The WHO Commission on Social Connection released the first systematic global synthesis on 30 June 2025. Its headline figure — 15.8% globally, roughly 1 in 6 — pools 23 datasets covering 153 countries and territories. 95% uncertainty interval: 11.4–21.4%.
Forest plot · Holt-Lunstad et al. (2015)
What the meta-analysis actually says.
Holt-Lunstad et al. (2015) pooled 70 prospective studies on roughly 3.4 million people. These are the fully-adjusted odds ratios for early mortality — the closed circle is the point estimate, the bar is the 95% confidence interval, k is the number of studies in the pool.
← drag or swipe to read the chart →
Effects are stronger in samples with average age under 65 — meaning lonely younger adults bear disproportionate risk. The 1.00 line marks no excess mortality.Holt-Lunstad et al. 2015
I·5
Movement I.5 — Geography
Where the loneliness lives.
Same species. Same nervous system. The share who currently feel lonely varies by a factor of more than two across world regions and by more than two across income brackets — strong empirical evidence that the soil is doing more work than the seed.
By WHO region · loneliness prevalence 2014–2023
African Region
24%
Eastern Mediterranean
21%
South-East Asia
18%
Region of the Americas
14%
Western Pacific
11%
European Region
lowest globally
10%
Source: WHO Commission on Social Connection, 2025.WHO 2025 The 14-percentage-point spread between Africa and Europe is far larger than any documented individual-level personality difference. Whatever else loneliness is, it is geographically variable.
By country income · the structural inequality of belonging
Low-income countries
24%
High-income countries
11%
People in low-income countries are more than twice as likely to report current loneliness as people in high-income countries — a finding the WHO commission specifically flags but which has been largely absent from the Western public conversation.WHO 2025
II
Movement II — Contagion
It moves between people.
Loneliness is contagious. It spreads through real social networks the way smoking, obesity, and happiness do — measurably, up to three degrees of separation.Cacioppo et al. 2009
ONE DEGREE
52%
if a direct connection is lonely, you are 52% more likely to be too
TWO DEGREES
25%
25% more likely if a friend of a friend is lonely
THREE DEGREES
15%
15% more likely at three degrees out — the effect disappears at four
“Loneliness occurs in clusters, extends up to three degrees of separation, and is disproportionately represented at the periphery of social networks.”
Modern economies optimize for thin ties because they are efficient. Workers who can move, consumers who can switch, citizens who can be replaced. The lonely person is not the system's failure mode — they are its predictable output.
Thick · few · slowDrag the years forward.Thin · many · fast
194019402025
Three series · same scrubber · annual data
U.S. one-person households
1940–2024 · % of all households
7.8in 1940
1970 · No-fault divorce begins to spread (CA 1970 → nationwide)
Time Americans spend socializing in-person
2003–2024 · minutes/day
60in 2003(scrubber at 1940)
2007 · iPhone launches — joinpoint #1
Americans who say “most people can be trusted”
1972–2024 · % of adults
46in 1972(scrubber at 1940)
1988 · Onset of accelerating decline
Sources: U.S. Census HH-1/HH-4 (households, 1940–2024); BLS ATUS socializing with friends (2003–2024, own axis — ATUS did not exist before 2003); GSS variable TRUST (1972–2024). Intermediate ATUS years are reconstructed from published joinpoint segments; see data.json for the full series.
You are not lonely because you are broken. You are lonely because the system requires you to be detachable.
Same window. Same country. Different evidence.
What moves with it.
Four U.S. indicators on the same 2005–2024 axis, each indexed to its 2005 value so the percentage change is directly comparable. As time spent in person with friends has collapsed, the share of prime-age adults living without a romantic partner has climbed, teen mental-health distress has climbed sharply, and the suicide rate has risen. The lines don’t prove causation. They show an era moving together.
← drag or swipe to read the chart →
Friend time, in personBLS American Time Use Survey, all U.S. adults. Annual mean minutes per day socializing in person with friends outside the household. The clearest continuous measure we have of physical disconnection.Kannan et al. 2023
Adults living unpartneredPew Research Center analysis of U.S. Census ACS data: share of prime-working-age adults (25–54) who are neither married nor living with a romantic partner. Cohabiting couples are counted as partnered — the cleanest available 'in a relationship' metric in published U.S. statistics.Fry et al. 2025
Teen persistent sadnessCDC Youth Risk Behavior Survey, U.S. high school students (grades 9–12). Share who report 'felt sad or hopeless almost every day for two weeks or more in a row, so that they stopped doing some usual activities' in the past year. Biennial since 1991.Centers 2024
Suicide rateCDC NCHS age-adjusted suicide rate per 100,000 U.S. standard population, all ages. Rose roughly 30% from 2002 to 2018 then plateaued. The 2023 Surgeon General's advisory explicitly named social disconnection as a contributor.Curtin et al. 2024
Each metric has its own confounders — cohabitation undercount for partnership data, methodology changes for teen surveys, pandemic shock for everything. The lines are placed on one axis so the era can be seen as an era, not so any single arrow can be drawn between them.
Methods & vintage
All four series were reconciled against their primary published sources on 2026-05-22. Endpoints rendered as hollow dots with a (p) suffix indicate provisional or carry-forward values where the primary source has not yet released a final number — see the per-series notes below.
Friend time, in person· vintage 2026-05-22 · 4 provisional points2003–2020 values are from Kannan & Veazie 2023 (SSM-Population Health), which applied joinpoint regression to BLS ATUS microdata and identified statistically significant slope changes in 2007, 2013, and 2019. 2021–2024 values are smoothed estimates of the post-COVID partial recovery, consistent with BLS ATUS annual summary releases; the underlying microdata for 2024 was published by BLS in June 2025. Excludes household members.
Adults living unpartnered· vintage 2026-05-22 · 1 provisional pointSource-published anchor points from Pew Research Center: 29% in 1990, 30% in 2000 (decennial census), 35% in 2010 (ACS), 38% in 2019 (ACS, peak), 36% in 2023 (ACS, January 2025 Pew update). The 2005 value is a linear interpolation between Census 2000 and ACS 2010 (no 2005 Pew analysis published). 2024 is a carry-forward of the 2023 ACS reading. The 2020 ACS sample was not released by Census due to COVID-19 data quality issues, so 2020 here is interpolated. Cohabiting partnerships not involving the household head (~2% of unpartnered) are under-counted by ACS — a known overestimate of the unpartnered share, applied consistently across years.
Teen persistent sadness· vintage 2026-05-22 · all points finalWeighted national prevalence from the CDC YRBS Data Summary & Trends Report (2013–2023) and earlier YRBS DSTR tables for 2005–2011. The 2021 wave was administered Sept–Dec 2021 in still-disrupted school contexts; the 2023 wave used a mixed mode design. Both methodology shifts likely affect comparability across the 2019–2023 segment, though the 2007–2019 trend is monotonic and unaffected by these issues. Independently confirmed by Mojtabai 2025 (BAPC 3.0% [95% CI 2.6–3.4%], n=119,654).
Suicide rate· vintage 2026-05-22 · 1 provisional point2009–2023 values match the CDC NCHS Data Brief 541 (December 2024) final NVSS mortality file exactly. 2005–2008 values are from earlier CDC NVSR Final Data reports. The 2024 value (13.7) is provisional from KFF’s analysis of CDC WONDER 2024; CDC NCHS has not yet released a final 2024 data brief. Rates are age-adjusted to the 2000 U.S. standard population using ICD-10 codes U03, X60–X84, and Y87.0.
Every value above (plus the underlying citations, source URLs, and indexing method) is exposed as machine-readable JSON at /essays/the-social-body/data.json for replication.
IV
Movement IV — Generational inversion
The curve flipped.
For decades, well-being in age was a U: high in youth, dip in middle age, rebound in old age. Blanchflower, Bryson & Xu (2025) showed that across 44 countries this pattern — “among the most striking, persistent in social science” — has been replaced by a monotonic decline. Loneliness is now highest in the young and lowest in the old.
← drag or swipe to read the chart →
c. 2010 · classic U-shape2024 · monotonic decline
Sources: Blanchflower, Bryson & Xu (2025), PLOS ONE — disappearance of the unhappiness hump across 44 countries.Blanchflower et al. 2025 WHO age bands (2025).WHO 2025 Cigna generational index (2025).Cigna 2025
The teenager is the new octogenarian. The data has reversed the cultural script.
V
Movement V — The Mirror
Three questions.
The UCLA 3-Item Loneliness Scale has three items — not three minutes. Each is a single sentence; most people finish in under a minute. Answer honestly. Your responses never leave this browser.
VI
Movement VI — The Sensor
You are not broken.
You are reading the room accurately.
The right word here might be sensor. Hunger isn't a body failing. It's a body asking. Pain works the same way. Loneliness too. The signal arrives so you'll do something about whatever set it off — and treating only the signal leaves the cause where it was.
Loneliness evolved to push you back toward the group. In the ancestral environment that worked. You walked to the fire. In the modern one the signal still fires, but the action it's asking for has been quietly priced out of the room. The fire is across six lanes of traffic. The fire wants $7 for the privilege of sitting near it. The fire is in a building you stopped living near after a job change. The fire is, increasingly, a notification.
So the signal becomes chronic. And a chronic social-pain signal, running long enough, begins to do to the body what chronic hunger does to it — inflammation, immune dysregulation, accelerated mortality. Holt-Lunstad's 2015 meta-analysis found social-deficit mortality effects on a par with the strongest comparator we know how to name out loud: smoking.Holt-Lunstad et al. 2015 The 2023 U.S. Surgeon General's advisory translated that into the line that stuck: roughly the same all-cause risk as fifteen cigarettes a day.Surgeon General 2023
The instinct, faced with this, is to tell the lonely person to put down the cigarette. Put down your phone. Find a club. Reach out. Some of that helps. All of it misses the part underneath. Nobody tells people living in a food desert to just eat better. The same patience is owed to people living in connection deserts.
“What prepares men for totalitarian domination in the non-totalitarian world is the fact that loneliness, once a borderline experience usually suffered in certain marginal social conditions like old age, has become an everyday experience of the ever-growing masses of our century.”
— Hannah Arendt, The Origins of Totalitarianism, 1951.Arendt 1951
Arendt published that sentence three years after a world war and seven decades before the Cacioppo lab put it under a microscope. In 2009, Cacioppo & Hawkley synthesized two decades of experimental work showing that lonely brains are measurably biased toward threat — they detect angry faces faster, interpret ambiguous social signals as more hostile, and remember social negatives more vividly, with elevated amygdala activation to negative social stimuli.Cacioppo et al. 2009 Modern neuroscience is confirming, from the bottom up, what Arendt argued from the top down: a population biased toward perceived threat is a population receptive to politics that identifies enemies and promises belonging. Few loneliness papers cite Arendt. Few political scientists cite Cacioppo. The two literatures are describing the same map from opposite ends.
Which suggests one of the more underdeveloped ideas in the literature: lonely people are not the system's failure mode. They are its most accurate sensor. A society that requires its participants to be detachable — geographically mobile workers, switchable consumers, replaceable citizens — applied to creatures evolutionarily wired to be attached, will produce the felt experience we are calling an epidemic. The lonely person is reading the room correctly. The room is the thing that needs the work.
Four experiments
What the evidence actually recommends.
Each is small. Each is grounded in primary research. None of them are 'just try harder.'
Experiment 01
One thick tie this week.
Pick one person. One real interaction — voice, walk, meal. Not a text. Not a like. The mortality literature finds that perceived isolation, not raw network size, carries the risk — one reliable bond counts.Holt-Lunstad et al. 2015
Experiment 02
One third place.
Klinenberg calls libraries, parks, public pools, and barbershops 'social infrastructure'. Find one within walking distance and become a regular. Belonging often begins somewhere ordinary enough to return to.Klinenberg 2018
Experiment 03
One transition support.
Loneliness peaks in transition — first job, new city, post-divorce, post-bereavement. If you are in one, name it. The data is unambiguous that the transitional are the most lonely, and the most reachable.WHO 2025
Experiment 04
One civic act.
Something local, not online. Cross-cutting ties — across class, age, religion — are the ones Putnam (2000) found protect democracies. The cure for loneliness and the cure for polarization may be the same prescription.Putnam 2000
People do not grow roots alone. We grow them into each other.
A note from the author
Why I made this.
I work in project management. Eight years, mostly creative operations. The job teaches you that when something ships late or breaks, the cause is almost never the person you'd name first. It's the conditions around them: the brief that was never clarified, the dependency nobody flagged, the meeting that should have happened in week two. After a while you stop looking for someone to blame and start looking at the setup.
I built this because I wanted to understand something that touches all of us, directly or sideways. I just came back from a month in Europe, and being somewhere that arranges daily life differently. How people eat, where they sit, what time things close, who you end up talking to without planning it. That made me notice how much of what we call happiness, meaning, how we treat each other, the planet, even our politics, is shaped by how connected we feel. Or don't. I'm still working that out, which is part of why I wanted to read the actual research instead of the takes about it.
Loneliness gets written about in two ways that both bother me: confident claims with no numbers, and numbers presented like the people inside them don't exist. I wanted to see the source material myself.
So I read it. The WHO published its first global synthesis recently. Holt-Lunstad's meta-analysis has been sitting there since 2015, with mortality data on three million people. Cacioppo mapped how loneliness spreads through a social network using Framingham data in 2009. Blanchflower documented the generational flip a few months ago. Young adults are now lonelier than the elderly in much of the rich world, which reverses what we thought we knew. None of this is hidden. The names are below.
What I tried to do was put it all on one desk. Arendt's 1951 paragraph next to the 2009 amygdala data she never got to see. Three U.S. time-series on a single scrubber. Your own UCLA-3 score plotted against the HRS population, scored in a browser that never sends your answers anywhere. The pieces existed. The arrangement didn't.
The deeper question for me is socioeconomic. Why systems produce the outcomes they produce when nobody inside them is choosing for that to happen. Loneliness turned out to be a clean example. We keep calling it personal, and meanwhile there's a decade of structural data sitting underneath.
Here's what I haven't figured out, and the paper that would change my mind would be the one that settles it: is the collapse in social trust in the GSS series causing young-adult loneliness, or is loneliness causing the trust collapse? The data is consistent with both. I'd rather say that than pretend I know.
I'm not an academic. I'm someone who reads carefully and wants to keep learning, and projects like this one are how I do it: trying to assemble something and finding out where my understanding breaks.
If any of this is useful, take it. If a number is wrong, tell me. Methods drawer is below.
If you skipped to this section, you are probably an epidemiologist, a data scientist, or a journalist trying to decide whether to cite the piece. Here is the model under the editorial.
The assumed causal model.
Loneliness is the exposure. Mortality is the outcome. Age, socioeconomic status, and baseline health are confounders — they cause both. Inflammation, sleep quality, and health behaviors are mediators — loneliness operates through them. Frailty creates a back-door (reverse) path that the literature has not fully resolved. Estimates from observational studies that adjust for confounders but not mediators identify a total causal effect; estimates that adjust for mediators (as some studies do) identify only the residual direct effect and will be biased toward null.
Hover (or tab through) any node to read its role in the model.
Dashed arrow: reverse causality (declining health reduces contact, which raises measured loneliness).
Conventions follow Greenland, Pearl & Robins (1999).Greenland et al. 1999 Adjusting for confounders (top row) is required; adjusting for mediators (middle) attenuates the total causal effect and is generally undesirable when the estimand is “does loneliness cause death?” The dashed back-door arrow encodes the reverse-causality problem that randomized trials — ethically impossible here — would close.
E-value sensitivity analysis.
VanderWeele & Ding (2017): the E-value is the minimum strength of association (on the risk-ratio scale) that an unmeasured confounder would need to have with both exposure and outcome to fully explain away an observed effect. For the headline mortality OR point estimates, E-values sit between 1.83 and 1.97; the loneliness lower bound (OR 1.04) yields E ≈ 1.24 — more fragile. Obesity sits near E ≈ 2.37; smoking near E ≈ 3.41. A confounder strong enough to nullify the loneliness–mortality association would need to be one we should already know about.
← drag or swipe to read the chart →
Living aloneOR 1.32 · E 1.97 (lower bound 1.54) Robust — requires a confounder roughly as strong as obesity to explain away.
Social isolationOR 1.29 · E 1.9 (lower bound 1.31) Moderately robust — confounder weaker than smoking but stronger than sedentary lifestyle.
LonelinessOR 1.26 · E 1.83 (lower bound 1.24) Sensitive — a moderate unmeasured confounder could plausibly nullify the effect.
Overall (pooled)OR 1.30 · E 1.92 (lower bound 1.59) Robust — requires a confounder roughly as strong as obesity to explain away.
Method: E(RR) = RR + √(RR · (RR − 1)).VanderWeele et al. 2017 Solid markers = point estimate; open markers = lower 95% CI bound. Reference lines = the published all-cause-mortality risk ratios of smoking, obesity, and sedentary lifestyle. A confounder strong enough to nullify any of these effects would itself be a major public-health story we should already have heard of.
Sample sizes — the evidence stack.
Every figure in the essay is anchored to a dataset of known size. The mortality meta-analysis pools 3.4 million participants. The HRS distribution rests on roughly 20,000 adults per wave. The contagion finding draws on 5,124 Framingham subjects — still the largest network-loneliness panel ever collected.
Source for the three-degrees-of-separation contagion finding.
N = 5.1k
Framingham Heart Study participants
10²10³10⁴10⁵10⁶10⁷
Full reproducibility manifest.
Every series, citation, forest-plot row, DAG node, and E-value is also exposed as machine-readable JSON at /essays/the-social-body/data.json. Each entry carries the canonical source URL and a retrievedAt timestamp. The endpoint is versioned and CORS-open; pull it into a notebook and re-run any of the analyses.
Each is somewhere I’m leaning on inference, on simplification, or on choices someone else made when the data was collected. The kind of thing worth knowing before you take a number from here and put it into something of your own.
Caveat 01
“Loneliness spreads up to three degrees of separation.”
The original network-contagion methodology cannot fully separate genuine social influence from people forming friendships with already-similar people (homophily) or being exposed to the same neighborhood / shock (shared environment). Cohen-Cole & Fletcher (2008) showed the same model produces statistically significant “network effects” for acne, height, and headaches — phenomena that cannot plausibly be contagious. VanderWeele’s (2011) sensitivity analysis concluded the loneliness and happiness findings are less robust to these alternative explanations than the obesity and smoking findings from the same lab.Cohen-Cole et al. 2008VanderWeele 2011
Caveat 02
“1 in 6 people are lonely globally.”
WHO 2025 is the first systematic global synthesis — which means there is no comparable earlier baseline. We cannot say from this number whether loneliness has risen, fallen, or stayed flat. WHO itself states: “Previous data are too limited to determine whether the rates of social isolation and loneliness have risen or fallen.” Country-level rates vary by a factor of more than two (Europe 10%, Africa 24%), driven partly by real environmental differences and partly by survey-instrument heterogeneity.WHO 2025
Caveat 03
“Lonely people have 26% higher mortality.”
The 1.26 odds ratio is the fully-adjusted random-effects pooled estimate, 95% CI [1.04, 1.53]. The lower bound just barely clears 1.00 — the effect is statistically real but the CI is wide, meaning the true effect could be anywhere from a 4% increase to a 53% increase. The unadjusted overall meta-analytic estimate is 1.51, considerably larger; the adjustment matters. Most importantly, this is a population-level association, not a destiny: many lonely individuals live long lives, and the pathway is mediated by behavior (sleep, diet, healthcare-seeking) and physiology (inflammation, immune dysregulation) that interventions can address.Holt-Lunstad et al. 2015
Colophon · primary sources
Every figure in this piece traces back here.
Arendt, H. (1951). The Origins of Totalitarianism.
Blanchflower, D. G., Bryson, A. & Xu, X. (2025). The declining mental health of the young and the global disappearance of the unhappiness hump-shape in age.
Cacioppo, J. T. & Hawkley, L. C. (2009). Perceived Social Isolation and Cognition.
Trends in Cognitive Sciences, 13(10), 447–454. Open source ↗
Cacioppo, J. T., Fowler, J. H. & Christakis, N. A. (2009). Alone in the Crowd: The Structure and Spread of Loneliness in a Large Social Network.
Journal of Personality and Social Psychology, 97(6), 977–991. Open source ↗
Centers for Disease Control and Prevention (2024). Youth Risk Behavior Survey — High school students reporting persistent feelings of sadness or hopelessness in the past year, 2005–2023.
CDC YRBS Data Summary & Trends Report (biennial). Open source ↗
Cohen-Cole, E. & Fletcher, J. M. (2008). Detecting implausible social network effects in acne, height, and headaches: longitudinal analysis.
Crowe, C. L., Domingue, B. W., Graf, G. H., et al. (2021). Associations of Loneliness and Social Isolation with Health Span and Life Span in the U.S. Health and Retirement Study.
Journals of Gerontology Series A, 76(11), 1997–2006. Open source ↗
Curtin, S. C., Garnett, M. F. & Hedegaard, H. (2024). Changes in Suicide Rates in the United States, 2002–2023 — Age-Adjusted Death Rates per 100,000.
CDC NCHS Data Briefs No. 509 (2024) and No. 541 (2025). Open source ↗
Fry, R. & Parker, K. (2025). Rising Share of U.S. Adults Are Living Without a Spouse or Partner (2021) and Share of U.S. Adults Living Without a Romantic Partner Has Ticked Down in Recent Years (2025).
Pew Research Center Social & Demographic Trends — analyses of U.S. Census Bureau decennial census and American Community Survey 1-year estimates. Open source ↗
Greenland, S., Pearl, J. & Robins, J. M. (1999). Causal Diagrams for Epidemiologic Research.
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. Open source ↗
Hughes, M. E., Waite, L. J., Hawkley, L. C. & Cacioppo, J. T. (2004). A Short Scale for Measuring Loneliness in Large Surveys.
Research on Aging, 26(6), 655–672 — UCLA 3-Item. Open source ↗
Hunt, M. G., Marx, R., Lipson, C. & Young, J. (2018). No More FOMO: Limiting Social Media Decreases Loneliness and Depression.
Journal of Social and Clinical Psychology, 37(10), 751–768. Open source ↗
Kannan, V. D. & Veazie, P. J. (2023). US trends in social isolation, social engagement, and companionship — nationally and by age, sex, race/ethnicity, family income, and work hours, 2003–2020.
SSM — Population Health, 21 (PMC9811250). Open source ↗
Klinenberg, E. (2018). Palaces for the People: How Social Infrastructure Can Help Fight Inequality, Polarization, and the Decline of Civic Life.
Silver, L., Keeter, S., Kramer, S., Lippert, J., et al. (Pew Research Center) (2025). Americans' Trust in One Another (and Why It's Declining): The share answering 'most people can be trusted' fell from 46% in 1972 to 34% in 2018.
The Cigna Group / Morning Consult (2025). The 2025 Loneliness in America Report — generational breakdown showing the youngest adults most lonely and Boomers least.
Cigna Group, Loneliness in America Report (2025 release). Open source ↗
U.S. Census Bureau (2024). America's Families and Living Arrangements: 2024.
U.S. Census Bureau Current Population Survey. Open source ↗
U.S. Census Bureau (2024). Historical Households Tables — HH-1 (Households by Type) & HH-4 (Households by Size), 1940–2024.
U.S. Census Bureau, Current Population Survey. Open source ↗
U.S. Surgeon General (Vivek Murthy) (2023). Our Epidemic of Loneliness and Isolation.
U.S. Department of Health & Human Services advisory. Open source ↗
VanderWeele, T. J. (2011). Sensitivity Analysis for Contagion Effects in Social Networks.
Sociological Methods & Research, 40(2), 240–255. Open source ↗
VanderWeele, T. J. & Ding, P. (2017). Sensitivity Analysis in Observational Research: Introducing the E-Value.
Annals of Internal Medicine, 167(4), 268–274. Open source ↗
World Health Organization (2025). From Loneliness to Social Connection: Charting a Path to Healthier Societies.
WHO Commission on Social Connection — flagship report. Open source ↗
World Health Organization (2025). Social connection linked to improved health and reduced risk of early death.
METHODOLOGY · All citation URLs re-verified 21 May 2026. Series sources: Census HH-1 (decennial 1940–2020) merged with CPS HH-4 (annual 2010–2024). BLS ATUS “social engagement with friends” anchored to Kannan & Veazie 2023 published values (60 min/day in 2003, 34 in 2019, 20 in 2020) with intermediate years reconstructed from their joinpoint segments (APC −4.38, +1.58 n.s., −6.89); 2021–2024 partial-recovery estimates from BLS news release. GSS variable TRUST shows actual wave years (waves occur ~biennially after 1994). Forest plot reproduces Holt-Lunstad 2015 Table 3 fully-adjusted random-effects pooled odds ratios with 95% CIs. UCLA-3 placement uses the published HRS distribution shape (Steptoe 2013, Crowe 2021); answers stay in the browser. Generational inversion combines Blanchflower, Bryson & Xu 2025 with WHO 2025 age bands. WHO regional dot lattices round prevalence to integers; the underlying point estimates and 95% uncertainty intervals (Africa 24.3% [20.4–29.0], EMR 21.0% [15.9–27.1], SEAR 18.3% [11.2–29.3], AMR 13.6% [10.2–18.6], WPR 11.0% [6.1–21.7], Europe 10.1% [8.2–12.5]) live in the WHO PDF.