Abstract
Depressive disorders affect roughly 13.3% of older adults worldwide and are associated with falls, functional decline, disability, and mortality; yet, despite a highly heterogeneous course, late-life depression is often treated as a static condition. To better characterize these dynamic courses, in four large prospective ageing cohorts (HRS, ELSA, SHARE, and CHARLS; 57,894 participants), we classified participants into five dynamic phenotypes based on depressive symptom states (CES-D or EURO-D) at the first four assessments after cohort entry, using rule-based trajectory classification, and employed a landmark design to examine their temporal predictive relationship with subsequent frailty index accumulation. Five phenotypes were reproducibly identified across all four cohorts: persistently healthy, remitting, late-onset, fluctuating, and persistent high burden, the last accounting for 7.9%-20.3% of participants. The persistent high burden phenotype (P5) consistently predicted faster non-cognitive frailty accumulation in the three cohorts amenable to landmark analysis (HRS 1.41, ELSA 1.25, SHARE 1.84 × 10-3/year; meta-analysis 1.42, 95% CI 0.65-2.19, I2 = 0), with a positive but concurrent longitudinal association in CHARLS (3.46 × 10-3/year), and showed significantly elevated odds of incident frailty (FI ≥ 0.25) in all three cohorts (OR 1.29-2.75). The fluctuating phenotype (P4) was associated with frailty mainly concurrently rather than temporally, and education was a directionally consistent predictor of phenotype membership across the four cohorts (OR 0.32-0.80). Persistent high burden thus represents a reproducible risk marker of accelerated frailty accumulation, underscoring the value of sustained management of older adults with chronically high symptom burden.