Cognitive function and its change over time: effects on depression trajectories in oldest-old adults

Using data from the Health and Retirement Study (1992-2020), this study investigates how cognitive function and its longitudinal changes influence trajectories of depressive symptoms among oldest-old adults.

Latent growth curve modeling (LGCM) was applied to examine changes in cognitive function and depressive symptoms across five time points (ages 90, 92, 94, 96, and 98+). The analytic sample consisted of 289 participants at baseline (age 90), with sample sizes ranging from 289 to 231 across subsequent age groups. In comparison, we conducted an additional LGCM longitudinal analysis for participants at an earlier baseline at age 60 (N = 17036), following them from 60 to 70 (n = 11915) and 80 years (n = 8942).

The unconditional LGCM (without covariates) showed that higher baseline cognitive function was significantly associated with lower baseline depressive symptoms (β = -0.304, p < 0.001), but did not predict changes in depression over time. In the conditional model (including gender, education, self-rated health, and activities of daily living as covariates), higher education was associated with higher cognitive functioning but also with higher depression at baseline. In contrast, better ADL functioning and better perceived health significantly predicted higher cognitive functioning and lower depression at baseline. The intercept of depression predicted its own slope (β = -0.934, p < 0.001), indicating that Individuals with higher initial depressive symptoms experience a slower increase (or less worsening) in depression over time. For the 60-80 group, the intercept of cognition significantly predicted the intercept of depression (β = -0.380, p < 0.001), consistent with the 90-98+ group. However, longitudinal associations differed across age groups, indicating distinct developmental patterns between mid-to-late and very late adulthood.

This study emphasizes that although cognitive and emotional health are interrelated at baseline, changes in one domain do not reliably predict changes in the other among the oldest-old adults population. Instead, education, physical functioning, and self-rated health emerged as key predictors of both cognitive and emotional well-being. These findings highlight the importance of multidimensional assessments and targeted support to promote resilience and quality of life in extreme old age.