Loneliness drives depression and poor health among older European adults, study finds

Loneliness may be one of the most powerful and overlooked forces shaping the health of older Europeans, and its effects cut across social class lines in ways that could reshape how governments fight health inequality. That is the central finding of a new study drawing on data from nearly 13,000 older adults across Europe, which shows that loneliness explains a substantial share of the differences in mental and physical health among people who share the same socio-economic circumstances, and that eliminating loneliness inequality could dramatically shrink the health gap between rich and poor.

The research, published in the open-access journal Heliyon by Daniel Howdon, Jochen O. Mierau, and Laura Viluma, takes a deliberately different angle from most studies of health inequality. Rather than asking why people of low socio-economic status are, on average, less healthy than those of high status, the team examined why health varies so much within each socio-economic group. This distinction matters because the variation within groups is considerable, and previous work has shown it is greatest within the lower socio-economic strata. In other words, even among people facing similar financial hardship, some remain remarkably healthy while others suffer from depression, chronic disease, and poor self-rated health. Understanding what protects the resilient ones, the authors argue, could offer policy targets that do not require the politically difficult task of redistributing income itself.

To investigate this, the researchers turned to the Survey of Health, Ageing and Retirement in Europe, known as SHARE, a longitudinal survey conducted every two years that tracks socio-economic indicators, lifestyle, and health in people aged 50 and over, along with their spouses. The team combined retrospective data from the third wave, called SHARELIFE, with contemporaneous data from the fifth wave, and after removing cases with missing information, their final analytical sample comprised 12,918 individuals. Because household income is frequently missing in SHARE and is provided as a series of imputed values, the regressions were carried out using multiple imputation methods, a statistical approach that accounts for the uncertainty introduced by missing data.

The study focused on three health outcomes measured at the same time as the survey: depressive symptoms, chronic conditions, and self-assessed health. Depression was operationalised through the EURO-D questionnaire, a validated instrument for measuring depressive symptoms in older Europeans, with individuals classified as depressed if they reported four or more symptoms, a cut-off that has been validated in prior literature. Chronic conditions were captured by a question asking whether respondents suffered from any long-term illness, disability, or infirmity. Self-assessed health was measured on a five-point scale from excellent to poor, and the researchers dichotomised it by flagging those who rated their own health as poor, following established practice in the health-inequality literature.

Socio-economic status was represented primarily by household income quintiles, with additional analyses using education levels grouped into three categories under the ISCED-97 classification. Loneliness, the study’s key explanatory variable, was defined by self-reported frequency: respondents who said they felt lonely “some of the time” or “often” were classified as lonely, distinguishing the subjective experience of loneliness from objective social isolation. This conceptual distinction is important. Social isolation, characterised by living alone or having few social ties, can be quantified objectively, whereas loneliness is the gap between the social contact a person actually has and the contact they desire. The two can diverge sharply, and the authors chose to focus on the subjective measure because it is the one most consistently linked to health harm.

The descriptive findings first confirmed what decades of research have established: a clear socio-economic gradient runs through every health measure. Average self-assessed health on the five-point scale was 3.36 in the lowest income quintile, where higher values indicate worse health, compared with 2.71 in the highest quintile. Depression prevalence fell steadily from 31 percent in the poorest group to 16 percent in the richest, and chronic conditions affected 58 percent of the lowest quintile against 43 percent of the highest. Loneliness followed the same gradient: 86 percent of the highest earners said they never felt lonely, against only 70 percent of the lowest quintile, while frequent loneliness was three times more common among the poorest. Crucially, the variability of depressive symptoms was also greater in lower income groups, with kernel density plots showing much fatter right tails in the distributions for poorer quintiles, meaning that low-income groups contain both many people with few symptoms and a substantial minority with severe symptom burden.

The core of the analysis used logistic regression to model the odds of each adverse health outcome, first with income and control variables for age, gender, and country of residence, and then adding loneliness. The results were striking. Even after controlling for socio-economic status, loneliness remained strongly associated with depression, and a dose-response pattern emerged: feeling lonely “often” raised the odds of clinically significant depressive symptoms more than feeling lonely only “some of the time.” Including loneliness in the models reduced the income coefficients, indicating that loneliness accounts for part of the association between poverty and depression, but its larger role was in explaining differences among people within the same income bracket. The pattern repeated for the other outcomes. Individuals who reported often feeling lonely had roughly four times higher odds of rating their own health as poor compared with those who never felt lonely, and similar associations held when education rather than income was used as the socio-economic indicator.

The authors also tested whether the loneliness–health relationship differed across income groups by estimating saturated models for each quintile. For depression, the point estimates rose slightly with deprivation, but overlapping 95 percent confidence intervals meant the differences could not be distinguished statistically, suggesting the harm of loneliness is broadly similar across the income spectrum. For chronic disease there was a hint that income partially buffers the impact of loneliness, with the association weaker among the wealthy, though again the imprecision of the estimates prevented firm conclusions. This suggests an intriguing asymmetry: money may compensate somewhat for loneliness in objective physical health, but not in mental health or self-assessed health, where the injury of feeling alone appears to operate independently of material means.

Perhaps the most eye-catching result came from a counterfactual analysis. The researchers modelled what would happen if the distribution of loneliness in the four lowest income quintiles were changed to match that of the highest quintile, where 86.4 percent of people report hardly ever or never feeling lonely, 10.77 percent some of the time, and just 2.83 percent often. Holding all other characteristics and coefficient estimates constant, and averaging over 1,000 probability draws to span possible within-group allocations, they found that the predicted probability of depression in the lowest income quintile would fall from about 31 percent to about 25 percent, a drop of roughly six percentage points, or a 19 percent reduction in the probability of reporting four or more depressive symptoms. Comparable reductions appeared for poor self-assessed health and chronic conditions. The authors stress that this exercise assumes a causal link that the observational data cannot prove, but it offers a provocative benchmark: if loneliness inequality could be eliminated, the socio-economic gap in depression would narrow substantially without anyone’s income changing by a cent.

The findings carry significant policy weight. A 2023 advisory from the US Surgeon General declared loneliness and isolation an epidemic, proposing six policy pillars to advance social connection, and the World Health Organization has named social isolation and loneliness one of the four main action areas of its 2020 to 2030 strategy. Previous research has estimated that loneliness poses a mortality risk comparable to smoking 15 cigarettes a day, and meta-analyses have tied it to depression, cardiovascular disease, stroke, and risky behaviours such as smoking and inactivity. Earlier work by one of the current study’s co-authors and colleagues found that loneliness explained between 21 and 51 percent of health inequalities across three separate metrics. The new study extends this literature by showing that loneliness’s greatest explanatory power lies within socio-economic groups rather than between them, and by including mental health outcomes, which have received far less attention than physical health in inequality research.

The mechanism, the authors suggest, fits within the science of resilience. Some individuals exposed to socio-economic adversity remain healthier than their circumstances would predict, and psychological and social resources such as strong social relations appear to buffer the effects of deprivation. Prior studies have shown that social relations moderate the effect of neighbourhood deprivation on mental health-related quality of life, and that psychosocial resources are stronger predictors of health in low socio-economic groups, precisely where they are most needed. Loneliness, as the subjective perception of deficient social connection, sits at the centre of this protective architecture.

The researchers are careful about limitations. SHARE data are self-reported and subject to recall and survey bias, the set of control variables was limited, and the counterfactual analysis rests on a strong causal assumption that the study design cannot verify. They call for validation with other datasets, a broader set of explanatory variables potentially identified through machine learning tools, larger samples for greater statistical precision, and corroboration in age groups beyond older adults, noting that while older people report loneliness more often, the mortality risk associated with it may actually be lower in this group than among the young.

Even so, the policy implication is hard to ignore. The extent to which health inequality between socio-economic groups can be reduced is often considered limited because its root causes seem difficult to modify. But loneliness within groups may be more tractable, and the evidence suggests that interventions targeting loneliness among low-income older adults could deliver large health gains and, as a side effect, reduce overall socio-economic health inequality. For a continent whose population is ageing rapidly, teaching policymakers to see loneliness not merely as an emotional hardship but as a modifiable determinant of health could prove one of the most consequential shifts in public health thinking of the decade.