Object-memory evaluation for prospective detection of dementia in normal functioning elderly: predictive and normative data.

In a prospective study of dementia in initially normal functioning elderly, a brief form of the Fuld Object-Memory Evaluation (OM) was administered to 474 cognitively normal community-residing volunteers aged 75-85 at baseline and annually thereafter. Seventy-two subjects later became demented. Memory test data from the last annual evaluation before cognitive change was noted were available for 56. Although the entire population recalled 7.28 (SD = 1.33) of the 10 objects on Trial 1 of the test at baseline, these 56 subjects recalled only 5.96 (SD = 1.85). When recall of 6 or fewer objects was used as a predictor, the OM test identified 32 of the 56 who subsequently became demented. Compared to an estimated base rate of 15% for dementia, the predictive value of a positive test (PV+) was 39%, and that of a negative test (PV-) was 89%. With a cutoff of 5 or fewer items recalled, the PV+ rose to 59% and the PV- was 94%. Although the OM test was only moderately sensitive to incipient dementia (.57), it was fairly specific (.84), and lowering the cutoff to 5 increased the specificity to .96. Memory testing would therefore seem to hold promise as a predictor of dementia in cognitively normal elderly.