key: cord-0306117-fm9vwoft
authors: Green, M. J.; Whitley, E.; Niedzwiedz, C. L.; Shaw, R. J.; Katikireddi, S. V.
title: Social contact and inequalities in depression and loneliness among older adults: A mediation analysis of the English Longitudinal Study of Ageing
date: 2020-07-02
journal: nan
DOI: 10.1101/2020.07.01.20143990
sha: eabcdfef143af631078233f4c9e5dd68da417708
doc_id: 306117
cord_uid: fm9vwoft

Background: Social contact, including remote contact (by telephone, email, letter or text), could help reduce social inequalities in depression and loneliness among older adults. Design: Cross-sectional survey. Participants: 8th wave of the English Longitudinal Study of Aging (2016/17), stratified by age (n=1,635 aged <65; n=4,123 aged 65+). Methods: Inverse probability weighted estimation of average effects of weekly in-person and remote social contact on depression (score of 3+ on 8-item CES-D scale) and two measures of loneliness (sometimes/often feels lonely vs hardly ever/never; and top quintile of UCLA loneliness scale vs all others). We also estimated controlled direct effects of education, partner status, and wealth on loneliness and depression under two scenarios: 1) universal infrequent (<weekly) in-person social contact; and 2) universal weekly remote social contact. Results: Weekly in-person social contact was associated with reduced odds of depression and loneliness, but associations with remote social contact were weak. Lower education raised odds of depression and loneliness, but differences were attenuated with infrequent in-person contact. Respondents living alone experienced more depression and loneliness than those living with a partner, and less wealth was associated with more depression. With universal infrequent in-person contact, differences narrowed among those aged under 65 but widened among those aged 65+. Universal weekly remote contact had little impact on inequalities. Conclusions: Reduced in-person social contact may increase depression and loneliness among older adults, especially for those aged 65+ who live alone. Reliance on remote social contact seems unlikely to compensate for social inequalities.

Many older adults experience depression and loneliness [1] [2] [3] [4] [5] , which are associated with lower quality of life, health-risk behaviours and poor physical health [5] [6] [7] [8] [9] [10] [11] , and are socially patterned, being more common among older adults from poorer socioeconomic backgrounds and who live alone [1-6, 12, 13] . Social mitigation responses to the worldwide Covid-19 pandemic, emphasising physical distance from others, may exacerbate these issues [14] [15] [16] .

Social contact with friends and family may contribute to inequalities in loneliness and depression. Social contact may be in-person or remote (i.e. via phone/internet etc.), but is distinct from loneliness, which is a perceived feeling of social isolation [1, 3, 9, 17] . One can feel lonely amid frequent social contacts, or not feel lonely with very few social contacts. Nevertheless, more frequent social contact is associated with less depression and loneliness [1, 3, 18, 19] , and is more common among older people who are socioeconomically advantaged or who live with a partner [18, 20, 21] .

The importance of digital communications has been increasing for some time, but remote contact has become especially salient during the covid-19 pandemic. Many countries have enacted 'lockdown' social mitigation measures, where reduced physical proximity to others is intended to slow the infection transmission rate, but such measures are likely to have psychological impacts [14, 15, 22, 23] . Identified as high risk for covid-19, older adults have been advised to follow stringent social distancing measures to avoid infection [14, 15] . Maintaining or increasing remote social contact has been promoted to mitigate impacts of social distancing [14, [22] [23] [24] , but remote contact has been less strongly associated with depression and loneliness than in-person contact [19] , so may not compensate adequately. Reducing in-person social contact or increasing remote social contact could both potentially narrow inequalities (by reducing social patterning of salutary factors) or widen them (if some benefit more from contact than others) [3] . Understanding the different contributions that inperson and remote social contacts make to inequalities in loneliness and depression is important in informing policy responses to improve health, especially with regards to mitigation of social distancing.

We use the most recent wave of data from the English Longitudinal Study of Aging, to estimate answers to the following questions:

1. What are the effects of in-person and remote social contact on loneliness and depression? 2. To what extent are inequalities in loneliness and depression affected by making inperson social contact infrequent for everyone? 3. To what extent are inequalities in loneliness and depression affected by making remote social contact frequent for everyone?

The English Longitudinal Study of Ageing (ELSA) is a large-scale, representative, longitudinal panel study of people aged 50 and over living in private households in England [25] . The core sample has been drawn from respondents to the Health Survey for England (HSE) since 1998. We focus on 7,223 core sample members who were interviewed at the most recent survey wave (wave 8: 2016/2017; 82.4% of those who were eligible for inclusion because they were still alive and living in the UK). We excluded 930 respondents who did . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted July 2, 2020. . https://doi.org/10.1101/2020.07.01.20143990 doi: medRxiv preprint not return a self-completion questionnaire, and a further 535 respondents with missing data on relevant variables (see Appendix 1) . This resulted in a final sample of 5,758 respondents (79.7% of the core sample members interviewed).

Details of the measures used are given in Appendix 2, but we describe the main measures briefly here. Our primary outcomes were depression and loneliness. Depression was measured with the Centre for Epidemiologic Studies Depression Scale (CES-D), using a validated threshold to establish probable psychiatric cases. Two measures of loneliness were included: 1) responses to the question "How often do you feel lonely?" (some of the time/often versus hardly ever/never); and 2) the UCLA loneliness scale, dichotomised to compare the loneliest quintile with the other four. Analyses for CES-D and the UCLA loneliness scale were repeated using continuous scores.

We considered three potential sources of inequality in loneliness and depression: 1) education, comparing low (completion of ordinary schooling or less) versus high (qualifications from continuing education); 2) partner status (living with or without a partner); and 3) household wealth, split into quintiles.

In-person social contact was coded to compare those who met up with children, other family, or friends at least weekly, versus those with less frequent contact. Remote contact was coded the same way but for contact by telephone, letter, email or text.

Potential confounders included are shown in Table 1 (detail in Appendix 2).

All models used weights to adjust for sampling and drop-out [25] . Preliminary analyses indicated interactions with age, so we stratified our analyses between those aged under 65 (the UK retirement age) and those aged 65 or more.

For research question 1 we separately estimated the effects of in-person and remote social contact on loneliness and depression, using inverse probability weighting (IPW) [26] . This is described more fully in Appendix 3 but adjusts for measured potential confounders (including education, partner status and wealth). Estimates represent average effects of in-person and remote social contact on depression and loneliness within the sample (assuming no residual confounding or reverse causation).

[ Figure 1 about here] For research questions 2 and 3, we assume social contact may mediate effects of either education, partner status or wealth on loneliness and depression outcomes, as shown in Figure 1 . We distinguish pre-and post-exposure confounders. Pre-exposure confounders (X) are potential common causes of the exposure, the mediator and the outcome. In contrast, post-exposure confounders (C) are potential common causes of the mediator and the outcome but may (or may not) be caused by the exposure. This distinction is important for estimating the effect of the exposure after intervention on the mediator [27] . Table 1 shows which variables were considered as pre/post confounders depending on the exposure in question. Given ambiguity regarding causal direction between partner status and socioeconomic . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 2, 2020. . variables (housing tenure, social class, economic activity and wealth) we performed sensitivity analyses with these variables re-positioned as pre-exposure confounders.

[ Table 1 about here] Total exposure effects were estimated using IPW (Appendix 3) to estimate the average effects of education, partner status or wealth on depression and loneliness, adjusting for preexposure confounders. These were compared to estimates of controlled direct effects (CDEs) for each exposure on loneliness and depression, which were estimated using inverseprobability weighted marginal structural models to adjust appropriately for pre-and postexposure confounders (Appendix 3) [27] . The CDE represents the estimated effect of the exposure under a hypothetical intervention to set the mediator (social contact) to the same value for all respondents. We estimated CDEs for two scenarios: 1) universal infrequent (<weekly) in-person social contact; and 2) universal weekly remote social contact. For wealth, we focus on results comparing the least to the wealthiest quintile.

Appendix 4 presents characteristics of the analytical sample by age. Respondents aged 65 or more (n=4,123) were more likely to be white, living alone, less educated, and have frequent in-person social contact than those aged under 65 (n=1,635). Appendix 4 also summarises characteristics of those who were excluded due to missing data, who were more likely to be at the extremes of the age range, from ethnic minority groups, living with a partner, less educated, less wealthy, have less frequent in-person and remote social contact, and were more likely to feel lonely or depressed. Table 2 shows proportions within each age group experiencing less than weekly in-person and remote social contact. Infrequent in-person social contact was more common in both age groups than infrequent remote social contact. Infrequent in-person social contact was most common among those with more education, who lived with a partner, or were wealthier. These differences were more pronounced in the 65+ age group, and the wealth difference was in the opposite direction for the under 65 age group. There was little social patterning of remote social contact by education, and the direction of patterning by partner status differed by age. Infrequent remote contact was more common in the lowest than the highest wealth quintile in the under 65 age group.

[ Table 2 about here] Table 3 shows estimates of the average effects of in-person and remote social contact on loneliness and depression. Weekly in-person social contact was associated with reduced odds of loneliness in both age groups, even after adjustment for confounders. Adjustment also revealed a borderline association between more frequent in-person social contact and reduced odds of depression in the 65+ age group. There were no clear associations with remote social contact after adjustment for confounders. Findings using continuous scores for depression and loneliness were similar (Appendix 5).

[ Table 3 about here] Table 4 shows ATE and CDE estimates of education, partner status and wealth on depression and loneliness for those both below and over the age of 65. Findings were similar with . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 2, 2020. . continuous measures of depression and loneliness (Appendix 6). Lower education was associated with more loneliness and depression in both age groups, and the CDE estimates suggested that universally infrequent in-person social contact would attenuate this inequality. CDE estimates of ensuring weekly remote contact for everyone indicated relatively little impact on inequalities in loneliness, but some minor attenuation of inequalities in depression.

[ Table 4 about here]

Living alone was associated with more depression and loneliness than living with a partner. CDE estimates for universal infrequent in-person social contact indicated that these differences would widen among those aged over 65 but narrow among those aged under 65. CDE estimates for ensuring weekly remote contact for everyone indicated relatively little impact on these inequalities. Findings for partner status were similar with socioeconomic confounders re-positioned as pre-exposure confounders (Appendix 7).

Being in the least wealthy compared to the wealthiest quintile was not clearly associated with loneliness except for the UCLA loneliness measure among the 65+ age group but was associated with greater odds of depression in both age groups. CDE estimates for universal infrequent in-person social contact indicated attenuation of this inequality in depression among those aged under 65 and the inequality in loneliness among those aged 65+, but increased inequalities in depression among those aged 65+. CDE estimates of ensuring weekly remote contact for everyone indicated relatively little impact on these inequalities, though there was some attenuation among the 65+ age group. Results for other wealth quintiles (Appendix 8) were similar, though there was not a clear linear wealth gradient.

In-person social contact was associated with reduced risk of depression and loneliness in older adults in England, while associations with remote social contact were relatively weak. We estimated effects of restricting everyone to infrequent (<weekly) in-person social contact and findings suggested narrower inequalities in depression and loneliness by education. Inequalities in depression and loneliness by partner status and in depression by wealth were impacted differently depending on age, with restricted in-person contact narrowing these inequalities in those aged <65 but widening them among adults aged 65+. Estimates for ensuring everyone received weekly remote contact indicated relatively little impact on inequalities.

Our findings are consistent with others showing associations between social contact and depression and loneliness [1, 3, 18] , with in-person contact more important than remote [19] . While others found social contacts concentrated among the more advantaged [18, 20, 21] , we found weekly in-person contact was less common among more advantaged adults, especially those aged 65+. Considering this and that in-person social contact was on average associated with less loneliness and depression, the narrowing of inequalities indicated by our CDE estimates (which allow for interaction between social contact and education) suggests heterogeneity in the impact of in-person social contact, with those of higher education deriving greater benefit. While we did adjust for close relationships with friends and other family members, this could be due to differences in the quality of contacts, or differences in frequency of contact beyond the weekly threshold used.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 2, 2020. . https://doi.org/10.1101/2020.07.01.20143990 doi: medRxiv preprint Our findings suggest that reductions in in-person social contact, e.g. under social mitigation measures related to covid-19, are likely to lead to increases in depression and loneliness in older adults. However, impacts may be experienced unequally and we estimated that those aged 65+ who live alone were particularly at risk for depression and loneliness under restrictions to in-person contact. This has been anticipated [14, 15] , with remote social contact advised to mitigate these impacts [22] [23] [24] . We found that remote contact, at least as experienced pre-covid-19, seemed unlikely to compensate, with little effect on depression or loneliness, or on inequalities in depression and loneliness by education, partner status or wealth (even when assuming no disparities in establishing regular remote contact).

Our estimates based on pre-covid-19 data may not necessarily generalise. For example, infrequent social contact under covid-19 mitigation measures, may have different effects to infrequent social contact experienced under other circumstances [28] . Moreover, the covid-19 situation may have prompted qualitative improvements in remote contact that alter its effect. For example, our measure of remote contact did not specifically ask about contact via videocalling, which is likely to have become increasingly relevant during the pandemic (though respondents may have included this as telephone contact). A number of suggestions for helping older adults manage social isolation focus on enhancing benefits from remote contacts, such as cognitive-behavioral interventions and techniques [29] , or interventions to set up remote 'telehealth' groups [30] . Our results, which estimate effects of remote contact as it was experienced pre-covid-19, should not be taken to mean that such efforts are likely to be ineffective, but highlight the importance of better understanding how benefit can be derived from remote contact, and how effective forms of remote contact might impact on inequalities (both issues likely to remain salient after the pandemic).

Additionally, we focus solely on mechanisms of social contact but there are several other mechanisms related to social mitigation of covid-19 that could impact inequalities in depression and loneliness among older adults, including bereavement, and anxieties or worries related to one's health or economic situation or those of loved ones [15] . Furthermore, while data were cross-sectional, our analyses assumed causal direction from exposure to social contact to depression and loneliness. While we may be confident that education preceded other measures, our results could also be accounted for by social contacts affecting wealth or partner status, by depression or loneliness affecting social contact, or by residual confounding.

In conclusion, reductions in in-person social contact could result in increased depression and loneliness among older adults. Adults aged 65+ who lived alone appeared especially vulnerable to reductions in social contact, with greater estimated increases in risk for depression and loneliness than similarly aged adults living with a partner. Inequalities in depression and loneliness by education could narrow as in-person contact is reduced, as those with more education seemed to derive more benefit from contact. Remote social contact seemed insufficient to mitigate adverse impacts or inequalities, and more attention is needed into how remote contact can become more efficacious, or to other strategies for addressing inequalities in depression and loneliness among older adults.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 2, 2020. . https://doi.org/10.1101/2020.07.01.20143990 doi: medRxiv preprint

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 2, 2020. . https://doi.org/10.1101/2020.07.01.20143990 doi: medRxiv preprint Depending on whether in-person or remote social contact was the mediator of interest, the other was included as a post-exposure confounder. b Since the causal direction between partner status and these socioeconomic variables was particularly ambiguous we performed sensitivity analyses where these variables were re-positioned as pre-exposure confounders. IADL=Instrumental Activities of Daily Living.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 2, 2020. . . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 2, 2020. . Table 1 . These estimates assume no residual confounding or reverse causation.

. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted July 2, 2020. . a ATE: Average Treatment Effect, i.e. the estimated average effect of each exposure, after adjusted for pre-exposure confounders (see Table 1 ). These estimates (and the CDE estimates) assume no residual confounding or reverse causation. b CDE: Controlled Direct Effect, i.e. an estimate of the effect of each exposure if social contact were set to the value shown in parentheses. CDE estimates allow for interactions between social contact and the exposure and additionally adjust for differences in post-exposure confounders that were not due to the exposure (see Table 1 ).

Loneliness and Isolation: Evidence Review. London: Age UK

Social Class, and Common Mental Disorders: Longitudinal Evidence from Three Cohorts in the West of Scotland

The relationship between wealth and loneliness among older people across Europe: Is social participation protective?

Late-Life Loneliness in 11 European Countries: Results from the Generations and Gender Survey

A tune in "a minor" can "b major": A review of epidemiology, illness course, and public health implications of subthreshold depression in older adults

Loneliness in the general population: prevalence, determinants and relations to mental health

Loneliness, social isolation, and behavioral and biological health indicators in older adults

Loneliness and social isolation as risk factors for coronary heart disease and stroke: systematic review and meta-analysis of longitudinal observational studies

Social isolation, loneliness and health in old age: a scoping review

The burden of mental disorders

Depression in older adults

Education and Risk for Late Life Depression: A Meta-Analysis of Published Literature

Depressive symptoms among older adults: The impact of early and later life circumstances and marital status

COVID-19 and the consequences of isolating the elderly

Mitigating the wider health effects of covid-19 pandemic response

Multidisciplinary research priorities for the COVID-19 pandemic: a call for action for mental health science. The Lancet Psychiatry

A Short Scale for Measuring Loneliness in Large Surveys: Results From Two Population-Based Studies

Loneliness, social relations and health and well-being in deprived communities

Does Mode of Contact with Different Types of Social Relationships Predict Depression in Older Adults? Evidence from a Nationally Representative Survey

The social capital of older people

Social Networks Among Men and Women: The Effects of Age and Socioeconomic Status

The psychological impact of quarantine and how to reduce it: rapid review of the evidence

Psychological effects of the Coronavirus disease-2019 pandemic

Intergenerational Digital Engagement: A Way to Prevent Social Isolation During The COVID-19 Crisis

English Longitudinal Study of Ageing: Waves 0-8

An Introduction to Propensity Score Methods for Reducing the Effects of Confounding in Observational Studies

Marginal Structural Models for the Estimation of Direct and Indirect Effects

Is this social isolation?-we need to think broadly about the impact of social experiences during covid-19

Strategies to Promote Social Connections Among Older Adults During "Social Distancing" Restrictions