The WHO's 871,000 Deaths Figure, and What Sits Behind It

The WHO Commission on Social Connection’s June 2025 report leads with a striking number: loneliness is linked to an estimated 871,000 deaths a year, roughly 100 an hour. The figure travels well. It has the shape of a fact that belongs on a slide. The question this note asks is narrower than whether loneliness is harmful — the underlying literature on that point is not in serious dispute — but whether the report’s own presentation of that number, and of technology’s role in producing it, is as careful about uncertainty as the evidence underneath it requires.

Where the number comes from, and what it inherits

The 871,000 figure is not a new primary estimate. It sits downstream of a body of association-based meta-analysis, most influentially Julianne Holt-Lunstad’s 2010 review of 148 studies covering 308,849 participants, which found that stronger social relationships were associated with a 50% greater likelihood of survival, and her 2015 follow-up, which reported odds ratios of 1.29 for social isolation, 1.26 for loneliness, and 1.32 for living alone as risk factors for early mortality. Those are solid, widely replicated numbers, and they carry appropriate statistical framing — odds ratios, confidence intervals, adjustment for health status.

What happens when that evidence is compressed into “871,000 deaths a year” is a different exercise. Converting a relative risk into an attributable death count requires assumptions about baseline mortality, the prevalence of loneliness in the population being modeled, and the causal structure connecting the two — assumptions the WHO report’s public-facing materials do not walk through. The number is presented as a settled global toll rather than as the output of a chain of estimates, each with its own margin of error. A reader working only from the report and the WHO Director-General’s remarks at its launch has no way to know how sensitive that figure is to the underlying prevalence assumptions, which the report elsewhere puts at roughly one in six people affected by loneliness worldwide — itself an estimate drawn from heterogeneous national surveys using different instruments.

This is worth setting against a genuine model of good practice in the same literature. The American Heart Association’s 2022 scientific statement on social isolation and cardiovascular and brain health reports comparably sized associations — about a 30% increased risk of heart attack, stroke, or death from either — and then does something the WHO report’s summary materials do not: it states explicitly, as a primary conclusion, that the absence of intervention trial evidence is the central gap in the field. The AHA statement treats its own numbers as a reason for more research, not as a settled basis for a global death toll. The WHO report treats a comparable evidence base as sufficient grounds for a specific, quotable count.

Technology gets a causal shortcut too

The same asymmetry shows up in how the report’s framing handles technology as a driver of loneliness. It is common, in coverage of loneliness policy, to treat declining face-to-face contact and rising smartphone and social media use as a single, obvious causal story: screens replaced sociability, therefore screens caused the loneliness rise. Jonathan Haidt’s 2024 book makes this case forcefully, arguing that the shift from play-based to phone-based childhood in the early 2010s produced a measurable rise in adolescent anxiety, depression, and loneliness. But Haidt’s own collaborative evidence review, compiled with Jean Twenge, includes dissenting studies alongside supportive ones — an unusual and useful acknowledgment that the causal claim is contested within the field that produced it, not merely by outside critics.

The WHO report does not extend that same care to its own references to technology. It gestures toward digital life as a contributing factor to disconnection without distinguishing displacement (technology substituting for in-person contact) from mediation (technology enabling contact that would not otherwise happen) or from confound (technology use rising alongside other social changes with independent effects on loneliness). A commentary from George Mason University’s public health program, published a month before this note, makes the more defensible move: it frames AI companionship specifically as a substitution risk to be watched, not a proven cause of loneliness, and it stops short of assigning it a causal share of any mortality estimate.

What the uncertainty actually implies for policy

None of this undermines the report’s central recommendation — that member states should treat social connection as a public health priority. The direction of the evidence supports that conclusion comfortably. But a policy document that asks governments to act on a number should show its work on that number, particularly when it is functioning as the headline rather than as one supporting statistic among many. A reader cannot currently tell whether 871,000 is a central estimate with wide bounds or a figure sensitive to a single set of prevalence assumptions.

A stronger version of this report would publish the attribution methodology alongside the topline figure, report a plausible range rather than a point estimate, and separate what the mortality literature actually measures — isolation and loneliness as correlates of death, adjusted for measurable confounders — from what a global death count implies, which is a causal claim the underlying studies were not designed to support. Until then, the number is best treated as an order-of-magnitude signal that social disconnection matters for survival, not as a figure precise enough to bear the weight it is being asked to carry.