Introduction Short‐form versions of the Montreal Cognitive Assessment (SF‐MoCA) are increasingly used to screen for dementia in research and practice. We sought to collate evidence on the accuracy of SF‐MoCAs and to externally validate these assessment tools. Methods We performed systematic literature searching across multidisciplinary electronic literature databases, collating information on the content and accuracy of all published SF‐MoCAs. We then validated all the SF‐MoCAs against clinical diagnosis using independent stroke (n = 787) and memory clinic (n = 410) data sets. Results We identified 13 different SF‐MoCAs (21 studies, n = 6477 participants) with differing test content and properties. There was a pattern of high sensitivity across the range of SF‐MoCA tests. In the published literature, for detection of post stroke cognitive impairment, median sensitivity across included studies: 0.88 (range: 0.70‐1.00); specificity: 0.70 (0.39‐0.92). In our independent validation using stroke data, median sensitivity: 0.99 (0.80‐1.00); specificity: 0.40 (0.14‐0.87). To detect dementia in older adults, median sensitivity: 0.88 (0.62‐0.98); median specificity: 0.87 (0.07‐0.98) in the literature and median sensitivity: 0.96 (range: 0.72‐1.00); median specificity: 0.36 (0.14‐0.86) in our validation. Horton's SF‐MoCA (delayed recall, serial subtraction, and orientation) had the most favorable properties in stroke (sensitivity: 0.90, specificity: 0.87, positive predictive value [PPV]: 0.55, and negative predictive value [NPV]: 0.93), whereas Cecato's “MoCA reduced” (clock draw, animal naming, delayed recall, and orientation) performed better in the memory clinic (sensitivity: 0.72, specificity: 0.86, PPV: 0.55, and NPV: 0.93). Conclusions There are many published SF‐MoCAs. Clinicians and researchers using a SF‐MoCA should be explicit about the content. For all SF‐MoCA, sensitivity is high and similar to the full scale suggesting potential utility as an initial cognitive screening tool. However, choice of SF‐MoCA should be informed by the clinical population to be studied.

No abstract