Perceived social isolation and cognitive functioning. Longitudinal findings based on the German Ageing Survey

There is a lack of longitudinal studies, which are both based on nationally representative samples and use standardized instruments to quantify social isolation. Thus, the purpose of this study was to determine the link between perceived social isolation and cognitive functioning longitudinally.

Objectives: There is a lack of longitudinal studies, which are both based on nationally representative samples and use standardized instruments to quantify social isolation.
Thus, the purpose of this study was to determine the link between perceived social isolation and cognitive functioning longitudinally. Results: Regressions showed that increases in perceived social isolation were associated with decreases in cognitive functioning. With regard to covariates, decreases in cognitive functioning were associated with increases in aging and worsening selfrated health, whereas changes in marital status, employment status, income, physical functioning, and physical illnesses were not associated with the outcome measure.
Conclusions: Based on a nationally representative sample and exploiting the panel data structure, the study findings extend current knowledge by showing that increasing perceived social isolation contributes to decreases in cognitive functioning among individuals aged 40 years and over longitudinally. Future longitudinal studies based on panel data methods are required to validate the study findings.

K E Y W O R D S
cognition, cognitive functioning, social disconnectedness, social exclusion, social isolation

| INTRODUCTION
It is widely acknowledged that there is an ongoing change in the demographic structure of industrialized countries, meaning that the number of individuals in late life is increasing continuously. It is also widely understood that age is associated with decreases in cognitive functioning.
Moreover, higher age is associated with increased social isolation. 1 Cognitive functioning 2 and social isolation 3 are both associated with subsequent morbidity and mortality, highlighting the importance of these factors.
Social isolation is linked to decreased social activities or social engagement, which, in turn, can stimulate cognitive functioning. 4 An association between social isolation and cognitive functioning has been demonstrated by recent cross-sectional 5 and longitudinal studies. 6,7 For example, it has been shown that social isolation is related to decreased cognitive function based on data from a Spanish nationally representative sample (n = 1691; individuals ≥50 years; two waves). Similar findings were made by Evans et al based on longitudinal data from the Cognitive Function and Ageing Study-Wales (n = 1524; individuals aged 65 years and over; two waves).
A recent systematic review on the longitudinal association between social isolation and cognitive functioning showed that these factors are weakly associated. 8 However, most of these longitudinal studies focused on social network and social activities without using standardized instruments (eg, using living alone as a surrogate for social isolation). 9 Thus, this review 9 also observed that there is a lack of longitudinal studies, which are both based on nationally representative samples and use standardized instruments to quantify social isolation. Furthermore, to date, we are only aware of one very recent study that has used linear fixed effects (FE) regression analysis to study the determinants of cognitive functioning over time. 10 This recent study investigated the longitudinal link between financial hardship and cognition using the FE regression analysis.
Using the FE regression analysis in studying the link between perceived social isolation and cognitive functioning over time has the advantage of reducing the problem of unobserved heterogeneity (eg, genetic disposition), which is a serious concern when analyzing data from large survey studies. 11   cohort-sequential design, which means that new baseline samples were introduced in the second, third, and fifth waves. While more than 10 300 individuals took part in the fifth wave, more than 6600 individuals were interviewed in the sixth wave. For example, in the most recent wave, the response rate was 63%, which is comparable to other survey studies performed in Germany. 12 Because perceived social isolation was only quantified from wave 5 onward, we used waves 5 and 6 in our study. Further details with regard to the DEAS study are provided by Klaus et al 13 Written informed consent was provided by all participants. Because the criteria for an ethical statement were not fulfilled (such as risk for the respondents or use of invasive methods), an ethics committee approval was not required.

| Dependent variable
The digit symbol test, adapted from the digit symbol substitution test, 14 was used to measure cognitive functioning. It measures perceptual motor speed and processing speed of visual perception and information. Ranging from 1 to 92, higher scores reflect better cognitive functioning. Numerous previous studies focusing on the determinants of cognitive functioning 15 have used it. Furthermore, it has been demonstrated that its psychometric characteristics are sound. 16

| Independent variables
A scale developed by Bude and Lantermann 17 was used to quantify perceived social isolation. It has four items (each item ranges from 1 [strongly agree] to 4 [strongly disagree]). The mean rating across all items was computed, with higher values corresponding to higher social isolation. In our study, Cronbach's alpha was .88.
Age, family status (married, and living together with spouse), others (married, and living separated from spouse, single, divorced, widowed), household net equivalent income (in Euro), as well as labor force participation (employed, retired, other [not employed]) were adjusted for in the analysis. Furthermore, self-rated health (from 1 = very good to 5 = very bad), physical functioning (using the "physical functioning" subscale of the SF-36; ranging from 0 [worst] to 100 [best]), and the number of physical illnesses (eg, diabetes; sum score ranges from 0 to 11 physical illnesses) were adjusted for in the analysis.

Key Points
• Link between social isolation and cognitive functioning was examined longitudinally In a sensitivity analysis, the number of important people in regular contact (ranging from 0 to 9), loneliness (using the widely used De Jong Gierveld scale 18 , ranging from 1 to 4; higher values correspond to higher loneliness), and depressive symptoms using the established

15-item version of the Center for Epidemiologic Studies Depression
Scale 19 , which ranges from 0 (which corresponds to no depressive symptoms) to 45 (reflecting severe depressive symptoms), were adjusted for.
It is worth noting that perceived social isolation is correlated with network size and loneliness. However, these are distinct concepts. 20 For example, individuals can perceive themselves as socially isolated without feeling lonely and while having a large network of frequent and close personal relationships (or vice versa). 21

| Statistical analysis
In large survey studies, it is almost impossible to control for This also means that factors that are constant over time (eg, sex or education) cannot be used as regressors in the FE regression analysis. Cluster-robust SEs were computed. 22 The significance level was set at P < .05. Stata 15.1 (StataCorp, College Station, TX) was used in the current study.

| Sample characteristics
We described observations included in the FE regression analysis in Table 1. In the analytical sample, 50.3% were women and average age was 65.0 years (SD: 10.7). The average perceived social isolation score was 1.6 (SD: 0.6), and the average cognitive functioning score was 64.5 years (SD: 13.3). Further details are described in Table 1.

| Regression analysis
Following the recommendations of Cameron and Trivedi, 23 we first tested whether our variables of interest have enough within variation to obtain precise estimates ("xttab" and "xttrans" commands in Stata).
Afterward, a series of FE regressions were conducted. In the main regression analysis (see Table 2) (n = 6420 observations), increases in perceived social isolation were associated with decreases in cognitive functioning (β = −1.13, P < .01). With regard to covariates, decreases in cognitive functioning were associated with increases in aging (β = −.31, P < .001) and worsening self-rated health (β = −.63, P < .01), whereas changes in marital status, employment status, income, physical functioning, and physical illnesses were not associated with the outcome measure.
In the sensitivity analysis, our main model was extended by adding the network size, loneliness, and depressive symptoms as covariates. However, the strength of the link between perceived social isolation and cognitive functioning was similar (β = −1.00, P = .01). Furthermore, it is important to note that decreases in cognitive functioning were not associated with changes in network size (β = .06, P = .32), changes in loneliness (β = −.08; P = .86), and changes in depressive symptoms (β = −.02, P = .59).
We also tested whether sex moderates the link between perceived social isolation and cognitive functioning; however, the interaction term (sex × social isolation) did not achieve statistical significance (P = .34).

| Main findings
The purpose of this study was to determine the association between perceived social isolation and cognitive functioning using a longitudinal approach based on data from a nationally representative sample of older adults in Germany. FE regressions revealed that increases in perceived social isolation were associated with decreases in cognitive functioning. With regard to covariates, decreases in cognitive functioning were associated with increases in aging and worsening selfrated health, whereas changes in marital status, employment status, income, physical functioning, and physical illnesses were not associated with the outcome measure.

| Previous research and possible explanations
Previous cross-sectional studies mainly showed a moderate association between social isolation and cognitive functioning. 5 Some longitudinal studies also showed a moderate association between social isolation at baseline and subsequent cognitive functioning. 8 We showed an intraindividual link between social isolation and cognitive functioning in our study and, therefore, extend previous knowledge.
The underlying mechanisms of this association are largely unknown. A possible explanation may be that social isolation is associated with infrequent social activities that stimulate cognitive functioning. 4 This is in line with the three hypotheses provided by (i) Active lifestyles (in terms of the absence of social isolation as well as mental and physical factors) can increase cognitive reserve (cognitive reserve hypothesis) by increasing the resilience of an individual to neuropathological damage. The underlying idea is, therefore, that individuals differ in the ability to cope with Alzheimer's disease (AD) pathology. The cognitive reserve hypothesis is supported by the robust finding of a link between low education and increased risk of dementia and AD. 24 According to Fratiglioni et al, 24 cognitive reserve might be facilitated by a better skill to use alternative brain networks as required or by more efficient use of brain networks.

| Strengths and limitations
This is one of a few longitudinal studies investigating the link between perceived social isolation and cognitive functioning. Data were drawn from a nationally representative sample of middle aged and older adults. A main challenge in survey studies-the problem of unobserved heterogeneity-was mitigated using the FE regression analysis. Both our main independent variable (perceived social isolation) and our outcome measure (cognitive functioning) were assessed using widely used scales. However, a published validation study (perceived social isolation) is missing. Several potential confounders were adjusted for in the analyses. Only a small sample selection bias has been identified in the DEAS study. 13 The possibility of selection bias cannot be entirely dismissed, Furthermore, decreases in cognitive functioning can lead to increases in social isolation 32  Observations 6420 Number of individuals 3210 Note: Beta-coefficients were reported; cluster-robust SEs in parentheses. *P < .05; **P < .01; ***P < .001; +P < . 10. lead to severely biased estimates. For this reasons, FE regressions were used in the present study.

| CONCLUSION
Based on a nationally representative sample and exploiting the panel data structure, the study findings extend current knowledge by showing that increasing perceived social isolation contributes to decreases in cognitive functioning among individuals aged 40 years and over longitudinally. Future longitudinal studies based on panel data methods are required to validate the study findings.

FUNDING INFORMATION
This research did not receive any funding from agencies in the public, commercial, or not-for-profit sectors.

CONFLICT OF INTEREST
The authors declare that they have no conflict of interest.

DATA AVAILABILITY STATEMENT
The data used in this study are third-party data. The anonymized datasets of the DEAS (1996,2002,2008,2011,2014, and 2017) are available for secondary analysis. The data have been made available to scientists at universities and research institutes exclusively for scientific purposes. The use of data is subject to written data protection agreements. Microdata of the DEAS are available free of charge to scientific researchers for nonprofitable purposes. The FDZ-DZA provides access and support to scholars interested in using DEAS for their research. However, for reasons of data protection, signing a data distribution contract is required before data can be obtained. Please see for further information (data distribution contract): https://www.dza.
de/en/fdz/access-to-data/formular-deas-en-english.html. Written informed consent was provided by all participants. The DEAS study is in accordance with the Helsinki declaration. Because the criteria for an ethical statement were not fulfilled (such as risk for the respondents or use of invasive methods), an ethics committee approval was not required.