What is the effect of frailty in combination with loneliness or social isolation on all-cause mortality in the UK Biobank?

Problem

Three health and social care related challenges for ageing populations are frailty (a state of reduced physiological reserve), social isolation (objective lack of social connections), and loneliness (subjective experience of feeling alone). These are associated with various adverse outcomes. The impact of combinations of these factors, however, is less clear. We examined how frailty in combination with loneliness or social isolation is associated with all-cause mortality using UK Biobank data.

Approach

UK Biobank participants aged 37-73 were recruited 2006-2010. Baseline data assessed frailty (frailty phenotype based on 5 criteria, 0=robust, 1-2=pre-frail, ≥3=frail and frailty index based on a count of deficits and categorised as robust (0-0.12), mild (0.12-0.24), moderate (0.24-0.36) and severe (>0.36) frailty), social isolation (a three-item scale, ≥2=socially isolated) and loneliness (two questions, 2=lonely). Each frailty measure was analysed separately in combination with loneliness and social isolation. Cox-proportional hazards models assessed association between frailty and loneliness or frailty and social isolation and all-cause mortality, adjusting for age, sex, ethnicity, socioeconomic status, smoking and alcohol intake.

Findings

461,047 with complete data were included (mean age 56.5, 251,604 female). Social isolation was more prevalent in frail participants (e.g. 21.1%, 3,282/15,564, of frail participants based on frailty phenotype) compared to pre-frail (11.1%, 19,017/172,053) and robust (7.1%, 19,359/273,430) participants. Findings were similar for loneliness (loneliness prevalence 13.7%, 6.4% and 3.2% in frail, pre-frail or robust participants, respectively). Frailty, social isolation, and loneliness were all more common in the context of higher socioeconomic deprivation. Social isolation was associated with increased mortality risk at all levels of frailty (frailty phenotype). Compared to robust non-isolated participants, hazard ratio (HR) was 1.29 (95% confidence intervals 1.22-1.37) for robust with social isolation. HRs for pre-frailty were 1.96 (1.87-2.06) with and 1.45 (1.41-1.49) without social isolation. HRs for frailty were 3.38 (3.11-3.67) with and 2.89 (2.75-3.05) without social isolation. As confidence intervals do not overlap, likely to have statistically significant additive interaction between frailty and social isolation on mortality risk. Loneliness was associated with increased mortality risk in robust participants (HR 1.14 (1.04-1.25)) and in pre-frail participants (HR 1.67 (1.56-1.79) with loneliness and 1.46 (1.42-1.50) without loneliness). However, for frail participants hazard ratios were similar with or without loneliness (2.94 (2.64-3.27) and 2.90 (2.76-3.04), respectively). Findings were similar when using the frailty index in place of the frailty phenotype.

Consequences

Social isolation is associated with greater risk of mortality across the frailty spectrum. Loneliness is also associated with mortality, with the greatest relative increase in risk seen in robust or pre-frail people. Pro-active identification of social isolation and loneliness within primary care, regardless of physical health status, may provide important opportunities for intervention and risk mitigation. Health and social care policies should consider the need to mitigate loneliness and social isolation.