Select Page

Table 17.25 Characteristics of the Glasgow Outcome Scale and Extended Version

 
Criterion Evidence
Reliability

Test-retest: κ ranged from 0.40-0.92 for the GOS and 0.40-0.87 for the GOSE. However, the retest period was lengthy, ranging from 3-6 months (Maas et al., 1983).

Interobserver Reliability: Jennett et al. (1981)) reported 95% agreement between observers using the original GOS. Agreement between assessment based on a mail-administered research questionnaire and assessment via interview by a psychologist was reported to be r=0.79 while agreement between a GP’s assessment and the psychologist interview was r=0.49 (Anderson et al., 1993) (TBI). Based on live interviews, κ=0.77 for GOS and 0.48 for GOSE. When ratings were based on previously recorded data, κ=0.58 for GOS and 0.49 for GOSE, and agreement between live and recorded data ratings was κ=0.77 for GOS and 0.53 for the GOSE (Maas et al., 1983) (TBI). 70% of GOS ratings were in perfect agreement while none differed by more than one category, and for the GOSE none differed by more than one category, with the most discrepancy seen in the middle categories ((Brooks et al., 1986); TBI)

Validity

Construct Validity: GOS ratings have been reported to be associated with results of neurological testing of motor tasks (p<0.001), psychomotor tests (p<0.05), assessments of memory variables (p<0.05), and attention variables (p<0.05) such that neuropsychological test performance decreased as a function of increased severity on the GOS rating scale (Satz et al., 1998) (TBI). Performance on cognitive tests 3 months post injury differed significantly (p<0.05) between outcome subgroups corresponding to GOS rating, demonstrating a clear gradation in cognitive scoring between groups in the expected direction, and this relationship was not as clear when the GOSE was used (Brooks et al., 1986).

Construct Validity (Known Groups): GOS scores could discriminate between groups based on categories of vocational recommendations (return to work, vocational training, supported work and continued remedial therapy (p<0.0001). GOS scores accounted for 76% variance between cell means (Mysiw et al., 1989) (TBI).

Concurrent Validity: Admission DRS scores correlated with initial Stover and Zeiger (SZ) ratings (r=0.92), discharge DRS scores correlated with discharge SZ scores (r=0.81), GOS scores (0.80) and EGOS scores (0.85) (Gouvier et al., 1987) (TBI). GOS ratings correlated with SF-36 subscale scores (r=0.51-0.68, p<0.01) (Jenkinson C. (1993) cited in ; Teasdale et al. (1998)(TBI), and GOS scores correlated with DRS ratings at admission to (r=0.50, p<0.01) and discharge from rehabilitation (r=0.67, p<0.01) (Hall et al., 1985)

Predictive Validity: GOS at discharge from rehabilitation significantly correlated with GOS 5-7 years after head injury (r=0.60, p<0.001) and with discharge destination (p<0.0001) (Massagli et al., 1996) (TBI).

Responsiveness From assessment 3 months post injury to 6 month assessment, 36% of patients demonstrated change in GOSE ratings while only 11% demonstrated change in category based on GOS ratings (p<0.05) (Levin et al., 2001)(TBI). From admission to discharge from rehabilitation, improvement shown by the DRS was significantly greater than that shown by the GOS (71% versus 33%, p<0.01) (Hall et al., 1985)(TBI).
Tested for ABI/TBI patients? Specific to head injury populations.
Other Formats

Structured Interview for the GOS/GOSE (Wilson et al. 1998; TBI): Improves reliability and removes limitations associated with scale ambiguity and lack of guidelines for administration. This method specifies criteria for separating the upper and lower bands of the upper 3 categories of the GOS. The structured interview consists of a series of questions regarding consciousness, independence (both at and away from home), social roles (work, social activities, leisure, relationships) and return to normal life (Wilson et al., 2000). The questionnaire focuses on aspects of social disability (effects on social and leisure activities and disruption to family and friendships) as originally described by Jennett et al. (1981). The structured interview format also allows for the inclusion of pre-injury disability status (Wilson et al., 2000) (TBI) and provides specific guidance regarding assignment to outcome category (Teasdale et al., 1998).

Reliability: Agreement between raters was reported to be 92% for the GOS and 78% for the GOSE when administered via structured interview, κw=0.89 and 0.85 for the GOS & GOSE respectively (Wilson et al., 1998).

Validity: Significant correlation reported between BI and GOS (rho=0.61, p<0.001) and between DRS scores and GOS ratings (rho=0.89, p<0.001)(Pettigrew et al., 1998) (TBI). When using the structured interview, Wilson et al. (2000) reported correlations with BI scores of 0.47 and 0.46 for the GOS and GOSE, respectively. GOS and GOSE ratings also correlated with DRS ratings (r=0.89 for both), Beck Depression inventory scores (r=0.61 & 0.64), GHQ scores (0.57 & 0.59), MOS SF-36 subscores (ranging from 0.41-0.67 and 0.47-0.71) and Neurobehavioural Functioning Inventory (NFI) scale scores, ranging from 0.33-0.57 and 0.37-0.63 for patient NFI ratings and 0.47-0.68 and 0.47-0.69 for NFI ratings obtained from friends or relatives. Levin et al. ((2001); TBI) reported that at 3 months post injury, GOS ratings were significantly associated (p≤0.05) with results on the CES-D, CIQ, Social Support questionnaire, and the paced auditory serial audition test (trial 1). GOSE ratings were significantly associated with results from the CIQ and the Paced auditory serial audition test (trial 1). In cases where both demonstrated linear association with scale scores (i.e. CIQ and the paced auditory serial audition test) GOSE ratings accounted for more of the variance in scale scores than GOS ratings (r2=0.35 versus 0.26 and 0.37 versus 0.19, respectively).

Telephone Administration (Structured Interview): Agreement between face-to-face-interview and telephone interview was reported to be κw= 0.92 for the GOSE. When GOSE scores were collapsed to GOS ratings, κw=0.92 and interobserver agreement was reported to be κw=0.84 and 0.85 ((Pettigrew et al., 2003); TBI).

Simple Postal Assessment (Hellawell et al. 2000; TBI): Using a simple, 4 question survey, inter-observer (GPs, family informants, experienced GOS raters) reliability was reported to range from κ=0.17 (between GP and experienced rater) to 0.61 (between GP’s and family informants).

Postal Questionnaires-based on the Structured Interviews for GOS and GOSE (Wilson et al. (1998) ((Wilson et al., 2002); TBI):-designed to be completed by the patient or a relative or caregiver of the patient or by the patient with the assistance of a significant other/caregiver. Questions are intended to discriminate between the categories of severe disability, moderate disability and good recovery (for the GOSE questionnaire, these are further subdivided into upper and lower bands). Return rates were reported to be 76% for the GOS questionnaire and 83% for the GOSE questionnaire. Test-retest reliability for the GOS was reported to be κw=0.94 and κw=0.98 for the GOSE. Agreement between GOS ratings assigned via postal questionnaire and telephone interview (using the structured interview) was reported as κw=0.67 while agreement using the GOSE questionnaire was higher (κw=0.92)

Edinburgh Extended Glasgow Outcome Scale ((Hellawell & Signorini, 1997) ; TBI): This scale is based on the GOS, but requires scoring for behavioural/emotional, cognitive, and physical functioning. Each patient is assigned a rating on each of these types of function. Descriptions are provided for each of the function types. Using retrospective data, interobserver agreement was reported as κ=0.20-0.55 for behavioural ratings, κ=0.56-0.63 for cognitive ratings, and 0.57-0.75 for physical ratings. Using current data, interobserver agreement for behavioural, cognitive and physical ratings was reported as κ=0.61, 0.62 and 0.73, respectively (Hellawell & Signorini, 1997).

Use by proxy? It is recommended that the best source of information be used, and that whenever possible, the information gained by interviewing close friends or family members be included (Wilson et al., 1998). Using a simple postal survey, GP’s and informants tended to rate patient outcome more positively than experienced GOS raters (Hellawell et al., 2000).