Loneliness, Isolation and Earlier Death: What the Studies Can and Can't Say
Published 2026-11-15
People who are socially isolated or feel lonely die earlier on average than others, in study after study. But the two differ: in a meta-analysis of 90 cohort studies, the link with death was larger for social isolation than for loneliness, and in several large studies the loneliness link largely disappeared once health and other circumstances were taken into account.
Headlines often say that loneliness shortens lives. For older people, and for the clubs, sheds, lunch clubs and befriending services that bring them together, the honest version matters more than the dramatic one. Feeling lonely is painful and worth acting on for that reason alone. The research shows that isolation and loneliness, and the poor health, low income and low mood that often travel with them, go along with shorter lives; how much is caused by loneliness itself, it cannot yet fully say. This study sets out what the studies show, how adjustment changes the estimates, and how the WHO’s deaths estimate was built.
In the largest pooled analysis, social isolation had a stronger link with earlier death than loneliness
The largest pooled analysis so far, published in 2023 by Wang and colleagues, combined 90 prospective cohort studies covering 2,205,199 adults aged 18 and over. In the general population, the pooled effect size for death from any cause was 1.32 for social isolation (95% confidence interval 1.26 to 1.39) and 1.14 for loneliness (1.08 to 1.20)[1]. Put plainly, people classed as isolated had about a third higher risk of dying during follow-up than people who were not, and people classed as lonely about a seventh higher, on average across studies. For death from cancer the figures were 1.24 for isolation and 1.09 for loneliness, and isolation was also linked with death from cardiovascular disease (1.34)[1].
These are associations between groups, measured in different ways across studies, not a forecast for any one person. Isolation is an objective state (few contacts, little taking part); loneliness is how people feel about their relationships.
Earlier reviews give different numbers, depending on the studies included. A 2015 meta-analysis by Holt-Lunstad and colleagues, covering studies from 1980 to February 2014, reported odds ratios of 1.29 for social isolation, 1.26 for loneliness and 1.32 for living alone, across studies that controlled for several possible confounders, and found no difference between objective and subjective measures[3]. They are summarised in our study of older people living alone. A 2018 meta-analysis of 35 articles (77,220 participants) that looked only at loneliness reported a pooled hazard ratio of 1.22 (1.10 to 1.35)[4]. A 2022 meta-analysis restricted to older adults found a hazard ratio of 1.10 (1.06 to 1.14) for loneliness, described by its authors as a small effect, with large differences between studies[5].
The WHO Commission on Social Connection summarised recent meta-analyses, among mostly older adults, as linking loneliness with a 9–22% higher risk of death[6]. Holt-Lunstad’s 2015 review also found social deficits more predictive of death in samples with an average age under 65, and that people’s health at the start influenced the results[3].
In studies of older people, the loneliness link is often there, but shrinks once health is counted
United States. In the Health and Retirement Study, 1,604 people over 60 (mean age 71) were asked in 2002 whether they felt left out, felt isolated or lacked companionship. Anyone who answered “some of the time” or “often” to any of the three was classed as lonely, a broad definition that took in 43% of the sample. Over six years, 22.8% of those classed as lonely died, against 14.2% of the rest. After adjusting for demographic factors, socioeconomic status, living situation, depression and a range of medical conditions, the hazard ratio was 1.45 (1.11 to 1.88)[7].
England. The English Longitudinal Study of Ageing followed 6,500 people aged 52 and over from 2004–2005, for a mean of 7.25 years. After adjusting for demographic factors and baseline health, the most isolated fifth still had a hazard ratio of 1.26 (1.08 to 1.48), while the hazard ratio for loneliness was 0.92 (0.78 to 1.09), no clear difference. The authors concluded that the effect of loneliness “was not independent of demographic characteristics or health problems”[8].
China. The Chinese Longitudinal Healthy Longevity Survey gives a similar picture. Among 35,254 people aged 60 and over (mean age 86.63), followed between 1998 and 2018, social isolation was linked with higher mortality (adjusted hazard ratio 1.22, 1.18 to 1.25). The loneliness link was not significant after adjustment for health indicators and low psychological wellbeing (1.01, 0.98 to 1.04), although among participants under 80 it was (1.15, 1.05 to 1.26)[9]. A later analysis of 8,956 people aged 65 and over in the same survey looked at change over a three-year interval: only persistent loneliness was linked with higher mortality (1.27, 1.17 to 1.38), while both transient (1.28) and persistent (1.39) isolation were[10].
Two patterns run through these studies: isolation keeps its link with death after adjustment more often than loneliness does, and in the Chinese study that measured change, loneliness that lasted was linked with mortality while loneliness that passed was not.
What researchers adjust for changes the answer, especially for loneliness
The clearest demonstration comes from UK Biobank, which followed 466,901 adults (mean age 56.5 at the start) for a mean of 6.5 years. Isolation was measured with a three-item scale and loneliness with two questions. Adjusting only for age, sex, ethnic origin and chronic disease, the hazard ratio for death was 1.73 for social isolation and 1.38 for loneliness. After further adjustment for socioeconomic factors, health behaviours, depressive symptoms, biological factors such as blood pressure, cognitive performance and self-rated health, it fell to 1.26 for isolation and 0.99 for loneliness[2]. In other words, the link between loneliness and death was almost entirely accounted for by differences that were measured at the start, such as disadvantage, poorer health and depressive symptoms.
That does not mean loneliness is harmless. Some of those factors may sit on the path between loneliness and death (loneliness may lead to low mood or giving up activity), so adjusting for them can hide part of a real effect. But the 1.38 cannot be read as the effect of loneliness alone, and the authors pointed to adverse socioeconomic conditions, unhealthy lifestyles and lower mental wellbeing as risks that policy could address[2].
Timing matters too. A UK Biobank analysis of 18,258 people aged 38 to 73 found that persistent loneliness (1.28, 1.02 to 1.61) and persistent isolation (1.42, 1.12 to 1.81) were linked with higher all-cause mortality over a median of 8.3 years[11]. Our study of loneliness, isolation and the heart looks at heart disease and stroke in detail.
Illness can cause loneliness as well as follow it, and genetic studies disagree with the cohorts
Two problems make it hard to say how much loneliness itself shortens lives. One is confounding: low income, disability, long-term illness and depression make people more likely both to feel lonely and to die sooner, and studies can only adjust for what they measured. The other is reverse causation: serious illness can lead to loneliness, as people who are unwell go out less and lose contact, and a study that measures loneliness once cannot fully separate the two directions.
Mendelian randomization uses genetic differences linked to a trait to test whether it plausibly causes an outcome. A 2024 UK Biobank study found that, during a median 12.2 years of follow-up, loneliness was associated with 30 of 56 diseases. But for 20 of the 26 diseases that could be tested genetically, including cardiovascular diseases, the genetic analysis did not support a causal link. Socioeconomic factors, health behaviours, depressive symptoms and other illnesses “largely explained” the associations, and the authors suggested loneliness “may serve as a potential surrogate marker rather than a causal risk factor for most diseases tested here”[12]. Genetic methods have their own assumptions, and this study looked at diseases, not death.
The WHO Commission takes a firmer view. Its 2025 report says the links between social disconnection and mortality “are probably causal”, because the evidence on social relationships and health meets the Bradford Hill criteria for weighing causation, while noting the genetic study’s finding that only six diseases appeared potentially causally linked to loneliness[6]. Both readings can be held at once: social connection as a whole very likely matters for health, and the specific contribution of feeling lonely, separate from isolation, illness and disadvantage, is less certain. Loneliness and low mood are closely tied; our study of loneliness and depression in later life covers that link.
The WHO’s estimate of deaths linked to loneliness is a model built on one hazard ratio
The WHO Commission on Social Connection estimated that about 871,000 deaths a year worldwide were linked to loneliness in 2014–2019. This is not a count of death certificates but a model, set out in the report’s second annex[6]:
- For each WHO region and four age groups (13–17, 18–29, 30–59 and 60 and over), the Commission took its estimate of the share of people who are lonely.
- It combined each share with a single hazard ratio, 1.14, from the Wang meta-analysis (the report says it came from 45 of the studies), using Levin’s formula for a population attributable fraction: the share of deaths that would not happen if the link were entirely causal and loneliness were removed.
- It applied those fractions to WHO’s Global Health Estimates of deaths for 2014–2019, and added up the results. Children under 13 were left out for lack of data.
The Commission lists the limits itself: the same hazard ratio was applied to every age group and region, although experiences of loneliness differ between age and cultural groups; the death figures run to 2019 while the loneliness estimates run to 2023; and “data on this topic are currently lacking”[6]. There is no estimate yet for social isolation, the measure with the larger association.
The model also inherits the problems above: if part of the 1.14 reflects illness or disadvantage rather than loneliness, the number of deaths caused by loneliness would be lower.
In the model, people aged 60 and over had the lowest estimated rate of loneliness of any age group, 11.8%, yet they account for about two-thirds of the modelled deaths (our sum of its regional figures), as most deaths come in later life. Their estimated loneliness ranged from about 8% in the WHO Region of the Americas and European Region to about 23% in the African Region[6].
A different, older comparison is often attached to these numbers: that weak social ties are as harmful as smoking 15 cigarettes a day. It traces back to a 2010 meta-analysis of 148 studies by Holt-Lunstad and colleagues, which pooled measures of the quality and quantity of social relationships, such as social integration and living arrangements, rather than loneliness alone, and reported a 50% increased likelihood of survival for people with stronger relationships[13]. It compared effect sizes; it did not test loneliness or establish an equivalence with smoking. Our study of common claims about loneliness traces that comparison in full.
What the evidence asks of us
The mortality evidence is strong enough to act on and uncertain enough to be described carefully. It points to a few priorities for governments, health services, funders and older people’s organisations.
- Treat isolation and loneliness as separate targets. The cohort evidence links social isolation with earlier death more consistently than loneliness[1][8]. Asking only “do you feel lonely?” misses people who see almost no one but do not call themselves lonely.
- Look at the circumstances around loneliness. Much of the loneliness link runs alongside disadvantage, poor health and depression[2][12]. Income, transport, accessible venues and support for long-term illness are part of the same picture.
- Pay attention to loneliness and isolation that last. Persistent loneliness and isolation were linked with higher mortality in Chinese and UK cohorts that measured change, and in the Chinese cohort passing loneliness was not[10][11].
- Use the WHO estimate as a reason to act, not as a precise toll. The 871,000 figure is a modelled estimate with stated limits[6]. It is not a count of deaths.
- Fund better data, especially outside Europe and North America. The WHO Commission’s five strategic areas include research and “measurement and data”, among them strengthening national monitoring and collecting more and better data globally[6].
Loneliness is worth reducing because of what it feels like, whatever it does to life expectancy. For more on what reduces loneliness and how strong that evidence is, see the Loneliness in Later Life research library.
Language: this piece says "older people" and treats loneliness as a common human experience that anyone can have, not a personal failing or an illness. It keeps apart three things that are often mixed up: loneliness (how people feel about their relationships), social isolation (how few social contacts they have) and living alone (a household arrangement). Where a study uses another term, such as "seniors" or "older adults", we keep it when quoting that study.
Why we're publishing this
Claims about loneliness and death often go further than the research. Older people’s charities, befriending services and professional bodies need figures they can cite without overstating them, and older people deserve not to be told, wrongly, that feeling lonely is a fatal condition. We have set out the main studies and their limits in one place. The case for social connection is theirs; our job is to show the evidence clearly.
How to read this data
Almost all the evidence here comes from observational cohort studies: people’s social lives were measured once (or occasionally twice) and deaths were counted over the following years. They show associations, not causes. A hazard ratio compares death rates over time between groups; an odds ratio compares odds, which is similar only when an outcome is uncommon; the Wang review’s abstract gives pooled “effect sizes” without naming the measure. A ratio of 1.14 means a 14% higher average risk in one group than another during the study, not a 14% chance of anything for any individual.
Loneliness was measured in different ways: three questions about feeling left out, isolated or lacking companionship in the Health and Retirement Study (anyone answering “some of the time” to any counted as lonely), two questions in UK Biobank, and short questionnaires or single questions elsewhere. Isolation was measured by contact with family and friends, taking part in organisations, or living alone. Each study adjusted for different factors. Most studies come from Europe, North America and East Asia; we found far fewer from Africa, South Asia or Latin America. Household surveys usually leave out care-home residents. The WHO deaths figure is a model, and its regional numbers reflect deaths in each region as well as estimated loneliness.
Use this data
Free to cite with attribution to Understanding Aging. u3a groups, men's sheds, senior centres, lunch clubs, befriending services, charities, campaigners and journalists are welcome to reuse these figures and charts. For the underlying figures as a spreadsheet, get in touch.
Important: This article summarises published research and statistics for general information. It is not medical or care advice and cannot be used to diagnose, treat or predict anything about an individual. If loneliness, low mood or isolation is affecting your health or someone else’s, a doctor or another health professional can help, and local services for older people can often put you in touch with groups and support nearby.
Sources
- Wang F, Gao Y, Han Z, et al. A systematic review and meta-analysis of 90 cohort studies of social isolation, loneliness and mortality. Nature Human Behaviour, 2023. europepmc.org/article/MED/37337095
- Elovainio M, Hakulinen C, Pulkki-Råback L, et al. Contribution of risk factors to excess mortality in isolated and lonely individuals: an analysis of data from the UK Biobank cohort study. The Lancet Public Health, 2017. europepmc.org/article/MED/28626828
- Holt-Lunstad J, Smith TB, Baker M, Harris T, Stephenson D. Loneliness and social isolation as risk factors for mortality: a meta-analytic review. Perspectives on Psychological Science, 2015. europepmc.org/article/MED/25910392
- Rico-Uribe LA, Caballero FF, Martín-María N, et al. Association of loneliness with all-cause mortality: a meta-analysis. PLoS One, 2018. europepmc.org/article/MED/29300743
- Schutter N, Holwerda TJ, Comijs HC, et al. Loneliness, social network size and mortality in older adults: a meta-analysis. European Journal of Ageing, 2022. europepmc.org/article/MED/36467548
- WHO Commission on Social Connection. From loneliness to social connection: charting a path to healthier societies. Report of the WHO Commission on Social Connection. World Health Organization, 2025. who.int/publications/i/item/978240112360
- Perissinotto CM, Stijacic Cenzer I, Covinsky KE. Loneliness in older persons: a predictor of functional decline and death. Archives of Internal Medicine, 2012. europepmc.org/article/MED/22710744
- Steptoe A, Shankar A, Demakakos P, Wardle J. Social isolation, loneliness, and all-cause mortality in older men and women. Proceedings of the National Academy of Sciences of the USA, 2013. europepmc.org/article/MED/23530191
- Yu B, Steptoe A, Chen Y. Social isolation, loneliness, and all-cause mortality: a cohort study of 35,254 Chinese older adults. Journal of the American Geriatrics Society, 2022. europepmc.org/article/MED/35229887
- Huang QM, Wei ST, Li ZH, et al. Associations of changes in social isolation and loneliness with all-cause mortality among older adults: a prospective cohort study. GeroScience, 2026. europepmc.org/article/MED/41160315
- Chen Y, Xue H, Nie Y, et al. Evaluation of changes in social isolation and loneliness with incident cardiovascular events and mortality. Journal of Epidemiology and Global Health, 2024. europepmc.org/article/MED/38801491
- Liang YY, Zhou M, He Y, et al. Observational and genetic evidence disagree on the association between loneliness and risk of multiple diseases. Nature Human Behaviour, 2024. europepmc.org/article/MED/39284978
- Holt-Lunstad J, Smith TB, Layton JB. Social relationships and mortality risk: a meta-analytic review. PLoS Medicine, 2010. europepmc.org/article/MED/20668659