Interobserver reliability: κ ranged from 0.39-0.79 (overall agreement = 90%, p=0.000) (Juarez & Lyons 1995). For each component, % agreement ranged from 83.8% (eye opening-right) to 98.7% (best motor response-left) and agreement was lowest for the eye opening component. Correlations between observers ranged from 0.855 (motor response-right) to 0.974 (best verbal response)(Fielding & Rowley, 1990). Low rates of disagreement were found (disagreement rating=0.089-0.187) and motor responses elicited by supraorbital stimulus had higher rates of disagreement than fingertip stimulation (p<0.01) (Teasdale et al. 1978). % agreement ranged from 55% (verbal)-74% (eye opening), with Spearman’s rho=0.587 (verbal) and -0.742 (motor) and κw=0.48 (verbal) to 0.72 (eye opening) ((Gill et al., 2004); TBI). Fielding and Rowley (1990) reported reliability of 98.6-100% among experienced nurses, 94.3%-96.2% among new graduates and 77.3%-100% among groups of student nurses. In a study by Gujjar et al. (2013) with a population more than half consisting of individuals with neurological conditions, it was found that the GCS has good to excellent inter-observer agreement.

Internal Consistency: Gujjar et al. (2013) also found the GCS had good to excellent internal consistency (Chronbach α test = 0.815).

Construct Validity: GCS 13-14 is associated with greater proportion of abnormality on CT and longer duration PTA than GCS 15 ((McCullagh et al., 2001) (TBI). Depth of coma as assessed by GCS is considered to reflect extent of brain damage. In a 1996 review, Prasad cited the following studies: good correlation between GCS and cerebral metabolic rate (Langfitt & Gennarelli, 1982), correlation with CSF enzymes (0.82-0.99; (Bakay & Ward, 1983); TBI) and good correlation with evoked potential abnormalities (no stat given) (Lindsay et al., 1981) (TBI). Mean values of serum enzymes LDH1 and CPK1 correlated with GCS scores within 72 hours of injury (r=0.89 for both), incidence of multiple trauma also correlated with GCS scores (p<0.01) (Bakay & Ward, 1983).

Concurrent Validity: GCS scores correlated with length of coma (r2=0.233, p<0.0001) (Katz & Alexander, 1994) (TBI).

Predictive Validity: GCS 13-14 and GCS 15 (mild head injury)were not predictive of neuropsychiatric outcome 6 months post-injury (McCullagh et al., 2001). On multiple regression, GCS was identified as a significant independent predictor of death (p<0.0001). However, the prognostic value of the GCS was noted to be affected by mechanism of injury and age of the patient ((Demetriades et al., 2004); TBI). Based on 10 years of head injury data (1992-2001), the GCS was significantly correlated with the GOS at 6 months post injury for each year of data from 1992-1996, but from 1997-2001, no significant association was reported ((Balestreri et al., 2004); TBI). GCS was predictive of survival (AUC=0.891) but only slightly more than the motor component score on its own (AUC=0.873), while the eye opening score did not add to the predictive accuracy of the GCS (Healey et al., 2003) (TBI). In predicting mortality, there was a significant association between total GCS scores and outcome such that on multivariate analysis, the motor and verbal components were associated with mortality while eye-opening was not. Additionally, among patients with total GCS >9, only the verbal component was significant on multi-variate analysis, whereas for patients with GCS ≤9, motor and verbal component scores were significantly associated with mortality and verbal score was a better predictor than motor score in this group (Teoh et al., 2000). Initial GCS scores were significantly associated with employment status at one year post-injury (p<0.05) (Cifu et al., 1997). Initial GCS was significantly associated with DRS scores, LCFS scores, FIM-motor and FIM-cognitive at admission to and discharge from rehabilitation, though correlations were low to moderate (r=0.16 to 0.37; all p<0.0005) (Zafonte et al., 1996)( TBI). Waxman et al. (1991) reported that, when taken immediately on arrival at hospital, reported correlations between GCS scores and GOS scores (r2 =0.16) as well as length of hospital stay (r2=0.08), length of intensive care stay(r2=0.05) and duration of ventilatory support (r2 =0.03) were low. However, correlations between GCS taken at 6 hours after hospital arrival and GOS scores were much stronger (r2=0.55). GCS assessed at 6 hours and change in GCS contributed significantly to the prediction of GOS (r2=0.71-model included GCS 6 hours, Initial severity score, number of abnormal CT findings & change in GCS score). GCS was predictive of GOS at 6 months (r2=0.135, p<0.001) but much less so at 12 months (r2=0.81, p<0.005) (Katz & Alexander, 1994). 95% of patients scoring higher than 7 on initial GCS had favourable GOS outcome, while 95% with GCS lower than 5 had unfavorable outcome. Prediction of outcome for patients with initial GCS of 5,6,7 was more difficult and 24 hours GCS scores were preferable among these middle band patients when patients had either improved or deteriorated into the range in which predictions were more accurate (Young et al., 1981). Best day 1 GCS was a significant predictor of 6 month outcome on the DRS (p<0.05; (Pastorek et al., 2004); TBI). Initial GOS was reported as a significant predictor of GOS score 6 months post injury (p<0.001) (Satz et al., 1998) (TBI). Pre-resuscitation GCS score correlated with survival for head injury patients (r=0.978, p<0.0001) and with functional outcome as assessed by the FIM at discharge (r=0.973, p<0.0001) (Udekwu et al., 2004) (TBI).