Multidomain functional trajectories and time-varying mortality risk: A 10-year follow-up from the Korean Longitudinal...

Functional decline in aging may proceed asynchronously across cognition, mood, and physical capacity, yet existing approaches often examine single domains or collapse function into composite indices, obscuring meaningful patterns. This study aimed to identify latent multidomain functional trajectories and examine their time-varying associations with mortality. Data were drawn from the Korean Longitudinal Study of Aging (KLoSA). Joint Growth Mixture Modeling was applied to cognitive function (Korean Mini-Mental State Examination, K-MMSE), depressive symptoms (Center for Epidemiologic Studies Depression Scale, CES-D), and physical dependence (activities of daily living/instrumental activities of daily living, ADL/IADL) across Waves 1-3 (2006-2010) in 6,398 community-dwelling adults. Cox proportional hazards and time-varying coefficient models estimated mortality risk during Waves 3-8 (2010-2020). Four trajectory classes emerged: Robust Functioning (74.4%), Physical Limitation (11.2%), Cognitive-Affective Decline (8.3%), and Multidomain Decline (6.2%). During follow-up, 1,137 deaths (17.8%) occurred. In fully adjusted models, all non-robust classes showed elevated mortality risk: Multidomain Decline (HR 2.60, 95% CI 2.16-3.14), Cognitive-Affective Decline (HR 1.45, 95% CI 1.20-1.76), and Physical Limitation (HR 1.37, 95% CI 1.15-1.62). Time-varying analyses revealed distinct temporal patterns: Cognitive-Affective Decline exhibited early-phase risk elevation (0-5 years: HR 1.97) that attenuated thereafter, whereas Physical Limitation showed persistent modest elevation. Multidomain Decline carried the highest hazard in both periods. Multidomain functional trajectories capture heterogeneity in both magnitude and timing of mortality risk. By revealing distinct temporal risk dynamics across configurations, integrated monitoring of cognition, mood, and physical function may help characterize multidomain vulnerability, with relevance for risk stratification pending further validation.