Fluid Intelligence Predicts Change in Depressive Symptoms in Later Life: The Lothian Birth Cohort 1936

Abstract

We examined reciprocal, time-ordered associations between age-related changes in fluid intelligence and depressive symptoms. Participants were 1,091 community-dwelling older adults from the Lothian Birth Cohort 1936 study who were assessed repeatedly at 3-year intervals between the ages of 70 and 79 years. On average, fluid intelligence and depressive symptoms worsened with age. There was also a dynamic-coupling effect, in which low fluid intelligence at a given age predicted increasing depressive symptoms across the following 3-year interval, whereas the converse did not hold. Model comparisons showed that this coupling parameter significantly improved overall fit and had a correspondingly moderately strong effect size, accounting on average for an accumulated 0.9 standard-deviation increase in depressive symptoms, following lower cognitive performance, across the observed age range. Adjustment for sociodemographic and health-related covariates did not significantly attenuate this association. This implies that monitoring for cognitive decrements in later life may expedite interventions to reduce related increases in depression risk.

Keywords: depression; dynamic; intelligence; lead-lag; longitudinal change.

Fig. 1.

Bivariate latent-change-score model. Fluid intelligence (GF) and depressive symptoms (DS) are modeled across 3-year age intervals (i.e., ages 70, 73, 76, and 79 years—indicated by subscripts). Unlabeled paths with single-headed arrows show regression effects fixed at 1. Unlabeled paths with double-headed arrows show freely estimated variances and covariances. Labels on paths show model-parameter constraints. Time-invariant covariates (e.g., sex, education) and time-varying covariates (e.g., cardiovascular disease) are not shown. Lagged correlations for Gf uniquenesses (r_is) are shown only for 3-year intervals (Lag 1), but the models also included correlated uniquenesses for 6-year (Lag 2) and 9-year (Lag 3) intervals. Slope = static linear change; i1 = Block Design; i2 = Matrix Reasoning; i3 = Spatial Span observed scores.

Fig. 2.

Mean trajectories of fluid-intelligence (GF) performance within subsamples of participants stratified by baseline levels of depressive symptoms (DS; a) and mean trajectories of estimated DS within subsamples stratified by baseline levels of GF (b). Stratification was by interquartile range: low = 0% to 25%, mid = 26% to 75%, high = 76% to 100%. Gray bands indicate 99% confidence intervals.

Fig. 3.

Vector field plot in which arrows show the direction and relative magnitude (longer = stronger effect) of expected changes in fluid intelligence and depressive symptoms. The dashed ellipse demarcates the 90% highest-probability-density region for the estimated true scores.