Clinical Question: How accurate is a rapid screening test for dementia?

Setting: Population-based

Study Design: Cohort (prospective)

Synopsis: The authors identified 344 subjects randomly drawn from a sample of 2,212 community-dwelling blacks older than 65 years. All 2,212 subjects were screened for dementia using a validated survey instrument, and 344 were oversampled to include more older and poor-scoring subjects. All of the 344 patients underwent a detailed evaluation to classify them as normal, cognitive impairment/not demented, or demented. The screening test began with subjects being asked to remember three items: apple, table, and penny. They were then asked three questions, with one point for each correct response: What year is this? What month is this? What is the day of the week? They were given one point for each of the three items that they could recall, for a total of six possible points. This was a well-designed study, with appropriate blinding of the evaluators, a prospective design, and all subjects undergoing the reference standard. At a cutoff of one or more errors, the test had a positive likelihood ratio (LR) of 1.9 and a negative LR of 0.03 for detecting patients with cognitive impairment or dementia. Thus, the test was highly effective at ruling out cognitive impairment if the patient had no errors, but one or two errors required further confirmation. Having three or more errors is strong evidence in favor of cognitive impairment (positive LR: 19.4). For the diagnosis of dementia, three or more errors had a positive LR of 7.4 and a negative LR of 0.13. The only limitation of the study is that the findings may not be generalizable to non-black populations, although the evaluation involving a mixed group of patients referred for dementia evaluation showed a similar performance.

Bottom Line: A brief screening instrument can rule out cognitive impairment if normal and is nearly diagnostic for dementia if a patient makes three or more errors. (Level of Evidence: 1a)