. Author manuscript; available in PMC: 2015 Apr 7.

Published in final edited form as: Dement Geriatr Cogn Disord. 2014 Apr 7;38(1-2):89–146. doi: 10.1159/000357839

Table 5.

The E-BEHAVE-AD [12]: conceptual framework, methodology, construct validity, reliability, criterion validity, and utility^*

A Conceptual framework
For numerous reasons, it is desirable not to rely solely on caregivers for the assessment of behavioral disturbances. Caregiver reports may be influenced by the emotional stress of caregiving, by the needs of the institution, by the burdens of caregiving, by the nature of the relationship with the subject, etc. Therefore, for both clinical and research purposes, it is desirable to have a direct assessment of behavioral disturbances in individuals with brain aging and the progression of AD. In addition, since it has been demonstrated that behavioral symptoms in patients with AD and, to some extent, with MCI and even SCI differ from those of other conditions, such as major depression or schizophrenia [22], it is desirable to have a rating instrument which can directly assess behavioral symptomatology in the brain aging to AD continuum.

B Methodology
Using the informant-based BEHAVE-AD as a model [1], an assessment instrument was developed which could be used for the direct evaluation of the subject. Six of the 7 BEHAVE-AD symptomatic categories were successfully adapted for the direct observation format. One category from the BEHAVE-AD, Diurnal Rhythm Disturbances, was not adapted because it was not possible to observe these symptoms in the context of a brief clinical interview. Twelve symptoms in the 6 behavioral symptomatic categories were identified. Up to 8 exemplars for each symptom are provided. The great majority of symptom exemplars are derived from the BEHAVE-AD. Seven of the 60 exemplars are derived from Teri et al. [128].
The 12 E-BEHAVE-AD items are scored on a 4-point severity scale. The general criteria for this scale are similar to that employed for the informant-based BEHAVE-AD assessment. Specifically, an item is scored as ‘0’ if no symptomatology with respect to the item is observed, as ‘1’ if the symptom is present in a mild form, as ‘2’ if the symptom is present in moderate form, such as with an emotional component, and as ‘3’ if the symptom is present in a severe form, such as with an emotional and physical component. Specific definitions appropriate to the individual items are provided.
In addition to a total score for the 12 items, the E-BEHAVE-AD has a global score assessment. This global score rating is also on a 4-point scale, which is similar to that employed for the individual E-BEHAVE-AD items. For example, a score of ‘0’ indicates no observable behavioral problems seen during the interview and ‘an absence of problems which are likely to be disturbing to the patient’s caregivers and/or disturbing or dangerous to the patient. Similarly, the ‘1’ = mild, ‘2’ = moderate, and ‘3’ = severe global scores are anchored to the criteria of disturbance to the caregivers and disturbance or danger to the subject (patient).
The E-BEHAVE-AD also contains instructions for the methodology of the assessment. For example, the interview should be conducted only with the rating clinician and the subject present, in the absence of caregivers or other informants. Also, the interview should be conducted as a conversation with the subject (patient). The conversation can be a dialogue and should include more general topics such as the subject’s activities or hobbies. The colloquy should last approximately 20 min.

C Construct validity
Since the E-BEHAVE-AD symptoms and scoring methodology are derived from the informant-based BEHAVE-AD, the construct validity features described for the BEHAVE-AD are also applicable for the E-BEHAVE-AD. These include the differentiating features from the cognitive and functional symptoms of AD. In addition, a criterion validity study was published describing the relationship between the observer-rated E-BEHAVE-AD symptoms and the informant-based BEHAVE-AD symptoms in 49 subjects. This study is described below.

D Reliability
Authors [Ref.]	Methodology	Findings
Auer et al. [12]	Twenty community-residing patients were evaluated by 2 raters. The sample comprised 11 women and 9 men, with a mean age of 73.2 years (SD = 8.0). Fifteen patients had AD, 1 had dementia associated with cerebrovascular disease, and 4 had dementia associated with diverse disturbances and/or etiologies. The mean MMSE score was 11.6 (SD = 9.4). On the GDS, subjects had the following severity distribution: GDS stage 4 (mild dementia), n = 3; GDS stage 5 (moderate dementia), n = 3; GDS stage 6 (moderately severe dementia), n = 13, and GDS stage 7 (severe dementia), n = 1. The 2 raters independently evaluated the patients on the E-BEHAVE-AD. One rater performed the interview while the other rater observed. The order of interviewer and rater was alternated. There were no communications between the raters with respect to scoring, and all scoring was done independently. Intraclass correlation was used to determine the interrater reliability between the raters.	For the total E-BEHAVE-AD, the ICC for the ratings of the 2 examiners was 0.97 (p < 0.01) using both raters-fixed and raters-random methods. For the individual BEHAVE-AD items, the ICCs ranged from 0.76 to 1.00 (there was perfect agreement for 3 of the 11 items for which a correlation coefficient could be computed) using the raters-fixed procedure. For one item, ‘hallucinations’, the correlation could not be computed because of absence of variance on the item since the symptom was not observed in the 20 subjects evaluated. All correlations for the 11 observed E-BEHAVE-AD symptoms were significant (p < 0.01). Using the raters-random procedure, the ICCs ranged from 0.73 to 1.00 for the 11 items for which correlations could be calculated. Also, using the raters-random procedure, all correlations were significant for the 11 E-BEHAVE-AD symptoms observed (p < 0.01). Hence, the E-BEHAVE-AD showed excellent interrater reliability.

E Criterion validity
Authors [Ref.]	Methodology	Findings
Auer et al. [12]	The BEHAVE-AD is an informant-based rating scale, whereas the E-BEHAVE-AD is a direct observation- based rating instrument. Depending in part on the perspective, including the needs of the subject (patient) and the needs of the persons who live with, or work with, or interact with the subject, either an informant- rated or a subject-based assessment instrument can be considered the standard for criterion validity. In addition, for other reasons, it was important to know the relationship between the observer-based subject symptomatology and the informant-based symptomatology. Therefore, Auer et al. [12] conducted a study of the relationship between symptomatic assessments using the informant-based BEHAVE-AD [1] and the direct observer-based E-BEHAVE-AD [12]. In accord with the standard rating scale procedures, information regarding the BEHAVE-AD was obtained from a knowledgeable informant with respect to the prior 2-week period, whereas information with respect to the E-BEHAVE-AD was obtained by a clinician observer in the course of a colloquy with the subject lasting approximately 20 min.	Pearson correlation coefficients were used to determine the agreement between the informant reports of symptomatology and the clinicians’ observation of symptomatology in each of the 6 symptomatic areas. Overall, for the 6 analogous symptomatic categories, there was a 0.51 correlation between the informant-based BEHAVE-AD symptomatic reports and the E-BEHAVE-AD symptomatic observations (p < 0.01). For one of the symptomatic categories, Hallucinations, correlations could not be computed since these symptoms were never observed by the clinician. For the remaining 5 symptomatic categories, significant (p < 0.01) relationships were observed for all symptomatic areas. The Pearson correlations ranged from 0.32 for Affective Disturbances to 0.44 for Aggressiveness.
Auer et al. [12]	A total of 49 subjects and informants (mean age, 72.5 years, SD = 8.4; 22 women and 27 men) were studied. All subjects were participants in research longitudinal and pharmacologic studies conducted at the New York University Medical Center’s Alzheimer’s Disease Center. Subjects were healthy apart from the presence of MCI (n = 5) or dementia (n = 39). The remaining subjects had no cognitive impairment or SCI (n = 5). Of the dementia subjects, 27 had probable AD [32], 6 had dementia associated with cerebrovascular disease, and 6 had dementia associated with diverse etiologies or processes. The mean MMSE score of the subject population was 18.3 (SD = 9.7). All ratings were obtained in the course of a single study visit by 2 different raters. One rater assessed the BEHAVE-AD based upon the informant reports, while the other rater assessed the E-BEHAVE-AD based upon subject observation. Each rater was unaware of the other rater’s evaluations. The 6 comparable BEHAVE-AD and E-BEHAVE-AD symptomatic categories were examined. Because the Diurnal Rhythm Disturbances symptomatic category of the BEHAVE-AD is not assessed with the E-BEHAVE-AD, this symptomatic category was excluded from the analyses. In addition, because the number of items assessed is not comparable in the BEHAVE-AD and the E-BEHAVE-AD symptomatic categories, the presence or absence of symptoms in each of the 6 comparable symptom categories was compared with the 2 assessment procedures.	In conclusion, there are significant relationships between behavioral disturbances observed in brain aging, MCI, and dementia (primarily AD and closely related conditions) using an approximately 20-min direct subject evaluation (the E-BEHAVE-AD) and an informant-based report of symptomatology over the prior 2 weeks. However, these methodologies are also, to some extent, measuring different constructs.

F Utility
The E-BEHAVE-AD can be very useful in providing an independent view of behavioral symptoms of dementia patients for clinical and research purposes. The major usage in worldwide studies of the E-BEHAVE-AD assessment until the present time has been in the context of the role of the E-BEHAVE-AD as part of the NYU-CIBIC-Plus assessment [13, 14]. In this context, the E-BEHAVE-AD has been used as part of a primary outcome measure in the pivotal trials associated with the initial approvals of 2 of the 3 most widely used medications for the treatment of AD, namely, rivastigmine (Exelon) [15, 16] and memantine (Namenda, Ebixa, Axura) [17]. However, there was no separate analysis of the CIBIC-Plus constituents in these studies. A recent study of a management program intervention examined the NYU-CIBIC-Plus component measures in terms of their contribution to efficacy [106]. The informant-based BEHAVE-AD component of the NYU-CIBIC-Plus was very sensitive to the intervention and showed very robust results. This observation of robust results in the BEHAVE-AD, albeit much greater in magnitude than in previous studies, is consistent with observations in previous studies which have found robust results on the BEHAVE-AD with nonspecific placebo interventions in clinical trials. However, the E-BEHAVE-AD assessment component of the NYU-CIBIC-Plus did not show a response to the management program. Hence, the E-BEHAVE-AD assessment is potentially a very useful independent source of relevant information with respect to the treatment of behavioral symptoms in dementia patients.