How the WHO Commission Handled Its Own Uncertainty

On 30 June 2025 the WHO Commission on Social Connection published From Loneliness to Social Connection: Charting a Path to Healthier Societies, reporting that one in six people worldwide is affected by loneliness and that loneliness is linked to an estimated 871,000 deaths a year — roughly 100 every hour. The Director-General’s remarks at the launch framed social connection as a determinant of health on par with other major public health priorities.

Two of those numbers do different kinds of work. “One in six” is a prevalence estimate drawn from survey instruments. “871,000 deaths” is a derivation: an epidemiological attribution built by combining prevalence with relative-risk estimates from observational cohorts. The report’s own wording is “linked to”. The rate — 100 an hour — reads as attribution. Both appear in the same document, and it is the second that travels.

This is not a complaint about a rounding error. It is a question about how a flagship report distributes its uncertainty, and where it chose to absorb uncertainty rather than display it.

Where 871,000 comes from

The relative risks underneath any such estimate come from a small number of meta-analyses. Holt-Lunstad’s 2010 review in PLoS Medicine, covering 148 studies and 308,849 participants, found stronger social relationships associated with a 50% increased likelihood of survival. Her 2015 review in Perspectives on Psychological Science reported odds ratios of 1.29 for social isolation, 1.26 for loneliness, and 1.32 for living alone, with effects persisting after adjustment for health status and larger in samples averaging under 65. The American Heart Association’s 2022 scientific statement, led by Crystal W. Cene, put the cardiovascular figures at a 29% increased risk of heart attack and cardiac death and a 32% increased risk of stroke.

These are robust associations, replicated across designs and populations. They are also, without exception, observational. The AHA statement was explicit that the absence of intervention evidence is the central research gap in the field — an unusual thing for a scientific statement to say about its own subject, and the most useful sentence in it.

A mortality attribution converts those odds ratios into a count. Doing so requires assuming that the association is causal at the population level, that the exposure is measured comparably across the countries contributing prevalence data, and that confounding by baseline health has been adequately handled. The first assumption is the one the underlying literature cannot yet supply. Holt-Lunstad’s 2021 review in the American Journal of Lifestyle Medicine argues that social connection should be treated as a modifiable protective factor alongside diet, exercise and smoking; the argument is reasonable, but “modifiable” is a hypothesis about interventions, not a finding from cohort studies.

The denominator problem

The prevalence input is not stable either, and this is where the report’s confidence is hardest to justify.

Gallup and Meta’s State of Social Connections, fielded in roughly 1,000 respondents aged 15 and over in each of 142 countries between June 2022 and February 2023, found 24% feeling very or fairly lonely and 49% reporting no loneliness at all. Gallup’s October 2024 release put the figure at more than one in five adults feeling lonely a lot of the previous day — a different question, a different reference period, a similar-looking number. The 2023 analysis of the same dataset noted that connectedness and loneliness are not simple inverses: 72% felt very or fairly socially connected, which does not arithmetically reconcile with the loneliness figure.

Against those, the WHO’s one in six is lower than most national estimates from high-income countries. The 2023 U.S. Surgeon General’s advisory reported that approximately half of U.S. adults experience loneliness. Loneliness in America 2025, published by the Cigna Group and Evernorth Research Institute in June 2025, put the figure at 57% — from fieldwork conducted 29 May to 14 June 2024, a year before publication. The CDC’s MMWR surveillance report published in June 2024 used 2022 data. Any global figure is therefore averaging instruments that are not measuring the same construct at the same time, a problem the 2023 BMC Public Health review of the field named directly: inconsistent measurement is the primary barrier to comparing findings across studies.

The Commission, established in November 2023 and co-chaired by Vivek Murthy and Chido Mpemba, could have reported its mortality estimate as a range conditioned on which prevalence instrument was used. It reported a point estimate and an hourly rate.

Technology is where the report had the least to work with

The Commission’s headline findings concern prevalence, mortality, and the concentration of loneliness among young people and in low-income countries. They do not resolve what is driving any of it. Yet the first question any health ministry implementing the report will be asked is whether phones and remote work are the cause, and the evidence available to answer that is markedly weaker than the evidence for the mortality association.

Consider what exists. Jonathan Haidt’s The Anxious Generation, published in March 2024, argues that the shift from play-based to phone-based childhood drove rising adolescent anxiety, depression and loneliness from the early 2010s, drawing on Jean Twenge’s generational research. The collaborative review documents Haidt and Twenge maintain alongside the book compile the supporting studies and include dissenting evidence; several researchers dispute the strength of the causal claim. That dispute is live, and a policy document that resolved it by assertion would be overreaching.

On adult work arrangements, Gallup’s State of the Global Workplace 2024 found loneliness reported by 25% of fully remote employees, 21% of hybrid, and 16% of fully on-site, with higher rates under 35. Those are cross-sectional differences with obvious selection: people who are already isolated may sort into remote roles. A February 2024 cross-sectional study of healthcare workers found something more interesting — workplace isolation and loneliness behave as separate constructs with different correlates, and perceived social support moderates the relationship between remote working and well-being. That finding complicates the simple reading of the Gallup gradient rather than confirming it.

The Cigna report’s own framing is instructive. It notes that Gen Z and Millennials report more loneliness than Gen X and Baby Boomers “despite greater technological connectedness”. The word despite smuggles in a causal expectation — that technological connectedness should reduce loneliness — which the survey did not test.

A report that named the technology question as unresolved, and said so as plainly as the AHA named its intervention gap, would have been more useful to ministries than one that emphasised the deaths figure. There is no way to tell from the headline findings whether the Commission did this; the figure that entered circulation was the body count.

One randomised trial, and a protocol that says so

Uncertainty about causes matters less if interventions work regardless. The intervention literature does not support that comfort.

Social prescribing is the most widely adopted response. The UK’s 2018 A Connected Society — the first national loneliness strategy — funded it and embedded loneliness measurement into the Office for National Statistics. The 2021 systematic review in Perspectives in Public Health found all nine included studies reporting positive individual impacts, with three reporting reduced use of GP, emergency, social work or inpatient services. Encouraging, and almost entirely uncontrolled.

A systematic review protocol posted in July 2025 states the position bluntly: the effectiveness of social prescribing for isolation and loneliness in older adults remains unclear despite growing adoption, and only one peer-reviewed randomised controlled trial exists in this area. The HEAL-HOA dual randomised trial, published in The Lancet Healthy Longevity in November 2024, tested volunteering and prosocial engagement against a control among lonely older adults in Hong Kong. It is valuable primarily because it is rare.

For digital-environment interventions specifically — screen-time policy, platform design regulation, phone-free schooling — the trial evidence available to a policymaker in August 2025 is thinner still.

The exposure that can be measured

The contrast worth drawing is with physical infrastructure. A study in Health & Place published this month tracked twelve categories of third place, the concept Ray Oldenburg introduced in 1989, across U.S. census tracts from 2010 to 2021, and found closures in every category between 2019 and 2021, concentrated in tracts with higher social vulnerability and in rural areas. Eric Klinenberg’s argument that social infrastructure measurably shapes rates of contact has an empirical footing that the technology hypothesis does not: coffee shops and libraries can be counted, geolocated, and linked to health records. American Time Use Survey evidence, as reported in February 2024, shows face-to-face socialising declining before the pandemic and continuing after it — a decline no single mechanism has yet been shown to explain.

A second edition of the Commission’s report would earn more by publishing the mortality estimate as an interval with its prevalence input and instrument stated, by separating isolation from loneliness in the attribution — the 2024 Scientific Reports analysis shows their relationship varies by age — and by declaring the technology question open. States can build libraries on weak causal evidence at low risk. Regulating platforms on the same evidence is a different proposition, and the report gave ministries no basis for telling the two apart.