The predictive value of dementia screening instruments in clinical populations

To determine the positive predictive value of the Mini-Mental Status Examination (MMSE), the Abbreviated Mental Test (AMT) and the Informant Questionnaire for Cognitive Decline in the Elderly (IQCODE) for dementia in different clinical settings.

Sensitivity and specificity of tests were compared against the criterion standard of clinical diagnosis made by a geriatrician or psychogeriatrician.

Two groups of patients were studied, a referred group of outpatients to a memory clinic (MC) and a random selection of patients referred to an aged care assessment team (ACAT).

Of 437 eligible MC patients, 299 were used in analyses of whom 216 subjects had dementia. Of 100 randomly selected ACAT patients, 78 agreed to be studied, and 32 subjects had dementia.

The correlation between the MMSE and AMT was 0.87 for the ACAT sample and 0.86 for the MC sample. The correlation between the MMSE and IQCODE was -0.65 for the ACAT sample and -0.56 for the MC sample. The correlation between the DMT and IQCODE was -0.62 for the ACAT sample and -0.54 for the MC sample. The AMT and MMSE had lower sensitivity and specificity in the MC sample compared with the ACAT sample. Using a sensitivity of 78% and a specificity of 88%, chosen on the basis of findings, the positive predictive value in clinical situations varied from 70% for patients over the age of 75 admitted to acute teaching hospitals to 84% for patients seen by ACATs, inpatient geriatric liaison patients or acute geriatric units. The positive predictive value for the general population aged 80-84 years was 45%.

The use of these tests is appropriate for the clinical situations described above. Their use in an unselected elderly population is not recommended as the tests are likely to produce more false positives than true cases of dementia.