Which Is Worse, Being Alone or Feeling Lonely

Key takeaways

• Across 148 studies and 308,849 people, stronger social relationships were associated with a fifty per cent higher likelihood of surviving the follow-up period.
• A 2013 cohort of 6,500 older adults found social isolation still predicted mortality after adjustment while loneliness did not.
• Holt-Lunstad's own 2015 meta-analysis found the opposite balance: isolation, loneliness and living alone all predicted mortality at comparable size, with no difference between objective and subjective measures.
• Poor social relationships were associated with a twenty-nine per cent higher risk of coronary heart disease and a thirty-two per cent higher risk of stroke.
• Nobody has yet shown that interventions targeting either one prevent those outcomes, which is the limit on all practical advice in this area.

Being alone and [feeling lonely](https://thewhyweproject.com/articles/why-can-you-feel-lonely-in-a-crowd) are different things, and for twelve years the field has been trying to work out whether they are different risks. The answer arrived at by the largest synthesis is that they are not, or not by much, which is awkward for the distinction and useful for you.

The headline evidence is as strong as anything in health psychology. Across 148 studies and 308,849 people, Julianne Holt-Lunstad, Timothy Smith and Bradley Layton found that having stronger social relationships was associated with a fifty per cent higher likelihood of surviving the follow-up period. That held across age, sex, how healthy people were at the start, what they eventually died of, and how long they were tracked.

So the general claim is settled. Which part of being connected does the work is not, and the two best attempts to answer it disagree with each other.

Julianne Holt-Lunstad

The 2010 meta-analysis is where the subject acquired its authority, and one detail inside it is the key to everything that followed. The association was not uniform across the ways researchers measured social connection. It was strongest for complex measures of social integration, meaning how many different kinds of relationship and social role a person actually has and participates in. It was weakest, and did not reach significance, for the simplest measure of all, whether somebody lives alone or with others.

That looked, for a while, like a clue. Living with another human is the most obvious form of not being alone, and in that analysis it predicted the least.

Five years later the same lead author went back to the question directly, and the clue did not survive. Holt-Lunstad, with Timothy Smith, Mark Baker, Tyler Harris and David Stephenson, pooled the studies that separately measured loneliness, social isolation and living alone. After controlling for confounds, all three predicted mortality at similar size: social isolation at twenty-nine per cent increased likelihood, loneliness at twenty-six, living alone at thirty-two. They report no difference between objective and subjective measures of isolation.

So the author whose work opened this article has, in her own later synthesis, found the split that the rest of this article is about to describe does not appear.

Andrew Steptoe

Three years after the first meta-analysis, and two before the second, one cohort study found something different. Andrew Steptoe, Aparna Shankar, Panayotes Demakakos and Jane Wardle followed 6,500 men and women aged fifty-two and over in the English Longitudinal Study of Ageing, measuring social isolation as contact with family and friends plus participation in civic organisations, measuring loneliness with a standard questionnaire, and tracking who died over the next seven years or so.

Both isolation and loneliness were associated with higher mortality, which is what everyone expected. Then they adjusted for demographic factors and for how healthy people were at the outset, and the two came apart.

Social isolation remained significantly associated with mortality after adjustment, with those in the most isolated fifth at roughly a quarter higher risk. Loneliness did not. And when loneliness was added into the model alongside isolation, the isolation association was unchanged, meaning loneliness was not the pathway through which isolation was doing its damage.

The authors' conclusion is careful. Both isolation and loneliness impair quality of life and wellbeing, and efforts to reduce isolation are likely to be more relevant to mortality.

Hold that against the 2015 meta-analysis and you have a real disagreement rather than a tidy finding. One cohort of 6,500 people, well measured over seven years, in which loneliness dropped out after adjustment. One synthesis of many cohorts, in which it did not.

Nicole Valtorta

The cardiovascular evidence cannot settle the split, and it is worth being precise about why, because it gets cited as though it could. Nicole Valtorta and colleagues searched sixteen databases for longitudinal studies and pooled sixteen datasets, covering 4,628 coronary heart disease events and 3,002 strokes, over follow-up periods running from three to twenty-one years. Poor social relationships were associated with a twenty-nine per cent increase in the risk of developing coronary heart disease and a thirty-two per cent increase in the risk of stroke. There was no difference between men and women.

Poor social relationships is a single pooled construct in that analysis. It does not separate the objective from the subjective, so it corroborates the general claim and says nothing at all about which half is doing the work. We include it because it establishes the magnitude in a second outcome, not because it takes a side.

The paper ends on the question the whole field is stuck on, and it is an honest place to stop. Nobody has shown that interventions targeting loneliness or isolation prevent either outcome. An association across hundreds of thousands of people is a strong reason to think something is there. It is not a demonstration that changing it changes anything.

What the split is for

The distinction is worth keeping even though the evidence does not resolve it, because the two things are measured differently and lived differently. Isolation is a fact about your week. How many people you spoke to, how many groups you belong to, whether anyone would notice a change in your routine. It is countable, it is boring, and it is the one the mortality evidence supports.

Loneliness is a fact about how that week felt. It is the gap between the connection you have and the connection you want, and a full diary is no protection against it. The Steptoe result does not say loneliness is harmless or imagined. It says loneliness is not what the mortality numbers were measuring, and the authors are explicit that it damages quality of life regardless.

One caution before anybody turns this into a personal prognosis. These are population associations from observational studies, not a forecast about a particular person, and the people in them differed in a hundred ways that no adjustment fully removes. The right use of a number like this is to decide what is worth an hour of your week. It settles nothing about you.

The reading we would defend

Our reading is that the meta-analysis beats the cohort on this question, and that anyone who tells you the science has settled which half matters is quoting one study and not the other.

A single cohort, however well run, can produce a null on a variable for reasons that have nothing to do with the world: how loneliness was measured in that questionnaire, who was left after adjustment, how much isolation and loneliness overlap in that particular sample. A synthesis across many cohorts is not immune to any of that, and it is less hostage to one of them. On the general question of what predicts dying sooner, the weight is with all three mattering at comparable size.

What we would still say is that isolation is the half you can act on directly, and that this is a practical point and not an evidential one. Join something with a fixed time that meets whether or not you feel like it. Keep a role that makes somebody expect you. Aim for more kinds of relationship, since complex social integration was the measure with the strongest association in the 2010 analysis, though that measure may be strongest partly because it is the best proxy for being well enough to have a varied life, which no adjustment removes.

Two ways this can go wrong

If you act only on isolation, you can end up with a calendar full of obligations that never touch how you feel, and no reading of this evidence makes that a success. The cohort that found loneliness dropping out after adjustment is not permission to treat loneliness as trivial, and the larger synthesis puts it almost level with isolation.

If you act only on loneliness, the risk is that the work stays private. Loneliness can be addressed entirely inside your own head, and sometimes should be. Isolation cannot, which is the whole of its practical advantage.

The honest answer to the question in the title is that nobody can tell you, because the two best studies disagree, and the larger one says the difference between them is small. What is not in doubt is that both matter and that only one of them is on your calendar.

Sources

Holt-Lunstad, J., Smith, T. B., & Layton, J. B. (2010). 'Social relationships and mortality risk: a meta-analytic review.' PLoS Medicine, 7(7), e1000316. 148 studies, 308,849 participants; a 50 per cent increased likelihood of survival for those with stronger social relationships, consistent across age, sex, initial health status, cause of death and follow-up period; strongest for complex measures of social integration and lowest for binary indicators of residential status. doi:10.1371/journal.pmed.1000316

Holt-Lunstad, J., Smith, T. B., Baker, M., Harris, T., & Stephenson, D. (2015). 'Loneliness and social isolation as risk factors for mortality: a meta-analytic review.' Perspectives on Psychological Science, 10(2), 227-237. Controlling for confounds, weighted average effects were social isolation 29 per cent, loneliness 26 per cent and living alone 32 per cent increased likelihood of mortality; the authors report no differences between measures of objective and subjective social isolation. doi:10.1177/1745691614568352

Steptoe, A., Shankar, A., Demakakos, P., & Wardle, J. (2013). 'Social isolation, loneliness, and all-cause mortality in older men and women.' PNAS, 110(15), 5797-5801. 6,500 participants aged 52 and over in the English Longitudinal Study of Ageing, mean follow-up 7.25 years; after adjustment for demographic factors and baseline health, social isolation remained significantly associated with mortality and loneliness did not; the association of isolation with mortality was unchanged when loneliness was included. doi:10.1073/pnas.1219686110

Valtorta, N. K., Kanaan, M., Gilbody, S., Ronzi, S., & Hanratty, B. (2016). 'Loneliness and social isolation as risk factors for coronary heart disease and stroke.' Heart, 102(13), 1009-1016. Sixteen databases searched, 16 longitudinal datasets pooled, 4,628 coronary heart disease events and 3,002 strokes over follow-up of 3 to 21 years; poor social relationships associated with a 29 per cent increase in incident coronary heart disease and a 32 per cent increase in stroke, with no differences by gender. doi:10.1136/heartjnl-2015-308790

A note on what this article does not claim: these are observational associations, the 2013 cohort and the 2015 meta-analysis disagree, and no intervention study has shown that reducing either isolation or loneliness prevents these outcomes.

Related concepts

• Social Isolation
• Loneliness and Mortality
• Reverse Causation

People

• Andrew Steptoe
• Nicole Valtorta
• Julianne Holt-Lunstad