What the WHO's 'One in Six' Loneliness Figure Actually Measured

In June 2025, the World Health Organization’s Commission on Social Connection published a figure that has since traveled well beyond public health circles: one in six people worldwide is affected by loneliness. Attached to it was a second, more alarming number — an estimated 871,000 deaths annually, roughly 100 an hour, attributable to loneliness and social isolation. Both numbers now appear routinely in reporting and advocacy as though they were a single clean measurement, comparable to a national unemployment rate or a case count. They are not. They are the output of a synthesis exercise, and understanding what went into that synthesis matters more than repeating the headline.

A composite estimate, not a single survey

The WHO figure did not come from one questionnaire administered to a representative sample of the world’s population, in the way a Gallup poll does. It is a pooled estimate drawn across the existing body of national and regional loneliness surveys, each of which used its own instrument, its own threshold for what counts as “lonely,” and its own sampling frame. The commission’s report and its accompanying materials frame this explicitly: the number is a synthesis, produced to give policymakers a workable global figure where none previously existed, not a harmonized measurement collected under a common protocol.

That distinction explains why one in six sits so far below other prevalence figures published around the same period. Gallup and Meta’s 2023 survey of roughly 1,000 people aged 15 and older in each of 142 countries — using a single question about feeling very or fairly lonely — found 24% of the world’s population lonely, close to one in four rather than one in six. Cigna’s 2025 U.S. survey put loneliness at 57% of American adults. The 2022 HINTS-6 data, analyzed and published in 2025, found that over 37% of U.S. adults reported moderate-to-severe loneliness, with about one in seven reporting severe loneliness alone. None of these four figures — one in six, one in four, 57%, 37% — is wrong. They are answers to different questions, asked with different instruments, at different thresholds, of different populations.

Where the instruments diverge

The Gallup measure is a single item: how often, if at all, do you feel lonely. Respondents choosing “very” or “fairly” lonely are counted. This is a coarse but comparable cross-national tool, precisely because its simplicity travels across 142 countries without requiring translation of a multi-item psychological scale.

Cigna’s instrument, by contrast, draws on a longer battery derived from established loneliness scales and captures a broader band of people who report loneliness at any frequency, which is part of why its figure runs so much higher than Gallup’s single-item snapshot. HINTS-6 categorizes respondents into none, mild, moderate, and severe loneliness states, which is a finer-grained approach than a binary “lonely or not” but produces a different denominator depending on where the line between “moderate” and “counted as lonely” is drawn. The AARP surveys of adults 45 and older use the 20-item UCLA Loneliness Scale, the closest thing this literature has to an academic gold standard, and even that produces a figure — roughly one in three — that sits in yet another band.

The WHO’s “one in six” pools across all of this heterogeneity. It is not that the pooling was done carelessly; it is that any single number produced by combining incompatible instruments will inevitably compress the range and land somewhere lower than the measures that ask about loneliness more expansively. A reader who treats one in six as the true global rate, and 24% or 37% as competing claims to be adjudicated, has misunderstood what each number is doing.

The mortality figure rests on a further layer of estimation

The 871,000-deaths claim compounds the measurement question. To produce an attributable-death estimate, researchers need a prevalence figure and a relative risk — the added probability of death associated with loneliness or isolation, drawn from the epidemiological literature. Holt-Lunstad’s 2015 meta-analysis in Perspectives on Psychological Science is one of the standard sources for that relative risk, reporting an odds ratio of 1.26 for loneliness and 1.29 for social isolation on early mortality, adjusted for health status. Multiply a chosen relative risk by a chosen prevalence estimate across a chosen population base, and an attributable-death figure falls out. Each of those three inputs carries its own uncertainty, and small changes in any one of them move the final number meaningfully. The 871,000 figure is therefore a modeled estimate built on other estimates, not a count of deaths in the way that, say, road traffic fatality statistics are counted.

None of this means the underlying concern is overstated. The direction of the evidence — that loneliness and isolation are associated with meaningfully elevated mortality risk, and that the association survives adjustment for baseline health — is well established across dozens of studies. What is not well established is a single, stable global prevalence number that can be tracked year over year the way a country tracks its inflation rate. The WHO report itself notes that loneliness rates are highest among young people and in low-income countries, a pattern broadly consistent with Gallup’s finding that adults aged 19-29 report loneliness at 27% against 17% for those 65 and older. The direction of these age findings agrees across sources even where the absolute levels do not, which is a more trustworthy signal than any single headline percentage.

What would actually settle this

A genuinely comparable global estimate would require a common instrument, administered with a consistent threshold, across a representative sample in every country, repeated on a fixed schedule. That does not exist, and building it is a multi-year undertaking rather than a synthesis exercise. Until it does, the honest use of any single loneliness prevalence figure is to state which survey produced it, which instrument was used, and what threshold counted as “lonely” — and to treat comparisons between one in six, one in four, and one in three not as contradictions to be resolved but as a reminder that the field has not yet agreed on what it is counting.