A GAMLSS approach to digital dementia screening and profiling using a two-step cognitive self-assessment tool in China

PURPOSE: To develop clinically actionable, education-stratified normative reference ranges for the two-step cognitive self-assessment (TCSA) tool, addressing China's urgent need for scalable cognitive screening in its rapidly aging population. METHODS: Using generalised additive models for location, scale and shape (GAMLSS) on a cohort of 40 588 community-dwelling adults (50-100 years), we developed reference curves for core cognition (primary tasks), multi-cognition domains (secondary tasks) and processing speed (TCSA-Response Time, TCSA-RT). Individuals with self-reported dementia or functional dependence assessed via activities of daily living scale were excluded. RESULTS: Smoothed percentile curves were established. A health equity gap manifested through 89% greater cognitive decline variability in low-education groups (σ intercept = -0.089 vs. -0.168). Education reduced annual decline by 8.7% (-0.042 vs. -0.046 points/year) while elevating baseline cognition (μ = 1.99). CONCLUSIONS: As the first large-scale normative framework in China's diverse aging population, it enables precision public health through education-specific screening thresholds for the digital TCSA tool.