Diagnosis of dementia using the mini-mental state examination in general practice and the community: Cochrane systematic review of diagnostic test accuracy

Talk Code: 

The problemRecent UK policy has encouraged case-finding of people with possible dementia in an attempt to increase the proportion of people with a formal diagnosis, which is currently around 50%. The mini-mental state examination (MMSE) is widely used in practice for identifying people with possible dementia but there is no robust review of its diagnostic test accuracy. We systematically reviewed the diagnostic test accuracy (DTA) of the MMSE.The approachWe searched for articles in six databases, combining free text and relevant controlled vocabulary terms. We contacted authors and translated papers where necessary. The target condition was any-stage all-cause dementia. We included studies in primary care or the general population. Reference standards were DSM, ICD-10 or CDR. The index test was the 30-item MMSE. Citation screening, data extraction and study risk of bias (using QUADAS-2) were done in duplicate. Meta-analysis was performed with metandi in Stata.Findings47,812 citations were retrieved with 23,502 duplicates. The full text of 384 possibly relevant citations was reviewed and we then excluded 305 papers. Studies were excluded at the full text stage because they were the wrong setting (132), the wrong study design (119), used an inappropriate reference standard (29), used the wrong index test (24) and only the title was available despite contacting authors (1). We included 45 studies (79 articles) in the meta-analysis, 37 asymptomatic general population; 7 primary care: 4 symptomatic, 2 asymptomatic, 2 mixed; 1 translated paper of uncertain setting.We judged that over half of studies were at low risk of bias, with the exception of flow and timing, where the risk of bias was “low” in 35% of studies and “unclear” in 43%. In four primary care studies using a cut-point of 19/30 indicating normal the MMSE had overall sensitivity 0.76 (95% CI 0.54 to 0.89) and specificity 0.88 (0.85 to 0.90); at the same cut-point in four community the sensitivity was 0.73 (0.58 to 0.84) and the specificity 0.94 (0.75 to 0.99). In 15 community based studies that used a cut point of 24/40 indicating normal the sensitivity was 0.85 (0.74 to 0.92) and the specificity 0.90 (0.82 to 0.95); there were too few primary care based studies with this cut point to perform meta analysis.ConsequencesThe MMSE may contribute to evaluating possible dementia in primary care and the community, though most studies were of asymptomatic people. Diagnostic accuracy of this single test alone is limited and appears inadequate to rule in or rule out disease: even at a threshold of 19/30 (lower than usually advised) the specificity is less than 90%.


  • Sam Creavin, Cochrane Collaboration, Oxford, UK
  • Anna Noel-Storr
  • Susanna Wisniewski
  • Sarah Cullum, Cochrane Collaboration, Oxford, UK