Social isolation, self-perceived loneliness, and the 7-year trajectories of depression and cognitive decline: a prospective cohort study

Social disconnectedness is a critical risk factor for late-life depression, yet the longitudinal associations of its objective (social isolation) and subjective (loneliness) components with depression and cognitive trajectories remain unclear. We analyzed 7-year data from 2,780 adults (≥ 45 years) in a prospective cohort. Linear mixed-effects models assessed trajectories of depressive symptoms (CES-D-10) and cognitive function (Z-score). Cox models evaluated incident depression risk in 2,211 baseline depression-free participants. A machine learning model (logistic regression with SHAP) identified and ranked predictors of 7-year incident depression. Higher baseline isolation predicted accelerated increases in depressive symptoms (Time×High Isolation: β = 0.21, p < 0.001) and faster cognitive decline. Both loneliness (aHR = 2.15, 95% CI: 1.80–2.56) and high isolation (aHR = 1.89, 1.52–2.35) independently predicted incident depression; co-occurrence conferred the highest risk (aHR = 3.05, 2.48–3.75). The model performed well (AUC = 0.821), with SHAP identifying loneliness as the foremost predictor, surpassing baseline depressive symptoms, functional impairment, and biomarkers. Social disconnectedness—particularly loneliness—is a progressive risk factor for depression and cognitive decline in aging adults. Its strong predictive power underscores loneliness as a key target for screening and interventions.