Contact Jennifer Bogner, Ph.D. Ohio Regional TBI Model System, at
Citation Bogner, J. (2000). The Agitated Behavior Scale. The Center for Outcome Measurement in Brain Injury. http://www.tbims.org/ combi/coglog ( accessed
The means and standard deviations for the Total Score and subscale scores are based on samples of persons with traumatic brain injury treated during the acute phases of recovery on an inpatient, rehabilitation unit. A prospective sample of all patients with brain injuries, regardless whether they were demonstrating agitation, revealed an overall mean ABS score of 21.01 and standard deviation of 7.35 for day shift nursing observations (Corrigan, 1989). For clinical purposes, we consider any scores (Total or converted subscale) 21 or below to be within normal limits; from 22 through 28 to indicate mild occurrence; 29 through 35 to indicate moderate; and more than 35 to be severe.
While norms based on a broader sampling of patients from other institutions will be desirable, the ABS remains quite usable for ipsative comparisons of the same individual from shift to shift, therapy to therapy, and/or day to day.
The initial validation studies (Corrigan, 1989) showed correlations between ratings conducted on the same day exceeded .70 for the Total score (subscale scores were not available at that time). Inter-rater reliability was recently re-examined (Bogner, et al, in press) due to the instrument's increased use with different populations, rated by individuals from various disciplines, based on a variety of observation periods.
With a sample of persons receiving acute rehabilitation for acquired brain injury, research assistants (psychology interns and a rehabilitation nurse) rated behavior based on 10-minute observation periods. The research assistants' ratings yielded a correlation coefficient for the Total score of .92. The correlation coefficients for the factors Disinhibition, Aggression, and Lability were .90, .91, and .73, respectively. The reliability of the Lability score is likely effected by the small number of items used in its calculation. When comparisons were made between the ratings made of the research assistants based on 10 minute observation periods and the ratings made by nurses based on an 8-hour shift, the correlation coefficients for the Total and subscale scores were much lower (.36 to .60). This difference reflects the fact that when ratings are made under different conditions, in different time intervals, by different types of raters, what is being measured is also different. Agitation has been known to change in intensity throughout the day, and in response to the regulation of stimulation. When serial monitoring is the goal, it is important that comparisons be made under comparable conditions.
Inter-rater reliability was also examined with a sample of individuals residing in a long-term care facility whose primary diagnosis was dementia. The ratings were again made by research assistants based on 10-minute observation periods. The correlation coefficient for the Total score was .91, while the coefficients for the factor scores were .87 for Disinhibition, .89 for Aggression, and .86 for Lability.
Examination of the internal consistency of the scale found Cronbach's alphas ranging from .83 to .92 in the original validation study (Corrigan, 1989). With the two samples examined by Bogner et al (in press), Cronbach's alphas ranged from .74 to .92. The relatively high internal consistency of the scale suggests that agitation, as measured by the ABS, is a unitary construct, with three facets being individually prominent at times.
The original development of the ABS demonstrated the content validity of items and concurrent validity of the Total Score. Subsequent studies have shown the ABS to be predictive of change in cognitive status (Corrigan & Mysiw, 1988) and able to differentiate confusion and inattention (Corrigan & Mysiw, 1988; Corrigan et al., 1992). Construct validity has been further substantiated by the identification of underlying factors (Corrigan & Bogner, 1994). The Total Score and three underlying subscales have proven stable over multiple samples (Corrigan & Bogner, 1994).