RESUMO
Background: The Addenbrooke's Cognitive Examination III (ACE-III) is widely used for evaluation of dementia. Aim: To assess the psychometric properties of ACE-III, analyzing its factor structure, its reliability (from an Item Response Theory [TRI] model) and its diagnostic usefulness. Material and Methods: We studied 1101 older people without cognitive impairment and 63 currently having a diagnosis and receiving treatment for dementia. Results: The presence of two factors for the Attention subscale (Orientation and Attention, separately) was suggested. The factorial analysis showed adequate adjustment in all the subscales, except for the new Attention subscale. In the TRI analysis, the Attention subscale presented a greater number of items with lack of fit compared to the other subscales. Using a proposed threshold of 66 points or less to identify cognitive impairment related to dementia, a sensitivity of 0.97 and specificity of 0.81 was obtained. Conclusions: ACE-III is a valid, reliable, and useful measure for the clinical detection of dementia. The combined use of Orientation and Memory subscales is proposed as an alternative and time-saving ACE-III indicator.