Social engagement and its links to cognition differ across non-Hispanic Black and White older adults

Abstract

Objective:

Racial inequalities in dementia have been linked to disparities in socioeconomic status, chronic diseases, and psychosocial stress. Less focus has been given to psychosocial protective factors. Previous studies suggest that social engagement promotes better cognitive aging, but few have examined whether social engagement or its associations with cognition vary across non-Hispanic Whites (NHW) and Blacks (NHB).

Method:

Participants included 465 adults (53% NHB) from the Michigan Cognitive Aging Project (M_age=63.59±3.15) who completed a comprehensive neuropsychological battery. Social engagement was operationalized as network size, frequency of social activity participation, and social support. Cognition was operationalized using factor scores corresponding to five domains: episodic memory, executive functioning, processing speed, language, and visuospatial functioning. Cross-sectional associations between social engagement and cognitive outcomes were examined using race-stratified regressions controlling for age, sex/gender, education, wealth, marital status, depressive symptoms, and chronic diseases.

Results:

There were no racial differences in social network size or social support. NHB participants reported less social activity participation than NHW participants. Social activity participation was positively associated with memory in NHW, but not NHB.

Conclusions:

These findings may suggest a threshold effect whereby NHB older adults are less likely to participate in social activities at the level needed to yield cognitive benefits. Lower social activity participation among NHB may reflect structural barriers and/or cultural differences in patterns of social engagement. This study highlights the need to improve measurement of and access to culturally relevant social activities for NHB to combat racial inequalities in cognitive aging.

Racial disparities in cognitive aging have been well documented, with non-Hispanic Black (NHB) older adults exhibiting worse performance on cognitive assessments and higher incidence of dementia than non-Hispanic Whites (NHW; Mayeda et al., 2016; Tang et al., 2001; Weuve et al., 2018). For example, in the domain of episodic memory, which has been shown to be a strong predictor of dementia risk (Boraxbekk et al., 2015; Welsh et al., 1991), several studies have demonstrated that NHB older adults obtain lower scores compared to NHW (Chen et al., 2021; Sol et al., 2020; Zaheed et al., 2021). Much research has focused on characterizing the greater prevalence of risk factors for cognitive impairment (e.g., lower socioeconomic status, discrimination) in NHB compared with NHW, but these risk factors only partially explain cognitive inequalities (Farina et al., 2020; Zahodne et al., 2017; Zahodne, 2021b). Less attention has been given to positive psychosocial factors that may reduce racial inequalities in cognition. Social engagement is a positive psychosocial factor that has been longitudinally associated with better late-life cognition (Bassuk et al., 1999; Glei et al., 2005; Holtzman et al., 2004; Zunzunegui et al., 2003). However, the role of social engagement as a protective factor for NHB and NHW remains understudied.

According to the social convoy model, social relations are multidimensional and shaped by personal characteristics, such as race (Antonucci et al., 2014; Kahn & Antonucci, 1980). Social engagement is an umbrella term that refers to multiple dimensions, including social network size, social activity, and social support. Longitudinal studies have demonstrated an association between larger social network size (i.e., number of family and friends; Cornoni-Huntley & Lafferty, 1986) and slower global cognitive (Bassuk et al., 1999; Holtzman et al., 2004; Zunzunegui et al., 2003) and episodic memory decline (Sharifian et al., 2020; Sörman et al., 2017) in older adulthood. A similar pattern has been observed between greater participation in social activities and less decline in the domains of episodic memory (Bosma et al., 2002; James et al., 2011), perceptual speed (James et al., 2011; Lövdén et al., 2005), visuospatial ability (James et al., 2011), and global cognition (Bosma et al., 2002; Glei et al., 2005; James et al., 2011). Furthermore, greater perceived social support (i.e., greater availability of help or emotional support; Berkman et al., 2000) has been associated with better global cognition at follow-up (Seeman et al., 2001), while satisfaction with social support was associated with less decline in episodic memory (Hughes et al., 2008). See Kelly et al., 2017 for a systematic review of social engagement and its differential effects on domains of cognitive functioning. However, most extant studies were conducted in largely white samples, and literature has yet to reach a consensus on how late-life social engagement may differ by race (Barnes et al., 2004) and whether associations between the dimensions of social engagement and cognition are moderated by race. In addition, measures of social activity typically include only six or fewer items, leaving the potential for this dimension to be incompletely characterized (Barnes, Mendes de Leon, Wilson, et al., 2004; Bosma et al., 2002; Bourassa et al., 2017; James et al., 2011; Krueger et al., 2009). Understanding racial differences in social engagement is a critical step toward developing interventions that target modifiable psychosocial factors to reduce cognitive disparities.

Findings from studies comparing the different facets of social engagement among NHB and NHW older adults have been mixed. For example, while NHB older adults have traditionally been characterized as having more frequent contact with social network members (e.g., extended family; Taylor et al., 1990) more recent studies suggest that NHB older adults may have smaller social networks (Ajrouch et al., 2001; Pugliesi & Shook, 1998) and lower levels of social engagement than NHW (Barnes, Mendes de Leon, Bienias, et al., 2004). Other studies have found no difference between NHB and NHW in social network size (Peek & O’Neill, 2001). Studies have also shown that the association between participating in more social activities and reduced global cognitive decline was stronger in NHW compared to NHB, but the association between social network size (Barnes et al., 2004) or social support (Kats et al., 2016) and global cognitive decline did not vary by race. However, another study observed a positive association between social network size and global cognition in NHB but not NHW older adults (Sharifian et al., 2018). Furthermore, Kats and colleagues (2016) found that among all NHW adults and among NHB women, having larger and more active social networks was associated with higher baseline global cognition, but there was no such relationship for NHB men. Inconsistencies in previous studies, which have only examined global cognition, suggest the need to clarify race differences in the association between multiple indicators of social engagement and multiple domains of cognition to shed light on which social factors may provide cognitive benefits for NHB compared with NHW older adults.

The present study seeks to address gaps in the extant literature on social engagement and cognitive aging disparities by using a more comprehensive measure of social activity participation and analyzing the effects of social engagement on multiple cognitive domains in a racially balanced sample of NHB and NHW older adults. Specifically, it aimed to examine (1) racial differences in three aspects of social engagement (social network size, social support, and social activities); and (2) the role of race as a moderator of independent associations between social engagement and cognitive performance across five domains (i.e., episodic memory, executive functioning, processing speed, language, and visuospatial functioning). Based on the social convoy model, which states that social relations and their impact on health can vary based on personal and situational characteristics, we predicted that the three indicators of social engagement and their associations with cognitive health would differ across NHW and NHB older adults. The overarching goal of the study was to more clearly characterize links between social engagement and cognitive aging among NHW and NHB older adults.

Method

Study Design

The Michigan Cognitive Aging Project (MCAP) is an ongoing longitudinal study of a regionally representative sample of adults ages 55+ without a dementia diagnosis at the time of study enrollment. Beginning in 2017, the cohort was recruited from Southeast Michigan through direct mailings using the 2016 voter registration list combined with census data. Inclusion criteria for the study were (1) age 55 or older; (2) English speaking. Exclusion criteria were having a self-reported diagnosis of dementia or memory impairment severe enough to interfere with daily activities. Although MCAP is collecting longitudinal data, at the time of the present study, only baseline data were available and thus all analyses are cross-sectional. The study protocol includes in person interviews and assessments to collect demographic, psychosocial, neuropsychological, physiological (e.g., height, weight, blood pressure), and biological data (e.g., blood spot). Self-reported race was dummy-coded into two mutually exclusive categories: non-Hispanic Black (NHB) and non-Hispanic White (NHW; reference group). Participants who did not self-identify as NHW or NHB (n = 34) were excluded in analyses for the current study. The current sample included 465 individuals. Ethical approval was obtained from the local institutional review board and all participants provided written informed consent.

Transparency and openness

We report how we determined our sample size, all data exclusions, all manipulations, and all measures in the study. Materials and analysis code for this study are available upon request. Statistical analyses were computed using SPSS version 27 (IBM Corp., Armonk, NY). This study’s design and its analyses were not pre-registered.

Measures

Cognitive Functioning

Cognitive functioning was quantified using factor scores generated by theory-based confirmatory factor analysis (CFA) as previously described (Zahodne, 2021a). The factor scores corresponded to five distinct cognitive domains: episodic memory, executive functioning, processing speed, language, and visuospatial functioning.

Episodic Memory.

Episodic memory functioning was assessed with seven indicators: the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) modified 10-Word List (Immediate, Delayed, and Recognition trials), the Craft Story task (Immediate and Delayed Recall trials), and the Benson Complex Figure (Delayed and Recognition trials).

Executive Functioning.

Executive functioning was assessed with three indicators: Number Span Backward, the Color Trails II, and the Stroop Color Word trial.

Processing Speed.

Processing speed was assessed with three indicators: the Symbol Digit Modalities Test, Color Trails I, and Stroop Color.

Language Functioning.

Language functioning was assessed by three indicators: Letter fluency, Animal fluency, and the Multilingual Naming Test (MINT).

Visuospatial Functioning.

Visuospatial functioning was assessed by three indicators: the Benson Complex Figure (Copy Trial), Judgement of Line Orientation (JLO), and the Montreal Cognitive Assessment (MoCA) Cube Copy.

Social Engagement

Social Network.

Social network size was evaluated using previously established self-report items (Cornoni-Huntley & Lafferty, 1986) querying the number of children, other relatives, and friends that respondents feel close to and talk to at least monthly. This measure has been widely used in previous studies (Barnes et al., 2004; Krueger et al., 2009; Mendes de Leon et al., 2001). Social network size was top-coded at 50 based on frequency distributions and in line with previous studies (e.g., Zahodne et al., 2019).

Social Support.

Perceived social support was measured with the question, “Do you feel that you can confide in them, that they are there when you need them?” in reference to four potential relationships (i.e., spouse/partner, closest child, closest other relative, and closest friend). Responses ranged from 1 (never) to 3 (always). A social support composite score was computed by averaging responses across the relationship types reported by the respondent. A higher score indicated greater perceived social support.

Social Activity.

For social activities, participants were asked how often they engaged in a list of 21 separate activities over the past month. The items were identical to those used in the Health and Retirement Study (HRS) Psychosocial Lifestyle Questionnaire (Hultsch et al., 1999; Jopp & Hertzog, 2010; Parslow et al., 2006). Responses were reverse coded so that frequency of participation in each activity ranged from 0 (never/nonrelevant) to 6 (daily). Using an approach adapted from Karp et al. (2006), each item was assigned a weight to indicate to what extent it involved social interaction, ranging from 0 (no social interaction) to 3 (high social interaction). In the Karp study, item weighting was validated independently by 13 cognitively intact older adult raters and was highly correlated with the authors’ original weights. For items that were not included in the study by Karp et al. (2006), co-authors (A.M.H. and A.Z.K.) independently assigned weights to each activity and then subsequently discussed with a larger group of 17 researchers (including L.B.Z., A.G.M, E.P.M., A.B.Z, & K.S.) to reach a consensus (Supplementary Table 1). A total of 15 activities involved social interaction and were included in the current study. Next, the frequency of engaging in each activity was multiplied by its social interaction score. The total social activity score is a sum of all weighted items, with higher scores indicating greater social activity participation.

Covariates

Analyses adjusted for the following covariates, all of which were self-reported at the time of study enrollment. Covariates were selected that could be conceptualized as confounding variables based on the existing literature demonstrating each having an association with both social engagement and cognition: age (years; English & Carstensen, 2014; Park et al., 2003); sex/gender (binary, reference group: men; Li & Singh, 2014; Thomas, 2011), completed years of education (0–20; Chapko et al., 2018; Glymour & Manly, 2008); marital status (binary; reference group: single/widowed/divorced/not currently partnered; (Harwood et al., 2000; Zaheed et al., 2021); wealth (assets minus debts; Cagney & Lauderdale, 2002; Rozanova et al., 2012). We additionally covaried for depressive symptoms (Glass et al., 2006; Wilson et al., 2004) and chronic disease burden (Meek et al., 2018; Nelson et al., 2020), which could be conceptualized as confounders and/or mediators of links between social engagement and cognition. Depressive symptoms over the past week were assessed with a 10-item version of the Center of Epidemiologic Studies Depression Scale (CES-D; Irwin et al., 1999; Radloff, 1977). Responses ranged from 0 (rarely or none of the time) to 3 (most or all the time). Responses were summed, and higher scores correspond to more depressive symptoms. Cronbach’s alpha in the current study was adequate (α = 0.825). Chronic disease burden was adapted from the HRS chronic disease health measure (Fisher et al., 2005) and was the sum of the presence of nine conditions: hypertension, diabetes, cancer or malignant tumor, chronic lung disease, heart problem, arthritis, dyslipidemia, eye problems, and hip fracture.

Statistical Analysis

Group differences were analyzed using Analysis of Variance (ANOVA) or Chi-Square tests as appropriate using two-tailed significance tests due to inconsistent findings in the previous literature. In addition to unadjusted group comparisons, separate models examined racial differences in each of the three social engagement variables, controlling for covariates. A p-value threshold of 0.05 was used for these descriptive analyses.

Given that the current study may not have been powered to detect higher-order effects and in line with recent recommendations for health disparities research (Ward et al., 2019), race-stratified hierarchical linear regression models were used to quantify the independent associations between the three social engagement variables (social network size, social activities, and social support) and one of the five cognitive outcomes (episodic memory, executive functioning, processing speed, language, visuospatial functioning), controlling for covariates. To address concerns of multiple comparisons, a Bonferroni-adjusted p value threshold of 0.01 (0.05 divided by five, the number of cognitive outcomes) was used for these analyses. Any result that did not meet the stricter threshold of 0.01 was considered exploratory for the purposes of generating hypotheses to be tested in future work.

Results

Participant characteristics are provided in Table 1. NHB older adults reported fewer years of education and less wealth, were less likely to be married/partnered, and had more chronic diseases and depressive symptoms compared with NHW older adults. Neither age nor sex/gender differed by race. NHB scored lower than NHW in all cognitive domains.

Table 1.

Participant Characteristics

	Full Sample	Non-Hispanic White	Non-Hispanic Black	Unadjusted		Adjusted
Variable (range)	N = 465	n = 217	n = 248	Effect Sizea	Group Comparison	Effect Sizeb	Group Comparison
Age (55 – 83)	63.59 (3.15)	63.69 (3.20)	63.50 (3.12)	0.03	NHW = NHB	-	-
% Women	56.6	55.8	62.9	0.07	NHW = NHB	-	-
% Married/Partnered	41.9	56.2	29.4	0.27*	NHW > NHB	-	-
Education (7 – 20 years)	14.19 (2.60)	15.21 (2.45)	13.29 (2.39)	0.40*	NHW > NHB	-	-
Chronic Disease Burden (0 – 8)	2.97 (1.73)	2.60 (1.65)	3.29 (1.73)	−0.20*	NHW < NHB	-	-
Depressive Symptoms (0 – 27)	8.16 (6.15)	6.99 (5.83)	9.22 (6.25)	−0.19*	NHW < NHB	-	-
Wealth (−120 – 950 ten thousands)	20.31 (63.99)	38.14 (88.12)	4.16 (15.50)	0.27*	NHW > NHB	-	-
Social Network Size (0 – 50)	8.28 (6.48)	7.96 (5.48)	8.56 (7.25)	−0.04	NHW = NHB	0.07	NHW = NHB
Social Activities (0 – 126)	55.92 (21.41)	58.24 (19.55)	53.95 (22.73)	0.10*	NHW > NHB	0.07	NHW = NHB
Social Support (1 – 3)	2.73 (0.36)	2.76 (0.35)	2.70 (0.37)	0.08	NHW = NHB	0.05	NHW = NHB
Episodic Memory (−2.99 – 2.28)	−0.03 (1.00)	0.52 (0.81)	−0.52 (0.90)	0.61*	NHW > NHB	0.44*	NHW > NHB
Executive Functioning (−3.02 – 2.56)	−0.02 (1.00)	0.59 (0.77)	−0.56 (0.88)	0.70*	NHW > NHB	0.52*	NHW > NHB
Processing Speed (−3.03 – 2.42)	−0.02 (1.00)	0.52 (0.81)	−0.50 (0.91)	0.59*	NHW > NHB	0.41*	NHW > NHB
Language (−2.40 – 2.74)	−0.01 (1.00)	0.61 (0.79)	−0.56 (0.85)	0.72*	NHW > NHB	0.53*	NHW > NHB
Visuospatial Functioning (−3.42 – 2.56)	−0.02 (1.00)	0.60 (0.75)	−0.58 (0.87)	0.72*	NHW > NHB	0.54*	NHW > NHB

Note. Values presented are means with standard deviations in parentheses or percentages, unless otherwise indicated.

^*p < 0.05 level.

^aCohen’s f was computed for continuous variables and Cramér’s V was computed for categorical variables.

^bCohen’s f was computed for variables of interest after adjusting for covariates.

NHB reported lower social activity scores than NHW, and this effect was attenuated after controlling for covariates (Table 1). In addition, NHW reported more frequent participation across a greater number of individual items (Supplementary Table 2). Social network size and levels of social support did not differ across racial groups. Social engagement variables were moderately correlated (rs ≤ 0.29), suggesting minimal risk of multicollinearity (Supplementary Table 3).

Race-stratified linear regression models revealed a positive association between social activity and episodic memory in NHW, but not NHB (Table 2; Figure 1). Among NHW, social activity was also positively associated with language (p = 0.011), but this finding does not meet the established p value threshold of 0.01 and therefore should be interpreted with caution (Supplementary Figure 1). Both NHW and NHB showed trends towards positive associations (0.01 < ps < 0.05) between social activities and cognitive domains of executive functioning and visuospatial functioning. Analyses within NHW also revealed a trend of negative associations (0.01 < ps < 0.05) between social support and each of the cognitive domains, but these associations between social support and all cognitive domains were positive and non-significant among NHB. There was no relationship between social network size and any cognitive domain within either racial group (Table 2; Supplementary Figure 1).

Table 2.

Results of Hierarchical Linear Regression Model Stratified by Race

		Episodic Memory		Executive Functioning		Processing Speed		Language		Visuospatial Functioning
Race	Variable	β	SE	β	SE	β	SE	β	SE	β	SE
Non-Hispanic White (n = 217)	Social Network Size	0.12	0.01	0.09	0.01	0.06	0.01	0.05	0.01	0.08	0.01
	Social Activities	0.19*	0.00	0.16+	0.00	0.12	0.00	0.19+	0.00	0.15+	0.00
	Social Support	−0.14+	0.15	−0.15+	0.15	−0.15+	0.16	−0.16+	0.15	−0.15+	0.15
	Adjusted R2		0.27		0.27		0.24		0.23		0.21
	R2 Change		0.06		0.05		0.03		0.05		0.04
	F Change		5.42*		4.07*		2.75+		3.92+		3.43+
Non-Hispanic Black (n = 248)	Social Network Size	0.02	0.01	−0.04	0.01	−0.04	0.01	−0.04	0.01	−0.05	0.01
	Social Activities	0.05	0.00	0.14+	0.00	0.12	0.00	0.13	0.00	0.15+	0.00
	Social Support	0.11	0.16	0.10	0.15	0.11	0.16	0.11	0.15	0.08	0.15
	Adjusted R2		0.23		0.24		0.23		0.23		0.20
	R2 Change		0.02		0.03		0.02		0.03		0.02
	F Change		1.42		2.37		2.21		2.44		2.18

Note. All models controlled for covariates. SE = standard error.

⁺p < 0.05 level

^*p < 0.01 level

Figure 1. Relationship between Social Engagement and Episodic Memory Across Racial Groups.

Note. Episodic memory is adjusted for covariates and the other social engagement variables. Each dot represents an individual participant. NHW= non-Hispanic White and NHB = non-Hispanic Black. +p < 0.05 *p < 0.01

Sensitivity Analyses

In order to determine whether the method of weighting each activity by its degree of social interaction drove our pattern of results, we performed a sensitivity analysis using the unweighted sum of frequency of participating in 15 social activities. The pattern of association was similar, but the effect size of the association between activity participation and episodic memory among non-Hispanic Whites was smaller and no longer significant (p = 0.02) in the model that included unweighted activity scores. In addition, to address concerns of reverse causation in which cognitive impairment may lead to lower engagement in social activities and the possibility that the sample could have included individuals with mild cognitive impairment, we performed another sensitivity analysis excluding participants whose memory performance was less than two SDs below the mean (n = 10). Patterns of association for the restricted sample were unchanged from the primary models.

Discussion

The present study observed racial differences in social activity participation, as well as in the link between social activity participation and episodic memory, but not for other cognitive domains. Specifically, NHB older adults reported less social activity participation than NHW, and greater participation in social activities was independently associated with better episodic memory performance in NHW, but not NHB older adults. There were no racial differences in social support or social network size, and neither of these indicators was significantly related to memory performance in either group. These results suggest that participating in social activities may represent a potential intervention target to improve cognition, but additional work is needed to examine whether cognitive benefits extend to NHB older adults.

Social Activity and Episodic Memory

The finding of a positive relationship between participation in social activities and cognition among NHW (Barnes, Mendes de Leon, Wilson, et al., 2004; Krueger et al., 2009; Peterson et al., 2020), and more specifically between social activities and memory (Bosma et al., 2002; James et al., 2011), is consistent with prior research. One potential mechanism underlying the beneficial effects of social activities on memory is cognitive stimulation (Vance & Crowe, 2006). Cognitive reserve theory suggests that neurophysiological characteristics, such as greater synaptic density, help to maintain cognitive performance despite underlying age- or disease-related brain pathology (Katzman et al., 1988). This theory emphasizes the importance of novelty to recruit and reinforce different brain networks. Engaging in a diverse array of social activities likely supports cognitive benefits by increasing exposure to environmental complexity, which encourages cognitive stimulation, supports brain plasticity, and fortifies cognitive reserve in older adulthood (Conroy et al., 2010; Fratiglioni & Wang, 2007; Tucker & Stern, 2011).

Another mechanism by which social activity may confer cognitive benefits is through promoting better mood and well-being. Social activities may buffer against the negative physiological impacts of stress on cognition (Cohen & Wills, 1985) such as elevated levels of glucocorticoids (Lupien et al., 1998). Stress and its physiological correlates are associated with worse memory function and reduced hippocampal volume in older adults (Lupien et al., 2009; Seeman & McEwen, 1996). Therefore, greater social activity has the potential to contribute to the preservation of late-life cognition by buffering stress. Importantly, the current study revealed racial disparities in social activity participation. Less frequent participation in social activities among NHB compared with NHW, as well as differential impact of social activities on episodic memory performance, may contribute to racial disparities in memory and dementia risk.

Racial Differences

In the current study, stratified models showed that more social activity was associated with better episodic memory in NHW, but not NHB, despite the smaller sample size of the NHW group. Recent work emphasizes the importance of examining both racial differences in the level of a risk/protective factor, as well as patterns of association in stratified models, when studying health disparities because a lack of interaction may cause one to overlook important contributing factors such as exposure prevalence (i.e., opportunities to engage in social activities; Ward et al., 2019). Indeed, the results of the current study are consistent with a threshold effect in that NHB older adults reported less social activity participation than NHW. Therefore, NHB participants may have been less likely than NHW to participate in social activities at the level needed to detect cognitive benefits.

Findings regarding racial differences in individual social activities may also align with studies that demonstrate the importance of activity variety in the formation of cognitive reserve. Peterson and colleagues (2020) showed that while leisure activity frequency was not associated with any cognitive domain in the pooled sample, leisure activity variety was positively associated with cognitive function in pooled sample and race-stratified models for NHW and NHB. Post-hoc analyses in the current study showed that NHW reported more frequent participation in a greater number of social activities compared with NHB (Supplementary Table 2). It could be the case that participating in a variety of different social activities confers cognitive benefits above and beyond the frequency of participation in any given activity.

The finding of lower overall social activity participation among NHB compared with NHW is consistent with previous studies that have shown NHB engage in less physical activity (Keadle et al., 2016; Paggi et al., 2016), volunteering (Hinterlong, 2006; Tang et al., 2012), and are less likely to participate in senior center activities (Pardasani, 2010) compared to NHW. One factor that may contribute to this disparity is the cultural relevance of activities. NHB have reported a lack of culturally diverse programming at senior centers, and expressed interest in certain activities such as creative writing and performing arts was limited to NHW (Pardasani, 2010). Employing more staff from racial and ethnic minority groups could facilitate increased participation through expanding the offerings of activities that are appealing to NHB older adults (Pardasani, 2004).

The absence of a relationship between activities and memory in the current study may also be due to an inadequate characterization of culturally relevant activities for NHB, compared to over-characterizing activities more appealing to NHW older adults. While Barnes et al. (2004) also found that the relationship between more frequent social activity and reduced global cognitive decline was stronger in NHW than NHB older adults, this association was still evident among NHB. Notably, the social activity measure used in that study included attendance of religious services and part-time or full-time work, neither of which were included in the leisure activity checklist used in the current study. Older NHB tend to report more religious activity than older NHW, which could represent a significant source of cognitive stimulation that was unaccounted for in the stand-alone leisure activity scale used in present study (Kim & McKenry, 1998; Kraal et al., 2019). The exclusion of religious activities from this commonly-used activities checklist may reflect cultural bias in measurement development. However, the present study included a more diverse array of activities in comparison to previous studies, including Barnes et al. (2004). While measurement bias may have contributed to the absence of a significant association between social activity and cognition among NHB, the fact that NHB reported less engagement in many of the queried activities still points to potential avenues to reduce racial disparities. Future studies should characterize which social activities are most popular and culturally relevant for NHB older adults in order to adequately assess whether they are associated with cognitive benefits. Preliminary evidence for positive associations between social activities and the cognitive domains of executive and visuospatial functional among NHB suggests that future studies with larger samples should examine whether social activities may be more beneficial for these non-memory domains in this group.

Beyond a lack of culturally relevant activities, previous research has revealed significant structural barriers to participation such as segregation, racial discrimination, and limited financial resources, which diminish opportunities for social engagement among NHB older adults (Mendes de Leon & Glass, 2004; Miner & Tolnay, 1998). Indeed, previous studies have found some measures of social activity to be correlated with socioeconomic status (Wilson et al., 2007), which may reflect the fact that those with higher socioeconomic status have more financial resources and/or more free time to engage in activities. Lower-income neighborhoods and neighborhoods with a higher proportion of NHB residents are less likely to have recreational facilities, such as fitness centers, membership sports, dance clubs, and public golf courses (Powell et al., 2006). In the present study, we found that the racial differences in social activity participation were slightly attenuated in analyses that adjusted for covariates such as education and wealth. In addition, wealth was correlated with social activities and social support among NHB, but not NHW. Education was associated with social activities among both groups. Future studies should include additional indicators of socioeconomic status such as occupational prestige, which may contribute to racial differences in social activity participation. Another study theorized that the historic institutional racism of art and cultural organizations, such as museums, limits usage among NHB older adults, resulting in a weaker association between arts and culture resources and cognitive function among NHB compared to NHW, despite NHB older adults having generally greater physical proximity to such spaces (Finlay et al., 2021). These results are consistent with findings from the current study that suggest structural inequalities may contribute to racial disparities not only through restricting the frequency of activity participation, but also due to greater vulnerability of NHB older adults to a variety of other hindering factors (e.g., racism or depressive symptoms). Indeed, a recent study found that depressive symptoms were more strongly related to lower cognitive leisure activity participation among NHB than NHW (Sharifian et al., 2021). Additional research is needed to characterize and eliminate barriers that disproportionately prevent NHB older adults from participating in activities that may be cognitively beneficial.

Social Network and Social Support

The lack of an association between social network size and episodic memory performance in the present study is consistent with several previous studies (Glei et al., 2005; Krueger et al., 2009; Seeman et al., 2001). Notably, other studies that found a main effect of social network size on cognitive function (Bassuk et al., 1999; Holtzman et al., 2004; Zunzunegui et al., 2003) did not control for other aspects of social engagement (e.g., support, participation in activities). Thus, the lack of a unique association involving social network size in the current study could suggest that other, related aspects of social engagement (e.g., social activities) are more proximally related to cognition.

Perceived social support was also not significantly associated with any cognitive domains in the current study. This finding is consistent with some previous studies (Bassuk et al., 1999; Dickinson et al., 2011) but not others (Holtzman et al., 2004; Krueger et al., 2009; Seeman et al., 2001). These inconsistent findings may point to individual differences in the need for social support and, by extension, its cognitive benefits. Because social support may confer cognitive benefits by buffering against the physiological and psychological effects of stress on cognition (Cohen & Wills, 1985), it is possible that social support is most beneficial for individuals who are experiencing major stressors like depression (Dickinson et al., 2011; Eisenberger et al., 2007). The current community-based study may not have observed reliable relationships between social support and cognition because of relatively low overall levels of depression. Preliminary evidence for negative associations between social support and multiple cognitive domains among NHW are consistent with other studies that included predominately NHW samples (Meister & Zahodne, 2021; Zahodne et al., 2019). Future studies should explore the possibility of reverse causation, as social support may increase among NHW older adults with worse cognitive functioning. Such studies should also consider racial differences, as all of the associations between social support and cognition among NHB were positive, though not significant.

Limitations and Future Directions

The present study is not without some limitations. The current findings are cross-sectional, and therefore we cannot draw conclusions about causal relationships. However, several previous longitudinal studies support prospective associations between greater social activity participation and reduced cognitive decline (Barnes et al., 2004; Bosma et al., 2002; Glei et al., 2005). In the current study, sensitivity analyses revealed that patterns of association persisted after excluding participants with impaired memory scores, providing some evidence against the alternate explanation that the findings are due to reverse causation where lower cognitive functioning leads to reduced social engagement. Future studies should include longitudinal measurement, though it should be noted that racial differences in cognitive level in the absence of strong evidence for racial differences in the rate of cognitive change suggests that cognitive level may be most relevant to the study of racial inequalities in dementia (Manly & Mungas, 2015). It is also important to consider how traditionally objective neuropsychological testing can be inherently culturally biased against NHB, leading to inaccuracies in observed test scores (Manly et al., 2004; Pedraza & Mungas, 2008). While subjective memory measures were not included in the present study, future studies should also consider the relationship between social engagement and self-report measures of memory functioning. Other limitations include using a single-item measure of perceived social support rather than distinguishing between different types of support such as emotional, instrumental and informational support, which may have differing patterns of association with cognition in older adulthood (Dickinson et al., 2011; La Fleur & Salthouse, 2017).

Strengths of the current study include the use of a more racially balanced sample relative to previous studies (Bassuk et al., 1999; Krueger et al., 2009), which eliminated the possibility that differing sample sizes were driving results from the stratified models. While examining sex/gender moderation was beyond the present scope, previous research observed that Black-White differences in the association between social networks and cognition varied by gender (Kats et al., 2016). Future studies with larger samples should examine the intersection of race and sex/gender in moderating the association between social engagement and memory. An additional strength was the use of an innovative method (adapted from Karp et al., 2006) for measuring social activity that accounts for different levels of social interaction (from low to high) as well as the frequency of participation, producing a more sensitive assessment. Results from sensitivity analyses demonstrated that weighting activities by their degree of social demand improved our ability to detect associations, suggesting that social interaction may be an important component of the protective cognitive effect of activities. We also included 15 different items for social activity compared to the 4–6 item measures of many previous studies. Other strengths include examining the independent effects of three social engagement variables on multiple domains of cognition and using a comprehensive set of covariates.

Conclusions

Findings from this research demonstrate the potential cognitive benefits of a socially active lifestyle in older adulthood and reveal important racial disparities. Intervention research focused on social activity participation as a means of improving older adults’ cognitive health must address racially patterned barriers to participating in social activities and consider cultural relevance.

Key Points.

Question:

Does race modify the association between social engagement and cognitive functioning among non-Hispanic black (NHB) and white (NHW) older adults?

Findings:

NHB older adults reported less social activity participation compared to NHW. Greater social activity participation was related to better memory performance among NHW, but not NHB older adults.

Importance:

Identifying and increasing access to culturally relevant social activities and reducing barriers to participation may help to address racial disparities in cognitive aging.

Next steps:

Future research should characterize cultural relevancy of social activities and structural barriers to participation for NHB older adults.