Journal of Rehabilitation Medicine 51-5 | Page 61

Concordance between self-awareness measures ficits (subscale 1), functional consequences of impair- ments (subscale 2) and realistic goal-setting (subscale 3). Subscale scores range from 0 (no impairment) to 3 (severe ISA), resulting in a total score of 0–9. The SADI assesses both the presence and nature of ISA across diverse areas of functioning. In a cluster analysis of long-term community outcomes after brain injury (6), SADI scores ≤3 distinguished between poor self- awareness (mean 3.9, standard deviation (SD) 1.7) and good self-awareness (mean 1.4, SD 1.1) groups, with the former group experiencing significantly poorer global psychosocial outcomes. Yet, administration time (20–30 min) may reduce feasibility of use in clinical practice. The Awareness Questionnaire (AQ) (7) is a brief (5–10 min) questionnaire that assesses ISA by compa- ring self-ratings and significant-other ratings of changes in sensory/motor, cognitive and behavioural/affective functions. In an inpatient sample, Sherer et al. (7) used logistic regression to identify cut-off points for ISA on the AQ, based on post-discharge employment status. They proposed the following discrepancy sco- res and clinical interpretations: < 20 mild or no ISA, 20–29 = moderate ISA, and > 29 = severe ISA. Given that most individuals demonstrate some level of ISA early after TBI (8), large discrepancies may be warran- ted to reliably indicate ISA in an inpatient setting. To guide assessment of self-awareness in the community this study aimed to examine concordance between the AQ and SADI for identifying ISA in individuals with long-term TBI. METHODS A file audit was conducted to retrieve AQ and SADI data for TBI participants involved in research on psychosocial outcomes (9) and rehabilitation (10, 11). Following ethics approval from hospital and university ethics committees, 150 participants with moderate- to-severe TBI (post-traumatic amnesia (PTA) > 24 h or Glasgow Coma Scale (GCS) score < 13) were recruited from metropolitan- based brain injury outpatient rehabilitation and community-based services (2004–17) across 3 studies. Of these, 30 were excluded due to non-traumatic aetiology (e.g. stroke), 15 sustained mild TBI, and 25 were missing significant-other data on the SADI and/or AQ. Eighty individuals with moderate-to-severe TBI, aged 18–57 years (M = 32.25, SD = 11.5 years), had previously completed the AQ and SADI within the same assessment session and were included in this study. They were predominantly male (81%) and the mean time since injury was 3.60 years (SD 2.9; range 0.25–10 years). The main cause of TBI was motor vehicle or other traffic accidents (65%), followed by falls (16%), sporting or work accidents (11%) and assaults (8%). The mean length of PTA was 59.13 days (SD 108.4 days) and GCS was 6.92 (SD 4.2). The AQ (7) asks respondents to compare a person’s abilities before and after TBI on 17 items (1 = much worse to 5 = much better), producing a total score of 17–85. Positive discrepancy scores (self-ratings minus significant-other ratings) indicate 377 under-reporting of impairments. The SADI (5) was administered by trained researchers who scored the interview independently of the AQ. A significant other (predominantly family members) completed a checklist concerning the person’s post-injury chan- ges to assist with scoring. Statistics Data were analysed using the Statistical Package for the Social Sciences (SPSS), version 24. Concordance between the SADI and AQ was examined using Spearman’s rho and by construc- ting receiver operating characteristic (ROC) curves to examine consistency in classification of ISA at different scores on each measure. For these analyses the total score on the SADI ≥ 2 to ≥ 6) was selected as the state variable due to the smaller range of possible scores than the AQ and previous research supporting the concurrent validity in the community setting (6). Sensitivity (true-positive rate) represents the proportion of individuals clas- sified as having ISA on the SADI and correctly identified by the AQ. Specificity (true negative rate) is the proportion classified as not having ISA on the SADI (< 3/9) and correctly identified by the AQ. Consistency of classification of ISA between the AQ and SADI was calculated, with scores > 0.80 considered “good” (10). RESULTS SADI total scores ranged from 0 to 9 (mean 2.5, SD 2.09) and AQ discrepancies ranged from 18 to –34 (mean 5.49, SD 9.96). Scores on the AQ and SADI were strongly associated (rho = 0.65, p < 0.001). Table I presents the ROC curve data for the SADI (state va- riable: ≥ 2 to ≥ 6) and AQ discrepancy scores ranging from ≥ 2 to ≥ 13. Scores in bold font represent the opti- mal balance between sensitivity and specificity. Rates of ISA were comparable at corresponding scores on the 2 measures, as follows: SADI ≥ 2~AQ discrepancy ≥ 3 (56–60% with ISA); SADI ≥ 3~AQ discrepancy ≥ 4 (45–48% with ISA); SADI ≥ 4~AQ discrepancy ≥ 9 (26–34% with ISA); SADI>5~AQ discrepancy ≥ 12 (13–26% with ISA). Table II presents the consistency in classification of ISA between the SADI and AQ, based on scores with optimal levels of sensitivity and specificity. SADI and AQ scores of ≥ 2 yielded very high rates of ISA (59–60%) and classification consistency was only 76.3%. Sensitivity, specificity and classification consistency improved to 80% for SADI ≥ 3 and AQ ≥ 4, with 45% and 48% classified as having ISA on each tool, respectively. As shown in Fig. 1, the area under the curve (AUC) in the ROC curve analysis was 0.88 (95% confidence interval (95% CI) 0.80~0.96). Clas- sification consistency was also good at higher scores on each measure, as follows: SADI ≥ 4~AQ discrepancy > 9 (83%); SADI > 5~AQ discrepancy ≥ 12 (84%). Slightly higher rates of ISA were identified by the AQ at each corresponding SADI score. J Rehabil Med 51, 2019