An analysis of dynamic, bidirectional associations between memory and verbal fluency with depressive symptoms in middle- and older-aged adults: A cohort study

Objectives: This study compared two approaches to analyzing bidirectional associations between aspects of cognition-specifically, verbal memory and fluency-and depression using multi-wave longitudinal data. The cross-lagged panel model (CLPM) does not distinguish between- versus within-person variation, whereas the random-intercepts CLPM (RI-CLPM) partitions variation into a stable, trait-like component that varies across individuals and a wave-specific deviation that varies within individuals. Methods: Observational study of 47,719 adults ages 45 to 95 from the Survey of Health, Ageing, and Retirement in Europe (mean age = 63 years, SD = 9 years at first assessment; 57 % female). Participants completed at least four of 8 biennial waves of data collection from 2004 and 2020. Depressive symptoms were assessed by the EURO-D. Cognition was assessed by animal fluency, and immediate and delayed word recall. Multi-group CLPMs and RI-CLPMs were fit on middle-aged (45-65 year) and older (65+ years) males and females. Results: The estimates from the CLPMs indicated bidirectionality in the associations between cognition and depression, whereas the RI-CLPMs provided clearer evidence that changes in depression may produce subsequent changes in cognition, rather than vice versa. Prospective associations were small and consistent across age, sex and cognitive measure. Limitations: Measures of executive function-a domain of cognition affected by aging and related to mood-were not available. Conclusions: CLPMs and RI-CLPMs address conceptually distinct questions about how verbal memory and fluency may be dynamically related to depression, and therefore, produce different inferences from the same data about how these aspects of cognition and depression are related.