How do components of social connection interact in their associations with all-cause and CVD mortality? A UK Biobank cohort analysis.

Problem

Social connection is a complex social phenomenon comprised of functional (e.g., loneliness), structural (e.g., social isolation), and quality (e.g., relationship strain) components. Each component of social connection is associated with higher all-cause mortality and cardiovascular disease. However, different components may interact in their combined associations with adverse health outcomes and could help identify higher risk groups. We aimed to explore associations between components of social connection – friends and family visit frequency (FFVF), participation in weekly group activities, living alone, and perceived loneliness – and all-cause and CVD mortality, and to examine how these components interact with one another to modify associations with adverse health outcomes.

Approach

Data: UK Biobank - 502,536 adults recruited 2006–10, age 37-73. Baseline self-reported exposures: three structural social connection components: FFVF (6 category ordinal variable), weekly group activity (yes/no), and living alone (yes/no); 2 functional components: frequency of ability to confide (6 category ordinal variable), and often feels lonely (yes/no). Outcomes: all-cause (ACM) and CVD mortality (CVDM) ascertained via linked national registries. Cox proportional hazard models, adjusted for sociodemographic and health confounders, used to examine combined associations and interactions for outcomes. Sensitivity analyses excluded those with prior CVD/cancer or who died within 2 years of recruitment.

Findings

Participants with full data (458,136 [91.2%]) were included. After median 12.6 years follow-up, there were 33,135 (7.2%) deaths, of which 5,112 (1.1%) were CVD deaths. Each component was independently associated with both outcomes. For FFVF, incrementally stronger associations (higher risk) were seen from a frequency of visits of less than monthly. In combined associations, compared to least isolated and not lonely, the association with outcomes generally strengthened stepwise with each additional component. Those with lowest FFVF, no weekly group activity, lived alone, and not lonely had strongest associations with ACM (HR 95%CI 2.34 [1.65-3.30]). However, there was considerable overlap of mortality estimates for those with lowest FFVF irrespective of other components. There was a significant interaction between FFVF and living alone for ACM; compared to highest FFVF, HRs (95% CIs) for lowest FFVF was 1.33 (1.22-1.46) in those not living alone and 1.77 (1.61-1.95) in those living alone.

Consequences

Each social connection component is important. However, lowest FFVF and living alone were associated with greatest mortality. The interaction between FFVF and living alone indicates that those with no friends or family contacts who also live alone are at particularly high risk of mortality and could benefit from targeted intervention. FFVF of less than monthly may represent a threshold effect which could inform interventions. While UK Biobank is a large and rich prospective data set it is not representative of the UK population. Associations may not be causal, although similar results for the sensitivity analyses add weight against reverse causality.