[Usefulness of the 'Mini-Mental State Test'for the Diagnosis of Dementia; Study of Criterion Validity in a Dutch Rural Population]

Objective: To establish the criterion validity of the MMSE for dementia in a community-based population.

Design: Descriptive.

Setting: Eight rural general practices and eight adherent institutions for long-term care near Zwolle, the Netherlands.

Methods: In the first (screening) stage, the MMSE was applied, in the second (diagnostic) stage the 'Cambridge examination for mental disorders of the elderly' (CAMDEX) was used. After exclusion of subjects with clinically relevant impairment of hearing and (or) vision, analyses were based on 2151 screened subjects. Based on a non-proportional, stratified randomized sampling procedure using the MMSE score as the stratification variable, 390 of the screened subjects were examined with the CAMDEX. Seventy-seven cases of DSM-III-R dementia were identified. Based on 390 paired observations the relationship between MMSE score and DSM-III-R dementia was modelled by logistic regression. After extrapolation of this relationship to the total screened population, sensitivity and specificity figures were calculated at several cut-offs of the MMSE, and Receiver Operating Characteristic (ROC) curves were constructed for the total population as well as for four subpopulations. (men, women, < 80, > or = 80 years).

Results: At the cut-off of 23/24, sensitivity was 0.76 and specificity 0.91 in the total population. However, the criterion validity of the MMSE varied substantially between the four sex-age combinations. Criterion validity was especially poor for women in the younger age range.

Conclusion: The MMSE has only limited value in clinical practice. Use of uniform cut-offs has to be rejected.