Dynamic Associations Between Social Interaction and Cognitive Function in Community‐Dwelling Older Adults: A Three‐Wave Comparison of Cross‐Lagged Panel and Random‐Intercept Cross‐Lagged Panel Models

ABSTRACT This study aimed to clarify the longitudinal relationship between social interaction and cognitive function in later life. Longitudinal data were analysed from a cohort of 340 older adults across three waves (2014–2020). We employed Bayesian Cross‐Lagged Panel Models (CLPM) and Random‐Intercept Cross‐Lagged Panel Models (RI‐CLPM) to examine the reciprocal association between the Index of Social Interaction (ISI) and cognitive impairment risk. Higher social interaction at baseline (T1, 2014) predicted a lower risk of later cognitive impairment. After partitioning out trait‐like differences, a robust within‐person protective effect remained: a positive deviation in an individual's social interaction at baseline (relative to their own average) significantly predicted a reduced risk of later cognitive impairment. Autoregressive effects of cognitive risk were primarily driven by between‐person stability. No significant cross‐lagged effects were observed for changes between later timepoints, potentially reflecting the impact of advanced ageing or the COVID‐19 pandemic during the final wave. Proactively increasing social engagement in the young‐old stage provides a significant protective effect against cognitive decline. These findings support social prescribing as a viable non‐pharmacological strategy and emphasize that interventions targeting dynamic behavioural changes may be more effective than those focusing solely on baseline social status.