[Detection of memory impairment using a recognition test for words and faces]

Fifty-two patients (mean age 78.3; sd: 5.5; range 70-94) who were admitted to psychogeriatric day care because of memory problems and other cognitive deficits were tested with the Dutch form of Warrington's Recognition Memory Test (RMT). Forty-five mentally normal elderly (mean age 79.5; sd: 5.7; range 69-92) were also tested in order to establish the sensitivity and specificity of the test for the detection of memory impairment. Mentally normal subjects were mainly recruited from homes for the aged and were rated by the staff as free from any symptoms of dementia or other psychiatric disease. Using cut-off scores at the 98%-specificity level, the sensitivity of the RMT was 54% for the verbal part (Recognition Memory for Words, RMW) and 73% for the non-verbal part (Recognition Memory for Faces, RMF). Selection of cut-off scores with high sensitivity (96%) resulted in specificities of 62% and 71% for RMW and RMF, respectively. The RMF appears to be a more valid test for the detection of memory impairment than the RMW. Selection of cut-off scores is contingent on knowledge of the pretest or prior probability of memory impairment. A cut-off score with high sensitivity is recommended if impairment is strongly suspected, whereas high specificity is needed to verify the presence of impairments for which there is little clinical evidence.