Intrinsic capacity impairment and incident dementia across three population-based cohorts: a harmonized individual participant data analysis

Abstract

Objectives To examine whether intrinsic capacity impairment predicts incident dementia across three countries and to evaluate its value for risk stratification. Design : Harmonized multicohort study with individual participant data meta-analysis. Setting and participants Community-dwelling adults aged ≥50 years without baseline dementia from CHARLS (n = 17,708), ELSA (n = 9,432), and HRS (n = 22,034). Measures Intrinsic capacity was assessed across five domains: cognition, locomotion, vitality, sensory, and psychological functioning. Multidomain impairment was defined as deficits in ≥3 domains. Results Pooled hazard ratios rose in graded fashion: 1.31 (95% CI 1.22–1.42) for one impaired domain, 1.65 (1.48–1.85) for two, and 2.41 (2.18–2.68) for ≥3 domains. A Psych-Cognitive impairment pattern approached Multidomain risk (HR 2.15 vs 2.68). Rapid IC decline predicted 42.3% cumulative dementia incidence over six years. Adding IC to a demographic model improved the C-index from 0.69 to 0.76 with net benefit on decision curve analysis. Associations persisted when the cognitive domain was excluded and with death as a competing event; psychological functioning was the most consistent independent domain. Conclusions Intrinsic capacity impairment robustly predicts incident dementia across diverse populations. Its low cost and universal applicability support integration into geriatric assessment and primary care for early dementia risk stratification.