UK Biobank study links social connections with reduced all-cause and cardiovascular mortality

In a recent study published in BMC Medicine, researchers examined the associations between functional and structural components of social connection and all-cause and cardiovascular disease (CVD) mortality using data from the United Kingdom (UK) Biobank.

They examined independent and combined associations to understand how these social components interacted.

Study: Social connection and mortality in UK Biobank: a prospective cohort analysis. Image Credit: antoniodiaz/Shutterstock.com

Background

Between 9.2 and 14.4% of the global population feels lonely, and 25% of adults might be socially isolated; these statistics indicating the extent of a lack of social connection raise concern.

There are interrelated structural (e.g., frequency of social visits with family and friends) and functional (e.g., perceived loneliness) components of social connection, and the deficit of even one could heighten the risk of all-cause and CVD mortality. 

Prior studies have shown independent associations between a functional or structural social connection component and a higher risk of all-cause mortality using a single-item measure. 

On the other hand, some studies used composite scales, e.g., the Berkman-Syme Social Network Index, to measure the structural component of social connection.

A meta-analysis of prospective studies examined the association between functional or structural social connection and all-cause mortality and even quantified them.

However, the observed effect sizes represented the aggregate effects of different measures, showing no accountability towards the strength of each measure and its impact on health. These analyses, however, failed to recognize potential synergistic interactions between functional and structural components.

Several mechanisms come into play when examining this association; for instance, reverse causality, where disabilities hinder people from forming or sustaining relationships. However, which social connection components are associated with mortality remains unclear, whether they vary with the assessment methods or how direct and indirect factors influence them. 

Poor immune function and neurodevelopmental impairment are some of the factors directly influencing this association, while substance abuse and poor mental or physical health indirectly affect this association.

Overall, there is a lack of research examining different components of social connection in one dataset to delineate all their effects, including independent, additive, and multiplicative effects. 

Insights into the health impact of different social connection components and their interactions could help guide policy to enhance social connectedness and improve health via targeted interventions.

About the study

In the present study, researchers invited 502,536 UK Biobank participants enrolled between 2006 and 2010 to visit one of the 22 assessment centers in England, Scotland, or Wales.

They collected their baseline data, which included their physical measurements and additional information collected using a questionnaire and an interview taken by a trained healthcare professional.

The team examined baseline data and all-cause and CVD mortality (adverse health outcomes), where the International Classification of Diseases (ICD) 10th revision codes I05 to I99, G45, G46, and Z86.7 defined CVD mortality.

Further, they measured their ability to confide in someone close and feelings of loneliness (two functional components) and the frequency of friends and family visits, weekly group activity, and living alone (three structural components).

Study covariates included self-reported sex, ethnicity, smoking status, alcohol intake, physical activity levels, the month of assessment, and 43 long-term health conditions. Besides, they included body mass index (BMI) as a continuous measure and postcode of residence at recruitment as a constant variable.

The researchers used a Cox proportional hazard model (time-to-event analysis) to examine the associations between social connection and mortality for all participants.

Given highly connected measures of social connection and the covariates, they detected potential multicollinearity using generalized variance inflation factors (GVIF) for all study variables.

They then examined the association between each functional component measure and adverse health outcomes separately, adjusting for all confounders.

They also investigated the combined association of these measures and their interactions concerning adverse health outcomes. They also created a new dichotomous ‘functional isolation’ variable and examined its associations with mortality.

Structural component analyses examined the association between each structural component measure and adverse health outcomes separately. Finally, the researchers also investigated the combined effect of functional and structural components.

Results

The main analysis encompassed 458,146 UK Biobank participants with a mean age of 56.5 years. Of these, 95.5% were of white ethnicity, and 54.7% were women. Of a total of 33,135 deaths during the average follow-up of 12.6 years, 1.1% were due to CVDs.

In general, participants reporting reduced social connection were more likely involved in unhealthy practices (e.g., smoking), were deprived socioeconomically, and belonged to a minority ethnicity. They also had higher BMI and more long-term health conditions.

Participants reporting functional measures of social connection, inability to confide in others, and feeling lonely showed a strong association with higher all-cause and CVD mortality, with respective hazard ratios (HRs) of 1.07 and 1.17 and 1.06 and 1.08, respectively.

Combining these measures formed a new dichotomous functional isolation variable, which also showed an association with higher all-cause and CVD mortality, with HRs of 1.08 and 1.16, respectively.

Fully adjusted models of associations between the frequency of friends and family visits and all-cause mortality showed that visits to friends and family less than once a month were associated with a much higher risk of all-cause mortality, with HRs for once every three months and never of 1.11 and 1.39, respectively.

The pattern was similar for CVD mortality, but associations were stronger and had wider confidence intervals (CIs).

Once-a-month visits to close ones provided maximum benefits, and once validated in other datasets, it could help identify which measures of social connection would be most beneficial to target via interventions.

Likewise, not engaging in weekly group activity and living alone increased the risk of all-cause and CVD mortality, HRs: 1.13 and 1.10 and 1.25 and 1.48, respectively, compared to those who engaged in group activities weekly and lived with at least one other.

Models of combined associations also showed that fewer friends and family visits heightened the risk of all-cause mortality regardless of whether participants reported engaging in a weekly group activity.

Furthermore, examining the combined associations between the two functional component measures and all-cause mortality when structural isolation was present showed that being unable to confide was associated with higher all-cause mortality regardless of feelings of loneliness (HRs: 1.41 vs. 1.38). 

However, in the absence of structural isolation, this association showed a greater difference between that reporting often feeling lonely and not (HR 1.16 vs. 1.07), highlighting the complexity and possible hierarchy in social connection components, especially for those who experienced numerous types of social disconnection.

Thus, it is critical to consider different measures when exploring the combined effects of all social components on health outcomes.

For instance, the authors observed that a lack of friends and family visits and living alone masked the lower risk of mortality associated with regular group activity.

Exploring this concept in other datasets could highlight intervention targets for the most isolated people in society.

Conclusions

So far, there is no standardized measure for social connection. However, the independent associations observed in this study between risks of living alone and mortality and its interactions with friends and family visits and weekly group activity point to further work ascertaining whether living alone could represent a simplified measure in studies examining social connection.

Those who live alone and show additional concurrent markers of structural isolation represent a population that could benefit from targeted support. Thus, policies and interventions that address different social connection components should target such high-risk groups.

Future studies should investigate the role of potential mediators (e.g., mental health problems) to elucidate the mechanistic pathways by which social disconnection causes mortality.