Three commonly used, brief cognitive tests erroneously identified dementia, resulting in more than a third of those screened being incorrectly classified, a retrospective analysis has concluded.

David Llewellyn, PhD, of the University of Exeter Medical School in the United Kingdom

Dr. David Llewellyn

The likelihood of a false-positive or false-negative result declined sharply when all three tests were given, however; only about 2% of patients were misclassified in all three, David Llewellyn, PhD, and his colleagues reported in Neurology: Clinical Practice.

The Mini Mental State Examination (MMSE), Memory Impairment Screen (MIS), and animal naming (AN) were susceptible to different measurement biases, wrote Dr. Llewellyn of the University of Exeter (U.K.).

Just one variable – an informant’s perception of the patient’s memory as unimpaired – consistently predicted inaccuracy in all three tests. Most of the patients in this category carried the diagnosis of cognitively impaired but not demented (CIND), a finding that has important clinical implications.

“These participants may be in the very early stages of conversion to dementia. ... Therefore, of those with low or borderline cognitive assessment results, reassessment to detect further decline may be appropriate.”

The study comprised 824 patients included in the Aging, Demographics and Memory Study, which is a subsample of the Health and Retirement Study. They completed the tests from 2001-2004, during which time they were a mean of 82 years old. A panel of experts adjudicated diagnoses, which they then parsed into all-cause dementia, CIND, or cognitively normal. The testing included a self and informant assessment of memory decline. The investigators also looked at 22 predictors of cognition, including patient characteristics, apolipoprotein E carriage (ApoE e4), and sociodemographic factors.

The prevalence of dementia was 35.3%; of the nondemented patients, 43% met the criteria for CIND. The team found that 35.7% of cases were misclassified by at least one test, 13.4% by two, and 1.7% by all three.

The MMSE was the least accurate, with a 21% misclassification rate, reflected in an 18.6% false-positive rate for those without dementia and a 2.4% rate of false-negative for those with dementia.

The MIS had a 16% misclassification rate, with a 9.5% rate of false-positive for those with no dementia and a 6.3% rate of false-negative for those without.

The AN had a 14% misclassification rate, with a 6.8% false-positive rate for those without dementia and a 7.7% false-negative rate for those with dementia.

For the MMSE, MIS, and AN, the number of participants with false-positives that met the criteria for CIND were 74.5%, 82.1%, and 82.1%, respectively.

In the final multivariate model, seven variables predicted misclassification, including black ethnicity for the MMSE; age, visual impairment, ApoeE4 noncarrier, and depression for the MIS; and no hyperlipidemia and normal informant memory assessment for the AN. Lower years of education and heart problems predicted misclassification on both the MMSE and AN.

An absence of informant-related poor memory predicted misclassification on all three tests.

“Failing to detect dementia can delay access to treatment and support, whereas false alarms lead to unnecessary investigations, causing pressure on health care systems,” Dr. Llewellyn said in a press statement. “Identifying people with dementia in a timely fashion is important, particularly as new methods of treatment come onstream. Our findings show that we desperately need more accurate and less biased ways of detecting dementia swiftly in clinic.”

The study was supported by the Halpin Trust, the Mary Kinross Charitable Trust, the Engineering and Physical Sciences Research Council, and the U.K. National Institute for Health Research. None of the authors reported any financial conflicts relevant to the work.

SOURCE: Llewellyn D et al. Neuro Clin Pract. 2019;1:1-9.