Answers are grounded in this paper and cite their sources.
BackgroundBrief cognitive screens could narrow India's dementia detection gap, but widely used instruments depend on reading, writing, or clock-drawing and lose accuracy in the large illiterate older population; Alzheimer's disease and related dementias drive most of this burden.ObjectiveTo derive and validate a fully verbal, literacy-fair candidate cognitive screening model.MethodsIn LASI-DAD Wave 1 (n = 4096) we derived a screen from prop-free verbal tasks (LASSO then forward selection) against algorithmic DSM-5 classification, compared discrimination (AUC) with reconstructed Hindi Mental State Examination (HMSE), Mini-Cog, Community Screening Instrument for Dementia (CSID) and Telephone Interview for Cognitive Status (TICS), performed leave-one-region-out validation and addressed incorporation bias using clinician Clinical Dementia Rating (CDR) ratings (n = 2527).ResultsThe sample was 57.5% illiterate; weighted prevalence was 26.7% (any impairment) and 9.0% (dementia). A four-item model-word-list learning, orientation to time, object naming, animal fluency ("Verbal-4")-reached AUC 0.757 for any impairment and 0.785 for dementia, indistinguishable from the HMSE, higher than the Mini-Cog, CSID and TICS (all p < 0.001), and similar across literacy (0.778 versus 0.776). In the non-random CDR-rated subsample, dementia discrimination was 0.872 against the clinician CDR versus 0.775 against the DSM-5 algorithm. Any-impairment discrimination attenuated in South India (0.688), analyses suggesting substantial case-mix and/or reference-standard contributions. A positive screen predicted mortality (adjusted HR 1.63, 95% CI 1.39-1.90).ConclusionsThis candidate screening model matched full-battery accuracy without a literacy gradient, held against a partially independent clinician standard, and carried independent prognostic information; it requires prospective calibration into an operational score before field use.
No takes yet. Share an insight, caveat, or question.
Jain et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: