1. Longitudinal Relationship of Cognitive Functioning with Depressive Symptoms in Older Adults Archana Jajodia, Ph.D. University of Southern California

2. Collaborators Archana Jajodia (first author) Andrew Revell John J. McArdle Lesley A. Ross Milton Strauss Vonetta Dotson

3. Aims Examine rates of change of symptoms of depression in older adults over time Examine rates of change of cognitive decline in executive functioning and memory Examine longitudinal relationship of symptoms of depression with cognitive decline in executive functioning and memory

4. Background Depression is related to cognitive decline in older adults  Executive functioning and memory Longitudinal evidence is less conclusive Direction of effects is not well-established Cognitive decline preceding depression has not been explored Delineating temporal relationships between cognitive functioning and depression may be key to identify mechanisms

5. Hypotheses Executive functioning and memory will decline with age and depression will increase Greater depression will be associated with  lower memory and executive functioning  greater decline in memory and executive functioning with age Lower executive functioning will predict subsequent increase in depression Higher depression will predict subsequent decline in memory

6. Measures Data from 15 time points Executive functioning  Fluency: animals, vegetables, fruits  Digit span backwards  Alphabet span  Number comparisons Memory  East Boston story: immediate, delayed recall  Word list learning: immediate, delayed, recognition Depressive Symptoms  CES-D (10 items)

7. Analytic Plan Univariate latent change models for  Depressive symptoms  Memory  Executive functioning Bivariate dynamic change models for  Memory and depressive symptoms  Executive functioning and depressive symptoms

8. X1X2X15 e X1  ex e X2  ex e X15  ex x1x2x15  x1  x14 … Building the Dynamic Latent Difference Score Model (McArdle et al., 2001) Step1. Modeling Latent Change

9. X1 1  x0  x1  x1  x0  x0,x1 x inter * x slope x inter X2X15 e X1  ex e X2  ex e X15  ex x1x2x15  x1  x14 x slope * xx xx … Step 2. Adding Slope and Intercept (Constant Change over Time Model)

10. xx X1 1  x0  x1  x1  x0  x0,x1 x inter * x slope x inter xx X2 xx X15 e X1  ex e X2  ex e X15  ex x1x2x15  x1  x14 x slope * xx xx … Covariates Step 3. Adding Proportional Change and Covariates (Dual Change over Time Model)

11. 1  y0  y0,y1  y1  y1  y0 y inter y inter * y slope * y slope Y1Y2Y15 e Y1  ey e Y2  ey e Y15  ey y1y15  y1  y14 yy yy yy yy yy y2 … Covariates Repeat for other Variable

12. xx X1  x,y 1  y0  y0,y1  y1  x0  x1  x1  x0  y1  y0  x0,x1  x1,y1  y0,x1  x0,y1  x0,y0 y inter y inter * y slope * y slope x inter * x slope x inter xx X2 xx X15 e X1  ex e X2  ex e X15  ex Y1Y2Y15 e Y1  ey e Y2  ey e Y15  ey y1y15 x1x2x15  x1  x14  y1  y14 yy yy yy x slope * yy xx xx yy y2 … … Covariates Bivariate Model

13. xx X1  x,y 1  y0  y0,y1  y1  x0  x1  x1  x0  y1  y0  x0,x1  x1,y1  y0,x1  x0,y1  x0,y0 y inter y inter * y slope * y slope x inter * x slope x inter xx X2 xx X5 e X1  ex e X2  ex e X5  ex Y1Y2Y5 e Y1  ey e Y2]  ey e Y4  ey y1y5 x1x2x5  x1  x4  y1  y4 yy yy yy  yx  xy x slope * yy xx xx yy y2 … … Covariates Adding Cross loadings

14. Results with Memory and Depressive Symptoms Proportional change (βs):  Memory t  ΔMemory t+1 = 0  CESD t  ΔCESD t+1 = 0.11 symptoms (11% increase, p- value = 0.001) Cross-loadings (γs):  Memory t  ΔCESD t+1 = 0.013 symptoms/unit score p-value = 0.007  CESD t  ΔMemory t+1 = -1.5 units/ symptom p-value = 0.00007 Model Fit:  χ 2 (552) = 1,215  RMSEA = 0.033

15. Results with Memory and Depressive Symptoms CorrelationsCESD- intercept CESD-slopeMemory- intercept CESD-slope-0.97 Memory- intercept -0.11ns Memory- slope 0.97-0.98ns

16. Results with Memory and Depressive Symptoms: Covariates Regression Weights CESD- intercept CESD-slopeMem- intercept Mem-slope Age at baseline 0.17ns-0.420.13 Femalens 0.15ns Educationns 0.18ns

17. Results with Executive Functioning and Depressive Symptoms Proportional change (βs):  ExecFn t  ΔExecFn t+1 = 0.07 (7% increase, p-value = 0.0000007)  CESD t  ΔCESD t+1 = -0.12 symptoms (12% decrease, p- value = 0.004) Cross-loadings:  ExecFn t  ΔCESD t+1 = -0.10 symptoms/unit score p-value = 0.0000006  CESD t  ΔExecFn t+1 = ns Model Fit:  χ 2 (552) = 1,227  RMSEA = 0.033

18. Results with Executive Functioning and Depressive Symptoms CorrelationsCESD- intercept CESD-slopeExecFn- intercept CESD-slopens ExecFn- intercept -0.140.33 ExecFn- slope ns -0.39

19. Results with Executive Functioning and Depressive Symptoms: Covariates Regression Weights CESD- intercept CESD-slopeExecFn- intercept ExecFn- slope Age at baseline 0.16-0.23-0.43ns Femalens 0.14ns Educationns 0.21-0.24

20. Summary of Results More depressive symptoms associated with poorer memory and executive functioning  Not with decline in cognition Executive functioning leads to changes in depressive symptoms over time Depressive symptoms lead to changes in memory over time

21. Questions & Further Work Cognitive measures – ok to use composites? Distinguishing late onset and early onset depression? Integrate history of major depression in models