Abstract
<jats:p>Detection and characterisation of dementia and Mild Cognitive Impairment (MCI) is essential for diagnosis and to support recruitment of patients into trials of disease-targeted therapies. Computerised systems offer a means of improving detection in community and primary care settings in a scalable way. This study presents further validation of the self-test PROTECT Cognitive Test System of eight assessments of memory, attention and executive function in 36,941 participants (35,822 healthy, 1046 MCI, 73 dementia). PROTECT shows robust separation of dementia and non-dementia participants (p < 0.001), and good discriminative ability for dementia (Area Under the Curve = 0.966) with 90.90% sensitivity and 87.80% specificity. An optimised detection algorithm robustly identified MCI (P<0.001) and predicted 24-month decline across cognitive domains in both amnestic MCI and non-amnestic MCI phenotypes compared with healthy controls. PROTECT cognitive data also correlated strongly with the plasma biomarkers p-tau217 and Neurofilament Light (n = 46). The system provides a means of improving dementia and MCI detection using self-testing, offering a scalable triage and monitoring tool for clinical pathways and trials.</jats:p>