 100

1,417 75.2 19.4 ( 15.2) 94 43.9 ( 19.5) 12 ( 4) 3.5 ( 4.6) 86 ( 15) 1,096 39.6 7.8 ( 1.5) 212 ( 66) 28.2 ( 14.4) 2.3 ( 2.3) 1.2 ( 0.7) 14.3

Dashes indicate parameters were not available, so the values were not applicable. The original model-generating population of the Emergency Trauma Score is shown in the second column. The available percentage of each predictor is presented as well. SD represented within parenthesis.

models ability to distinguish between trauma patients who will or will not survive. It was expressed by calculating the area under receiver-operating curve (AUROC) with 95% CIs. Calibration plots were used to visualize predicted and actual mortality. In the discrimination and calibration analysis, the EMTRAS was applied as it was presented in the original article, that is, a 12-point scale based on recorded predictors. Regression coefcients using the EMTRAS predictors in their continuous form were reestimated in each validation cohort and compared with the coefcients of the original generating cohort (Table 3 in Raum et al [1] and Critical Care Medicine

M.R.Raum, personal communication, 2011). A risk score was calculated based on the original coefcients following the formula: 0.4035 + 0.0397 Age 0.1722 GCS 0.0639 base excess 0.0334 prothrombin time. Coefcients were reestimated for the EMTRAS predictors in the two validation cohorts using logistic regression with the risk score used as offset. In this way, the difference between the original and newly estimated coefcient could be tested for statistical signicance. Signicance was attributed to a p value of 0.05. All analysis was performed using the Statistical Package for Social Sciences version 18.0 (IBM, Chicago, IL).
www.ccmjournal.org

85

Joosse et al

RESULTS
During the years 20042010, a total of 3,001 patients were admitted to center A after presentation at the trauma resuscitation room. Two hundred patients died in hospital (overall mortality 6.7%). Of the 1,417 patients who were admitted to center B after presentation at the trauma resuscitation room during the period 20062011, a total of 203 patients died in hospital (overall mortality 14.3%). Table 2 shows demographic data and prognostic variables in both validation cohorts. The original population in which the EMTRAS has been developed (EMTRAS population) is included for comparison. Both validation cohorts were comparable for sex, injury mechanism, and age with the original EMTRAS population. The mean ISS was 12.8 in center A and 19.4 in center B. Both mortality and the distribution of the EMTRAS predictor values corresponded with a less severely injured population as compared with the original EMTRAS population. This was expected as the eligibility criteria for EMTRAS evaluation were extended from patients with ISS of 16 or higher (original EMTRAS) to all potentially severe trauma patients presented at the trauma resuscitation room (present study). Physiologic variables, except GCS and intubation status, were not routinely recorded in the trauma registry of center B. Data from medical records were deemed unreliable for analysis, and therefore, these data are not presented in Table2. Predictors for EMTRAS calculation were complete for 2,353 patients (78%) in the center A and for 1,096 patients (77%) in the center B (Table2). The EMTRAS demonstrated good discrimination for both validation cohorts. The AUROC for cohort A was 0.94 (0.93 0.96) and for cohort B 0.92 (0.900.94), whereas the AUROC for the original EMTRAS population was 0.81. Receiveroperating curves for both cohorts are presented in Figure1. Subgroup analysis of patients with an ISS more than or equal to 16 resulted in an AUROC of 0.90 for center A and 0.89 for center B. Subgroup analysis of patients with an ISS less than 16 resulted in an AUROC of 0.94 for center A and 0.82 for center B. In both cohorts, predicted mortality was systematically too high compared with actual mortality in patients with low-to-medium expected risk (< 25%) (Fig.2). When selecting patients with an ISS more than or equal to 16, calibration was better in the lower expected risk range (Fig.3), whereas for patients with an ISS less than 16, the EMTRAS overestimated mortality across the full range of EMTRAS scores (Fig.4). Reestimation of the logistic regression coefcients using the EMTRAS predictors in their continuous form in the validation cohort resulted in signicant changes of several predictors compared with the original coefcients (Table 3). The coefcient for GCS changed signicantly in both validation cohorts, whereas the coefcient for the prothrombin time changed signicantly only for center A. The direction of change indicated to a stronger association between the predictor and mortality, which corresponds with the increase in AUROCs found in both validation cohorts.
86
www.ccmjournal.org

DISCUSSION
The results of this study show that the EMTRAS for early prediction of mortality performs well in two external validation cohorts. The EMTRAS discriminates even slightly better between those who will survive and who will not than in the original study. Predicted mortality was systematically too high compared with actual mortality in patients with low-tomedium expected risk (< 25%). Calibration improved in the lower expected risk range after exclusion of patients with ISS less than 16. EMTRAS is a feasible tool to estimate mortality risk at an early stage in potentially severe trauma patients, as approximately 80% of the patients has complete data to calculate EMTRAS. The strength of the EMTRAS can be explained by the fact that each predictor alone is strongly related to mortality and that the model was developed in a large cohort producing robust estimates. Age represents an independent predictor of mortality in trauma patients (911). Age classied in ve categories improved predictive performance over age as a continuous linear variable (12). The age categories in that study (1544, 4554, 5569, 7079, 80+) included an extra category for septuagenarian but were otherwise much comparable to the EMTRAS age categories. Initially described as an assessment tool for head-injured patients (13, 14), the GCS has become an essential component of trauma severity systems. In the revised trauma score (15), the GCS carries the most weight compared with the respiratory rate and systolic blood pressure. It has been suggested that physiologic data, like GCS, are often missing (16). In our study, data for GCS were missing in approximately 10% of the cases for center A, whereas there were no missings for center B. The

Figure 1. Receiver-operating characteristic curves together with areas under the curve estimates with 95% CIs for the Emergency Trauma Score in center A (closed circle) (n = 3,001 patients) and center B (closed square) (n = 1,417 patients). January 2014 Volume 42 Number 1

Clinical Investigations

Figure 2. Calibration plot of the Emergency Trauma Score in trauma centers A (continuous line, closed circle) and B (dashed line, closed triangle). Each dot represents expected and observed mortality for patients with similar Emergency Trauma Score (EMTRAS) (i.e., EMTRAS 0 and EMTRAS 1). The diagonal line indicates perfect calibration.

Figure 4. Calibration plot of the Emergency Trauma Score in trauma centers A (continuous line, closed circle) and B (dashed line, closed triangle) for patients with an Injury Severity Score less than 16. Each dot represents expected and observed mortality for patients with similar Emergency Trauma Score (EMTRAS) (i.e., EMTRAS 0 and EMTRAS 1). The diagonal line indicates perfect calibration.

GCS cannot be obtained reliably in intubated or obtunded patients. Using the motor component of the GCS (17, 18) only partly eliminates this problem as most intubated patients are sedated as well. In this study, we used the last GCS obtained before intubation.

Figure 3. Calibration plot of the Emergency Trauma Score in trauma centers A (continuous line, closed circle) and B (dashed line, closed triangle) for patients with an Injury Severity Score more than or equal to 16. Each dot represents expected and observed mortality for patients with similar Emergency Trauma Score (EMTRAS) (i.e., EMTRAS 0 and EMTRAS 1). The diagonal line indicates perfect calibration.

Both base excess and prothrombin time reect the lifethreatening condition called the lethal triad (19, 20). Base excess reliably reects acute acid-base derangements in trauma patients (21). An outcome prediction model for trauma patients incorporating base excess outperformed existing models in trauma patients admitted to the intensive care (22). Improved discrimination of models incorporating base excess was conrmed in a Dutch trauma population (23). Most early trauma deaths are related to uncontrolled hemorrhage (24). Coagulopathy is invariably related to hemorrhagic shock and is an independent predictor of mortality (25). Both laboratory variables were available in the majority of trauma patients, also in those who were less severely injured (approximately 90%). Modern analyzers can report prothrombin time within 30 minutes, which falls within the trauma resuscitation room stay of the majority of trauma patients. The strength of this study is that it addresses some of the shortfalls of the original EMTRAS publication as noted by Esposito (2) in his ne commentary. As suggested, the performance of EMTRAS in patients with an ISS less than 16 was investigated and imputation of key values was applied. EMTRAS is intended to be a prospective tool, and we envisioned that all potentially severe trauma patients would benet from early EMTRAS outcome prediction. Therefore, we choose to evaluate all adult trauma patients treated at the trauma resuscitation room. Additionally, subgroup analyses were performed for patients with an ISS below or above 16. Missing values in prognostic models lead to a reduction in statistical power and may lead to biased results (26). In our study, multiple imputation was applied on predictors necessary
www.ccmjournal.org

Critical Care Medicine

87

Joosse et al

TaBle 3. Original Coefcients and Reestimated Coefcients for Emergency Trauma Score Predictors in the Two Validation Cohorts
Validation Cohorts Predictor Original Cohort Center A p Center B p

Age Glasgow Coma Scale Base excess Prothrombin

a

0.0397 0.1722 0.0639 0.0334

0.038 0.308a 0.087 0.046

a

0.725 < 0.001 0.180 0.007

0.049 0.340a 0.093 0.035

0.103 < 0.001 0.196 0.769

Signicant change. p values indicate the difference between the original coefcient and the reestimated coefcient in the validation cohort.

for EMTRAS calculation, and therefore, no cases were lost for analysis. This study is the rst validation of the EMTRAS in an external database. How should the poor calibration for patients with an ISS below 16 be explained? Either there is a structural overestimation of mortality or an under registration of deceased patients. The possibility of under registration or incorrect registration of mortality could be excluded as the relevant administrative procedure is strictly monitored. Therefore, we sought to nd explanations in our data that contribute to a high EMTRAS but that are not related to an increased mortality risk due to trauma. A few cases of patients with epileptic seizures combined with minor injuries have been identied that can explain a low GCS that usually should not lead to a fatal outcome. Other causes of impaired consciousness not caused by neurotrauma are alcohol intoxication and hypothermia. Inevitably, some patients who had been sedated and intubated because of dangerous agitation will have been incorrectly coded with a GCS of 3. Finally, oral anticoagulant therapy was not evaluated in this study. In some cases, the prothrombin time may have been prolonged due to anticoagulant therapy and not as a result of traumatic hemorrhage or tissue injury. Several limitations need to be taken into account when interpreting the results of this study. Our primary aim was to validate the EMTRAS in an external dataset. About one third of the patients included in this study were treated in center B. The same center also contributes patients to the German Trauma Registry. The EMTRAS was originally developed in a cohort of the German Trauma Registry during the period 1993 until 2003. Therefore, theoretically we conducted a partly temporal validation of the EMTRAS. However, the contribution of center B to the original EMTRAS population was approximately 1%, so we consider this study to be a truly external validation. Inclusion criteria for patients judged as potentially severe trauma patients were not identical for both institutions. In center A, all patients presented at the trauma resuscitation room were included, whereas in center B, only code red patients were included. For patients not classied as code red, no full trauma team was activated and no extensive laboratory tests were routinely performed. Therefore, these patients were not eligible for EMTRAS calculation. This resulted in the inclusion of relatively more severely injured patients for center B. The
88
www.ccmjournal.org

ratio of ISS more than or equal to 16 versus ISS less than 16 was approximately 1:2 for center A and 1:1 for center B. For practical reasons, inclusion criteria for those amenable for EMTRAS outcome prediction should coincide with (eld) triage criteria or trauma team activation criteria. However, this study demonstrates that even in the presence of a national ambulance protocol that states criteria for level 1 trauma center referral, in-hospital trauma team notication criteria vary between institutions. This may become more problematic when these criteria will be applied to emergency medical systems in various countries with different prehospital triage protocols and in-hospital trauma team activation protocols (27, 28). Early outcome prediction using the EMTRAS of a broader range of trauma patients, including those with normal physiologic variables, would increase missing data for a selected group of patients. Presumably, many institutions will not routinely perform laboratory tests like base excess and prothrombin time on hemodynamically stable trauma patients. The clinical value of the EMTRAS has not been established with this study. In general, discrimination and calibration of the EMTRAS seems suitable for prognostication of trauma patients. The practical value of the EMTRAS as a tool for clinical decision making needs to be determined in further studies. The potential of the EMTRAS lies in the selection of patients who may benet of damage control surgery, who need admission to a monitored setting or in counseling patient and family.

CONCLUSIONS
This study validates the EMTRAS for early outcome prediction of trauma patients in two external cohorts. The performance of the EMTRAS, examined by discrimination and calibration, was excellent. Eligibility criteria for early outcome prediction with the EMTRAS based on the ISS is undesirable, as the anatomic nature of the ISS prevents early calculation during the resuscitation process.

ACKNOWLEDGMENTS
We thank Maarten W. Nijsten, MD, PhD, Department of Critical Care, and Jan B. Hulscher, MD, PhD, Department of Surgery, both from University Medical Center Groningen, for their kind cooperation in this study and their valuable comments on the manuscript.
January 2014 Volume 42 Number 1

Clinical Investigations

REFERENCES

1. Raum MR, Nijsten MW, Vogelzang M, et al; Polytrauma Study Group of the German Trauma Society: Emergency trauma score: An instrument for early estimation of trauma severity. Crit Care Med 2009; 37:19721977 2. Esposito TJ: Commentary on Raum et alEmergency trauma score. Crit Care Med 2009; 37:21222123 3. Baker SP, ONeill B: The Injury Severity Score: An update. J Trauma 1976; 16:882885 4. PHTLS: Prehospital Trauma Life Support. Sixth Edition. St. Louis, Msoby, PHTLS, 2007 5. American College of Surgeons Committee on Trauma: Advanced Trauma Life Support for Doctors. Chicago, IL, American College of Surgeons, 2008 6. Rubin DB, Schenker N: Multiple imputation in health-care databases: An overview and some applications. Stat Med 1991; 10:585598 7. Janssen KJ, Vergouwe Y, Donders AR, et al: Dealing with missing predictor values when applying clinical prediction models. Clin Chem 2009; 55:9941001 8. Moons KG, Donders RA, Stijnen T, et al: Using the outcome for imputation of missing predictor values was preferred. J Clin Epidemiol 2006; 59:10921101 9. Kuhne CA, Ruchholtz S, Kaiser GM, et al; Working Group on Multiple Trauma of the German Society of Trauma: Mortality in severely injured elderly trauma patientsWhen does age become a risk factor? World J Surg 2005; 29:14761482 10. Morris JA Jr, MacKenzie EJ, Damiano AM, et al: Mortality in trauma patients: The interaction between host factors and severity. J Trauma 1990; 30:14761482 11. McGwin G Jr, MacLennan PA, Fife JB, et al: Preexisting conditions and mortality in older trauma patients. J Trauma 2004; 56:12911296 12. Schluter PJ: Trauma and Injury Severity Score (TRISS): Is it time for variable re-categorisations and re-characterisations? Injury 2011; 42:8389 13. Teasdale G, Jennett B: Assessment of coma and impaired consciousness. A practical scale. Lancet 1974; 2:8184 14. Jennett B, Teasdale G, Galbraith S, et al: Prognosis in patients with severe head injury. Acta Neurochir Suppl (Wien) 1979; 28:149152

15. Champion HR, Sacco WJ, Copes WS, et al: A revision of the trauma score. J Trauma 1989; 29:623629 16. West TA, Rivara FP, Cummings P, et al: Harborview assessment for risk of mortality: An improved measure of injury severity on the basis of ICD-9-CM. J Trauma 2000; 49:530540 17. Healey C, Osler TM, Rogers FB, et al: Improving the Glasgow Coma Scale score: Motor score alone is a better predictor. J Trauma 2003; 54:671678 18. Eftekhar B, Zarei MR, Ghodsi M, et al: Comparing logistic models based on modied GCS motor component with other prognostic tools in prediction of mortality: Results of study in 7226 trauma patients. Injury 2005; 36:900904 19. Thorsen K, Ringdal KG, Strand K, et al: Clinical and cellular effects of hypothermia, acidosis and coagulopathy in major injury. Br J Surg 2011; 98:894907 20. Danks RR: Triangle of death. How hypothermia acidosis & coagulopathy can adversely impact trauma patients. JEMS 2002; 27:6168 21. Juern J, Khatri V, Weigelt J: Base excess: A review. J Trauma Acute Care Surg 2012; 73:2732 22. Kroezen F, Bijlsma TS, Liem MS, et al: Base deficit-based predictive modeling of outcome in trauma patients admitted to intensive care units in Dutch trauma centers. J Trauma 2007; 63:908913 23. de Jongh MA, Verhofstad MH, Leenen LP: Accuracy of different survival prediction models in a trauma population. Br J Surg 2010; 97:18051813 24. Sauaia A, Moore FA, Moore EE, et al: Epidemiology of trauma deaths: A reassessment. J Trauma 1995; 38:185193 25. Brohi K: Diagnosis and management of coagulopathy after major trauma. Br J Surg 2009; 96:963964 26. Sterne JA, White IR, Carlin JB, et al: Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls. BMJ 2009; 338:b2393 27. Rehn M, Lossius HM, Tjosevik KE, et al; Rogaland Trauma System Study Collaborating Group: Efcacy of a two-tiered trauma team activation protocol in a Norwegian trauma centre. Br J Surg 2012; 99:199208 28. Fung Kon Jin PH, van Olffen TB, Goslings JC, et al: In-hospital downgrading of the trauma team: Validation of the Academic Medical Center downgrading criteria. Injury 2006; 37:3340

Critical Care Medicine

www.ccmjournal.org

89