Misdiagnosis using brief cognitive screening instruments in a population-representative sample

Abstract

The Affordable Care Act requires physicians to screen older patients for cognitive impairment. The Mini-Mental State Exam (MMSE) has been used widely for this purpose, though frequently a 'one-size-fits-all' score < 24 is used to indicate dementia. Because many factors influence cognitive test performance, a single MMSE cutpoint risks both over and under diagnosis of cognitive impairment. We examined performance of the standard MMSE cutpoint in relation to clinical diagnosis in a population-representative sample that reflects primary care throughout the U.S. Data were from the Aging Demographics and Memory Study (ADAMS), a cognitive aging study including participants aged 70 and older enrolled in the Health and Retirement Study. Participants received a standardized diagnostic protocol with extensive in-home clinical and neuropsychological assessment to determine consensus diagnosis of normal cognition; cognitive impairment, not demented (CIND); or dementia. Our sample included individuals with baseline diagnosis of normal cognition (n = 224) or CIND (n= 187), with 2 to 4 follow-up assessments spanning up to 5.9 years. Using multivariate regression, we assessed the contributions of demographic and genetic factors to MMSE score. MMSE scores ranged from 10–29 for CIND and 17–30 for cognitively normal. At baseline, 12.7% classified as cognitively normal had an MMSE < 24, while 50.6% diagnosed as CIND scored < 24. Overlap of MMSE scores between diagnoses of CIND and normal cognition was 90%. Higher age, less education, male gender, and ethnic minority status was associated with lower MMSE scores in both cognitively normal and CIND. Presence of at least one APOE e4 allele was associated with lower MMSE scores in normal cognition, but not CIND. Combined, these individual differences accounted for 36%-46% of variance in MMSE score, but their contribution did not differ among those who remained stable cognitively, who declined cognitively, or who improved cognitively. Findings in a heterogeneous community sample highlight considerable overlap between normal and impaired cognition on brief cognitive screening instruments. This underscores that performance on these measures should not be interpreted in isolation but are a useful tool for tracking change over time and establishing current cognition in the context of the full clinical history.