Back

Accuracy of pattern-based dementia diagnostic protocols: Using longitudinal data to infer etiology of Alzheimer's disease and related dementia as compared to stroke or normal aging

Clouston, S.; Hanes, D. W.; Danesh Yazdi, M.

2025-02-27 epidemiology
10.1101/2025.02.26.25322969 medRxiv
Show abstract

IntroductionThe goal of this study was to compare accuracy for pattern-recognition protocols to prospectively identify dementia due to Alzheimers disease and related dementias (ADRD) or stroke. MethodsWe used data from cognitively unimpaired respondents who completed at least 5 cognitive assessments during waves 3-12 of the Health and Retirement Study (HRS), a longitudinal study of older US residents. Participants were assessed at wave 12 (in 2019) for cognitive status. Patterns of participants cognitive decline were analyzed to differentially identify ADRD and stroke and were compared against self-reported and objective diagnoses of amnestic, executive, and multidomain mild cognitive impairment (MCI) and ADRD. We reported sensitivity/specificity to detect new-onset dementia at the final wave of observation. ResultsAfter applying inclusion/exclusion criteria, 43 (1.69%) of cognitively unimpaired participants developed dementia, while 165 (6.49%) developed amnestic MCI. Patterns of cognitive decline consistent with ADRD affected 372 (14.6%) of respondents, while patterns of cognitive decline consistent with stroke were evident in 917 (36.1%) participants. ADRD- consistent cognitive declines were evident in 75.8% and 76.7% cases of amnestic MCI and dementia, respectively, though only 24.5% reported receiving a clinical diagnosis of dementia. Sensitivity/Specificity of ADRD was 94.3%/87.0% when detecting dementia without stroke. DiscussionThis study implies that we can reliably use longitudinal patterns of cognitive decline to differentially diagnose ADRD from Stroke in most participants with dementia.

Published in Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring · not in our set (fewer than 10 published preprints to learn from) · training set

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.