A diagnostic test accuracy study investigating GP clinical impression and brief cognitive assessments for dementia in primary care, compared to specialised assessment

BACKGROUND: Many health systems are interested in increasing the number of uncomplicated and typical dementia diagnoses that are made in primary care, but the comparative accuracy of tests is unknown. OBJECTIVE: Calculate diagnostic accuracy of brief cognitive tests in primary care DESIGN: We did a diagnostic test accuracy study in general practice, in people over 70 years who had consulted their GP with cognitive symptoms but had no prior diagnosis of dementia. The reference standard was specialist assessment, adjudicated for difficult cases, according to ICD-10. We assessed 16 index tests at a research clinic, and additionally analysed referring GPs clinical judgement. RESULTS: 240 participants had a median age of 80 years, of whom 126 were men and 132 had dementia. Sensitivity of individual tests at the recommended thresholds ranged from 56% for GP judgement (specificity 89%) to 100% for MoCA (specificity 16%). Specificity of individual tests ranged from 4% for Sniffin’ sticks (sensitivity 100%) to 91% for TUG (sensitivity 23%). The 95% centile of test duration in people with dementia ranged from 3 minutes for 6CIT and TAC, to 16 minutes for MoCA. Combining tests with GP judgement increased test specificity and decreased sensitivity: e.g., MoCA with GP Judgement had specificity 87% and sensitivity 55%. CONCLUSIONS: Using GP judgement to inform selection of tests was an efficient strategy. Using IQCODE in people who GPs judge as having dementia and 6CIT in people who GPs judge as having no dementia, would be a time-efficient and accurate diagnostic assessment.