Objective: To compare a family informant's report of memory loss in an older family member to standardized clinical diagnoses of cognitive impairment.
Setting: Duke Established Populations for Epidemiologic Studies of the Elderly (EPESE), a 10-year longitudinal study of community dwellers aged 65 and greater in five counties of North Carolina.
Participants: A stratified random sample of potentially demented participants was selected from the second wave of the Duke EPESE using responses to a brief cognitive screen. A neuropsychological battery was administered to these participants, and their family informants were asked whether they recognized memory loss in the participant. One hundred fifty-seven participants completed the full evaluation and also had an available family informant.
Main outcome measures: Family informant's report of memory loss (yes, no, sometimes) compared to expert consensus diagnosis of cognitive impairment or dementia.
Results: There was poor concordance between the clinical diagnoses of cognitive impairment or dementia and the family informant's recognition of memory loss (kappa=-0.05; P=.74). When informants reported memory loss, 30% of participants were found not to have a cognitive loss. Among participants in whom family informants reported no memory loss, 75% were diagnosed with dementia or cognitive impairment (sensitivity, 0.70, 95% confidence interval [CI], 0.61 to 0.78; specificity, 0.24, 95% CI, 0.13 to 0.40; positive predictive value, 70%; negative predictive value, 25%).
Conclusions: Asking family members about memory loss in a patient may be an unreliable strategy to detect dementia.