Strong social ties are one of the things most consistently linked to a long, healthy life. Across 148 studies and more than 300,000 people, those with stronger relationships were about 50% more likely to be alive when follow-up ended. Poor social relationships also track with higher rates of heart disease and dementia. That evidence is observational: it follows people over years and never assigns anyone more company. Part of the raw risk is that sick people withdraw before the illness kills them.
Isolation and loneliness are measured separately and behave differently: isolation is being alone, loneliness is feeling alone. Both matter. Loneliness responds to help. The approaches that address the thoughts behind it work better than simply arranging more contact, and volunteering and shared purpose track with living longer too.
Findings & Outcomes
Being Alone and Feeling Alone
Isolation is scored as a count, who you live with, how often you see people, whether you belong to anything. Loneliness is the gap between the relationships you have and the ones you want. A person can live alone, see friends weekly, and feel content. Another can be surrounded by family and still feel unreachable.
Loneliness is usually the result of a circumstance, and most of those circumstances sit outside anyone's control:
- grief, and the loss of the routine that went with a person
- caring for someone full time
- chronic illness and pain
- hearing loss that makes group conversation exhausting
- a move to a new city, or shift work
- a low income that closes off the ordinary ways adults keep friendships going
- the years of caring for young children
What the Evidence Shows
Later reviews put numbers on the risk. Poor social relationships track with roughly a quarter to a third higher risk of death. The same work finds about 29% more coronary heart disease and 32% more stroke, plus more later dementia. All of it comes from watching large groups over years, so it shows a pattern but cannot prove a cause.
The reason for caution is that illness makes people withdraw before it shortens a life. Depression itself dampens interest in other people, so low mood can cut someone off before it does any other harm. Dementia changes the brain for years before it is diagnosed, and those early changes make company harder.
A study that measures relationships once and counts deaths later can mistake this for isolation causing death, when the illness came first.
The largest cohort shows how much of the apparent risk is really underlying illness. In nearly 467,000 UK adults, the link from isolation to death fell from a hazard ratio of 1.73 to 1.26 after full adjustment and stayed there. The felt experience of loneliness fell to a null once the same adjustments were made.
Dementia divides the same way. Isolation kept predicting risk after heavy adjustment. About three-quarters of the loneliness link ran through depressive symptoms measured alongside it.
The Research & Studies
Everything here is based on the research we have collected and checked, sorted into groups and ordered with the strongest evidence first. Click any claim to open the studies behind it.
Social Connection
Stronger relationships tracked with 50% higher odds of survival across 148 studies
Across 148 studies and more than 300,000 people, those with stronger relationships were about 50% more likely to be alive when follow-up ended. It is a large association and it is observational: the studies followed people, they did not give anyone more friends, and illness itself pushes people into isolation. The signal is bigger when relationships are measured in several dimensions at once than when it comes down to whether someone lives alone.
Odds ratio 1.50 (95% CI 1.42 to 1.59); mean age 64, mean follow-up 7.5 years. Multidimensional measures of social integration gave 1.91, while the crude single question of living alone gave 1.19 with an interval crossing 1. Odds of survival is not the same quantity as a reduction in the risk of death, and the two are swapped constantly in retellings. Illness, depression and poverty all shrink a network before they shorten a life, so which way the causation runs is unresolved.
Who this may not transfer to:Roughly balanced by sex, and the effect was independent of it.
The measures that carried the strongest signal counted participation and belonging over time, which is why the practical unit is a repeating slot in the week, not a raw headcount.
The study · 1
Holt-Lunstad et al., social relationships and mortality risk, a meta-analytic review · PLoS Med 2010;7(7):e1000316
Isolation, loneliness and living alone each tracked with 26 to 32% higher odds of death
Pooling studies from 1980 to 2014, people who were socially isolated, lonely, or living alone were roughly a quarter to a third more likely to die over follow-up. This review found no difference between the objective measures and the felt ones, and the effect was actually stronger below age 65 than above it.
Weighted average effects with confounders controlled: social isolation OR 1.29, loneliness OR 1.26, living alone OR 1.32, corresponding to 29%, 26% and 32% higher likelihood of death. Objective and subjective measures did not differ. Results were stronger in samples with a mean age under 65 than in older ones, and baseline health status influenced the size of the association, which is the reverse-causation problem surfacing inside the meta-analysis itself.
Who this may not transfer to:Gender was tested as a moderator and the results were consistent across it, as they were across follow-up length and world region.
The study · 1
Holt-Lunstad et al., loneliness and social isolation as risk factors for mortality · Perspect Psychol Sci 2015;10(2):227-237
After full adjustment the isolation-death link held at 1.26, while loneliness fell to a null
In nearly 467,000 UK adults, being socially isolated still predicted earlier death after accounting for health and circumstance, while the feeling of loneliness did not once the same adjustments were made.
Social isolation hazard ratio 1.73 (95% CI 1.65 to 1.82) adjusted only for age, sex, ethnicity and chronic disease, falling to 1.26 (1.20 to 1.33) after full adjustment. Loneliness ran 1.38 (1.30 to 1.47) minimally adjusted and 0.99 (0.93 to 1.06) fully adjusted, a measured null. The full adjustment included depressive symptoms, which is plausibly a step on the path from loneliness to death, not a confounder, so the loneliness null is a ceiling on the direct effect, not a measure of the whole one.
Who this may not transfer to:Men and women, with sex adjusted for in every model. UK Biobank volunteers are healthier and less deprived than the UK population, which tends to shrink observed associations.
The structural side of a social life, who you actually see and belong to, is the more robust thing to work on, since it is what survives adjustment.
The study · 1
Elovainio et al., contribution of risk factors to excess mortality in isolated and lonely individuals, UK Biobank · Lancet Public Health 2017;2(6):e260-e266
The most isolated older adults were about 26% more likely to die, and loneliness added nothing to the model
In 6,500 older English adults, the most isolated fifth were about 26% more likely to die, and adding loneliness to the model did not change that, so loneliness was not the channel carrying isolation's effect.
After adjustment for demographics and baseline health, the most isolated fifth had a hazard ratio for death of 1.26 (95% CI 1.08 to 1.48). Loneliness in the same model gave 0.92 (0.78 to 1.09) and left the isolation estimate unchanged when added. Isolation was measured once at baseline from contact with family and friends plus civic participation, and a single measurement cannot capture a social life that shifts with health over seven years.
Who this may not transfer to:Men and women aged 52 and over. The published hazard ratios are pooled across sex, not stratified.
The study · 1
Steptoe et al., social isolation, loneliness, and all-cause mortality in older men and women · Proc Natl Acad Sci USA 2013;110(15):5797-5801
Losing a spouse raised the risk of death about 23%, more for men than for women
Losing a spouse raised the risk of death about 23% on average, and the effect was larger for men than for women, with the gap shrinking as people got older.
Mean hazard ratio 1.23 (95% CI 1.19 to 1.28) among high-quality adjusted estimates, comparing widowed with married people; 1.27 (1.19 to 1.35) for men against 1.15 (1.08 to 1.22) for women. Drawn from 1,381 estimates across 124 publications and more than 500 million person-records. Magnitude also varied with sample size, region and level of adjustment, and a spouse's illness often precedes their death by years of caregiving, so the exposure period is not clean.
Who this may not transfer to:Estimates are reported separately for men and women, which is unusual in this literature and is why this row carries the sex finding for the page.
Bereavement is the one place where the effect splits clearly by sex, with men after losing a spouse the higher-risk group, so this is where extra contact and routine are worth deliberately rebuilding.
The study · 1
Shor et al., widowhood and mortality, a meta-analysis and meta-regression · Demography 2012;49(2):575-606
Talking therapy produced a small-to-medium drop in loneliness, g 0.43, across 28 trials
Talking therapies, most often group cognitive behavioral therapy delivered weekly, produced a small-to-medium drop in loneliness across 28 trials.
Loneliness fell against control groups with a pooled Hedges' g of 0.43 across 28 randomized trials (N = 3,039), most of them face to face, group based and weekly. Heterogeneity between trials was considerable and study quality was mixed on the Cochrane risk-of-bias tool. Type of intervention only approached significance as a moderator, so the pooled figure does not pin down which therapy does the work.
Who this may not transfer to:Participants came from a diverse range of cultures, age groups and populations; no sex-stratified pooled effect is reported.
These therapies treat the expectation of rejection and the effort that company takes, the same target as the social-cognition approach that performed best in the trial comparisons, which is why they are the route to reach for when contact itself feels unsafe, not just tiring.
The study · 1
Hickin et al., the effectiveness of psychological interventions for loneliness · Clin Psychol Rev 2021;88:102066
Targeting loneliness-driving thoughts cut it most, about 0.6 SD, while adding contact barely moved it
When loneliness is treated directly, the approaches that work best are the ones that change how a person reads social situations, not the ones that simply put more people in the room. Across the strongest trials the average change was small, and 14 of 20 showed none at all.
Across 20 randomized comparison studies the mean reduction in loneliness was 0.198 standard deviations (95% CI 0.08 to 0.32). Within that group, addressing maladaptive social cognition reduced loneliness by 0.598, about 0.6, (4 studies), enhancing social support by 0.162 (12 studies), improving social skills by 0.017 (2 studies, p = 0.90) and increasing opportunities for contact by 0.062 (2 studies, p = 0.67). This is a moderator analysis, not a head-to-head trial, it reached significance at p = 0.05 exactly, the winning cell holds four studies, and effect sizes shrank steadily as designs got stronger.
Who this may not transfer to:The trial samples were mostly female, and the analysis reports no reliable sex-stratified effect. An earlier reading that women showed smaller loneliness reductions could not be confirmed in the paper and one passage indicates gender composition did not moderate the result, so no sex difference is claimed here.
If contact feels frightening, not merely effortful, that appraisal is the target with the best evidence, and a therapist working in cognitive behavioral terms is the version of that with trials behind it. Putting a lonely person in a room with others is the approach that has failed most reliably.
The study · 1
Masi et al., a meta-analysis of interventions to reduce loneliness · Pers Soc Psychol Rev 2011;15(3):219-266
Video calls made little to no measurable difference to loneliness in three nursing-home trials
Video calls made little to no measurable difference to loneliness in the three small nursing-home trials that tested them. For someone whose only alternative is no contact at all, though, a call is still contact.
Mean difference 0.44 points on the 20 to 80 point UCLA Loneliness Scale at three months (95% CI 3.28 better to 2.41 worse), 0.34 at six months and 2.40 at twelve, against usual care across 3 cluster quasi-randomized trials totaling 201 nursing-home residents. Certainty was downgraded three levels for study limitations, imprecision and indirectness, so this is closer to an absence of usable evidence than a demonstrated null. No included study reported social isolation as an outcome, and the population does not resemble most readers.
Who this may not transfer to:All participants were nursing-home residents with a mean age of at least 65. The review does not report a sex breakdown.
The useful distinction is between contact at the same time and contact that is not: a call or video call involves turn-taking and repair, so it works best as a supplement to seeing people, not a full replacement for it, especially when in-person contact is out of reach.
The study · 1
Noone et al., video calls for reducing social isolation and loneliness in older people, Cochrane rapid review · Cochrane Database Syst Rev 2020;5:CD013632
Heart And Vascular
Poor social relationships tracked with 29% more heart disease and 32% more stroke
Poor social relationships were tied to about a 29% higher risk of heart disease and a 32% higher risk of stroke across 16 long-term datasets.
Pooled relative risk 1.29 (95% CI 1.04 to 1.59) for incident coronary heart disease and 1.32 (1.04 to 1.68) for stroke, over 4,628 coronary events and 3,002 strokes with follow-ups of 3 to 21 years. Both lower bounds sit near 1.04, so the pooled estimates are consistent with much smaller effects than those percentages suggest. The exposure mixed loneliness and isolation measures, and subclinical heart disease reduces activity and participation years before an event is recorded.
Who this may not transfer to:Subgroup analyzes did not identify any differences by gender. All included studies were set in high-income countries.
The study · 1
Valtorta et al., loneliness and social isolation as risk factors for coronary heart disease and stroke · Heart 2016;102(13):1009-1016
Cognition
Socially isolated people had about 26% higher dementia risk after heavy adjustment, loneliness none
In UK Biobank, socially isolated people had about a 26% higher risk of later dementia even after heavy adjustment, while loneliness did not predict it once depression was accounted for. Three-quarters of the loneliness-to-dementia link ran through depressive symptoms.
Social isolation hazard ratio 1.26 (95% CI 1.15 to 1.37) after adjustment for demographic, socioeconomic, biological, cognitive, behavioral and psychological factors including loneliness and depression; loneliness gave 1.04 (0.94 to 1.16), with 75% of its association attributable to depressive symptoms. Mean follow-up 11.7 years across 462,619 people. The dementia prodrome runs longer than most cohorts follow, so this length of follow-up does not remove reverse causation, and the supporting brain-imaging arm is cross-sectional.
Who this may not transfer to:Men and women, with sex adjusted for in every model and no sex-stratified estimate published. Same cohort as the mortality adjustment row, so the two are not independent replications of each other.
The study · 1
Shen et al., associations of social isolation and loneliness with later dementia · Neurology 2022;99(2):e164-e175
A poor social network tracked with 59% higher dementia risk across 2.3 million people
Pooling 33 cohorts and 2.3 million people, a poor social network raised dementia risk about 59% and weak social support about 28%, and over follow-ups of ten years or more good social engagement was modestly protective. Loneliness on its own did not reach significance.
Poor social network RR 1.59 (95% CI 1.31 to 1.96), poor social support RR 1.28 (1.01 to 1.62), and in studies following people 10 years or more, good social engagement RR 0.88 (0.80 to 0.96). Loneliness gave RR 1.38 with an interval of 0.98 to 1.94 that crosses 1, so a real loneliness effect is not excluded, only unconfirmed at this power. The social support pooling showed substantial heterogeneity (I-squared 55.5%), meaning the studies behind it disagree.
Who this may not transfer to:The review does not report a sex breakdown across its 33 included studies, and no sex-stratified pooled estimate is given.
The study · 1
Penninkilampi et al., the association between social engagement, loneliness, and risk of dementia · J Alzheimers Dis 2018;66(4):1619-1633
Longevity And Mortality
People who volunteer were about 22% less likely to die over follow-up
People who volunteer were about 22% less likely to die over follow-up across five pooled cohorts, and reported better mood and life satisfaction. The catch is that being well enough to volunteer is itself a sign of health, and trials did not confirm the mood benefits the cohorts showed.
Pooled relative risk of death 0.78 (95% CI 0.66 to 0.90) across five cohort studies. Cohorts also showed favorable effects on depression, life satisfaction and wellbeing but not on physical health, and the experimental studies did not confirm those wellbeing benefits. Five cohorts is a thin base for a mortality estimate, and healthy-volunteer selection is at its most literal here: being mobile, well and organized enough to commit to regular unpaid work is already a marker of health.
Who this may not transfer to:No sex-stratified mortality estimate is reported, and the review does not give an overall sex breakdown of the pooled cohorts.
Giving and shared activity is the route with a mortality cohort behind it, so where you have the capacity, doing something for others is a reasonable thing to lean toward, read as an association, not a guarantee.
The study · 1
Jenkinson et al., is volunteering a public health intervention? Systematic review and meta-analysis · BMC Public Health 2013;13:773
Mood & stress
Over five years, loneliness came before worsening low mood, not the other way around
Over five years, feeling lonely came before a worsening of low mood, while low mood did not come before loneliness. It is some of the clearest evidence that at least part of the harm runs from loneliness to illness, not the other way around.
In cross-lagged panel models across 229 adults aged 50 to 68 in Chicago, loneliness predicted subsequent change in depressive symptoms while depressive symptoms did not predict subsequent change in loneliness. The ordering held after accounting for gender, ethnicity, education, physical functioning, medications, social network size, neuroticism, stressful life events, perceived stress and social support. Cross-lagged models establish temporal precedence rather than causation, the loneliness and depression scales share item content, and an unmeasured factor changing before both could produce the same ordering.
Who this may not transfer to:Men and women. Gender was tested as a possible influence on the cross-lagged association and did not account for it.
The study · 1
Cacioppo, Hawkley and Thisted, perceived social isolation makes me sad, 5-year cross-lagged analyzes · Psychol Aging 2010;25(2):453-463
How It Works
There are three likely mechanisms.
Anatomy of the Practice
1The stress response left switched on
Feeling unsafe among others, or having no one to share a load with, keeps the body low-level braced. Over years that shows up as higher blood pressure, higher stress hormones, and higher markers of inflammation in isolated and lonely people. This is association, not proof: the same illnesses that raise these markers also shrink their social circles. It is still the most likely way isolation acts on the body.
2Health behaviors and a safety net
People who see others regularly tend to eat, move, sleep, and take medicines more reliably. Someone is also there to notice when they are unwell and get them to care sooner. Much of the survival difference may run through these ordinary mechanics rather than the feeling itself, one reason the effect shrinks once behavior and circumstance are accounted for.
3Isolation is often the result, not the cause
Illness, depression, low income, and the long prodrome of dementia all pull people into isolation years before they cause harm. That confounding weakens every estimate. The cleaner longitudinal studies show loneliness precedes worsening mood. That sequence is the strongest evidence available.
What Helps
When loneliness is treated head-on, the approaches split sharply. Across the strongest trials the average change was small, and most individual studies showed none.
What works depends on what the approach targets. The interventions that helped targeted the thoughts behind loneliness, the expectation of rejection, the habit of reading company as unsafe. Ones that only increased contact did several times worse. Putting a lonely person in a room with others is the approach that has failed most reliably.
The best-tested version is talking therapy, most often group cognitive behavioral therapy delivered weekly, which produced a small-to-medium drop in loneliness across 28 trials. It retrains those same thoughts, so it is the first thing to try when being around people feels frightening rather than just tiring.
For most people the problem is simply too few people in the week, with no fear involved. What the mortality studies tracked was regular, repeating contact, and steadiness counted for more than volume. When starting feels heavy, the guide to doing hard things is built for exactly that.
Volunteering is the only self-directed route with a mortality cohort behind it, and it carries the added pull of giving something. Healthy people are the ones well enough to volunteer, which inflates the finding.
Ways to Do It
Start at whichever rung fits your week, and keep it going.
Pick one person you can say the true thing to, and make contact. Choose a call over a typed message where you can, because a call lets you both respond in the moment. Depth counts more than headcount in the loneliness measures, so one person you can be direct with does more for the feeling than ten acquaintances. If you already have that person, contact them more often, and you need not go looking for more people.
Something that repeats, with the same people: a weekly class, a choir, a team, a standing walk, a shift at the same place. Repetition is what turns an acquaintance into a familiar face. One reliable slot beats three you keep canceling. For a caregiver or someone housebound, a short call at the same time each day counts as the slot.
Shared activity asks less of you than conversation does, since doing something side by side removes the need to produce small talk, which is the barrier most people name. A sport, a garden, a workshop, a rehearsal, or volunteering. Pick the thing you would actually keep going back to.
The Chinese Medicine View
The Heart is said to house the Shen, the whole of a person's mental and emotional presence, read in the eyes and in how someone holds a conversation. Joy and a sense of belonging nourish the Shen; their absence depletes it, and connection keeps the spirit settled.
When Heart Blood or Heart Yin runs short, the Shen loses its root: sleep breaks, the person startles easily, and company grows tiring where it once restored. These are recognized Heart patterns in the clinic.
The tradition counts seven emotions as internal causes of illness, each with its own effect on the Qi. Sorrow disperses the Qi and is paired with the Lung, so prolonged grief depletes Lung Qi and shows as a weak voice, shallow breathing, and frequent sighing. Overthinking knots the Qi and burdens the Spleen, the pattern often assigned to rumination after a loss.
Two classical presentations name this ground. Visceral agitation describes a person sorrowful, given to weeping for no reason they can name, treated with Gan Mai Da Zao Tang. Lily Disease describes someone who wants to eat but cannot, wants to rest but cannot settle, often following a shock or a febrile illness.
A quiet person with a small circle who sleeps well, eats well, and has energy is not a patient. The tradition holds back from pushing a depleted person (after illness, childbirth, surgery, or bereavement) into social effort. The counsel is rest, warmth, food, and time, with activity added back gradually. The two frames answer different questions, and neither one validates the other. The research describes hundreds of thousands of people but cannot say what one person should do. The tradition describes one person in detail but has never measured a population.
Cautions For This Practice
Everything to be aware of is here, in one place. This practice suits most healthy people; a few situations call for real care.
When withdrawal is depression
Losing interest in people, alone or alongside low mood, poor sleep, loss of appetite, or a sense that nothing is worth doing, is a core symptom of depression. It responds to treatment. If that fits, talk to a doctor before working on your calendar. The [depression guide](/go/integrative/condition/depression) covers the options and what each one is worth.
When contact brings dread
Feeling fear before seeing people, physical symptoms at the thought of it, or planning your days to avoid it, is closer to anxiety than to loneliness. It is treatable, and the approach that changes how social situations are read has the best trial evidence behind it. The [anxiety guide](/go/integrative/condition/anxiety) is the place to start.
After a bereavement
Getting out more does not fix grief. The raised death risk after losing a spouse is largest for men, and in the first months after the loss. Rebuilding a little routine and regular contact at your own pace helps; being told to move on does not. If low mood, hopelessness, or not eating persist for weeks, that has tipped into something worth taking to a doctor.
If you are thinking about harming yourself
Do not sit with this alone. Call your local emergency number or the crisis line your country publishes, and if someone is nearby, tell them and ask them to stay.
Sudden withdrawal in an older adult
In an older person, withdrawal that comes on suddenly (with memory changes, getting lost, or losing the thread of conversations) needs a medical assessment. It can be an early sign of something treatable.
New hearing loss
Untreated hearing loss is a common, treatable reason people quietly stop joining group settings, and it is missed for years. Check it if crowded rooms have started to feel like too much work.
Start slow, be smart, read the research, and consult a professional if you have any concerns. This is here to inform your choice, not make it for you.
Common Questions
Does talking to people online count?
It counts when it happens at the same time as the other person. A call, a video call, or a game played alongside someone involves turn-taking and repair; reading about people online involves neither. Direct trial evidence that video calls lower loneliness is thin and drawn from nursing homes, so it works best as a supplement to seeing people in person.
Will making more friends help me live longer?
No trial has shown that, and none could easily be run: you cannot randomly assign people friendships and wait to count the deaths. The survival figures only tell us who is already connected; they say nothing about what happens when someone makes new friends. That gap has never been tested in a trial.
Is loneliness really as bad as smoking 15 cigarettes a day?
That number is not in the research. The 2010 meta-analysis compared its effect against other risk factors and put the size on par with smoking as a risk factor, and above obesity or physical inactivity. No count of cigarettes appears in the paper; the 15-a-day framing came from press coverage.
Explore Related
Other pages this one connects to, by the evidence they share, the outcomes they touch, and the ground they cover.
How this connects
- May help with
All 13 sources on this page independently checked and cross-referenced.
Thomas Dehli, Founder & Editor, Sacred Lotus
Sacred Lotus has published Chinese medicine reference material since 2001. Integrative pages are held to the same standard as the herb and formula library: cite the source, grade the claim at its real strength, and say where the research has not looked. This page is educational and it is not medical advice. Last reviewed and updated August 9, 2026.
Evidence strength
How confidently the research supports a claim. Strength describes the evidence, not our endorsement.