Every other article in this section is about something you do: how you sleep, how you move, what you eat. This one is about something that mostly happens to you, and it has effect sizes comparable to the classic risk factors.
That combination — large effect, limited individual control — is why it gets written about less than it should.
The numbers
The reference point is a meta-analytic review by Julianne Holt-Lunstad and colleagues, published in 2015, drawing on more than 70 independent studies.
After statistically controlling for a range of confounders — age, sex, socioeconomic status, health status, physical activity and smoking — the pooled effect sizes were:
- Social isolation: odds ratio 1.29, a 29% higher likelihood of death
- Loneliness: odds ratio 1.26, a 26% higher likelihood
- Living alone: odds ratio 1.32, a 32% higher likelihood
The authors’ own conclusion was that the influence of both objective and subjective social isolation on mortality risk is comparable with well-established risk factors.
Set that beside numbers from earlier articles in this section. In the cardiorespiratory fitness analysis, smoking carried a hazard ratio of 1.41 and diabetes 1.40. Social isolation at 1.29 is not in a different league from those.
The finding that changes how you read it
Two results inside that meta-analysis matter more than the headline.
Objective and subjective isolation performed almost identically. Social isolation — measured by counting contacts, group memberships, marital status — and loneliness, measured by asking people how they feel, produced effect sizes of 1.29 and 1.26. The authors found no meaningful difference between them.
This is not intuitive. You would expect the objective measure to be more reliable, since it does not depend on self-report or mood. Instead, how connected a person feels predicts mortality about as well as how connected they demonstrably are.
Which means the problem cannot be solved by arithmetic. A person surrounded by people can carry the same risk as someone who sees nobody. Adding contacts to a diary does not necessarily address it.
The effect was stronger in people under 65. Social deficits were more predictive of death in samples with an average age below 65 than above it.
That runs against the assumption that this is an old-age problem. Loneliness in your forties and fifties appears to carry more predictive weight, not less.
What might be doing the damage
Several mechanisms, probably operating together, and none of them fully established.
Behaviour. Isolated people exercise less, eat worse, drink more, adhere less well to medication, and are less likely to seek medical attention early. If you live alone, nobody notices the mole on your back, nobody says you look unwell, and nobody suggests you see someone about the cough.
Stress physiology. Loneliness is associated with altered cortisol patterns, higher inflammatory markers and elevated blood pressure. Chronic activation of the stress response has downstream effects on the cardiovascular and immune systems.
Sleep. Lonely people sleep worse, with more fragmented sleep, which links this article back to everything in the sleep section.
Reverse causation, which cannot be dismissed. Illness causes isolation. People become unwell, stop going out, lose contacts, and then die of the illness that started the sequence. Studies adjust for baseline health, and the meta-analysis found initial health status did influence the findings — an honest acknowledgement that some of this runs the other way.
What the evidence does not support
Two things are worth stating clearly, because this topic attracts overstatement.
The heterogeneity in this literature is very high. Holt-Lunstad’s analysis reported an I² of 97.8%, and other meta-analyses in the field report similar figures. That means the studies disagree with each other substantially, and pooled estimates should be treated as approximate rather than precise. Some individual studies find the association disappears when health indicators and mental health are fully adjusted for.
“Equivalent to 15 cigarettes a day” is a claim to be careful with. It is widely repeated and traces to comparisons in this literature, but the underlying comparisons are between odds ratios drawn from different study designs and populations. The general point — that this sits alongside major risk factors — is supported. The specific cigarette count is a rhetorical device.
What actually helps
The unsatisfying truth is that the intervention literature here is much weaker than the epidemiology. We know the association is real. We know much less about what fixes it.
What has some support:
Shared activity beats arranged socialising. Interventions built around doing something together — a class, a choir, a walking group, volunteering — tend to fare better than those built around meeting to talk. The activity provides a reason to attend and a subject other than the loneliness itself.
Depth over count. Since subjective loneliness predicts as strongly as objective isolation, the number of interactions is not the target. One or two relationships where you can say something true matter more than a wide network of pleasant acquaintances.
Structural contact. Regular, scheduled, low-effort contact outperforms occasional intense contact, because it does not require anyone to initiate. A standing Tuesday phone call, a weekly game, a fixed thing.
Address the mechanism when you cannot address the cause. If isolation is not fixable in the short term — bereavement, geography, caring responsibilities — the downstream mechanisms still are. Movement, sleep and medical follow-up are the routes through which much of the risk is thought to operate.
The transitions worth watching
Loneliness tends to arrive with life events rather than gradually.
Retirement removes the largest source of incidental daily contact most people have, usually without a replacement structure prepared in advance. This is the one that catches people out, because it is anticipated as a relief and experienced as a subtraction.
Bereavement, particularly of a partner, removes both the relationship and the person who organised the social life. In many couples one person maintains the calendar; when that person dies, the calendar stops.
Children leaving, moving house, and the death of the friend who was the group’s organiser all have the same shape.
The common feature is that none of them feels like a health event. Nobody leaves a retirement party thinking about mortality risk. But if the effect sizes above are anywhere near right, these transitions deserve the same deliberate planning that people give to their pension.
This article is for general information and is not medical advice. If you are struggling with loneliness or low mood, speaking to your GP is a reasonable step — this is something they take seriously. If you need someone to talk to now, Samaritans can be reached free on 116 123, at any hour.
Sources
- 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;10(2):227–237. https://pubmed.ncbi.nlm.nih.gov/25910392/
- Full text: https://journals.sagepub.com/doi/full/10.1177/1745691614568352
- Holt-Lunstad J, Smith TB, Layton JB. Social relationships and mortality risk: a meta-analytic review. PLoS Medicine, 2010;7(7):e1000316.
- A meta-analysis of the association between loneliness and all-cause mortality in older adults, including discussion of heterogeneity across the literature. Psychiatry Research, 2023. https://www.sciencedirect.com/science/article/abs/pii/S0165178123003803
- The association between loneliness, social isolation and all-cause mortality in a nationally representative sample of older women and men. Aging & Mental Health, 2021. https://www.tandfonline.com/doi/full/10.1080/13607863.2021.1976723
This article is for general information and is not medical advice. If a health problem is affecting your daily life, speak to your GP.
Comments
We read every comment. Be kind, stay on topic, and please don't ask for medical advice — we can't give it.