Linking Depressive Symptoms and Cognitive Functioning: The Mediating Role of Leisure Activity

Abstract

Objectives:

Greater depressive symptoms are associated with cognitive decline in older adulthood, but it is not clear what underlying factors drive this association. One behavioral pathway through which depressive symptoms may negatively influence cognitive functioning is through activity engagement. Prior research has independently linked greater depressive symptoms to both lower leisure and physical activity and independently linked both lower leisure and physical activity to lower cognition. Therefore, depressive symptoms may negatively influence cognition by reducing engagement in beneficial leisure and/or physical activities that help to maintain cognition.

Methods.

The current study examined associations between depressive symptoms, leisure activity, physical activity, and global cognitive functioning using longitudinal data from the Washington Heights-Inwood Columbia Aging Project (WHICAP; n = 5,458 older adults). A multilevel structural equation model estimated the between-person and within-person effects of depressive symptoms on global cognition through leisure and physical activity.

Results.

Leisure activity, but not physical activity, mediated the association between depressive symptoms and global cognition between- and within-persons. When individuals reported high depressive symptoms, they also reported fewer leisure activities, which was associated with lower global cognition.

Conclusion.

These findings highlight behavioral pathways through which depressive symptoms may negatively influence cognitive functioning. Findings support the view that perhaps depressive symptoms act as a risk factor for cognitive impairment by reducing leisure activity engagement.

In older adulthood, depressive symptoms have been linked to lower cognitive functioning. Longitudinal studies have demonstrated that late life depressive symptoms are associated with faster memory decline (Lohman et al., 2013; Zahodne, Stern & Manly, 2014) as well as higher rates of incident mild cognitive impairment and dementia (Richard et al., 2013). Despite prior research demonstrating a prospective association between depressive symptoms and cognitive functioning, mediators of the association between depressive symptoms and cognitive functioning in later life are not yet clear. In the current study, we specifically examined modifiable behavioral pathways that may underlie the relationship between depressive symptoms and cognitive functioning.

One potential pathway through which depressive symptoms could increase risk of late-life cognitive impairment is through a decrease in activities that help to preserve functioning in older adulthood. Depressive symptoms have been associated with a loss of interest or pleasure in potentially-rewarding activities, a decrease in motivation to engage in these activities (Scarapicchia et al., 2014), and reduced energy for leisure activities (Leibold, Holm, Raina, Reynolds & Rogers, 2014). For example, in a cross-sectional study, nursing home residents with depression engaged in fewer social activities that can lead to positive reinforcements from others, as compared to residents without depression (Achterberg et al., 2003). Depressive symptoms are also prospectively associated with greater risk for a sedentary lifestyle and decreased physical exercise (Roshanaei-Moghaddam, Katon & Russo, 2009; Scarapicchia et al., 2014).

Reductions in activity engagement due to depressive symptoms may have negative implications for cognitive functioning. In particular, engagement in leisure activities, such as playing cards or board games and visiting with friends, has been found to be a protective factor against age-related cognitive declines. Greater participation in leisure activities has been associated with better cognitive functioning (Hultsch, Hertzog, Small & Dixon, 1999; Ihle et al., 2019; Mueller, Raymond & Yochim, 2013) and lower risk of dementia (Jonaitis, et al., 2013; Scarmeas, Levy, Tang, Manly & Stern, 2001; Verghese et al., 2003; 2006). Leisure activities may benefit cognitive functioning through cognitive enrichment (Valenzuela, Sachdev, Wen, Chen & Brodaty, 2008), social and/or physical stimulation (Hertzog, Kramer, Wilson & Lindenberger, 2009) that helps to reduce the negative effects of common age-related neuropathology (Wilson & Bennet, 2003). For example, older adults who were taught how to juggle in an intervention study showed significant increases in gray matter volume in brain regions involved in complex visual processes compared to the control group (Boyke, Driemeyer, Gaser & Büchel & May, 2008), suggestive that leisure activities may help to maintain brain functioning that supports cognition.

Further, engagement in physical activity, such as walking or running, has also been independently linked to cognitive functioning in older adulthood (Calamia, et al., 2018; Ferencz et al., 2014; Scarmeas et al., 2009). Indeed prior intervention research has shown a protective effect of engaging in aerobic exercise for cognitive health in older adulthood (Erickson et al., 2011). Cross-sectional work has also found self-reported physical activity to buffer the negative effects of genetic risk (i.e., presence of risk alleles PICALM, BIN1, & CLU) on episodic memory (Ferencz et al., 2014). Higher self-reported physical activity has also been linked to lower Alzheimer’s disease risk (Scarmeas et al., 2009). Physical activity may benefit cognitive functioning by reducing risk of diseases that may negatively affect cognition (Warburton, Nicol & Bredin, 2006) as well as by improving cardiorespiratory functioning, which is associated with improvements in cognition (Rogers, Meyer & Mortel, 1990).

Overall, depressive symptoms have been linked to lower leisure and physical activity engagement (Achterberg et al., 2003; Leibold et al., 2014; Roshanaei-Moghaddam et al., 2009; Scarapicchia et al., 2014), and lower leisure and physical activity engagement has been linked to worse cognitive functioning (Hultsch et al., 1999; Jonaitis et al., 2013; Mueller et al., 2013; Scarmeas et al., 2001; Verghese et al., 2003; 2006). However, it is unclear whether leisure and/or physical activity engagement mediate the relationship between depressive symptoms and cognitive functioning. Therefore, the goal of the present study was to examine the mediating role of activity engagement in the relationship between depressive symptoms and cognitive functioning.

To assess the underlying behavioral pathways linking depressive symptoms and cognitive functioning, we examined both within-person and between-person associations between depressive symptoms, leisure activity, physical activity, and global cognition using a multilevel structural equation model approach (MSEM; Preacher, Zyphur & Zhang, 2010). While multilevel modeling is frequently used to model intraindividual variability in time-varying variables, the use of MSEM allows for the examination of both within-person and between-person differences in time-varying variables. That is, individuals may have overall differences at the between-person level (i.e., individual differences) as well as within-person fluctuations across waves (i.e., intraindividual variability). We hypothesized that leisure activity and physical activity would each independently mediate the relationship between depressive symptoms and global cognition at the within-person and between-person levels.

Methods

Participants and Procedure.

Data from the Washington Heights-Inwood Columbia Aging Project (WHICAP; Manly et al., 2005; Tang et al., 2001) were used. The WHICAP is a prospective, community-based longitudinal study of individuals at least 65 years old in Northern Manhattan. These individuals were initially identified to participate through Medicare records or a commercial marketing company in three waves: 1992, 1999, and 2009. Participants were subsequently follow-up every 18 to 24 months and participants had, on average 3 time points of available data (SD = 1.90; range 1 – 13, 69% > 1 time point). In the current study, data from waves in which participants met diagnostic criteria for dementia (see Stern et al., 1992 for detailed information) were excluded due to concerns regarding the validity of their self-report data (i.e., depressive symptoms and activity engagement). The final sample of the study was 5,458 individuals. All participants received and signed informed consent and were reimbursed for their participation for each assessment. Data collection procedures were approved Columbia University’s Institutional Review Board (IRB). Secondary data analyses were approved the University of Michigan’s IRB.

Measures.

Depressive Symptoms.

Depressive symptoms were measured across time points using the short version of the Center for Epidemiologic Studies-Depression Scale (CESD; Irwin, Haydari & Oxman, 1999). Participants answered Yes (1) or No (0) to 10 items assessing depressive symptoms such as “I felt that everything I did was an effort” and “I felt depressed.” Scores were summed across items, and depressive symptoms could range from 0 (no depressive symptoms present) to 10 (all depressive symptoms present).

Physical Activity.

Engagement in physical activities were measured with the Godin Leisure Time Exercise Questionnaire (Godin & Shepard, 1985), which assessed the amount of light, moderate and vigorous activities the participant engaged in over the past 14 days. For light, moderate and vigorous exercise, participants reported (a) whether they engaged in this activity by reporting Yes (1) or No (0), (b) the number of times, and (c) the total number of minutes. Based on previous research (i.e., Scarmeas et al., 2009), a summary score of their total physical activity engagement was calculated for each individual using the following formula: number of minutes x number of times x coefficients (9 for vigorous, 5 for moderate, 3 for light exercise, which correspond to the metabolic equivalent). This metabolic equivalent (MET) expresses the energy cost consumption during specific physical activities as multiples of resting metabolic rates.

Leisure Activity.

Engagement in leisure activities was assessed with a check list that measured whether or not the participant engaged in 13 different social and recreational activities within the past two weeks (Scarmeas et al., 2001). Items specifically asked participants if they (a) spent time on any hobby, (b) gone for a walk, (c) gone to visit friends, (d) received visits at home, (e) done physical exercise, (f) gone to the movies, (g) read a magazine, (h) watched TV, (i) done volunteer work, (j) played cards, (k) gone to a club or center, (l) gone to classes, (m) gone to church. Participants reported Yes (1) if they engaged in the activity or No (0) if they did not. The total number of activities was summed, and scores could range from 0 to 13.

Global Cognition.

Cognitive functioning was assessed with a comprehensive neuropsychological battery encompassing episodic memory, language, visuospatial and speed/executive functioning and has been described in previous research (i.e., Siedlecki, Manly, Brickman, Schupf, Tang, & Stern, 2010; Stern et al., 1992). Episodic memory composite scores included immediate, delayed and recognition trials from the Selective Reminding Test (Buschke & Fuld, 1974). Language scores included measures of naming, letter and category fluency, verbal abstract reasoning, repetition, and comprehension. Visuospatial scores included recognition and matching trials from the Benton Visual Retention Test (Benton, 1955), the Rosen Drawing Test (Rosen, 1981), and the identities and Oddities subtest of the Dementia Rating Scale (Mattis, 1976). Speed/Executive Functioning scores included both trials of the Color Trails test. Cognitive domain composite scores were derived by converting cognitive variables into Z-scores and averaging them for domain in the larger WHICAP sample. The four cognitive domains were highly correlated (.51 < r < .72). Consistent with previous research (see Gu et al., 2015; Wilson, Boyle, James, Buchman & Bennett, 2015; Wilson, Rajan, Barnes, Weuve & Evans, 2016), a composite across all 4 domains was created by averaging the four domain scores to represent global cognition (α = .81).

Covariates.

Analyses controlled for main effects of age, race, sex/gender, education, activities of daily living (ADLs), and recruitment year. Sex/gender was self-reported at the initial WHICAP visit. Race was self-reported by participants and subsequently dummy-coded into two variables representing African Americans and Caribbean Hispanics in reference to non-Hispanic Whites. Education was represented by the number of years of school (0 – 20). Activities of daily living were assessed using the Blessed Functional Activity scale and higher scores represented greater impairment. Finally, recruitment year (1992, 1999 & 2009) was controlled for to account for cohort differences. Age and ADLs were assessed across waves and thus, were controlled for at both the within-person and between-person levels. Sex/gender, race, ethnicity, education, and recruitment year were controlled for at the between-person level.

Statistical analysis.

In order to address our research questions and due to the nested structure of the data (time points within individuals), a multilevel structural equation model (MSEM) was conducted using Mplus, Version 8 (Múthen & Múthen, 2007) following recommendations by Preacher and colleagues (Preacher et al., 2010). Syntax for this model can be found in Supplementary Material. MSEM, unlike traditional multilevel modeling, combats conflation of within-person and between-person variance by allowing Level 1 variables (i.e., time-varying variables) to be decomposed at both the within-person and between-person level. Analyses corresponded to the 1-(1,1)-1 design in which the predictor (depressive symptoms), mediators (physical activity, leisure activity), and outcome (global cognition) were all assessed at Level-1 (Preacher et al., 2010). Indirect effects were quantified based on the products of constituent paths as recommended by Hayes (2009), and thus, were calculated through the computation of the products of a*b (a = coefficient estimate of the association between the independent variable and the mediating variable, b = the coefficient estimate of the relationship between the mediating variable and the outcome variable) (Preacher et al., 2010).

Of our sample, 87.17% had a baseline visit, 55.50% had a first follow-up, 33.73% had a second follow-up, 17.31% had a third follow-up, 10.39% had a fourth follow-up and 6.82% had a fifth follow-up. Importantly, missing data were managed with full information maximum likelihood with robust standard errors.

Results

Means and standard deviations for all variables at baseline are described in Table 1. Correlations between variables of interest and intraclass correlations (ICC) are described in Table 2. The ICC was calculated initially with a fully unconditional model with no predictors to assess the partition of variance at the within- and between-person levels. ICCs revealed sufficient variance at the within- and between-person levels for a multilevel model approach across Level-1 variables (i.e., time-varying; see Table 1). Subsequently, a multilevel structural equation model (MSEM) was conducted and the standardized coefficients, standard errors, and indirect effects are listed in Table 3. Standardized within and between-person mediation pathways from the MSEM are depicted in Figure 1.

Table 1.

Means and Standard Deviations at Baseline.

	M	SD	Range
Age	76.28	6.70	60–109
Female (%)	67.00	-	-
Non-Hispanic Black (%)	29.50	-	-
Hispanic (%)	44.00	-	-
1999 Cohort (%)	44.10	-	-
Education (years)	10.60	4.97	0–20
Activities of Daily Living	1.22	1.48	0–8
Global Cognition	0.29	0.60	−2.87–1.83
Depressive Symptoms	1.78	2.05	0–10
Leisure Activity	7.04	2.51	0–13
Physical Activity (MET)	1363.03	2461.16	0–45960

Table 2.

Intraclass Correlations and Correlations for Variables of Interest at Baseline

	ICC	1	2	3	4
1. Global Cognition	.79
2. Depress	.49	−.18***
3. Leisure	.54	.40***	−.27***
4. Physical	.39	.23***	−.05**	.26***

Note. ICC = intraclass correlation, Depress = depressive symptoms

^*= p < .05

^**= p < .01

^***= p <.001.

Table 3.

MSEM Mediation predicting Global Cognition from Depressive Symptoms and Activity Engagement.

Parameter	Estimate	SE	Estimate/SE	95% CI
Between-Persons Effects
Intercept	−.29***	.07	−4.45	[−.42, −.17]
DEPRESS → LEIS	−.23***	.04	−5.91	[−.30, −.15]
DEPRESS → PHYS	−.11*	.06	−1.98	[−.21, .−.001]
DEPRESS → COG	.05	.03	1.74	[−.01, .10]
LEIS → COG	.16***	.02	7.59	[.12, .20]
PHYS → COG	.04	.02	1.74	[−.01, .09]
LEIS ←→ PHYS	.35***	.07	5.10	[.21, .48]
ADL → DEPRESS	.66***	.02	27.55	[.61, .70]
ADL → LEIS	−.34***	.04	−8.78	[−.42, −.26]
ADL → PHYS	−.15	.08	−1.80	[−.32, .01]
ADL → COG	−.17***	.03	−5.72	[−.23, −.11]
Age → DEPRESS	−.04*	.02	−2.12	[−.08, −.01]
Age → LEIS	−.12***	.02	−6.22	[−.15, −.08]
Age → PHYS	−.31***	.05	−6.00	[−.42, −.21]
Age → COG	−.20***	.02	−12.90	[−.23, −.17]
SEX → DEPRESS	.12***	.02	7.53	[.09, .15]
SEX → LEIS	.03*	.02	1.98	[.00, .06]
SEX → PHYS	−.10**	.04	−2.71	[−.17, −.03]
SEX → COG	.06***	.01	4.91	[.03, .08]
EDU → COG	.52***	.01	36.40	[.49, .54]
Cohort → COG	.05***	.01	3.96	[.02, .07]
Black → COG	−.21***	.01	−15.60	[−.23, −.18]
HISP → COG	−.24***	.02	−14.56	[−.28, −.21]
LEIS Indirect Effect	−.05***	.01	−4.46	[−.06, −.03]
PHYS Indirect Effect	−.01	.00	−1.24	[−.02, .01]
MEM Residual Variance	.46***	.01	42.51	[.44, .48]
Within−Person Effects
DEPRESS → LEIS	−.08***	.01	−6.71	[−.11, −.06]
DEPRESS → PHYS	.00	.01	0.09	[−.03, .03]
DEPRESS → COG	−.04**	.01	−2.97	[−.06, −.01]
LEIS → COG	.11***	.01	10.11	[.09, .13]
PHYS → COG	.03***	.01	2.85	[.01, .04]
LEIS ←→ PHYS	.11***	.01	8.92	[.08, .13]
ADL → DEPRESS	.10***	.01	6.76	[.07, .13]
ADL → LEIS	−.06***	.01	−4.81	[−.08, −.04]
ADL → PHYS	.00	.01	0.82	[−.03, .03]
ADL → COG	−.07***	.01	−5.73	[−.10, −.05]
Age → DEPRESS	.02	.01	1.72	[−.00, .05]
Age → LEIS	−.28***	.01	−23.24	[−.31, −.26]
Age → PHYS	.00	.01	0.06	[−.03, .03]
Age → COG	−.40***	.01	−36.22	[−.43, −.38]
LEIS Indirect Effect	−.01***	.01	−5.56	[−.01, −.004]
PHYS Indirect Effect	.00	.00	0.09	[.00, .00]
MEM Residual Variance	.79***	.01	82.35	[.77, .81]

Note. Standardized coefficients are reported. DEPRESS = depressive symptoms, LEIS = leisure activities, PHYS = physical activities, COG = Global Cognition, SEX = sex/gender

^*= p < .05

^**= p < .01

^***= p < .001

Figure 1.

MSEM 1-(1, 1)-1 Mediation Model. Standardized estimates are reported. For simplicity, covariates are not depicted. Note. DEPRESS= depressive symptoms, LEIS = leisure activity, PHYS = physical activity, COG = global cognition.

Within-Person Effects.

As shown in Table 3, leisure activity significantly mediated the relationship between depressive symptoms and global cognition at the within-person level. When individuals reported higher depressive symptoms, individuals also reported lower leisure activity engagement, which was associated with lower global cognition. Depressive symptoms were not associated with physical activity, but physical activity was positively associated with global cognition. The within-person indirect effect for physical activity as a mediator was nonsignificant. After accounting for mediators and covariates, a significant direct effect of depressive symptoms on global cognition still remained at the within-person level. On waves when individuals reported higher depressive symptoms, individuals also reported lower global cognition. In addition, main effects of age and ADLs were significant. As persons aged, global cognition decreased. Further, as ADL limitations increased, global cognition decreased. This model accounted for 21% of the within-person variance in global cognition, and 14% of the association between depressive symptoms and global cognition at the within-person level as explained by leisure activity.

Between-Person Effects.

As shown in Table 3, leisure activity significantly mediated the relationship between depressive symptoms and global cognition at the between-person level. Across waves, individuals who reported more depressive symptoms also had lower overall engagement in leisure activities and in turn, lower overall global cognition. Depressive symptoms were associated with less physical activity, but physical activity was not associated with global cognition.

Significant effects of age, sex/gender, ADLs, education, recruitment year, race and ethnicity were found. Older age and greater ADL limitations were associated with lower global cognition. Non-Hispanic Whites reported higher global cognition scores than African Americans and Caribbean Hispanics. A more recent recruitment year, being female and higher education were associated with higher overall global cognition. This model accounted for 55% of the between-person variance in global cognition, and 20.83% of the association between depressive symptoms and global cognition at the between-person level was explained by leisure activities.

Sensitivity Analyses

Although all four cognitive domains that comprised the global cognition score were highly correlated and showed good internal consistency, it may be the case that the individual cognitive domains may show a distinct pattern of associations with depressive symptoms and activity engagement. Therefore, we conducted sensitivity analyses by running a MSEM model separately for each of the four domains (memory, language, visuospatial functioning and speed/executive functioning). Importantly, leisure activity, but not physical activity significantly mediated the relationship between depressive symptoms and each cognitive domain. Across cognitive domains, the within-person and between-person indirect effects were similar in size (see Supplementary Table 1).

Additionally, ADLs were included as a covariate in our primary analytic model to control for potential confounding by mobility/health; however, ADLs may also be a downstream effect of cognitive decline. Therefore, we conducted a sensitivity analysis replacing ADLs with chronic illness burden, another commonly used index of health (see Supplementary Table 2). Chronic illness burden was a time-varying covariate (i.e., measured across waves) and represented by the number of self-reported chronic conditions out of a list of 10 potential conditions. Importantly, results were nearly identical to those controlling for ADLs.

As 2-items of the leisure activities scale specifically ask about physical activities (i.e., gone for a walk, physical exercise), we also conducted a sensitivity analysis with a recoded leisure activities score excluding these two items (see Supplementary Table 2). Findings revealed identical to those previously reported. Further, a sensitivity analysis was conducted using Bayesian credible intervals as an alternative method for significance testing of indirect effects (Múthen & Asparouhov, 2012). Importantly, the pattern of findings was identical those previously reported using inferential methods (i.e., p-values; see Supplementary Table 2).

Finally, our main analytic model excluded data from waves in which participants had a consensus diagnosis of dementia as our measures of depressive symptoms and activity engagement were self-reported. However, as the exclusion of data in which participants had a consensus diagnosis of dementia may truncate the range of the dependent variable, a sensitivity analysis was conducted that included all data from participants regardless of dementia status (see Supplementary Table 2). The pattern of associations was consistent with that previously reported.

Discussion

We examined whether engagement in leisure and physical activities mediated the relationship between depressive symptoms and global cognition in older adulthood. Overall, we found that at both the within-person and between-person levels, leisure activity, but not physical activity, mediated the relationship between depressive symptoms and global cognition. This pattern of associations was found for global cognition as well as when each cognitive domain (memory, language, visuospatial and speed/executive functioning) was examined separately. After accounting for covariates and mediators, a significant direct effect of depressive symptoms on global cognition remained at the within-person level, suggestive that other mechanisms may be at play that link depressive symptoms to cognitive functioning in older adulthood.

Disengagement from Leisure Activities

Our findings in the current study are consistent with previous work that found that depressive symptoms cause fatigue and decreased motivation to engage in activities (Achterberg et al., 2003; Leibold et al., 2014), and engagement in leisure activities was beneficial for cognition in older adulthood (Ihle et al., 2019; Mueller et al., 2013, Wang et al., 2006). This study extended previous work by estimating the indirect effect of depressive symptoms on global cognition through leisure activity engagement. Our finding that leisure activities mediated the association between depressive symptoms and global cognition at the between-person level are consistent with prior research demonstrating that older adults who engage in leisure activity not only better maintain their cognitive performance (Mueller et al., 2013), but also have reduced risk for Alzheimer disease (Scarmeas et al., 2001). These between-person results could support either the view that depressive symptoms function as a risk factor for cognitive decline or that they reflect a prodrome of cognitive decline such that depressive symptoms are an early sign of neurodegenerative disease (Jorm, 2000).

Further, we found that leisure activities mediated the association between depressive symptoms and memory at the within-person level and therefore, depressive symptoms could potentially operate as a risk factor through a reduction in activity engagement. Within individuals and across time, higher depressive symptoms were associated with lower engagement in leisure activities that may provide cognitive, social, and/or physical stimulation to help maintain cognitive functioning (Hertzog et al., 2009). Due to our design, we cannot conclude whether depressive symptoms acts as a risk factor or a prodrome of neuropathology. Future investigations should further examine the link between depressive symptoms and cognitive functioning in older adulthood, which could reflect risk factor or prodromal interpretations.

Linking Physical Activity to Depressive Symptoms and Episodic Memory

Physical activity engagement was not found to mediate the relationship between depressive symptoms and cognitive functioning, conflicting with our original hypothesis. Physical activity engagement, however, significantly predicted global cognition at the within-person level. Thus, higher physical activity engagement was associated with higher global cognition. This finding coincides with previous research linking exercise with better cognition (Calamia, et al., 2018; Ferencz et al., 2014; Scarmeas et al., 2009). For example, when examining the impact of objective physical activity on episodic memory, a greater number of steps per day was associated with better visual episodic memory in older adults, but not younger adults (Hayes et al., 2015). Therefore, disengagement from physical activity in older adulthood may have detrimental effects on cognitive functioning in later life.

When examining the relationship between depressive symptoms and physical activity, no significant within-person effect emerged, contrasting with our hypothesis that an increase in depressive symptoms would reduce physical activity. We did, however, find that overall depressive symptoms were associated with lower physical activity between-persons. These mixed findings may due to the nature of how physical activity was measured. Participants were asked to report the frequency (times, minutes) of physical activity and may be influenced by response bias. Prior research has found that participants have problems accurately recalling their absolute amount of physical activity (Prince et al., 2008) and construing activities as being physical. For example, an older adult who goes shopping with friends may view this activity as being inherently social and report it as a leisure activity but may not perceive it as being physically engaging (i.e., walking around stores). Therefore, a more objective measure of physical activity may help to better clarify the nature of these relationships. For example, in a prospective study, objective measures of physical activity (volute of expiratory oxygen at peak exertion) was associated with subsequent cognitive functioning, but self-report measures of physical activity were not (Barnes, Yaffe, Satariano, & Tager, 2003).

It may be the case that individuals are less prone to socially desirable responding when reporting engagement in leisure activities (i.e., doing hobbies, visiting friends) compared to physical activity. Additionally, the distinctions between leisure and physical activity on cognitive functioning may be due to the differential pathways in which these unique activities influence cognition (see review; Cheng, 2016). Engagement in leisure activities may promote cognitive functioning via cognitive enrichment mechanisms (Valenzuela et al., 2008) whereas physical activity may benefit cognition through reduced cardiovascular disease risk (Warburton, Nicol & Bredin, 2006). Prior intervention studies have shown immediate effects of engagement in cognitively-stimulating leisure activities such as juggling (Boyke et al., 2008). Therefore, it may be the case that leisure activities have more immediate benefits to cognitive health whereas engagement in physical activity (i.e., reduced disease risk) may require longer durations for the benefits to accrue. Future research should examine activity engagement across the life course and its relation to cognitive health in older adulthood.

Overall, the differential effects of leisure and physical activity in explaining the relationship between depressive symptoms and cognitive functioning may be due to leisure activities being more directly and profoundly affected by depressive symptoms, as seen by correlation coefficients at baseline (see Table 2). However, there may also be self-reported bias that distinguish self-reported leisure and physical activity engagement.

Limitations and Future Directions

Although the current study sheds light on potential mediators of the relationship between depressive symptoms and cognitive functioning, there are some notable limitations. First, we examined self-reported measures of activity engagement. As previously described, bias exists in self-reported physical activity such as social desirability (i.e., over-reporting; Adams et al., 2005). This may potentially explain why physical activity engagement was not a significant mediator of the association between depressive symptoms and cognitive functioning. Future studies should further assess how depressive symptoms affect objective assessments of physical activity and their link to cognitive functioning in older adulthood.

Second, future work should include clinical assessments of depression. The current community sample reported relatively low levels of depressive symptoms and therefore may not represent individuals with clinically-diagnosed depression. Although scores on the CESD have been associated with clinical diagnoses of major depressive disorder (i.e., Haringsma, Engels, Beekman & Spinhoven, 2004), there may be a qualitative difference between depressive symptoms and the presence of clinical depression. However, it is notable that a recent meta-analysis of prospective studies examining the link between depression and incident dementia reported comparable results when depression was operationalized using continuous measures of depressive symptomatology versus categorical ratings of clinical depression (Cherbuin, Kim & Anstey, 2015). Further, our findings highlight the robustness of the effects, as subclinical levels of depressive symptoms were still significantly associated with activity engagement and cognition in this community sample.

Finally, while the current study used longitudinal data to examine how within-person variability in depressive symptoms were associated with within-person variability in cognitive functioning, the use of concurrent measures within each wave cannot disentangle directionality. Although prior research has shown that depressive symptoms influence activity engagement (i.e., Leibold et al., 2014; Roshanaei-Moghaddam et al., 2009; Scarapicchia et al., 2014), it may also be the case that activity engagement influences depressive symptoms (Poelke et al., 2016; Lee, Lee, Brar, Rush & Jolley, 2014) and in turn, cognitive functioning. Due to the time scale of the current study (i.e., approximately 2 years between each assessment), lagged models were not viewed as theoretically appropriate for the modeling of the association between depressive symptoms and activity engagement. Thus, replication of our findings with more micro-longitudinal designs (i.e., weekly and/or daily assessments) as well as with the use of more experimental designs are necessary in order to verify the hypothesized causal effects of depressive symptoms on activity engagement and cognition.

Conclusion

In conclusion, this study extended previous work by estimating an indirect path of depressive symptoms to cognitive functioning through activity engagement within individuals. We found that leisure activity, but not physical activity, significantly mediated the relationship between depressive symptoms and global cognition. While the current study does not demonstrate causal associations, it is possible that depressive symptoms in older adulthood act as a risk factor for memory decline by prompting disengagement from beneficial activities that help to preserve global cognition.

Public Significance Statement.

The current study highlights the behavioral pathways in which depressive symptoms may negatively influence cognitive functioning in later life. Specifically, when individuals feel depressed, they may disengage from beneficial leisure activities that help to promote cognitive functioning. Future research should investigate whether intervening on these behavioral pathways perhaps helps to alleviate the negative cognitive consequences of depressive symptoms in older adulthood.