Figure shows the inclusion procedure of the TBI patients for both cohorts. In the derivation study 10 healthy, 10 orthopedic (upper or lower leg fracture) and 11 neurological controls (10 patients suffering from a sensorimotor ischemic stroke without aphasia and 1 patient with a benign cerebral tumor) were included. For the validation study we recruited 20 healthy individuals and 10 patients with traumatic orthopedic injuries. The different control groups were collapsed into one single control group per cohort. Table demonstrates the demographic and clinical characteristics of TBI patients and control subjects. Within the derivation study, patients and controls differed considerably in gender and age (Table a). For the validation cohort differences were less apparent except for the age difference of nine years between mild TBI patients and controls (Table b). The first PTA assessment in the moderate/severe patients included in the derivation cohort was executed at a mean of 17,3
days (standard deviation [sd] 14,9) post-injury, for the validation cohort the mean was 8,6 (sd 10,1) days.
Inclusion flow diagram of the derivation cohort (Cohort I) and the validation cohort (Cohort II). Sep., September; Oct., October; Dec., December; TBI, traumatic brain injury; mTBI, mild traumatic brain injury; Mod., moderate; Sev., severe.
Table shows the sensitivity and specificity of individual items administered to the derivation cohort on test days one and two. Moderate/severe TBI patients performed poorer on all items compared to controls except for name, date of birth, residence, marital status, offspring and occupation (not shown in Table ). In mild TBI, only time of day (margin 30 minutes) on day one and mode of transport on days one and two showed a significant difference between patients and controls.
Discriminative value: Sensitivity and specificity of individual test items on first two days of admission
No significant differences in test performance on the memory items were found between the picture and word groups (data not shown). Both item groups were combined for further analyses. The 24-hours free recall and recognition scores did not differ significantly between mild TBI patients and controls (Table ), whereas moderate/severe TBI patients performed significantly poorer than controls. The 24-hours free recall of memory items proved to be more difficult than recognition for both patients and controls, as demonstrated by higher sensitivity and lower specificity values.
A correlation was found between name of hospital
and city of hospital
0.61; p <0.0001) and month
, day of week
and period of day
(0.50 <r <0.56; p <0.0001). Not surprisingly, date
and date of admission
showed a correlation in the mild TBI patients (r
0.67; p <0.0001).
For our PTA scale (Table ), to be validated in the second part of this study, we selected the items age, name of hospital, time, day of week, month, mode of transport and recall of three words after 24 hours based on the following arguments. For the use of a new scale at an ED, we chose items with discriminative value from at least day one in mild TBI patients: time and mode of transport. Both items showed high specificity (>95%). We selected age since it showed significant differences in sensitivity and specificity in moderate/severe TBI patients on at least two consecutive days.
The items name of hospital, day of week and month were selected because of high specificity and significant differences in test performances between moderate/severe TBI and controls. Year demonstrated discriminative value only on day one, hence it was excluded. Items date and admission date were not selected since specificity was low (≤80%). Period of day was not selected because it correlated with time. Finally, city of hospital was excluded as it correlated with name of hospital and proved significant only on day one in the moderate/severe TBI patients.
To test anterograde amnesia, we selected for our new PTA scale words instead of pictures as memory items, as they were not inferior to pictures and might be more practical, especially at ED settings. We preferred recall rather than recognition based on its superior sensitivity.
Table demonstrates the sensitivity and specificity of each individual item of our proposed PTA scale as found in the validation cohort on the three primary administrations (TBI patients and control subjects grouped together). Additionally, data on three-word recognition are given. In the validation cohort specificity of the item time of day was lower at the first (86% versus 96%) and 24-hours administration (89% versus 100%). The other items were never failed by the control subjects.
Sensitivity and specificity of each individual item of the proposed PTA scale
The Area Under the Curve (AUC) of free recall was larger than that of recognition. Consequently, free recall demonstrated better discriminative values: First administration: 0.71 (95% confidence interval [CI]: 0.62-0.79) versus 0.62 (95% CI: 0.52-0.72); second administration: 0.76 (95% CI: 0.61-0.92) versus 0.60 (95% CI: 0.42-0.77); third administration: 0.87 (95% CI: 0.79-0.96) versus 0.71 (95% CI: 0.58-0.84).
Performances on our proposed PTA scale and existing PTA scales showed strong and significant correlations, except for the modest, nevertheless significant, correlations between performances on our PTA scale and the GOAT (Table ).
Spearman correlation coefficients demonstrating the relationship between our proposed PTA scale and existing PTA scales
Table compares the sensitivity and specificity of the proposed PTA scale with current PTA scales, based on the sum scores of each individual scale (Table ). Our new PTA scale had largely comparable AUC’s to the existing PTA scales.
Sensitivity, specificity and AUC’s (ROC analysis) of our proposed PTA scale and existing PTA scales