Article Text
Abstract
Background Between 1997 and 2017 the number of middle-aged people living alone in the UK increased by 53% and loneliness is now recognised as an important policy area. We aimed to understand the interrelationships between loneliness, living arrangements and emotional support in predicting suicidal thoughts and behaviours.
Methods Between 2006 and 2010 sociodemographic and health data were collected from 5 00 000 participants, aged 40–69, in UK Biobank. These data were linked to hospital admission records for self-harm and suicidal ideation until March 2015, and records for death by suicide until February 2016. Additionally, in 2016–2017, 1 50 000 participants completed an online questionnaire which probed thoughts of self-harm, self-harm behaviour and attempted suicide. Exposures assessed were baseline measures of self-reported loneliness, living arrangements and emotional support (frequency of confiding). Deaths by suicide and hospital admissions were investigated with Cox proportional hazards models and logistic regression was used for self-report outcomes. Analyses were adjusted for socio-demographic factors including deprivation and employment, and multimorbidity.
Results In adjusted analyses loneliness was the risk most consistently related to all outcomes including death by suicide, (hazard ratio (HR) 1.75, 95% CI 1.22 to 2.51), hospital admissions (HR 4.41, 95% CI 2.50 to 7.76) and self-reported suicide attempts (HR 5.38, 95% CI 3.35 to 8.63). After adjustment, not living with a partner was associated with increased risks of hospital admissions and dying by suicide, but not with the self-report measures of suicidality. Not living with a partner had a stronger relationship with death by suicide for men (HR 2.08, 95% CI 1.36 to 3.18) than for women (HR 1.16, 95% CI 0.59 to 2.31). After adjustment, emotional support was associated with the self-report but not administrative measures of suicidality. For example, when people with the least emotional support were compared to those with the most, the odds ratio was 3.00 (95% CI 1.71 to 5.28). Statistical interactions indicated that people who were lonely despite living with a partner had the highest risk of contemplating self-harm and that loneliness eliminated the protective effects of living with a partner for deaths by suicide.
Conclusion Loneliness was the strongest predictor of suicide risk irrespective of living arrangements and emotional support and loneliness explained the relationships between living arrangements and self-reported suicidality but not suicidality as indicated by administrative records. A limitation is that selective participation may bias the results for self-reported suicidality; however, the large sample size enabled participants to be followed up using death records, which are less vulnerable to bias.