Loneliness and objective social isolation in observed depressive symptom transitions across six ageing cohorts

Abstract Background Loneliness and objective social isolation may have different associations with changes in depressive symptoms. We examined transitions into and out of elevated symptom states across six ageing cohorts. Methods Adults aged 50–110 years in CHARLS, ELSA, HRS, KLoSA, MHAS and SHARE contributed 276,045 intervals between consecutive eligible observations. Cohort models related starting loneliness and a three component isolation index to subsequent symptom state using a complementary log-log link, nominal interval offset and participant clustered standard errors. Exponentiated coefficients were pooled using random effects meta-analysis and interpreted as cloglog association ratios (CARs), not observed recovery speeds. Results Per one cohort standard deviation higher loneliness, pooled CARs were 1.43 (1.27–1.61) for onset and 0.82 (0.78–0.85) for recovery. Corresponding isolation estimates were 1.03 (0.98–1.08) and 0.98 (0.93–1.02) per component. After correcting an overlapping-item score and adjusting for baseline symptom severity in four cohorts, the loneliness recovery CAR was 0.87 (0.83–0.91), with substantial heterogeneity and a prediction interval crossing 1 (0.69–1.10). Restricted-wave analyses were directionally similar.