Abstract
Objective: social support has demonstrated cross-sectional associations with greater cognitive function and a protective effect against cognitive decline in older adults, but exploration of its temporal role in cognitive ageing from mid-life to older adulthood has been limited. We aimed to quantify the associations of social support, assessed at mid-life, with cognitive function in mid-life and with cognitive decline into late life among African Americans and Caucasians.
Methods: data from the community-based, prospective Atherosclerosis Risk in Communities (ARIC) cohort of 15,792 biracial participants were examined for baseline and longitudinal associations of mid-life social support with global cognition at mid-life and with 20-year change in global cognition, respectively, stratified by race. Interactions with sociodemographic and cardiometabolic covariates were additionally explored within each race group. Social support was ascertained using two metrics: interpersonal support and social network.
Results: interpersonal support was directly associated with greater global cognition at baseline in both race groups. Social network was directly associated with greater global cognition at baseline among Caucasians and African American females, but it was not significantly associated with global cognition in African American males. Neither mid-life social support measure was associated with 20-year change in global cognition.
Conclusions: higher levels of social support were moderately associated with greater multi-dimensional cognitive function at mid-life, but mid-life social support was not associated with temporal change in global cognitive function over 20 years into late life. Prospective studies with time-dependent measures of social support and cognition are needed to better understand the role of social engagement in ageing-related cognitive functioning.
Keywords: older people, cognition, social, support, cognitive, psychosocial, social support
Objective
Cognitive outcomes among adults approaching older age are influenced by vascular exposures [1–4], sociodemographic variables [5–7] and psychosocial factors [8–13]. Social support is a particular psychosocial construct that may decelerate the cognitive ageing process, stemming from both its positive, cross-sectional associations with greater cognitive function [9–13] and protective associations against steeper cognitive decline reported among older adults [11, 14–16].
Models explaining the link between social support and cognition indirectly through social mechanisms have been proposed [17, 18]. In addition, a direct neurological correlation between social support and increased grey matter volume has been observed, albeit in younger adults [19]. The temporal relationship between social support and cognition from mid-life to late life, however, is not well understood primarily because of the paucity of prospective data with cognitive and social support information collected prior to the seventh decade of life (i.e. 60–69 years of age).
We assessed whether social support is associated with cognitive function at mid-life, and whether mid-life social support is protective against faster cognitive decline into older adulthood, in efforts to clarify the temporal relationship between social support and ageing-related cognition. To achieve this, we estimated the associations of two metrics of social support ascertained at mid-life—interpersonal support and social network—with global cognitive function in mid-life and with 20-year change in global cognitive function among African American and Caucasian men and women aged 48–67 from the large, population-based Atherosclerosis Risk in Communities (ARIC) Study.
Methods
Study population
The prospective ARIC Study [20] enrolled 15,792 participants aged 45–64 during 1987–89 (Visit 1). Participants were recruited using a probability sample from four US communities: Forsyth County, NC; Jackson, MS; suburbs of Minneapolis, MN and Washington County, MD. ARIC participants were then followed up at clinical exams in 1990–92 (Visit 2), 1993–95 (Visit 3), 1996–98 (Visit 4) and 2011–13 (Visit 5). The study was approved by the Institutional Review Boards at each study centre.
Cognitive assessments
The Digit Symbol Substitution Test (DSST) [21], Delayed Word Recall Test (DWRT) [22] and Word Fluency Test (WFT) [23] of cognition were administered at ARIC Visits 2, 4 and 5. The DSST is a test of executive function and processing speed, the DWRT measures verbal learning and immediate memory and the WFT assesses executive function and expressive language. All tests are validated, standardised instruments commonly used in epidemiologic studies of dementia and cognitive function. Together, the tests include most of the measures recommended in the Uniform Data Set executed in 2005 across all National Institute on Aging-sponsored Alzheimer's Disease Centers [24].
A z-score was calculated for each test to standardise participants' cognitive scores to their baseline scores at Visit 2 (also baseline for the purposes of our study). This was done by first calculating the mean and standard deviation of scores on the DSST, DWRT and WFT for the entire study population at baseline. Then for each test, the baseline population mean was subtracted from each participant's raw score at subsequent visits, and this difference was divided by the baseline population standard deviation. A global z-score was created at Visits 2, 4 and 5 by averaging individual z-scores across the three tests among participants with three non-missing scores at each visit. This global z-score serves as a standardised, multi-domain measure of cognitive performance used to represent global cognitive function.
Social support measures
Participants' social support was also assessed at baseline (Visit 2) using the short form of the Interpersonal Support Evaluation List [ISEL-SF; 25] as well as the Lubben Social Network Scale [LSNS; 26]. Both questionnaires are psychometrically valid and routinely applied instruments to measure perceived social support [27, 28]. The 16-item ISEL-SF is used to measure an individual's perception of his or her appraisal support, tangible assets, belonging support and self-esteem support. Each question of the ISEL-SF is scored on a 4-point rating scale (definitely true, probably true, probably false and definitely false; scored 0–3). The total (summed) score is an aggregate index of social support, with higher scores indicating greater levels of perceived interpersonal support [28]. Established cut-offs for ISEL-SF scores do not exist, so we categorized total scores into quartiles based on the distribution of the subgroup of interest (e.g. Caucasian participants).
The LSNS is a self-assessed measure of the size and availability of one's active social network of family, friends and peers, consisting of 10 questions on a 0–5 rating scale. Total scores range from 0 to 50 and are classified based on established levels of risk for social isolation [29]: socially ‘isolated’ (≤20), ‘high risk’ for isolation (21–25), ‘moderate risk’ for isolation (26–30) and ‘low risk’ for isolation (≥31). We collapsed the first two of these categories together to form a category designated as ‘isolated/high risk’, as <5% of participants scored within the ‘isolated’ range.
Analytic samples
Participants were excluded from analyses if they were not of African American or Caucasian race (n = 42), were African American in Minneapolis (n = 20) or Washington County (n = 30), had missing global z-scores at baseline (n = 83), had missing education level (n = 20) or had any missing responses on the ISEL-SF (n = 237) or LSNS (n = 900). This resulted in analytic samples for cross-sectional analyses of social support measures with global z-score at baseline to be composed of 13,782 participants for ISEL-SF and 13,119 participants for LSNS. The study populations for longitudinal analyses included those participants from cross-sectional analyses who repeated multi-domain cognitive testing at Visits 4 and 5 (N = 5,411 for ISEL-SF and N = 5,195 for LSNS).
Statistical models and covariates
All analyses were stratified by race, given the reported racial differences in rates of cognitive decline [30]. Within each race group, we also tested for modification of cross-sectional and longitudinal associations between social support measures and cognitive outcomes by each baseline covariate (indicated by P < 0.10 for the interaction term).
Generalised linear models (GLMs) were used to estimate cross-sectional mean differences in global z-score across categories of each social support measure, with the least level of the social support measure (i.e. first quartile for ISEL-SF and ‘isolated/high risk’ group for LSNS) as the referent category. Fully adjusted GLMs controlled for the following potential confounders at baseline: age (continuous), sex, study centre, highest education level (<high school, high school, >high school), cigarette smoking (current, former, never), alcohol consumption (current, former, never), prevalent hypertension (mean resting blood pressure >140/90 mmHg or self-reported use of antihypertensive medication) and prevalent diabetes (fasting glucose of 126 mg/dl, non-fasting glucose of 200 mg/dl, a self-reported history of diabetes or treatment for diabetes at Visit 2).
Subsequently, generalised estimating equation (GEE) models were used to estimate longitudinal mean differences in 20-year change in global z-score by categories of each social support measure, with the lowest category of each social support measure as the reference group. Fully adjusted GEE models controlled for the aforementioned baseline covariates, time on study and interactions between these covariates and time on study. GEE models were chosen due to their ability to account for intraindividual correlation of cognitive test scores through repeat visits. Statistical analyses were performed in SAS, version 9.3 (SAS Institute Inc., Cary, NC, USA).
This work was supported by the National Heart, Lung, and Blood Institute contracts (HHSN268201100005C, HHSN268201100006C, HHSN268201100007C, HHSN268201100008C, HHSN268201100009C, HHSN268201100010C, HHSN268201100011C and HHSN268201100012C). Neurocognitive data are collected by U01 HL096812, HL096814, HL096899, HL096902, HL096917 with previous brain MRI examinations funded by R01-HL70825. Financial sponsors did not play a role in the design, execution, analysis or interpretation of data, or writing of the study.
Results
Characteristics of participants (N = 13,119) involved in the examination of the cross-sectional association between social network (as measured by LSNS) and global z-score at baseline are summarised in Table 1. Baseline distributions of covariates across categories of ISEL-SF were similar as those across LSNS, so we only present characteristics by LSNS. Furthermore, while the highest proportion of participants scored in the middle categories on the ISEL-SF or LSNS, for brevity, we report estimates of mean difference in cognitive outcomes only between those categories representing the highest versus lowest categories of social support. Of note, associations between the middle relative to lowest level of the social support measures and cognitive outcomes were generally similar or slightly depreciated in magnitude.
Table 1.
Baseline characteristicsa by race group and Lubben Social Network Scale (LSNS) category (N = 13,119)
| African Americans |
Caucasians |
|||||
|---|---|---|---|---|---|---|
| Low risk | Moderate risk | Isolated/high risk | Low risk | Moderate risk | Isolated/high risk | |
| (n = 916) | (n = 1,265) | (n = 909) | (n = 2,902) | (n = 4,570) | (n = 2,557) | |
| Age, years, mean (SD) | 56.6 (5.8) | 56.1 (5.8) | 55.9 (5.7) | 57.6 (5.7) | 57.2 (5.7) | 57.0 (5.7) |
| Female, n (%) | 576 (62.9) | 853 (67.4) | 560 (61.6) | 1,563 (53.9) | 2,587 (56.6) | 1,237 (48.4) |
| Study centre, n (%) | ||||||
| Forsyth County, NC | 98 (10.7) | 139 (11.0) | 119 (13.1) | 897 (30.9) | 1,423 (31.1) | 840 (32.9) |
| Jackson, MS | 818 (89.3) | 1,126 (89.0) | 790 (86.9) | – | – | – |
| Minneapolis, MN | – | – | – | 1,063 (36.6) | 1,650 (36.1) | 870 (34.0) |
| Washington County, MD | – | – | – | 942 (32.5) | 1,497 (32.8) | 847 (33.1) |
| Education level, n (%) | ||||||
| <High school | 365 (39.8) | 466 (36.8) | 314 (34.5) | 500 (17.2) | 651 (14.3) | 418 (16.3) |
| High school | 281 (30.7) | 358 (28.3) | 248 (27.3) | 1,271 (43.8) | 2,213 (48.4) | 1,109 (43.4) |
| >High school | 270 (29.5) | 441 (34.9) | 347 (38.2) | 1,131 (39.0) | 1,706 (37.3) | 1,030 (40.3) |
| Cigarette smoking, n (%) | ||||||
| Current | 224 (24.5) | 324 (25.6) | 225 (24.8) | 505 (17.4) | 979 (21.4) | 622 (24.3) |
| Former | 255 (27.9) | 381 (30.2) | 271 (29.8) | 1,153 (39.7) | 1,857 (40.7) | 1,075 (42.1) |
| Never | 436 (47.6) | 558 (44.2) | 412 (45.4) | 1,244 (42.9) | 1,731 (37.9) | 860 (33.6) |
| Alcohol consumption, n (%) | ||||||
| Current | 295 (32.2) | 441 (34.9) | 311 (34.3) | 1,769 (61.0) | 2,933 (64.2) | 1,714 (67.0) |
| Former | 253 (27.7) | 386 (30.5) | 321 (35.4) | 489 (16.8) | 816 (17.9) | 443 (17.3) |
| Never | 367 (40.1) | 437 (34.6) | 275 (30.3) | 644 (22.2) | 817 (17.9) | 400 (15.7) |
| Hypertension, n (%) | 432 (47.5) | 591 (47.1) | 432 (47.8) | 688 (23.8) | 1,052 (23.1) | 611 (24.0) |
| Diabetes, n (%) | 239 (26.3) | 298 (23.9) | 224 (24.9) | 377 (13.0) | 493 (10.8) | 280 (11.0) |
| Global z-score, mean (SD) | −0.73 (1.03) | −0.68 (1.05) | −0.69 (1.01) | 0.27 (0.86) | 0.28 (0.85) | 0.15 (0.88) |
aCharacteristics appearing in this table are for analytic sample using LSNS as the exposure.
Social support and cognition in mid-life
In baseline models adjusted for sociodemographic and additionally for cardiometabolic covariates, higher levels of interpersonal support and social network were each significantly associated (P < 0.05) with greater multi-dimensional cognitive function in both race groups (Table 2). Estimates of cross-sectional associations between interpersonal support and global z-score, controlled for age, sex, and study centre, were attenuated by 43 and 33% upon adjustment for education level in African Americans and Caucasians, respectively. Estimates from fully adjusted models indicate a difference of 32 and 22% of one standard deviation in the baseline global z-score of our analytic samples of African Americans (1.03) and Caucasians (0.86), respectively, between the highest and lowest quartiles of interpersonal support.
Table 2.
Race-stratified estimates, with 95% confidence intervals (CIs) and P values, of mean differences in global z-score of multi-domain cognitive function (global z-score) at baseline and in 20-year change in global z-score by social support metric
| African Americans |
Caucasians |
|||||||
|---|---|---|---|---|---|---|---|---|
| N | Estimate | 95% CI | P value | N | Estimate | 95% CI | P value | |
| Interpersonal support (highest versus lowest quartile) | ||||||||
| Cross-sectional (Baseline) | ||||||||
| Model 1 | 3,285 | 0.60 | (0.51, 0.69) | <0.001 | 10,497 | 0.30 | (0.26, 0.34) | <0.001 |
| Model 2 | 3,285 | 0.34 | (0.27, 0.42) | <0.001 | 10,497 | 0.20 | (0.16, 0.24) | <0.001 |
| Fully adjusted | 3,220 | 0.33 | (0.25, 0.41) | <0.001 | 10,450 | 0.19 | (0.15, 0.23) | <0.001 |
| Longitudinal (20-year change) | ||||||||
| Fully adjusted | 1,012 | −0.01 | (−0.14, 0.12) | 0.84 | 4,373 | 0.01 | (−0.05, 0.05) | 0.95 |
| Social network (‘low risk’ versus ‘isolated/high risk’) | ||||||||
| Cross-sectional (Baseline) | ||||||||
| Model 1 | 3,090 | 0.01 | (−0.08, 0.09) | 0.93 | 10,029 | 0.11 | (0.07, 0.15) | <0.001 |
| Model 2 | 3,090 | 0.07 | (0.00, 0.15) | 0.05 | 10,029 | 0.12 | (0.08, 0.16) | <0.001 |
| Fully adjusted | 3,029 | 0.08 | (0.01, 0.15) | 0.03 | 9,981 | 0.12 | (0.08, 0.16) | <0.001 |
| Longitudinal (20-year change) | ||||||||
| Fully adjusted | 976 | 0.01 | (−0.11, 0.13) | 0.87 | 4,194 | −0.03 | (−0.09, 0.03) | 0.29 |
Model 1: adjusted for age, sex and study centre;
Model 2: adjusted for age, sex, study centre and education level;
Fully adjusted: adjusted for age, study centre, sex, education level, cigarette smoking, alcohol consumption, hypertension and diabetes.
In fully adjusted models, the level of active social network was modestly (∼10% of one standard deviation in baseline global z-score between the highest and lowest categories) associated with global z-score at baseline in each race group. The estimate of this association became statistically significant among African Americans in fully adjusted models. Conversely, subsequent control for covariates only inappreciably affected the estimate of the cross-sectional association between social network and global z-score among Caucasians.
Social network (‘low risk’ versus ‘isolated/high risk’) was significantly associated with greater global z-score at baseline in African American females (mean difference = 0.17, 95% CI: 0.08, 0.26) but not in African American males (mean difference = −0.07, 95% CI: −0.19, 0.05). There was no other evidence of interaction detected.
Mid-life social support and 20-year cognitive change
In fully adjusted models, the mean 20-year decline in global z-score was −0.85 (−0.98, −0.73) with SD = 0.06 among African American participants and −0.99 (95% CI: −1.06, −0.93) with SD = 0.03 among Caucasian participants. Longitudinal associations between social support metrics and 20-year change in global z-score were not observed in African Americans or Caucasians (Table 2). Lastly, no evidence of effect modification by any covariates was detected in longitudinal analyses.
Conclusions
Using prospective data from the population-based ARIC cohort, this study examined associations of mid-life interpersonal support and social network with global cognition at mid-life and with change in global cognition into late life. Higher level of social support was moderately associated with greater global cognitive functioning at mid-life but did not predict change in global cognitive function into older adulthood.
The associations between social support and global cognition observed at mid-life in this study, and reported at late life in other studies [10–16], may point to an influence of social engagement on cognition in different life epochs, although such a hypothesis is not supported by the literature. Conversely, these cross-sectional associations may represent reverse or bi-directional causality.
The absence of longitudinal associations in this study suggests the lack of a relationship between mid-life social support and cognitive decline into late life. Yet reliance on a single measurement of this psychosocial exposure, assessed only at baseline, may have failed to ascertain time-varying associations. For instance, changes in social support systems that occur between mid-life and older age may have the strongest effect on cognitive trajectories thereafter.
Bias attributable to selective cohort attrition over the course of the long cohort follow-up may also contribute to explain the lack of longitudinal associations. Participants with less social support in mid-life did in fact have a higher incidence of co-morbidities, disability and other factors associated with death or dropout over the extended follow-up period in these data. Their underrepresentation in analytic samples may have diminished the ability of this study to detect significant protective effects for mid-life social support against steeper cognitive decline, depending on whether this group's cognitive decline was sufficiently accelerated in comparison to that of those participants who were included in analytic samples. To counterbalance for this predicted underrepresentation, we applied inverse-probability-of-attrition weights to longitudinal models. These weights, however, did not lead to any marked change in longitudinal estimates.
While social support in mid-life would not appear to be protective against cognitive decline into late life, higher level of social support was moderately associated with greater cognitive function at mid-life. Further clarification of the roles of social support in cognitive ageing will require consideration of changes in the strength, duration and timing of social networks across the life course. Prospective cohorts initiated at mid-life or earlier, characterized with both time-varying social support and cognitive measures, would assist in further elucidating the temporal relationship between social engagement and cognitive ageing.
Key points.
We investigated the temporal role of social engagement in cognitive ageing using population-based data.
Social support was modestly associated with greater cognitive function at mid-life.
Mid-life social support did not predict cognitive decline into older adulthood.
Disentanglement of the temporal relationship between social engagement and cognitive ageing necessitates prospective studies with time-repeating measures of social support and cognition.
Conflicts of interest
None declared.
Acknowledgements
The authors thank the staff and participants of the ARIC study for their important contributions.
References
- 1.Yaffe K, Vittinghoff E, Pletcher MJ et al. Early adult to mid-life cardiovascular risk factors and cognitive function. Circulation 2014; doi:CIRCULATIONAHA-113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Debette S, Seshadri S, Beiser A et al. Midlife vascular risk factor exposure accelerates structural brain aging and cognitive decline. Neurology 2011; 77: 461–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Rawlings AM, Sharrett AR, Schneider AL et al. Diabetes in midlife and cognitive change over 20 years: a cohort study. Ann Intern Med 2014; 161: 785–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Haring B, Wu C, Coker LH et al. Hypertension, antihypertensive treatment, sodium intake and cognitive decline. Circulation 2015; 131: A50. [Google Scholar]
- 5.Mortamais M, Portet F, Brickman AM et al. Education modulates the impact of white matter lesions on the risk of mild cognitive impairment and dementia. Am J Geriat Psychiat 2014; 22: 1336–45. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Amieva H, Mokri H, Le Goff M et al. Compensatory mechanisms in higher-educated subjects with Alzheimer's disease: a study of 20 years of cognitive decline. Brain 2014; 137: 1167–75. [DOI] [PubMed] [Google Scholar]
- 7.Gottesman RF, Rawlings AM, Sharrett AR et al. Impact of differential attrition on the association of education with cognitive change over 20 years of follow-up the ARIC neurocognitive study. Am J Epidemiol 2014; 179: 956–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Aggarwal NT, Wilson RS, Beck TL et al. Perceived stress and change in cognitive function among adults aged 65 and older. Psychosom Med 2014; 76: 80–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Richard E, Reitz C, Honig LH et al. Late-life depression, mild cognitive impairment, and dementia. JAMA Neurol 2013; 70: 383–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Lee BK, Glass TA, James BD, Bandeen-Roche K, Schwartz BS. Neighborhood psychosocial environment, apolipoprotein E genotype, and cognitive function in older adults. Arch Gen Psychiatry 2011; 68: 314–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Shankar A, Hamer M, McMunn A, Steptoe A. Social isolation and loneliness: relationships with cognitive function during 4 years of follow-up in the English Longitudinal Study of Ageing. Psychosom Med 2013; 75: 161–70. [DOI] [PubMed] [Google Scholar]
- 12.Yeh SCJ, Liu YY. Influence of social support on cognitive function in the elderly. BMC Health Serv Res 2003; 3: 9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Sims RC, Hosey M, Levy SA, Whitfield KE, Katzel LI, Waldstein SR. Distinct functions of social support and cognitive function among older adults. Exp Aging Res 2014; 40: 40–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Barnes LL, De Leon CM, Wilson RS, Bienias JL, Evans DA. Social resources and cognitive decline in a population of older African Americans and whites. Neurology 2004; 63: 2322–6. [DOI] [PubMed] [Google Scholar]
- 15.Bassuk SS, Glass TA, Berkman LF. Social disengagement and incident cognitive decline in community-dwelling elderly persons. Ann Intern Med 1999; 131: 165–73. [DOI] [PubMed] [Google Scholar]
- 16.Fratiglioni L, Wang HX, Ericsson K, Maytan M, Winblad B. Influence of social network on occurrence of dementia: a community-based longitudinal study. Lancet 2000; 355: 1315–9. [DOI] [PubMed] [Google Scholar]
- 17.Holtzman RE, Rebok GW, Saczynski JS, Kouzis AC, Doyle KW, Eaton WW. Social network characteristics and cognition in middle-aged and older adults. J Gerontol B Psychol 2004; 59: P278–84. [DOI] [PubMed] [Google Scholar]
- 18.Seeman TE, Lusignolo TM, Albert M, Berkman L. Social relationships, social support, and patterns of cognitive aging in healthy, high-functioning older adults: MacArthur studies of successful aging. Health Psychol 2001; 20: 243–55. [DOI] [PubMed] [Google Scholar]
- 19.Che X, Wei D, Li W et al. The correlation between gray matter volume and perceived social support: a voxel-based morphometry study. Soc Neurosci 2014; 9: 152–9. [DOI] [PubMed] [Google Scholar]
- 20.The ARIC Investigators. The Atherosclerosis Risk in Communities (ARIC) Study: design and objectives. Am J Epidemiol 1989; 129: 687–702. [PubMed] [Google Scholar]
- 21.Wechsler D. WAIS-R Manual: Wechsler Adult Intelligence Scale-Revised. New York: Psychological Corporation; 1981. [Google Scholar]
- 22.Knopman DS, Ryberg S. “A verbal memory test with high predictive accuracy for dementia of the Alzheimer type”. Arch Neurol 1989; 46: 141–5. [DOI] [PubMed] [Google Scholar]
- 23.Lezak MD. Neuropsychological Assessment. New York: Oxford University Press; 2004. [Google Scholar]
- 24.Morris JC, Weintraub S, Chui HC et al. The Uniform Data Set (UDS): clinical and cognitive variables and descriptive data from Alzheimer Disease Centers. Alz Dis Assoc Dis 2006; 20: 210–6. [DOI] [PubMed] [Google Scholar]
- 25.Cohen S, Mermelstein R, Kamarck T, Hoberman HM. Measuring the Functional Components of Social Support. Social Support: Theory, Research and Applications. Netherlands: Springer, 1985; 73–94. [Google Scholar]
- 26.Lubben JE. Assessing social networks among elderly populations. Fam Community Health 1988; 11: 42–52. [Google Scholar]
- 27.Payne TJ, Andrew M, Butler KR, Wyatt SB, Dubbert PM, Mosley TH. Psychometric evaluation of the interpersonal support evaluation list–short form in the ARIC study cohort. SAGE 2012; 2: 1–8. [Google Scholar]
- 28.Lubben J, Gironda M. Measuring Social Networks and Assessing Their Benefits. In: Phillipson C, Allan G, Morgan D, eds. Social Networks and Social Exclusion. Sociological and Policy Perspectives. Hampshire, UK: Ashgate; 2004. [Google Scholar]
- 29.Rubinstein RL, Lubben JE, Mintzer JE. Social isolation and social support: an applied perspective. J Appl Gerontol 1994; 13: 58–72. [Google Scholar]
- 30.Sisco S, Gross AL, Shih RA et al. The role of early-life educational quality and literacy in explaining racial disparities in cognition in late life. J Gerontol B Psychol 2014; 0: gbt133. [DOI] [PMC free article] [PubMed] [Google Scholar]