Alzheimer's Disease Assessment Scale: reliability and validity in a multicenter clinical trial.

Psychometric characteristics of the Alzheimer's Disease Assessment Scale (ADAS) were examined on the basis of data from 440 patients with dementia of the Alzheimer type that were collected before treatment in a multicenter clinical drug trial. Coefficients of internal consistency of above .80 for the cognitive (ADAS-Cog) and the noncognitive section (ADAS-Noncog) indicated a high degree of homogeneity of item contents within the two assessment domains. Test-retest reliability was estimated to be .93, .98, and .96 for ADAS-Cog, ADAS-Noncog, and the total score (ADAS-Total), respectively. Reliably detectable individual changes, which were derived from the reliability estimates, were 7, 3, and 8 points for ADAS-Cog, ADAS-Noncog, and ADAS-Total, in that order. Factor analysis and correlations with MMSE, SKT, and NOSGER scores support the validity of the ADAS-Cog and ADAS-Noncog scores with regard to the cognitive and the noncognitive assessment domains. The ADAS summary scores, almost all of the cognitive items, and some of the noncognitive items discriminated significantly between stages of severity of dementia, as classified independently by MMSE and SKT scores.