Confirmatory factor analysis of the Mattis Dementia Rating Scale

Mattis Dementia Rating Scale (MDRS) test results for a heterogeneous sample of 185 patients, aged 47-95, were subjected to confirmatory factor analysis. We assessed the goodness-of-fit of one- to five-factor models, using conceptual groupings of items based upon prevailing neuropsychological practice. A two-factor model (verbal/nonverbal) fit the data as well as or better than other multi-factorial models, including the five-subtest scheme. In each model, the factors were highly intercorrelated, suggesting that they were largely overlapping but separable. The MDRS Conceptualization and Memory subtests correlated at +1.00, indicating that they are not separable. In addition, two items (name writing, imitating movements) did not load on any factors, due to restricted variance. These findings indicate that a more parsimonious interpretation of test results, using the verbal/nonverbal dimensions, may be warranted, particularly for screening purposes in heterogeneous populations. Suggestions are provided for improving the utility of the MDRS.