Sage Journals: Discover world-class research

Abstract

Social connection is associated with memory function in aging adults. However, little understanding exists of the pathways by which two distinct aspects of social connection—social isolation and functional social support—are related to each other and to memory. Using baseline and three-year follow-up data from the Tracking Cohort of the Canadian Longitudinal Study on Aging (n = 12,834), we explored whether functional social support mediated the relation between social isolation and memory in community-dwelling adults aged 45–85 years. We found the association between social isolation and memory operated through, not independently of, functional social support. We did not find effect modification by age group or sex. Health professionals can use gerontological care appointments to evaluate whether social relationships meet the functional social support needs of aging adults.

Keywords

cognitive function CLSA memory social connection

What this paper adds

• Examines the mediating role of functional social support on the association between social isolation and memory among middle-aged and older adults who live in the community.

• Generates novel findings supporting the need to consider functional social support when studying the effect of social isolation on memory.

Applications of study findings

• Health practitioners and primary care providers should assess whether aging adults believe their social relationships – regardless of quantity – meet their emotional, tangible, or affectionate functional social support needs.

• The identification of unmet needs can trigger connections with health and social support services, such as through day center programs.

• These programs should focus on addressing unmet functional support needs, rather than simply growing the size of social networks.

Background and Objectives

Memory

Memory is a complex neural process in which the brain encodes, consolidates, and retrieves information (Zlotnik & Vansintjan, 2019). Episodic memory is the ability to remember personal experiences and events such as recalling whether one took a medication in the morning (Struble & Sullivan, 2011). Declines in episodic memory function are an early symptom of major neurocognitive disorder (Raposo Pereira et al., 2024). Working memory is a component of executive function relating to the temporary storage and processing of information (Drag & Bieliauskas, 2010). Measures of working memory have been found to be key indicators of other cognitive functions, such as reasoning and problem solving, and executive functioning in activities of daily life (Kane et al., 2007).

Social Support

Structural and functional aspects of social support are modifiable factors that can promote memory (Mogic et al., 2023). Structural social support is assessed by the size of an individual’s social network (e.g., the number of persons in the network and the frequency of contact with these persons) and the frequency of participating in a range of social activities (de Jong Gierveld et al., 2006). Social isolation (SI) is the absence of social networks and the lack of participation in social activities (Menec et al., 2019). Functional social support (FSS) refers to an individual’s perception of the degree to which they can rely on members of their social network for support in times of need (Sherbourne & Stewart, 1991). SI increases with age (Ran et al., 2024) and is associated with poor memory function (Piolatto et al., 2022) and reduced FSS (Gable & Bedrov, 2022; Santini et al., 2020). Higher levels of FSS are associated with better memory function (Ma et al., 2024; Mogic et al., 2023).

Theoretical Frameworks

Several theoretical frameworks help to explain the intertwined nature of SI, FSS, and memory (Figure 1). A common theme of these theories is the notion that cognitive function is stimulated by social participation in activities spanning the gamut of leisure and work. Cunha et al.’s (2024) recent meta-analysis of 15 studies found a significant pooled association between social participation and reduced cognitive decline. Fernández et al. (2023) theorized that the cognitive demands from social participation draw upon memory and verbal fluency more so than other cognitive domains such as numeracy or temporal orientation.

Figure 1

Conceptual Model

According to the cognitive enrichment (“use-it-or-lose-it”) hypothesis, minimizing SI by engaging in positive behaviors, including staying connected with others through social activities, is key to maintaining memory function throughout the aging process (Hertzog, 2009). Social engagement preserves cognitive function through exposure to novel stimuli such as diversity of ideas, information, activities, verbal and nonverbal social cues, faces, and speech patterns (Thoits, 2011). Similarly, the stress hypothesis suggests increased social engagement can reduce psychological stress and thereby preserve memory function (Souza-Talarico et al., 2011). In rodents, SI is associated with prolonged neuroendocrine stress responses leading to neuronal changes (e.g., loss of dendritic spines and neuronal cell death). These structural changes are accompanied by deficits in learning and memory performance, mirroring the cognitive impairments linked to stress in human studies (Mumtaz et al., 2018). The cognitive reserve hypothesis suggests that individuals may build resilience to the cognitive impacts of neurodegeneration throughout the life course, based on accumulated levels of cognitive stimulation through factors such as larger and more diverse social networks and engagement in frequent social activities (Sharifian et al., 2022). Individuals exposed to greater cognitive stimulation can form more extensive and efficient neural networks that compensate for age-related changes in neuropathology (Barulli & Stern, 2013).

These hypotheses also explain the association between FSS and memory. Meaningful social connections with persons who can provide FSS amplify the cognitive benefits of social engagement (Kelly et al., 2017). Further, supportive interpersonal relationships, evidenced by increased availability of FSS, offer coping resources to manage stressful events (Baum et al., 2020), while large social networks without concomitant FSS may be insufficient to reduce stress (Cohen & Wills, 1985). Similarly, individuals with large social networks and higher levels of FSS have been shown to display greater cognitive reserve (Saito et al., 2018).

The association between SI and FSS is predicated on the fact that some level of social network size is a precondition for receipt of FSS (Hülür, 2022). Empirically, Santini et al. (2020) showed that higher structural support/lower SI is associated with stronger FSS. However, FSS is not dependent on social network size in a dose-response fashion (Gurung et al., 2003), as seen in situations where individuals who are socially isolated report high FSS because they can rely on help from as few as one or two social network members (Henning-Smith et al., 2019). Conversely, individuals might have large social networks, but they may report low FSS because they do not believe their social network members will provide help when needed. Although both SI and FSS have been linked to memory, the potential mediating effect of FSS on the relation between SI and memory has been understudied in the literature. Following a systematic literature search, we found 10 articles that regressed cognitive function, and 11 articles that regressed memory function, on SI and FSS as independent variables (Endresz, 2024). However, none of these articles explored FSS as a mediator.

We propose FSS as the mediator because one must have some individuals in their social network from whom to obtain FSS. Therefore, FSS is in the causal pathway between SI (including social networks) and memory. The negative impact of high SI on memory may be ameliorated by strong FSS from even one person, whereas the benefits of low SI on memory may be unrealized when a large social network cannot alleviate unmet needs (Zahodne et al., 2019). We examined whether FSS mediates the association between SI and memory in community-dwelling, middle-aged and older adults.

Research Design and Methods

Data Source

The Canadian Longitudinal Study on Aging (CLSA) is a population-based, panel study of middle-aged and older Canadians (Raina et al., 2019). Participants were between 45 and 85 years of age at baseline (t₀) and were recruited into pre-defined provincial, age, and sex sampling strata. Stratified sampling was later expanded to enroll more persons with less than high school education. Participants in the CLSA’s Tracking Cohort (n = 21,241) were recruited across all 10 provinces and are being followed up at approximately three-year intervals via computer-assisted telephone interviews. Complete information about sampling and recruitment is available elsewhere (CLSA, 2023). The present study utilized data collected at t₀ (2010–2015) and the first follow-up timepoint (t₁: 2015–2018). We obtained ethics approval from the University of Waterloo’s Office of Research Ethics (file # 44733). The CLSA obtained informed consent from all participants prior to enrollment.

Study Measures

Outcome Variable

The main outcome variable was memory function at t₁. A modified version of the Rey Auditory Verbal Learning Test (RAVLT) was used to measure participants’ immediate (RAVLT I) and delayed (RAVLT II) recall memory (Tuokko et al., 2017). CLSA investigators modified the original RAVLT to fit within the time constraints of the participant interviews by eliminating an interference list recall and reducing the number of learning trials from five to two (Tuokko et al., 2017). The modified RAVLT is therefore a measure of episodic and working memory.

During the computer-assisted telephone interviews, participants listened to a recorded list of 15 words and had 90 seconds immediately thereafter to recall as many of these words as possible (RAVLT I). Five minutes later, interviewers asked participants to again recall as many of the words as possible, but this time within 60 seconds without hearing the recorded list again (RAVLT II). Participants’ responses to the RAVLT I and II administrations were audio recorded and later scored by trained CLSA staff.

The CLSA combined the raw scores from the RAVLT I and II administrations into a latent construct variable that had a mean score of 100 and a standard deviation of 15. This variable represented participants’ overall memory, with higher values indicating better performance across the RAVLT I and II administrations. Details of the procedure for devising the overall memory score are reported elsewhere (CLSA, 2022). We used this memory score to quantify memory function at t₀ and t₁. We elected not to study RAVLT I and II separately because of the close correlations between the raw scores.

Exposure Variable

The main exposure variable was SI at t₀. It was measured using Menec et al.’s (2019) index, which ranges from 0 to 5, with higher scores representing greater SI. The index is described in Appendix A. We followed Menec et al.’s (2019) recommendation and dichotomized scores at a cut point of 2, with persons scoring 2–5 classified as socially isolated and those scoring 0–1 classified as not socially isolated. This coding allowed us to classify individuals as socially isolated if they met at least half of the criteria for SI on the index.

Mediator Variable

The mediator variable was FSS at t₁. FSS scores were derived from the 19-item Medical Outcomes Study-Social Support Survey (MOS-SSS) (Sherbourne & Stewart, 1991), which generates a score for overall FSS that encompasses multiple subtypes of FSS (i.e., emotional/informational, tangible, affectionate, and positive social interactions). The CLSA utilized the RAND scoring formula (RAND Corporation, n. d.) to transform all question responses into a FSS score ranging from 0 to 100, with higher scores representing greater FSS. Since participants in the CLSA generally reported high levels of FSS, the scale scores were left skewed. To account for left skewness, we dichotomized the scale scores at the median (88.2 at t₀ and 89.5 at t₁) to create “high” and “low” FSS groups.

Covariates

Covariate selection drew upon relevant literature concerning SI, FSS, and memory (Hülür, 2022; Kelly et al., 2017; Meister & Zahodne, 2022; Zahodne et al., 2019; Zuelsdorff et al., 2019). The selected covariates included sociodemographic, health, and lifestyle variables. Sociodemographic variables included t₀ age group (45–54 years, 55–64 years, 65–74 years, and 75 years or older [groups based on sampling strata]), sex (male/female), province of residence (one of the ten Canadian provinces), education (highest level of educational attainment), and income (annual household income) (see Table 1). Health status variables included t₀ depressive symptoms (Center for Epidemiologic Studies Depression Scale – 10-item (Radloff, 1977): severe versus not severe depressive symptoms based on Andresen et al.’s (1994) cutoff of 10 points or more out of 30), chronic conditions (yes/no), and functional impairment status (yes/no). Lastly, lifestyle behavior variables included t₀ smoking status (non-user, occasional user, daily user) and alcohol consumption (not at all, occasionally, regularly).

Table 1

Sample Characteristics: Overall and by Social Isolation Status at Baseline in the Canadian Longitudinal Study on Aging Tracking Cohort (n = 12,384)

Characteristic	Total (n = 12,834)	Not socially isolated (n = 10,202)	Socially isolated (n = 2,632)
Characteristic	n (%)	n (%)	n (%)
Sex (t₀)
Male	6,182 (48.2)	4,923 (48.3)	1,259 (47.8)
Female	6,652 (51.8)	5,279 (51.7)	1,373 (52.2)
Age (t₀)
45–54 years	3,973 (30.1)	3,326 (32.6)	647 (24.6)
55–64 years	4,287 (33.4)	3,370 (33.0)	917 (34.8)
65–74 years	2,721 (21.2)	2,118 (20.8)	603 (22.9)
75 years or older	1,853 (14.4)	1,388 (13.6)	465 (17.7)
Education (t₀)
Less than high school	875 (6.8)	660 (6.5)	215 (8.2)
High school diploma	1,680 (13.1)	1,355 (13.3)	325 (12.3)
Some post-secondary education	930 (7.3)	719 (7.0)	211 (8.0)
Post-secondary degree/diploma	9,349 (72.9)	7,468 (73.2)	1,881 (71.5)
Province of residence (t₀)
Alberta	1,235 (9.6)	1,015 (9.9)	220 (8.4)
British Columbia	1,437 (11.2)	1,085 (10.6)	352 (13.4)
Manitoba	896 (7.0)	724 (7.1)	172 (6.5)
New Brunswick	826 (6.4)	638 (6.3)	188 (7.1)
Newfoundland and Labrador	711 (5.5)	586 (5.7)	125 (4.7)
Nova Scotia	964 (7.5)	781 (7.7)	183 (7.0)
Ontario	3,007 (23.4)	2,452 (24.0)	555 (21.1)
Prince Edward Island	679 (5.3)	542 (5.3)	137 (5.2)
Québec	2,239 (17.5)	1,702 (16.7)	537 (20.4)
Saskatchewan	840 (6.5)	677 (6.6)	163 (6.2)
Total annual household outcome (t₀)
< $20,000	573 (4.5)	283 (2.8)	290 (11.0)
$20,000 to < $50,000	3,132 (24.4)	2,254 (22.1)	878 (33.4)
$50,000 to < $100,000	4,569 (35.6)	3,760 (36.9)	809 (30.7)
$100,000 to < $150,000	2,271 (17.7)	2,005 (19.7)	266 (10.1)
≥ $150,000	1,622 (12.6)	1,450 (14.2)	172 (6.5)
Missing	667 (5.2)	450 (4.4)	217 (8.2)
Functional status (t₀)
No assistance required	11,626 (90.6)	9,355 (91.7)	2,271 (86.3)
Assistance required ≥1 activity	1,152 (9.0)	812 (8.0)	340 (12.9)
Missing	56 (0.4)	35 (0.3)	21 (0.8)
Chronic conditions (t₀)
No chronic conditions	1,255 (9.8)	1,081 (10.6)	174 (6.6)
≥1 chronic condition	11,549 (90.0)	9,095 (89.1)	2,454 (93.2)
Missing	30 (0.2)	26 (0.3)	4 (0.2)
Depressive symptoms (t₀)
Not severe	10,907 (85.0)	8,853 (86.8)	2,054 (78.0)
Severe	1,093 (14.8)	1,328 (13.0)	575 (21.8)
Missing	24 (0.2)	21 (0.2)	3 (0.1)
Current smoking status (t₀)
Non-smoker	7,700 (60.0)	6,164 (60.4)	1,536 (58.4)
Occasional smoker	214 (1.7)	155 (1.5)	59 (2.2)
Daily smoker	953 (7.4)	679 (6.7)	274 (10.4)
Missing	3,967 (30.9)	3,204 (31.4)	763 (29.0)
Alcohol consumption (t₀)
Non-drinker	1,369 (10.7)	998 (9.8)	371 (14.1)
Occasional drinker	1,924 (15.0)	1,466 (14.4)	458 (17.4)
Regular drinker	9,152 (71.3)	7,428 (72.8)	1,724 (65.5)
Missing	389 (3.0)	310 (3.0)	79 (3.0)
Functional social support (t₀)
Low	6636 (51.7)	4905 (48.1)	1731 (65.8)
High	6198 (48.3)	5297 (51.9)	901 (34.2)
Functional social support (t₁)
Low	6653 (51.8)	4960 (48.6)	1693 (64.3)
High	6181 (48.2)	5242 (51.4)	939 (35.7)

Notes: Chi-square <0.05 in bolded font; frequencies shown are column %; t₀ = baseline; t₁ = follow-up; Not severe depressive symptoms $<$ 10; Severe depressive symptoms ≥10.

Statistical Analysis

Regression Analysis

First, we regressed t₁ memory score on t₀ SI status, controlling for FSS at t₀ and t₁, and memory at t₀. Based on CLSA recommendations, the “base” model also included age group, sex, and province as covariates to address the CLSA’s complex survey design (CLSA, 2023). Second, the remaining t₀ covariates (i.e., sociodemographic, health status, lifestyle behaviors) were added to the base model to create an “adjusted” model.

Mediation Analysis

Following the approach of Imai and Yamamoto (2013) and Tingley et al. (2014): (1) FSS_t1 was regressed on SI_t0 to obtain ${\hat{β}}_{{SI}_{t 0}}$ for the “a” path of the mediation model, or the effect of the main exposure on the mediator; (2) memory_t1 was regressed on SI_t0 and FSS_t1 to obtain ${\hat{β}}_{{FSS}_{t 1}}$ for the “b” path, or the effect of the mediator on the outcome, controlling for the main exposure. We used $\hat{β}$ to symbolize the estimated and unstandardized coefficients in the regression models. These models were adjusted for all covariates. The “a” path model was also adjusted for FSS_t0 and the “b” path model was also adjusted for FSS_t0 and memory_t0, as recommended by Hayes (2022), since associations in regression models with t₁ variables as outcomes may be inflated by not controlling for t₀ values of these variables.

To complete the mediation analysis, the “a” and “b” path models were used to calculate $\hat{β}$ values for the “ab,” “c-prime,” and “c” paths. The “ab” path is the indirect effect of SI on memory that “passes through” or is mediated by FSS; the “c-prime” path is the direct effect of SI on memory, where FSS is treated as a covariate; and the “c” path is the total effect of SI on memory, or the combined indirect and direct effects.

The mathematical calculations to obtain the $\hat{β}$ ’s for the “ab,” “c-prime,” and “c” paths were implemented using R v4.3.0 (The R Foundation for Statistical Computing, Vienna, Austria) and the mediation package (Tingley et al., 2014), which also generated 95% confidence intervals (CI) around the $\hat{β}$ ’s via the Monte Carlo sampling method, White’s heteroskedasticity-consistent estimator for the covariance matrix, and 10,000 simulations (Yzerbyt et al., 2018). We assessed the presence of mediation using the joint significance test, which requires the $\hat{β}$ ’s for both the “a” and “b” paths to be statistically significant for mediation to be present (Yzerbyt et al., 2018).

Since the “a” path of the mediation model was computed using logistic regression, given the outcome (FSS_t1) was dichotomous, we used Kenny’s procedure (2024) to rescale ${\hat{β}}_{{S I}_{t 0}}$ and the 95% CI from the logistic to the linear scale to match the “b” path, whose $\hat{β}$ ’s and CIs were obtained through multiple linear regression (memory score was continuous).

Moderated Mediation Analysis

We stratified the sample into four age groups (45–54, 55–64, 65–74, ≥75 years) and repeated the mediation analysis within each stratum to assess possible effect modification. For each of the five mediation paths, Cuzick’s (2005) forest plot method was used to check for effect modification, which was deemed present if all the stratum-specific 95% CIs excluded the unstratified $\hat{β}$ . This analysis was repeated to test sex as an effect modifier.

Missing Data and Sensitivity Analyses

Participants were removed from the analysis if they had missing data on SI at t₀, or FSS or memory at t₀ or t₁. Participants with missing covariate data were retained in the analytical sample by creating “missing” response categories for all instances of missing covariate data (primary analysis). To assess the potential impact of missing data, we explored whether SI_t0 status, memory_t0 scores, and FSS_t0 scores differed among those who dropped out of the CLSA post-t₀, compared to those who returned for t₁ (non-dropouts). A simple logistic regression model was utilized to obtain the odds of dropping out among socially isolated versus non-socially isolated persons at t₀. Mean memory scores and median FSS scores at t₀ were compared across dropouts versus non-dropouts using the t-test and the Mann–Whitney U test, respectively.

Two sensitivity analyses were carried out to assess the robustness of our procedure for handling missing data. First, we repeated all the analyses using complete case analysis on all variables, which meant removing participants with missing data not just on SI, FSS, and memory, but on all covariates as well. Second, multiple imputation (MI) was employed to impute values for covariates that possessed high levels of missingness (>2%). The imputations were conducted using nine imputation cycles (each yielding one imputed dataset) and predictive mean matching (PMM) in R’s mice package. The mediation analysis was repeated on each of the nine imputed datasets and the relevant $\hat{β}$ s for each of the five paths across the nine datasets were combined using Rubin’s Rules (Rubin, 1987), whose equations were programmed into an Excel spreadsheet and independently double-verified for accuracy. The resulting single set of combined $\hat{β}$ s served as the final mediation result from the MI procedure. Appendix B graphically depicts our analysis plan.

Results

Sample Size and Characteristics

In the Tracking Cohort, 17,052 of the 21,241 participants at t₀ (80.3%) provided t₁ information. After removing participants who had missing SI information at t₀ and those who had missing FSS and memory information at t₀ or t₁, 12,834 out of 17,052 participants (75.4%) remained in the analytical sample (Figure 2). Table 1 shows the distribution of the analytical sample’s characteristics, both overall and stratified by SI at t₀. Descriptive information about the sample is described in Appendix C.

Figure 2

Derivation of Analytical Sample

Multivariable Linear Regression Analysis

In the base model, SI status at t₀ had a small and statistically significant, inverse association with memory at t₁, indicating the average memory score among socially isolated persons was lower than the average score among non-socially isolated participants ( $\hat{β}$ = −0.75; 95% CI: −1.32, −0.18). After adjusting for the remaining covariates, the effect of SI remained negative, but was no longer statistically significant ( $\hat{β}$ = −0.13; 95% CI: −0.68, 0.45). The full output for the regression analyses with the base and adjusted models can be found in Appendix C (Figures C1 and C2).

Mediation Analysis

Figure 3 shows the results of the mediation analysis. On the “a” path, SI at t₀ significantly and negatively impacted FSS at t₁, after adjusting for all covariates, such that the odds of having high compared to low FSS decreased by 36% in the socially isolated versus not socially isolated group (OR = 0.64; 95% CI: 0.58, 0.70). After converting the odds ratio from the “a” path to the linear scale, the $\hat{β}$ was −0.06 (95% CI: −0.08, −0.04). On the “b” path, FSS at t₁ was significantly and positively associated with t₁ memory after adjusting for all covariates ( $\hat{β}$ = 0.59; 95% CI: 0.09, 1.10), suggesting that the average memory score was 0.59 points higher in individuals with high compared to low levels of FSS. Since the $\hat{β}$ ’s from the “a” and “b” paths were both statistically significant, the effect of SI on memory was mediated by FSS, according to the joint significance test (Yzerbyt et al., 2018).

Figure 3

Mediation Model: Social Isolation, Functional Social Support, and Memory. Notes: p < 0.05 in Bolded Font; Adjusted for Baseline Functional Social Support, Baseline Memory, Baseline Sociodemographic Factors, Health Status, and Lifestyle Behaviors. T₀ = Baseline; T₁ = Follow-Up

SI at t₀ impacted memory scores at t₁ indirectly through FSS at t₁ (“ab” path). On the “ab” path, memory scores at t₁ decreased on average by 0.03 points (95% CI: −0.06, −0.01) in socially isolated participants versus non-isolated participants, as mediated by FSS at t₁ and adjusted for all t₀ covariates. The direct effect of SI on memory (“c-prime” path) was not significant—though still inverse—after adjustment for all covariates (with FSS treated as a covariate in this pathway [ $\hat{β}$ = −0.13; 95% CI: −0.68, 0.45]). The direct effect is the same result that we obtained from the multivariable regression analysis above. The total effect of SI on memory (“c” path [ $\hat{β}$ = −0.16; 95% CI: −0.72, 0.41]) yielded a non-statistically significant result. We did not find evidence of moderated mediation by age group (Figure 4) or sex (Figure 5).

Figure 4

Forest Plot: Moderated Mediation Analysis by Age Group. Notes: Adjusted for Baseline Functional Social Support, Baseline Memory Score, Baseline Sociodemographic Factors, Health Status, and Lifestyle Behaviors; Vertical Line Represents the Unstratified Regression Coefficient

Figure 5

Forest Plot: Moderated Mediation Analysis by Sex. Notes: Adjusted for Baseline Functional Social Support, Baseline Memory Score, Baseline Sociodemographic Factors, Health Status, and Lifestyle Behaviors; Vertical Line Represents the Unstratified Regression Coefficient

Differential Dropouts Over Follow-Up and Sensitivity Analysis

On average, participants who were socially isolated at t₀ had 42% higher odds of dropping out before t₁ than those who were not isolated at t₀ (OR = 1.42; 95% CI: 1.31 to 1.53). Similarly, those who dropped out had slightly lower average memory scores at t₀ than those who did not drop out. While median FSS scores were roughly the same between dropouts and non-dropouts, the Mann–Whitney U test suggested a statistically significant difference in median FSS score between dropouts and non-dropouts, likely due to the large sample size (see Appendix D).

When comparing the results of the primary analysis to the complete case and multiple imputation sensitivity analyses, findings were generally consistent with some minor differences (Figure 6; Appendix E). Some $\hat{β}$ 's moved closer to the null and were no longer statistically significant, although their directions of effect remained unchanged. $\hat{β}$ ’s whose directions of effect did change were non-significant in all analyses.

Figure 6

Forest Plot: Mediation Model Results Comparing Primary and Sensitivity Analyses. Notes: Adjusted for Baseline Functional Social Support, Baseline Memory, Baseline Sociodemographic Factors, Health Status, and Lifestyle Behaviors. T₀ = Baseline; T₁ = Follow-Up

Discussion and Implications

In this population-based panel study of middle-aged and older Canadians, we observed that the association between SI and memory was mediated by FSS, which mitigated the adverse effect of SI on memory. Although the mediation effect was small, individuals who were socially isolated performed better on the RAVLT if they reported high levels of support (high FSS) compared to low levels of FSS. Our study provides novel findings about the potential mediating effect of FSS on memory.

Only two previously published articles bear some resemblance to our study. However, neither article looked specifically at the mediating effect of FSS on SI and memory. The first article (Yang et al., 2020) was a cross-sectional analysis of 7,410 participants in the China Health and Retirement Longitudinal Study, all aged 60 years or over. The authors found loneliness to partially mediate the association between SI and cognitive function. Loneliness is one’s subjective perception of a gap between what they want and what they have in terms of the quantity and quality of interpersonal relationships (de Jong Gierveld, 1987). The second article (Santini et al., 2020) was a longitudinal cohort study of 3,005 older adults, aged 57 to 85 years, from the National Social Life, Health, and Aging Project. The authors investigated whether perceived isolation mediated the association of social disconnectedness with depression and anxiety. While they found a non-significant direct relation between social disconnectedness and the two outcomes, they also reported that perceived isolation mediated the relation, such that social disconnectedness predicted higher amounts of perceived isolation, which in turn predicted greater symptoms of depression and anxiety.

In contrast to our statistically non-significant direct effect, the literature has generally reported strong and statistically significant associations between SI and memory (Hülür, 2022; Peng et al., 2022; Zahodne et al., 2019; Zuelsdorff et al., 2019). Multiple reasons may explain the discrepant findings. Our missing data analysis revealed that participants who were socially isolated at t₀ had higher odds of dropping out of the CLSA and were therefore not included in the analytical sample. The participants who dropped out after t₀ also had lower FSS and memory scores on average compared to the individuals in our analytical sample. Attrition on all three variables may have biased the effect size toward the null. Further, during the recruitment interview, CLSA staff excluded potential participants who appeared to be cognitively impaired, which led to an overrepresentation of cognitively healthy participants in the sample, thereby likely biasing the regression results toward the null. On the other hand, the SI index employed in our study contained a larger number of items relative to SI measures utilized in other research (Hülür, 2022; Meister & Zahodne, 2022; Peng et al., 2022; Zahodne et al., 2019; Zuelsdorff et al., 2019). The results of our work may therefore provide a more comprehensive assessment of the relationship between SI and memory because the index strongly aligns with the objective nature of social isolation (Menec et al., 2019). In addition, we included a robust set of covariates in our regression models, which may have minimized confounding bias relative to previous studies that adjusted for fewer covariates.

Strengths and Limitations

This study has multiple strengths. In addition to the comprehensive SI index and robust covariate set described above, the CLSA’s sampling frame permitted us to include both middle-aged and older adults from across all 10 Canadian provinces in our analyses. This allowed us to apply the results of the study to a broad target population. Most importantly, while many studies assessed the effects of SI or FSS on memory, we are unaware of any published studies that explored the mediation effects of FSS on the relationship between SI and memory.

Some limitations should be noted. CLSA participants were generally healthier, more highly educated, and had higher incomes than average (Raina et al., 2019); therefore, the results of this study may only apply to the subset of middle-aged and older adults who share the characteristics of the analytical sample. Attrition and exclusion of persons with overt cognitive impairment could have led to small effect sizes. Further, at the time of analysis, only two timepoints of CLSA data were available to test for mediation. The two timepoints were also spread 3 years apart, which may not be sufficient to observe clinically relevant changes in memory. We did not include loneliness in the analysis because it is a different construct from SI and FSS (Taylor et al., 2023); the CLSA also did not collect loneliness data at t₀. Future research drawing upon additional CLSA follow-up timepoints may add to our novel findings, as increased memory impairment and the emergence of neurocognitive disorders over time will expand researchers’ ability to elucidate the intertwined nature of SI, FSS, and memory. Given the role of social participation in the theoretical frameworks discussed earlier, a specific focus on the interplay between social participation, FSS, and memory may also be warranted as further follow-up data become available from the CLSA.

Policy and Practice

Health practitioners and primary care providers should consider using regular gerontological care appointments to evaluate whether social relationships meet the functional social support needs of aging adults. Examples of such needs are emotional, tangible, or affectionate support needs. The identification of unmet needs can trigger connections with health and social support services, such as the day center programs offered by the Province of Québec’s centers locaux de services communautaires (local community service centers) (CIUSSS de l’Ouest-de-l’Île-de-Montréal, 2025). This approach aligns with the World Health Organization’s (2021) call for social prescribing as a means of reducing SI and the formation of the WHO Commission on Social Connection (2025) to prioritize social connections (SI and FSS) as a public health priority.

Conclusion

Our study showed that SI is associated with memory indirectly through FSS in middle-aged and older adults. This evidence suggests the association between SI and memory at least partially operates through, not independently of, FSS. Given the limitations described above, the true mediation effect could be much larger, especially in target populations that are more heterogeneous with respect to health and the social determinants of health. Health policy makers and practitioners should consider FSS when they evaluate SI and memory.

Supplemental Material

Supplemental Material - The Association Between Social Isolation, Functional Social Support, and Memory: A Mediation Analysis of the Canadian Longitudinal Study on Aging

Supplemental Material for The Association Between Social Isolation, Functional Social Support, and Memory: A Mediation Analysis of the Canadian Longitudinal Study on Aging by Nicole Endresz, Suzanne L. Tyas, Colleen J. Maxwell, and Mark Oremus in Journal of Applied Gerontology.

Footnotes

Acknowledgments

This research was made possible using the data/biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA), as well as the following provinces, Newfoundland and Labrador, Nova Scotia, Quebec, Ontario, Manitoba, Alberta, and British Columbia. Funding for the Canadian Longitudinal Study on Aging (CLSA) is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference: LSA 94473 and the Canada Foundation for Innovation, as well as the following provinces, Newfoundland and Labrador, Nova Scotia, Québec, Ontario, Manitoba, Alberta, and British Columbia. This research has been conducted using the CLSA Baseline Tracking Dataset v4.0 and Follow-up 1 Tracking Dataset v3.1, under Application Number 2209026. The CLSA is led by Drs. Parminder Raina, Christina Wolfson, and Susan Kirkland. The time and commitment of the participants to the CLSA study platform is gratefully acknowledged, without whom this research would not be possible. The opinions expressed in this manuscript are the author’s own and do not reflect the views of the CLSA.

ORCID iD

Mark Oremus

Ethical Considerations

We obtained ethics approval from the University of Waterloo’s Office of Research Ethics (file # 44733); the CLSA’s Data and Sample Access Committee approved access to the Tracking Cohort dataset.

Consent to Participate

The CLSA obtained written informed consent from all participants prior to enrollment.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Velux Stiftung Foundation: 1190 and the Canadian Institutes of Health Research: MM1-174917.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data Availability Statement

Data are available from the CLSA () for researchers who meet the criteria for access to de-identified CLSA data.

Supplemental Material

Supplemental Material for this article is available online.

References

Andresen

E. M.

Malmgren

J. A.

Carter

W. B.

Patrick

D. L.

(1994). Screening for depression in well older adults: Evaluation of a short form of the CES-D (center for epidemiologic studies depression scale). American Journal of Preventive Medicine, 10(2), 77–84. https://doi.org/10.1016/S0749-3797(18)30622-6

Barulli

Stern

(2013). Efficiency, capacity, compensation, maintenance, plasticity: Emerging concepts in cognitive reserve. Trends in Cognitive Sciences, 17(10), 502–509. https://doi.org/10.1016/j.tics.2013.08.012

Baum

Taylor

S. E.

Singer

J. E.

(Eds.), (2020). Handbook of psychology and health Volume IV, social psychological aspects of health (1st ed.). Routledge.

CIUSSS de l’Ouest-de-l’Île-de-Montréal . (2025). Day centres. https://www.ciusss-ouestmtl.gouv.qc.ca/en/care-and-services/seniors-andor-those-experiencing-a-loss-of-independence/day-centres

CLSA . (2022). Derived variables – cognition (COG) normative data (tracking assessment). https://www.clsa-elcv.ca/wp-content/uploads/2023/06/dv_cog_norms_trm_v1.0_2022apr04-1.pdf

CLSA . (2023). Sampling and computation of response rates and sample weights for the Tracking (telephone interview) participants and Comprehensive participants. https://www.clsa-elcv.ca/wp-content/uploads/2023/06/clsa_datasupportdoc_sampwgts_bl_fu1_v2.0_2023feb13_final_002.pdf

Cohen

Wills

T. A.

(1985). Stress, social support, and the buffering hypothesis. Psychological Bulletin, 98(2), 310–357. https://doi.org/10.1037/0033-2909.98.2.310

Cunha

Voss

Andrade

Delerue-Matos

(2024). Is formal social participation associated with cognitive function in middle-aged and older adults? A systematic review with meta-analysis of longitudinal studies. Behavioral Sciences, 14(4), 262. https://doi.org/10.3390/bs14040262

Cuzick

(2005). Forest plots and the interpretation of subgroups. Lancet (London, England), 365(9467), 1308. https://doi.org/10.1016/S0140-6736(05)61026-4

10.

de Jong Gierveld

(1987). Developing and testing a model of loneliness. Journal of Personality and Social Psychology, 53(1), 119–128. https://doi.org/10.1037/0022-3514.53.1.119

11.

de Jong Gierveld

van Tilburg

Dykstra

P. A.

(2006). Loneliness and social isolation. In Vangelisti

A. L.

Perlman

(Eds.), The Cambridge handbook of personal relationships (1st ed., pp. 485–500). Cambridge University Press. https://doi.org/10.1017/CBO9780511606632.027

12.

Drag

L. L.

Bieliauskas

L. A.

(2010). Contemporary review 2009: Cognitive aging. Journal of Geriatric Psychiatry and Neurology, 23(2), 75–93. https://doi.org/10.1177/0891988709358590

13.

Endresz

(2024). The association between social isolation, functional social support, and memory: A moderated mediation analysis of the Canadian longitudinal study on aging [masters dissertation. University of Waterloo]. UWSpace. https://dspacemainprd01.lib.uwaterloo.ca/server/api/core/bitstreams/44400870-967b-4de5-abc4-4b241ac5d739/content

14.

Fernández

García-Mollá

Oliver

Sansó

Tomás

J. M.

(2023). The role of social and intellectual activity participation in older adults' cognitive function. Archives of Gerontology and Geriatrics, 107, Article 104891. https://doi.org/10.1016/j.archger.2022.104891

15.

Gable

S. L.

Bedrov

(2022). Social isolation and social support in good times and bad times. Current Opinion in Psychology, 44, 89–93. https://doi.org/10.1016/j.copsyc.2021.08.027

16.

Gurung

R. A. R.

Taylor

S. E.

Seeman

T. E.

(2003). Accounting for changes in social support among married older adults: Insights from the MacArthur Studies of Successful Aging. Psychology and Aging, 18(3), 487–496. https://doi.org/10.1037/0882-7974.18.3.487

17.

Hayes

A. F.

(2022). Introduction to mediation, moderation, and conditional process analysis: A regression-based approach (3rd ed.). The Guilford Press.

18.

Henning-Smith

Moscovice

Kozhimannil

(2019). Differences in social isolation and its relationship to health by rurality. The Journal of Rural Health: Official Journal of the American Rural Health Association and the National Rural Health Care Association, 35(4), 540–549. https://doi.org/10.1111/jrh.12344

19.

Hertzog

(2009). Use it or lose it: An old hypothesis, new evidence, and an ongoing case study. In Bosworth

H. B.

Hertzog

(Eds.), Aging and cognition: Research methodologies and empirical advances (pp. 161–179). American Psychological Association. https://doi.org/10.1037/11882-008

20.

Hülür

(2022). Structural and functional aspects of social relationships and episodic memory: Between-person and within-person associations in middle-aged and older adults. Gerontology, 68(1), 86–97. https://doi.org/10.1159/000514949

21.

Imai

Yamamoto

(2013). Identification and sensitivity analysis for multiple causal mechanisms: Revisiting evidence from framing experiments. Political Analysis, 21(2), 141–171. https://doi.org/10.1093/pan/mps040

22.

Kane

M. J.

Brown

L. H.

McVay

J. C.

Silvia

P. J.

Myin-Germeys

Kwapil

T. R.

(2007). For whom the mind wanders, and when: An experience-sampling study of working memory and executive control in daily life. Psychological Science, 18(7), 614–621. https://doi.org/10.1111/j.1467-9280.2007.01948.x

23.

Kelly

M. E.

Duff

Kelly

McHugh Power

J. E.

Brennan

Lawlor

B. A.

Loughrey

D. G.

(2017). The impact of social activities, social networks, social support and social relationships on the cognitive functioning of healthy older adults: A systematic review. Systematic Reviews, 6(1), 259. https://doi.org/10.1186/s13643-017-0632-2

24.

Kenny

D. A.

(2024). Mediate. https://davidakenny.net/cm/mediate.htm

25.

Liao

(2024). Social support and cognitive activity and their associations with incident cognitive impairment in cognitively normal older adults. BMC Geriatrics, 24(1), 38. https://doi.org/10.1186/s12877-024-04655-5

26.

Meister

L. M.

Zahodne

L. B.

(2022). Associations between social network components and cognitive domains in older adults. Psychology and Aging, 37(5), 591–603. https://doi.org/10.1037/pag0000672

27.

Menec

V. H.

Newall

N. E.

Mackenzie

C. S.

Shooshtari

Nowicki

(2019). Examining individual and geographic factors associated with social isolation and loneliness using Canadian Longitudinal Study on Aging (CLSA) data. PLOS ONE, 14(2), Article e0211143. https://doi.org/10.1371/journal.pone.0211143

28.

Mogic

Rutter

E. C.

Tyas

S. L.

Maxwell

C. J.

O’Connell

M. E.

Oremus

(2023). Functional social support and cognitive function in middle- and older-aged adults: A systematic review of cross-sectional and cohort studies. Systematic Reviews, 12(1), 86. https://doi.org/10.1186/s13643-023-02251-z

29.

Mumtaz

Khan

M. I.

Zubair

Dehpour

A. R.

(2018). Neurobiology and consequences of social isolation stress in animal model: A comprehensive review. Biomedicine & Pharmacotherapy, 105, 1205–1222. https://doi.org/10.1016/j.biopha.2018.05.086

30.

Peng

Burr

J. A.

Han

S. H.

(2022). Cognitive function and cognitive decline among older rural Chinese adults: The roles of social support, pension benefits, and medical insurance. Aging & Mental Health, 27(4), 1–9. https://doi.org/10.1080/13607863.2022.2088693

31.

Piolatto

Bianchi

Rota

Marengoni

Akbaritabar

Squazzoni

(2022). The effect of social relationships on cognitive decline in older adults: An updated systematic review and meta-analysis of longitudinal cohort studies. BMC Public Health, 22(1), 278. https://doi.org/10.1186/s12889-022-12567-5

32.

Radloff

(1977). The CES-D scale: A self report depression scale for research in the general population: West Publishing Co.

33.

Raina

Wolfson

Kirkland

Griffith

L. E.

Balion

Cossette

Dionne

Hofer

Hogan

Van Den Heuvel

E. R.

Liu-Ambrose

Menec

Mugford

Patterson

Payette

Richards

Shannon

Sheets

Taler

V.,

Young

(2019). Cohort profile: The Canadian longitudinal study on aging (CLSA). International Journal of Epidemiology, 48(6), 1752–1753j. https://doi.org/10.1093/ije/dyz173

34.

Ran

Wei

Yang

(2024). Prevalence of social isolation in the elderly: A systematic review and meta-analysis. Geriatric Nursing, 58, 87–97. https://doi.org/10.1016/j.gerinurse.2024.05.008

35.

RAND Corporation. (n.d.).

Social support survey instrument scoring instructions. https://www.rand.org/health-care/surveys_tools/mos/social-support/scoring.html

36.

Raposo Pereira

George

Dalla Barba

Dubois

La Corte

(2024). The memory binding test can anticipate alzheimer’s disease diagnosis at an early preclinical stage: A longitudinal study in the INSIGHT-preAD cohort. Frontiers in Aging Neuroscience, 16, Article 1414419. https://doi.org/10.3389/fnagi.2024.1414419

37.

Rubin

D. B.

(1987). Multiple imputation for nonresponse in surveys (1st ed.). Wiley. https://doi.org/10.1002/9780470316696

38.

Saito

Murata

Saito

Takeda

Kondo

(2018). Influence of social relationship domains and their combinations on incident dementia: A prospective cohort study. Journal of Epidemiology and Community Health, 72(1), 7–12. https://doi.org/10.1136/jech-2017-209811. https://jech.bmj.com/content/72/1/7

39.

Santini

Z. I.

Jose

P. E.

York Cornwell

Koyanagi

Nielsen

Hinrichsen

Meilstrup

Madsen

K. R.

Koushede

(2020). Social disconnectedness, perceived isolation, and symptoms of depression and anxiety among older Americans (nshap): A longitudinal mediation analysis. The Lancet. Public Health, 5(1), e62–e70. https://doi.org/10.1016/S2468-2667(19)30230-0

40.

Sharifian

Zaheed

A. B.

Morris

E. P.

Sol

Manly

J. J.

Schupf

Mayeux

Brickman

A. M.

Zahodne

L. B.

(2022). Social network characteristics moderate associations between cortical thickness and cognitive functioning in older adults. Alzheimer's & Dementia: The Journal of the Alzheimer's Association, 18(2), 339–347. https://doi.org/10.1002/alz.12383

41.

Sherbourne

C. D.

Stewart

A. L.

(1991). The MOS social support survey. Social Science & Medicine, 32(6), 705–714. https://doi.org/10.1016/0277-9536(91)90150-B

42.

Souza-Talarico

J. N. D.

Marin

M. F.

Sindi

Lupien

S. J.

(2011). Effects of stress hormones on the brain and cognition: Evidence from normal to pathological aging. Dementia & Neuropsychologia, 5(1), 8–16. https://doi.org/10.1590/S1980-57642011DN05010003

43.

Struble

L. M.

Sullivan

B.-J.

(2011). Cognitive health in older adults. The Nurse Practitioner, 36(4), 24–34. https://doi.org/10.1097/01.NPR.0000394724.43311.16

44.

Taylor

H. O.

Cudjoe

T. K. M.

Lim

M. H.

(2023). The state of loneliness and social isolation research: Current knowledge and future directions. BMC Public Health, 23(1), 1049. https://doi.org/10.1186/s12889-023-15967-3

45.

Thoits

P. A.

(2011). Mechanisms linking social ties and support to physical and mental health. Journal of Health and Social Behavior, 52(2), 145–161. https://doi.org/10.1177/0022146510395592

46.

Tingley

Yamamoto

Hirose

Keele

Imai

(2014). Mediation: R package for causal mediation analysis. Journal of Statistical Software, 59(5), 1–38. https://doi.org/10.18637/jss.v059.i05

47.

Tuokko

Griffith

L. E.

Simard

Taler

(2017). Cognitive measures in the Canadian longitudinal study on aging. The Clinical Neuropsychologist, 31(1), 233–250. https://doi.org/10.1080/13854046.2016.1254279

48.

World Health Organization . (2021). Social isolation and loneliness among older people: Advocacy brief. https://iris.who.int/bitstream/handle/10665/343206/9789240030749-eng.pdf?sequence=1

49.

World Health Organization . (2025). WHO commission on social connection. https://www.who.int/groups/commission-on-social-connection

50.

Yang

Wang

Edelman

L. S.

Tracy

E. L.

Demiris

Sward

K. A.

Donaldson

G. W.

(2020). Loneliness as a mediator of the impact of social isolation on cognitive functioning of Chinese older adults. Age and Ageing, 49(4), 599–604. https://doi.org/10.1093/ageing/afaa020

51.

Yzerbyt

Muller

Batailler

Judd

C. M.

(2018). New recommendations for testing indirect effects in mediational models: The need to report and test component paths. Journal of Personality and Social Psychology, 115(6), 929–943. https://doi.org/10.1037/pspa0000132

52.

Zahodne

L. B.

Ajrouch

K. J.

Sharifian

Antonucci

T. C.

(2019). Social relations and age-related change in memory. Psychology and Aging, 34(6), 751–765. https://doi.org/10.1037/pag0000369

53.

Zlotnik

Vansintjan

(2019). Memory: An extended definition. Frontiers in Psychology, 10, 2523. https://doi.org/10.3389/fpsyg.2019.02523

54.

Zuelsdorff

M. L.

Koscik

R. L.

Okonkwo

O. C.

Peppard

P. E.

Hermann

B. P.

Sager

M. A.

Johnson

S. C.

Engelman

C. D.

(2019). Social support and verbal interaction are differentially associated with cognitive function in midlife and older age. Neuropsychology, Development, and Cognition. Section B, Aging, Neuropsychology and Cognition, 26(2), 144–160. https://doi.org/10.1080/13825585.2017.1414769

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.90 MB

The Association Between Social Isolation,Functional Social Support,and Memory: A Mediation Analysis of the Canadian Longitudinal Study on Aging

Abstract

Keywords

What this paper adds

Applications of study findings

Background and Objectives

Memory

Social Support

Theoretical Frameworks

Research Design and Methods

Data Source

Study Measures

Outcome Variable

Exposure Variable

Mediator Variable

Covariates

Statistical Analysis

Regression Analysis

Mediation Analysis

Moderated Mediation Analysis

Missing Data and Sensitivity Analyses

Results

Sample Size and Characteristics

Multivariable Linear Regression Analysis

Mediation Analysis

Differential Dropouts Over Follow-Up and Sensitivity Analysis

Discussion and Implications

Strengths and Limitations

Policy and Practice

Conclusion

Supplemental Material

Supplemental Material - The Association Between Social Isolation, Functional Social Support, and Memory: A Mediation Analysis of the Canadian Longitudinal Study on Aging

Footnotes

Acknowledgments

ORCID iD

Ethical Considerations

Consent to Participate

Funding

Declaration of conflicting interests

Data Availability Statement

Supplemental Material

References

Supplementary Material