Using Glasgow Coma Score at Admission to Predict One Year Mortality in Injured Admitted to the Trauma Centre of a North Indian State Lacking an Organized System of Trauma Care

Aim: To investigate whether Glasgow Coma Score at admission can be used to predict one year mortality in a resource constrained scenario lacking prehospital care. Study Design: Prospective observational study Place and Duration of Study: Study was conducted at the King George Medical University trauma centre between January 1 st , 2010 to December 31 st , 2010 Methodology: Information regarding factors known to affect mortality was recorded. Patients were Original Research Article Verma et al.; ARRB, 7(4): 241-248, 2015; Article no.ARRB.2015.125 242 followed up till death or for a period of one year since injury. Patients that died due to reasons other than injury and those with incomplete data sets were excluded from analysis. Three patients died due to reasons other than injury. Logistic regression, actuarial survival analysis, Cox proportionate Hazard model were used to identify predictors of one year mortality. Results: We enrolled 572 patients (478 males, 94 females, mean age 40.81±16.3 years). One hundred forty three patients died during follow up (140 due to injury and 3 due to other reasons). Thirty six patients had incomplete data sets. Kaplan Meir survival curve showed two distinct phase of mortality namely before and after 6th day of injury. Mortality within six days of injury was predicted by age, ISS, APTT, Glasgow Coma Score at admission and cervical spine injury. Mortality after 6 day was predicted by Glasgow Coma Score at the time of admission. Last death occurred 8 weeks after injury. Conclusion: Glasgow coma score at the time of admission is a valid predictor of one year mortality in trauma care systems lacking in pre-hospital care and an organized system of trauma care. Patients with low Glasgow Coma Score at admission continue to die 8 weeks after injury and therefore should be managed in the hospital for at least 8 weeks.


INTRODUCTION
The Glasgow Coma Score was first introduced in the 1970s with the objective of providing a simple, practical and reliable method to record the level of consciousness of patients and monitor change [1] without resorting to invasive or specialist techniques. GCS was developed to standardize the reporting of neurologic findings and to provide an objective measure of the level of function of comatose patients [2]. The original intention behind the Glasgow Coma Score was to describe the level of consciousness 6 hours post-head injury in order to monitor the patient and prognosticate the outcome. However it is commonly used to predict patient outcome following trauma. A number of studies have been published that describe the validity of the admission or Emergency Department collected Glasgow Coma Sore as a predictor of survival post-trauma [3][4][5][6][7][8]. Glasgow coma scale score has been demonstrated to predict two week mortality [9]. In a resource constrained scenario with inadequate number of trauma beds available at the only trauma centre of Uttar Pradesh, many patients are discharged as soon as the post injury life threatening period is over. Since the effects of traumatic brain injury on mortality are known to last for many years after the initial insult, there is possibility of them subsequently dying at home [10]. This effect on mortality in TBI patients is maximal in the first year after injury. This study was conducted prospectively as a part of MD-Phd thesis entitled "Predictors of one year mortality in adult injured patients admitted to the trauma centre of CSSMU (Now KGMU)" The objective of this study was to investigate whether Glasgow Coma Score can be used as predictor of one year mortality in the injured.

Study Area
The study was conducted in the Trauma Centre, KGMU.

Study Design
Prospective Observational Study.

Population
Source Population -Adult Injured patients from Lucknow and adjoining areas. Target Population -Adult injured patients from Lucknow and adjoining areas. Sampling Frame -Adult injured patients admitted to CSMMU trauma centre.

Sample Size
Two hundred and sixty patients are required for the necessary power using the thumb rule of 10 patients per variable studied but to increase the representativeness of the sample, and to check the representativeness of our sample we consecutively recruited, subject to written informed consent, patients admitted on Mondays and 8 randomly selected Wednesdays and Saturdays. It was planned that if patients admitted on different days of a week are similar, they shall be included in the final analysis.

Inclusion Criteria
Injured patients more than 18 years of age of either sex admitted to the trauma centre of CSMMU.

Exclusion Criteria
Injured patients less than 18 years of age and burn patients admitted to the trauma centre were excluded All adult injured patients fulfilling the inclusion criteria, subject to written informed consent and were assessed for the following factors -1) Age 2) Sex 3) Preexisting chronic conditions -coronary artery disease, Chronic obstructive pulmonary disease, hypertension, renal disease, diabetes mellitus, malignancies and any other disease 4) Injury Severity Score 5) Glasgow coma score at the time of admission 6) Referral from peripheral hospital 7) Blood sugar level at admission 8) Hypotension at admission 9) Coagulopathy at admission -Activated Plasma Thromboplastin Time 10) Blood transfusion -number and time to 1 st blood transfusion 11) Iatrogenic mistakes -Due to lack of an ongoing audit and record keeping there was no means to asses patients for aetrogenic mistakes other than missed injuries. 12) Number of Surgeries performed on the patient 13) Socioeconomic status -Below poverty line card 14) Time since injury to admission to trauma centre GCS score at admission was assigned by the casualty medical officer who is the first person to attend to an injured at KGMU trauma centre. If intubation was required, it was done after assigning the GCS score.

Statistical Analysis
A standardized questionnaire in which item analysis had been done for inter-observer and intra-observer disagreement was used to collect data. Data was collected on Microsoft XL and stata was used for analysis. Steps for analysis included descriptive analysis, bivariate analysis, and logistic regression analysis to identify predictors of one year mortality. Patients with incomplete data sets and those that died due to reasons other than injury were excluded from the analysis. Time trend analysis of mortality was done using Actuarial Survival Analysis and Cox Proportionate Hazard Model to identify the factors that predict one year mortality.

RESULTS
During the study period, 592 eligible patients were admitted to the KGMU trauma centre. Of these 572 consented to the study. Fifty seven percent (327/572) of patients were referred admitted and 43% (245/572) were directly admitted patients. Patients were predominantly male (83.5%) with median age 38 years (range 18 years to 92 years), and a median ISS of 9.
One hundred thirty patients (24.40%) had a Glasgow Coma Score less than 9. Fifty one patients (9.56%) had a Glasgow Coma Score between 9 and 12. Three hundred fifty three (66.25%) had a Glasgow Coma Score ranging between 13 and 15. Mean time to admission to trauma centre after sustaining injury was 54.22±185.2 hours. The demographic and clinical characteristics as well as mortality of the patients admitted on different days of the same week were found to be similar Table 1. During follow up one hundred forty three patients died (24.96%) (3 due to unrelated cause and hence excluded from analysis).
Of the 569 patients, Glasgow Coma Score was not recorded in 35 patients due to quadriplegia (26 patients) or injury to maxillofacial region (9 patients). Data on ISS was not recorded in 1 patient. In one of patients with maxillofacial injury (in which Glasgow Coma Score was not recorded), time to admission since injury was also not recorded. Patients with incomplete data set were excluded from analysis Bivariate analysis identified higher age in years (p=0.005), a higher ISS (p=0.000), higher respiratory rate (RR) at admission (p=0.000), lower systolic (p=0.000) and diastolic blood pressures (p=0.000) at admission, higher time to first blood transfusion (hours) (p=0.0005), lower GCS (p=0.001) and higher duration of hospital stay (p=0.001) to be significantly associated with patients that died. Presence of CAD (p=0.0010), COPD (p=0.0171), hypertension (p=0.003) coagulopathy (p=0.000) was found to be significantly associated with mortality. Multivariate analysis of data on 533 patients using logistic regression and cox proportionate hazard analysis identified age, ISS, Glasgow Coma Score, raised APTT, and systolic blood pressure at admission to be valid predictors of one year mortality. Abnormal respiratory rate (> 26 or ≤ 10) and presence of cervical spine injury were found to be significant predictors on logistic regression but not significant on cox proportionate hazard analysis. Bootstrapping of the logistic regression model and cox proportionate hazard model identified age, ISS, APTT, and Glasgow Coma Score at admission to be significant predictors of one year mortality. However systolic BP and cervical spine injury were found to be insignificant on bootstrapping.
Kaplan Meier survival curve showed two distinct phases of mortality, namely within 6 days of sustaining injury and another after more than 6 days of sustaining injury Fig. 1.
Of the 140 deaths, 86 occurred within 6 days of sustaining injury while 54 occurred after 6 th day.
Week wise distribution of mortality is shown in Table 2. Since mortality showed two distinct phases, predictors for both the phases were identified using logistic regression. On logistic regression, mortality within six days of injury was found to be predicted by age, ISS, APTT, and Glasgow Coma Score at admission. Mortality after 6th day was found to be predicted exclusively by Glasgow Coma Score at the time of admission. Bootstrapping of the model for mortality after 6 day confirmed Glasgow Coma Score to be the only valid predictor of mortality after 6 Table 3 summarizes the injuries in the 54 patients that died after 6 th day of injury. Abnormal Glasgow Coma Score was seen in 45 of these patients (38 due to head injury  Since mortality showed two distinct phases, predictors for both the phases were identified using logistic regression. On logistic regression, njury was found to be predicted by age, ISS, APTT, and Glasgow Coma Score at admission. Mortality after 6th day was found to be predicted exclusively by Glasgow Coma Score at the time of admission. Bootstrapping of the model for mortality after 6 th firmed Glasgow Coma Score to be the only valid predictor of mortality after 6 th day. Table 3 summarizes the injuries in the 54 day of injury. Abnormal Glasgow Coma Score was seen in 45 of these patients (38 due to head injury, 7 due to hypoxemia and shock either due to chest injury or injury to extremities). In five patients of cervical cord injury the Glasgow Coma Score could not be recorded. Normal Glasgow Coma Score was recorded in 4 patients (1 intertrochanteric, 1 fracture neck femur, 1 massive laceration of urinary bladder and 1 thoracic cord injury). Thirty five of the thirty eight head injury patients had a Glasgow Coma Score < 9. None of these (including the referred) were intubated prior to arrival to the trauma centr the 54, 32 arrived on 2 nd day after injury, 11 arrived on 45 th day after injury and 10 arrived on 7 th day after injury. In one patient time since injury to admission to trauma centre was not recorded. hypoxemia and shock either due to chest injury or injury to extremities). In five patients of cervical cord injury the Glasgow Coma Score could not be recorded. Normal Glasgow Coma Score was recorded in 4 patients (1 e neck femur, 1 massive laceration of urinary bladder and 1 thoracic cord injury). Thirty five of the thirty eight head injury patients had a Glasgow Coma Score < 9. None of these (including the referred) were intubated prior to arrival to the trauma centre. Of day after injury, 11 day after injury and 10 arrived on day after injury. In one patient time since injury to admission to trauma centre was not Cumulative frequency %

DISCUSSION
Our results validate Glasgow Coma Score at admission to be a valid predictor of mortality in injured. Admission or Emergency department collected Glasgow Coma Score is a known valid predictor of survival post trauma [3][4][5][6][7][8]. These studies were done on diverse trauma populations and used a variety of statistical tools. The motor component of the Glasgow Coma Score recorded during pre-hospital care has been reported to be efficacious in predicting mortality in trauma victims [6].
In our study 54 deaths occurred after 6 th day. A significant finding of the study is that after 6th day of injury Glasgow Coma Score at admission is the only significant predictor of one year mortality. Glasgow Coma Score within 4 hours of injury has been reported to predict two week mortality [10]. In our study 110 patients (78.5%) died within two weeks of injury and another 30 died after two weeks up to 16 th week (87.14% mortality occurred within 3 weeks and 97.85% mortality occurred within 8 weeks). This suggests that the predictive value of Glasgow Coma Score at the time of admission extends beyond two weeks reported by other studies [10]. Predictive power of Glasgow Coma Score beyond 6 th day can be explained by a very high percentage of patients having an abnormal Glasgow Coma Score (45 out of 54) in this group of 54 that died after 6 th day, and a vast majority (38 of 54) of these being head injuries.
A very important finding of our study is that of these 54 deaths, 28 were head injury patients having a Glasgow Coma Score of < 8that arrived without any kind of airway management which is contrary to current recommendations. Secondary brain injury due to hypoxia, hypercarbia, hypotension hypocarbia, hypoglycemia, and hyperglycemia causes further injury to the brain, worsening the outcome. Lack of airway support and delayed arrival to the trauma centre can be the cause of delayed mortality. However we cannot conclude that on the basis of the results of our study.
Presence of paralysis, use of sedative medications and coexisting injuries may lead to not so accurate mortality prediction using Glasgow Coma Score [11]. Another impediment to using Glasgow Coma Score in mortality prediction is pre-hospital care incorporating advanced life support system protocols, use of paralytics in the field and during transport resulting in lack of accurate early postresuscitation Glasgow Coma Score [11]. These changes in pre-hospital management may account for the results reported in a study [11] which reported subsets of patients wrongly classified as severe. The same changes in prehospital management may account for significant correlation between Glasgow Coma Score and 6month Glasgow Outcome scale reported between 1992 and 1996 and subsequent loss of correlation for the period 1997-2001 [12]. Very good predictive power as demonstrated in our study may be due to lack of pre-hospital care, lack of use of sedatives and lack of intubation facilities in peripheral hospitals [13] and during transport [14].

LIMITATIONS OF THE STUDY
Glasgow Coma Score is an ordinal score and therefore the difference between unit scores is not consistent. The difference only states that the lower score is worse. Another problem with the score is that it is skewed towards the motor component. Summing up the three components assumes an equal weight age for the three components, thus leading to loss of information as the same score can be made with different combinations of the components. It has been reported that the patients with same score may have different mortality risks due to difference in the GCS profile making up the score [15]. Intermediate levels of scores (7 to 11) have been reported to have lower discriminative value [16]. But even a low GCS does not always predict the outcome of severe HT. Injured children in the absence of ischemic hypoxic injury with Glasgow Coma Scores of 3 to 5 have been reported to recover independent function [17]. Another problem with Glasgow Coma Score is lack of reliability of initial scores. These limitations of the Glasgow Coma Score can be addressed by correlating it to other mortality prediction scores, in addition to the model used in our study. A significant limitation of our study is that GCS score at the time of admission was not correlated against other well established outcome prediction scores used for traumatic brain injury.

CONCLUSION
Glasgow Coma Score at the time of admission is a valid predictor of one year mortality in trauma care systems lacking in pre-hospital care and an organized system of trauma care. Patients exhibiting a low Glasgow Coma Score at admission should not be discharged as soon as the immediate life threatening period is over. Rather, they should be managed in the hospital for a period of at least 8 weeks. Studies focusing on Glasgow Coma Score as predictor of mortality should consider following patients up to at least 3 weeks and preferably up to 8 weeks.