The Cost of Loneliness Is Not Yet a Number

Three claims from the Surgeon General’s advisory of 2 May travelled further than the rest: that approximately half of U.S. adults report experiencing loneliness, that the mortality risk of social disconnection is comparable to smoking up to 15 cigarettes a day, and that the response should be organised around six pillars of a national strategy. All three are epidemiological or programmatic. None of them is a price.

That is not a criticism of the advisory. It is a description of the state of the field. Loneliness has accumulated a substantial body of evidence on prevalence and a genuinely strong body of evidence on mortality risk, and almost nothing on what disconnection costs a health system or an economy, or what a reduction in it would be worth. Cost figures nevertheless circulate in policy discussion, usually with the confident air of an established finding. The purpose of this piece is to set out what would be needed to produce such a figure defensibly, and to show which of those inputs the published literature actually supplies.

Four inputs, and how many the literature provides

A credible cost estimate for a population health risk factor needs four things. A prevalence base rate, defining how many people are exposed. An attributable risk, linking exposure to a costed outcome. A unit cost for that outcome. And a counterfactual — an intervention of known effect, so that the estimate describes something avoidable rather than merely something present.

The literature supplies the second of these reasonably well. It supplies the first inconsistently, the third barely, and the fourth almost not at all.

The base rate moves by twenty-five points depending on the question asked

Cigna’s 2020 report put 61% of U.S. adults at sometimes or always lonely, up seven points on the prior year. The Harvard Making Caring Common survey, published in February 2021, put serious loneliness at 36%, including 61% of adults aged 18 to 25. The AARP Foundation’s national survey of 3,020 adults aged 45 and older found one in three lonely. The National Academies concluded in 2020 that roughly a quarter of adults aged 65 and older are socially isolated. The Surgeon General’s advisory says approximately half.

These are not competing measurements of one quantity. AARP used the 20-item UCLA Loneliness Scale, which is why its figure is comparable to the academic literature and why it is lower than instruments that ask a single direct question about feeling lonely. Harvard’s threshold was serious loneliness, a deliberately narrower construct than sometimes lonely. The National Academies figure is not a loneliness figure at all: social isolation is a structural property of a person’s network, countable in contacts and memberships, while loneliness is a subjective state, and the two are independently predictive of mortality. A cost model that treats the 25% isolated among older adults and the 33% lonely among the over-45s as the same population is double-counting or substituting, and there is no way to tell which without the underlying items.

The date matters as much as the instrument. The Cigna figure was published in January 2020 and describes pre-pandemic fieldwork. The Harvard figure describes an October 2020 sample taken in the depths of pandemic restriction. Anyone constructing a time series from published prevalence numbers is mostly constructing a series of question wordings and field dates.

The practical consequence: a national cost model multiplied by a 61% base rate returns a figure roughly 70% larger than the same model run at 36%, before a single assumption about health effects has been made.

The attributable risk is the strongest link in the chain

Here the evidence is comparatively firm. Holt-Lunstad’s 2010 meta-analysis in PLoS Medicine, covering 148 studies and 308,849 participants, found that stronger social relationships were associated with a 50% increased likelihood of survival. The 2015 follow-up 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 running stronger in samples averaging under 65. The American Heart Association’s 2022 scientific statement, led by Crystal W. Cene, put the cardiovascular figures at roughly 30% increased risk of heart attack, stroke, or death from either: 29% for heart attack and cardiac death, 32% for stroke.

Two features of these estimates complicate costing. First, they are observational. Adjustment for baseline health status reduces but does not eliminate the possibility that illness produces isolation rather than the reverse, and the meta-analytic literature is explicit that it describes association.

Second, mortality is an awkward outcome to cost. A risk factor that shortens life does not necessarily raise lifetime health expenditure; in some models it lowers it. The morbidity pathway — the AHA’s finding that isolation is associated with worse prognosis in people who already have coronary heart disease or stroke, including recurrent stroke — is the more plausible route to genuine excess spending, but the statement reports relative risk, not utilisation or cost.

The only quantified utilisation evidence rests on three studies

The 2021 systematic review in Perspectives in Public Health examined the impact of social prescribing on loneliness. All nine included studies reported positive individual impacts. Three reported reductions in GP, emergency, social worker, or inpatient service use.

Three studies is where the economic case for the most widely adopted intervention currently stands. The companion systematic review published in the International Journal of Environmental Research and Public Health the same year found increases in self-esteem and self-confidence as the key outcomes, and noted limited trial evidence and substantial heterogeneity across programmes. The 2022 qualitative meta-synthesis in BMC Health Services Research found that participants described benefit extending beyond social contact into restored meaningful participation and purpose, and that structured, purposeful group activity appeared more effective than contact alone. That last finding is useful for programme design and unusable for costing, because meaning has no unit price and the synthesis does not report one.

The workplace channel is the only place a quantified productivity effect appears. Cigna reported that lonely workers miss work twice as often due to illness and five times as often due to stress. Those ratios are the raw material for an absenteeism cost estimate, and they are also cross-sectional, self-reported, drawn from a single sponsored survey, and dependent on that survey’s own loneliness threshold — the same 61% instrument that sits at the top of the prevalence range. Multiplying an unstable base rate by an unvalidated absence ratio produces a large number with no confidence interval.

The AHA named the gap explicitly

The most important sentence in the 2022 statement, for present purposes, is its identification of the absence of intervention evidence as the central research gap. Without an intervention of known effect, there is no denominator for cost-effectiveness. Cost per quality-adjusted life year requires knowing what a given expenditure buys in effect terms; the loneliness literature can currently say what exposure predicts but not what remediation delivers.

This has not stopped governments from acting, and it should not have. The UK published the first national loneliness strategy in October 2018, embedding loneliness measurement into the Office for National Statistics and funding social prescribing. Japan created a cabinet post in February 2021, with Tetsushi Sakamoto appointed, and the two governments held the first bilateral loneliness ministerial meeting that June. None of those commitments rested on an economic model. They rested on prevalence and mortality evidence, which is the honest basis, and Holt-Lunstad’s 2021 argument that social connection belongs alongside diet, exercise, and smoking as a modifiable protective factor is the strongest available framing for that decision. The National Academies, and the clinician-facing commentary on its report in the American Journal of Geriatric Psychiatry, made the same move: recommend routine assessment, and be specific about what assessment would require in practice, rather than promising savings.

What would make the economic case real

Four things, in rough order of tractability.

Trials that pre-register health service utilisation as a primary outcome and measure it through linked administrative records rather than participant recall. Self-reported reductions in GP visits are the weakest form of the strongest available claim.

Reporting of cost per unit change on a named instrument — a specified movement on the UCLA scale or the ONS measures — rather than aggregate national totals. National totals cannot be checked; a cost per scale point can.

Separate treatment of isolation and loneliness throughout. The 2015 meta-analysis gives them separate odds ratios for a reason, and a model that uses one as a proxy for the other is not conservative in a known direction.

Longitudinal designs. The JACSIS study of Japanese national data across 2020 and 2021 can speak to change over the pandemic precisely because it followed the same people; almost every prevalence figure quoted in cost arguments is a cross-sectional snapshot from a differently worded survey.

Until those exist, the defensible position is narrow and worth stating plainly: the direction of the effect is well established, the magnitude of the avoidable cost is unknown, and any government funding this at scale is accepting an unpriced bet. The UK’s decision to put loneliness questions into the national statistical apparatus in 2018 is the single step most likely to eventually settle the price, and it will take another decade of that series before it can.