Brief memory tests are moderately accurate screening tools, not diagnoses: a meta-analysis of the MMSE found pooled sensitivity of 79.8% and specificity of 81.3% in memory clinics, and a review of 34 articles found the MoCA better than the MMSE at telling mild cognitive impairment from normal cognition (mean AUC 0.883 vs 0.780). How accurate a score is depends on the setting, the person's education and background, and the cut-off used.
What the studies found
A 2009 meta-analysis of 34 dementia studies and five MCI studies looked at the MMSE in different settings. In memory clinics the pooled sensitivity was 79.8%, specificity 81.3%, positive predictive value 86.3% and negative predictive value 73.0%; in non-clinical community settings the pooled sensitivity was 85.1%, specificity 85.5%, positive predictive value 34.5% and negative predictive value 98.5%. The authors concluded that the MMSE offers modest accuracy, with best value for ruling out dementia in community and primary care, and had very limited value for diagnosing MCI against healthy controls.
A 2019 systematic review of 34 articles compared the MoCA and the MMSE directly. More than 80% of the articles showed the MoCA to be superior to the MMSE in discriminating between people with mild cognitive impairment and no cognitive impairment; the area under the curve varied from 0.71 to 0.99 for the MoCA and 0.43 to 0.94 for the MMSE, and the mean AUC was 0.883 for the MoCA versus 0.780 for the MMSE. Both tests were found to be accurate in the detection of Alzheimer's disease.
Another systematic review, covering 20 MoCA and 13 MMSE studies, reported that the best cut-off for detecting MCI was 24/25 for the MoCA (n = 9350; sensitivity 80.48%, specificity 81.19%; AUC 0.846) and 27/28 for the MMSE (n = 882; sensitivity 66.34%, specificity 72.94%; AUC 0.736).
The MoCA's originally suggested cut-off score of 26/30 led, in some studies, to an inflated rate of false positives, particularly for those of older age or lower education. A 2018 meta-analysis of nine studies (from 304 identified) found that a cut-off of 23/30 yielded the best diagnostic accuracy for separating healthy cognitive ageing from possible MCI, and that it lowered the false-positive rate.
Using National Alzheimer's Coordinating Center data on 3895 people (80.7% non-Hispanic White, 15.0% non-Hispanic Black and 4.2% Hispanic), a 2018 study found optimal MoCA cut-offs for detecting MCI of 25 among non-Hispanic Whites, 24 among Hispanics and 23 among non-Hispanic Blacks. Optimal cut-offs for detecting dementia were 19 among non-Hispanic Whites and 16 for both non-Hispanic Blacks and Hispanics, and lower educational attainment produced lower optimal cut-offs.
What remains uncertain
The reviews do not agree on one cut-off: the original MoCA threshold of 26/30, 24/25 and 23/30 all appear, and the 2018 study found that the best cut-off varied by race, ethnicity and education. A score from one test is a screening result, and the library holds no study that gives a single score at which a diagnosis can be made.
- Studies
- A re-examination of Montreal Cognitive Assessment (MoCA) cutoff scores
- Is the Montreal Cognitive Assessment (MoCA) test better suited than the Mini-Mental State Examination (MMSE) in mild cognitive impairment (MCI) detection among people aged over 60? Meta-analysis
- Optimal cutoffs for the Montreal Cognitive Assessment vary by race and ethnicity
Accuracy also depends on where the test is used. The 2009 meta-analysis found a positive predictive value of 34.5% in community settings against 86.3% in memory clinics, and reported only limited evidence on the MMSE for MCI, with five robust studies comparing MCI with healthy subjects.
The review of 34 articles drew on studies with a wide range of results (MoCA AUC 0.71 to 0.99), so the averages above hide large differences between studies. A clinician who knows the person can say what a result means for them.
This page describes what published studies in alzhub's library report. It is not medical advice and does not apply to any one person. Where a decision would follow, a clinician who knows the person is the one who can say what a result or a choice means for them.
Questions
- Which is more accurate, the MoCA or the MMSE?
- For mild cognitive impairment, a review of 34 articles found the MoCA superior, with a mean AUC of 0.883 against 0.780 for the MMSE, and found both tests accurate for detecting Alzheimer's disease.
- Is a low MMSE or MoCA score a diagnosis?
- No. The records describe these as screening tools. A meta-analysis found MMSE positive predictive value of 34.5% in community settings and 86.3% in memory clinics, so what a score means depends on the setting.
- Does education change MoCA scores?
- Studies report that it does. A 2018 analysis of 3895 people found that lower educational attainment produced lower optimal cut-offs, and that optimal cut-offs differed by race and ethnicity.
Read on
Last updated