BACKGROUND

Alzheimer's disease (AD) is often associated with agitation and aggression, which may impair function, impede care, and be a major source of stress for caregivers. The Cohen-Mansfield Agitation Inventory (CMAI) is often used to assess agitation and aggression. In its original, nursing-home version, it is a 29-item, caregiver-informed, clinician-administered 7-point scale that assesses the frequency of various agitation or aggressive behaviors. However, the instruction manual advises against the use of the total score in favor of a domain-based analysis. This recommendation has been followed in both clinical trials and practice. Because the CMAI is comprehensive and easy to administer, we sought to determine the validity of its total score as a single construct for assessing agitation and aggression in patients with AD.

METHODS

We used a previously conducted factor analysis of the CMAI scores from two risperidone trials in patients with dementia (N = 648), and a follow-up analysis of the subset of patients with psychosis of AD (N = 479), to examine, using vector analysis and an effect-size-versus-signal-to-noise ratio analysis, whether the total CMAI score could confidently be used as a global measure of agitation and aggression in AD.

RESULTS

Our findings suggest that the CMAI items from the dataset analyzed load into 4 clusters, which cover about 50 % of the total data variance. Surprisingly, items with the lowest signal-to-noise ratio (hitting, performing repetitious mannerisms, aimless pacing or wandering) had the strongest response to treatment (and vice versa), and belonged to different factors. The further observation that many items were spread among the factors, instead of primarily measuring a single factor or domain, suggests that there is a continuum of symptoms, and separating them into domains requires separating very similar items that measure two or more domains.

CONCLUSIONS

These findings suggest that assessing agitation and aggression via CMAI domains instead of the total score is likely to miss important behavioral signals. Using total CMAI score in clinical trials and practice, along with the assessment of individual items, is warranted.