Results

FACTOR ANALYSIS OF THE NEUROBEHAVIOURAL RATING SCALE: REVISED

Factor analysis of the NRS-R was performed using the maximum likelihood (ML) procedure with oblique rotation for the total sample (n=210). The relation between sample size and number of variables was sufficient with more than five patients/variable.41 There were no missing data points. Review of the literature regarding previous factor analyses of the NRS suggested that the number of factors necessary to adequately characterise the major components ranged between four1 and six.11 Additionally, Vanier et al 18 reported a five factor solution in their sample for the French version of the NRS-R.

The current ML analysis yielded five factors for which scores were formed by summing the ratings of items with loadings ⩾0.38 on a given factor. This correlation criterion is in line with Stevens42 who suggests adjusting the minimum level of significance of factor loadings according to sample size rather than an arbitrary 0.3 or higher. The value at α=0.01 for a significant two tailed correlation is ⩾0.38 in a sample of 210 patients. The five factor model accounted for 93% of the total variance.

As shown in table 2, factor I (executive/cognition) comprised items which included difficulties in mental flexibility, difficulties in planning, difficulties in oral comprehension, inaccurate self appraisal, and memory difficulties. Factor II (positive symptoms: distortions or exaggerations of normal functions) was made up of items such as unusual thought content, suspiciousness, hyperactivity/agitation, and disinhibition. Factor III (negative symptoms: a diminishment or absence of functions normally present) was derived from items that reflected emotional withdrawal, blunted affect, and hostility. Factor IV (mood/affect) included items measuring depressive mood, anxiety, irritability, mental fatigability, and lability of mood. Factor V (oral/motor) consisted of items rating difficulties in articulation, difficulties in oral expression, and motor slowing.

Three items loaded significantly on two factors. These multiple loading items included “mental fatiguability” (mood/affect and oral/motor), “hostility” (positive symptoms and negative symptoms) and “lability of mood” (positive symptoms and mood/affect). The internal consistency (Cronbach's coefficient α) of the factors was acceptable ranging from 0.62 (negative symptoms) to 0.88 (executive/cognition). The interfactor correlations were minimal to moderate ranging from 0.19 (positive symptoms and oral/motor) to 0.61 (executive/cognition and oral/motor). Four items did not load significantly on any factor including “attention,” “reduced alertness,” “guilt,” and “hallucinations.” Although these items are not part of the five factor solution, the clinician may wish to retain these items in the patient's examination as they provide important information which may have clinical relevance for specific patients.

For descriptive purposes, the frequency distribution of the NRS-R ratings is presented in table 3. Deficits found in at least 50% of the patients were generally from the items on the executive/cognition factor, including difficulty in planning, mental flexibility, impaired self appraisal, and conceptual disorganisation. In addition, over 50% of the sample demonstrated evidence of some degree of depressive mood. The NRS-R total score (simple sum of the 29 item scores) is included in the following analyses to demonstrate the utility of the NRS-R as a single composite measure of neurobehavioural disturbance in addition to the factor scores.

RELATION OF DEMOGRAPHIC VARIABLES AND INJURY SEVERITY TO NRS-R SCORES

There were no significant effects of age, education, or sex or any interactions of these variables on the NRS-R total score or any of the five factor scores. The NRS-R total score and all five factor scores obtained at 6 months after injury were correlated with the admission (postresuscitation) GCS score. Modest but significant associations were found between the GCS score and the NRS-R total score (r=−0.20, p=0.005), executive/cognition (r=−0.18, p=0.01), positive symptoms (r=-0.17, p=0.01), and oral/motor factor scores (r=−0.24, p =0.0008).

RELATION OF GLOBAL OUTCOME SCALES TO NRS-R FACTOR SCORES

One way analysis of variance (ANOVA) for group differences (based on groups formed by GOS ratings at 6 months after injury) indicated significant differences for the NRS-R total score (F (2,207)=116.5, p<0.0001) and each of the five factor scores (all at p<0.0001). Post hoc comparisons (using Tukey's procedure to control type I error) between these groups (1v 2, 1 v 3, and 2v 3) indicated significant differences in the NRS-R total score for each of the three comparisons. In a similar fashion, one way ANOVAs were calculated within each of the five factor scores for between group comparisons (fig 1). Significant differences were found for the three between group comparisons for the executive/cognition, mood/affect, and oral/motor factors such that more severe dysfunction or deficit was associated with greater levels of disability measured on the GOS. Similar results were found for the positive symptoms and negative symptoms factor scores except for the good recovery versus moderate disability comparisons.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Mean comparisons between groups based on GOS ratings 6 months after injury (n=210). Higher scores indicate greater impairment. All between group comparisons within each factor are significant at p<0.017 unless otherwise indicated.

The NRS-R total score correlated significantly with concurrent GOS (r=0.72) and DRS (r=0.74) scores at 6 months. Similar significant correlations were found with each of the five factor scores and the GOS (r=0.28 to 0.74) and DRS (r=0.26 to 0.74) scores. This suggests that the NRS-R total score is sensitive to global outcome compared with these two widely recognised outcome measures.

VALIDATION WITH NEUROPSYCHOLOGICAL CONSTRUCTS

The NRS-R factor scores were correlated significantly with specific domains of neuropsychological functioning which included verbal35 36) and visual memory,37 38 speed dependent visuomotor tracking,25 39 40 manual dexterity,26 27 and speeded language production.24 Table 4 illustrates the correlations between the NRS-R factor scores and the neuropsychological domain factor scores. The strongest correlations of the mood/affect factor were with neuropsychological domains known to be sensitive to affective disorders (for example, depression) including memory and rapid visual-motor tracking.43-47

SENSITIVITY TO CHANGE DURING RECOVERY

To determine the sensitivity of the NRS-R to changes during recovery, the scores of patients who had completed the NRS-R at both the 3 and 6 month assessments (n=165) were compared. The mean difference between the NRS-R total score at 3 and 6 months was significant (p<0.0001). Additionally, pairedt tests comparing mean differences for each of the five factor scores disclosed significant differences for executive/cognition (p<0.0001), negative symptoms (p<0.006), and oral/motor (p<0.004). Only the positive symptoms and mood/affect factor scores did not change significantly between 3 and 6 months.

As a more stringent test of the sensitivity of the NRS-R, pairedt tests for differences in NRS-R scores were calculated for patients whose GOS score did not change from 3 to 6 months (n=105). Results indicated that the NRS-R total score changed significantly, as well as the executive/cognition and oral/motor factor scores (table 5). In this analysis, the proportion of those whose score changed on the NRS-R total score was 92.3%, whereas 81.9% changed on the executive/cognition factor score, and 60.0% changed on the oral/motor factor score. This dissociation implies that although a measure of global outcome (GOS) did not reflect a change in the patient's overall level of recovery, certain factors of the NRS-R were sensitive to changes over time in specific domains of neurobehavioural functioning.

Discussion

Clinician based instruments allow a trained clinician or psychometrician to make objective assessments of operationally defined classes of behaviour. This makes possible the tracking of behavioural recovery of patients with head injury and allows multiple centres to rate patient behaviour using a standard procedure in multicentre clinical trials. Ambiguity in making NRS-R ratings was reduced by specifically defining each rating level (absent to severe) for each item based on behaviours found during the semistructured interview. However, the clinician is only able to make ratings based on a relatively brief interaction with the patient. For example, the patient may have been hostile to hospital staff or had experienced hallucinations previously in the same day, but if these behaviours were not witnessed at the time of the interview, they would not be recorded in the ratings of the NRS-R. In a clinical setting the NRS-R as a clinician based instrument can be one source of information about a patient's pattern of improvement and residual behavioural deficits when combined with reports from relevant others and the patient's self appraisal. This provides the clinician with a basis from which to determine the accuracy of the patient and family perspective, which can be important for rehabilitation planning.

Instruments used in clinical trials must be able to measure outcome for all or nearly all patients enrolled in order to be viable. Using neuropsychological tests as a primary outcome measure is problematic as many severely impaired patients are not able to comply with all of the tests. In the current study, only 42.5% of the patients (in the study group) were able to fully complete the 2 hour neuropsychological battery 6 months after injury. In this study, the completion rate for survivors was 97.6% for the GOS and 96.1% for the DRS, and the completion rate for the NRS-R was 74.9%. Although neuropsychological data can be very useful in characterising deficits after closed head injury, other measures which require less cooperation from the patient are likely to be more useful in conducting clinical trials. For these reasons, the NRS-R is well suited as a secondary outcome measure in that it quantifies important features of neurobehavioural recovery and places minimal demands on the patient (compared with formal neuropsychological assessment) to carry out basic tasks of reasoning, temporal orientation, memory, language comprehension and expression, and mental calculation. Although certainly not a substitute for a comprehensive neuropsychological evaluation, the NRS-R provides complimentary information about behavioural disturbances which are otherwise difficult to quantify, and which neuropsychological testing alone does not provide.

Interrater reliability for the NRS-R was not consistently high across all items as reported in the study by Vanier et al 18 (for example, difficulty in planning). This issue becomes of greater concern as the number of examiners and study centres increases. However, training examiners using videotaped administrations of the NRS-R (initial study training workshops), annual review, observation of testing procedure, and stringent adherence to the scoring guidelines can facilitate maximisation of interrater and test-retest reliability.

At least modest convergent and divergent validity for the factor scores of the NRS-R was shown in that the neuropsychological domain scores were more substantially related to the executive/cognition factor than to other factors. Slightly less robust correlations were also found between the oral/motor factor and the specific neuropsychological domains of speeded visual-motor and manual dexterity abilities as would be expected. Further, the positive symptoms and negative symptoms factors were virtually unrelated to the neuropsychological domains.

Although the NRS-R has a strong factor reflecting executive and general cognitive functioning, this should not be used as a substitute for a detailed cognitive assessment of executive and intellectual functioning. This does not rule out the use of the NRS-R in pharmacological trials—for example, to screen for cognitive declines or improvements as a result of therapeutic agents or other interventions. The NRS-R factor scores could be used as a profile of cognitive, affective, and neuropsychiatric domains demonstrating areas of improvement or decline for specific types of neurobehavioural sequelae. It is possible that this could greatly improve the design of clinical trials which target specific types of outcome and expand the scope of characterising outcome by a global categorisation.

The NRS-R has demonstrated sensitivity to sequelae which contribute to morbidity. Items on the positive symptoms factor are important features of neurobehavioural outcome and have been shown to affect long term outcome after closed head injury.48 49 Additionally, behaviours associated with the negative symptoms and mood/affect factors were only moderately associated with GOS and DRS scores, but have long term implications for rehabilitation and community re-entry.50 51

There are striking differences between the factor solution for the NRS reported by Levin et al 1 and the current NRS-R. One possibility is the larger sample size of the current study. However, several other plausible reasons exist. Firstly, the NRS was validated on a heterogeneous group of patients with closed head injury encompassing the full range of injury severity and widely variable degrees of chronicity. Conversely, the current NRS-R study was based on a fairly homogenous sample of patients with severe closed head injury and tightly defined end points of 3 and 6 months after injury. Another possibility is the change in the Likert scale used to rate each item; the NRS-R has four rating levels ranging from absent to severe instead of the more complicated seven point scale of the NRS. Based on previous experience with the NRS and recovery from head injury, two new items were added: “irritability” and “difficulties in mental flexibility”. Both items represent important aspects of neurobehavioural outcome. Mental inflexibility is considered a hallmark symptom of frontal lobe dysfunction and has been hypothesised to be a potential obstacle to resuming adequate social functioning and employment.51-57 Similarly, irritability has often been documented in studies based on reports by relatives of patients recovering from severe closed head injury.58 Combining the items “tension” and “anxiety” to form “anxiety” seems to be an improvement given the ambiguity of the tension item and the weak loading it had on the somatic/anxiety factor of the NRS. The division of the “inattention” item into separate components of “reduced alertness” and “attention” may have had an effect on the factor structure of the current solution by distinguishing impaired alertness or general arousal from focused concentration on a task. Both features are important considerations, but “reduced alertness” is likely to be more meaningful at early stages of recovery whereas “attention” may have greater implications in the postacute phase.

Two minor differences were noted between the results of the current factor analysis and those reported by Vanier et al.18 These differences likely stem from two principal sources. The first is that, as in the study by Levinet al,1 the full range of head injury severity was included in the. investigation of Vanieret al as opposed to a homogeneous sample of severe head injuries in the current study. Secondly, both the original study of Levin et al and that of Vanieret al had wide ranging time points after injury when assessments occurred. In the current study, all patients were tested within relatively narrow windows at about 3 and 6 months after injury. Despite the differences in severity of closed head injury and follow up intervals after injury, the factor solutions found in this multicentre clinical trial and those reported by Vanieret al are quite similar. For instance, the executive/cognition factor has seven out of eight of the items loading on the intentional behaviour factor of Vanier et al. The positive symptoms factor is identical to that of Vanieret al except for the inclusion of “irritability”. The current results generally seem to confirm the factor solution reported by Vanier et al.

Given the differences between the factor structure in this study and the one reported by Vanier et al, we suggest using the following guidelines for the appropriate use of the NRS-R factor scores, as both solutions are useful. Studies investigating the full range of head injury severity and heterogeneous chronicity would be advised to use the factor solution reported by Vanieret al.18 Studies of a homogeneous group of severe head injuries using about 6 months as a consistent end point should consider using the factor structure reported in this study.

The factor structure reported here fits well with the known major cognitive, emotional and behavioural sequelae of severe closed head injury, and Luria's three units model of functional organisation of the brain59 60 as described by Vanieret al.18 The executive/cognition factor encompasses important areas of cognition often impaired in closed head injury (memory and temporal orientation) as well as major elements of executive dysfunction (planning, mental flexibility, self appraisal, conceptual disorganisation, and initiative or motivation) which have been shown to contribute significantly to long term disability53-57 and failure to return to work,54 and to impose serious emotional burden on family members.61-64 The positive symptoms factor includes items which are important not only in the acute phase, but have been shown to persist up to 10 or 15 years after njury.61 65-68 the negative symptoms factor reflects reduced initiation and lack of emotional variation occasionally seen in damage to the frontal lobes. The mood/affect factor could be useful in tracking emotional reactions as a patient's level of self awareness increases or in studying closed head injury related mood disorders. As demonstrated by the factor's correlation with neuropsychological domains, the oral/motor factor taps both verbal output as well as deficits in psychomotor speed and dexterity which are commonly seen in severe closed head injury.69 70