Countries Are Not Measuring the Same Thing

The American Heart Association’s scientific statement, published on 4 August, does something in its title that most coverage of it does not: it separates objective from perceived social isolation. The headline finding — roughly a 30% increased risk of heart attack, stroke, or death from either, with 29% for heart attack or coronary death and 32% for stroke — is a pooled estimate drawn across studies that operationalised those two things in different ways, sometimes interchangeably. That is not a flaw in the statement, which is explicit about the gap. It is a property of the underlying literature, and it becomes acute the moment anyone tries to compare one country’s prevalence figure to another’s.

Isolation, loneliness, and living alone are three exposures, not one

Holt-Lunstad’s 2015 meta-analysis remains the cleanest demonstration that these are distinct. It reported odds ratios for early mortality of 1.29 for social isolation, 1.26 for loneliness, and 1.32 for living alone — three measures, three estimates, each surviving adjustment for baseline health status. The effects were larger in samples averaging under 65 than in older ones. Her earlier 2010 review, across 148 studies and 308,849 participants, put the survival advantage associated with stronger social relationships at 50%.

Isolation is a structural fact about a network: how many people, how often, in what roles. Loneliness is a subjective state, which Cacioppo and Patrick characterised in 2008 as an aversive signal analogous to hunger — an internal alarm, not a headcount. A person can be structurally isolated and not lonely, and densely connected and severely lonely. If both are independently predictive, then a survey that measures one and a survey that measures the other are not two readings of the same quantity, and averaging them tells you nothing.

Within the United States, the same construct produces figures from 12% to 61%

The US has no harmonised national instrument, and the consequences are visible in the published record.

The AARP Foundation’s 2018 survey of 3,020 adults aged 45 and older found one in three lonely, using the 20-item UCLA Loneliness Scale. The report flagged its own choice of instrument as the reason its figure is comparable to the academic literature — a telling thing to have to say.

Cigna’s report, published in January 2020 and therefore describing pre-pandemic fielding, put the share of US adults who sometimes or always feel lonely at 61%, up seven points year over year, with 73% among 18- to 22-year-olds. That is a self-report frequency threshold, not a scale score.

Harvard’s Making Caring Common survey, published in February 2021 with pandemic-period data, reported 36% of Americans in “serious loneliness,” including 61% of adults aged 18 to 25. Two reports, twelve months apart, 61% and 36% — but the 61% figures refer to different populations under different definitions.

The National Academies’ 2020 consensus report estimated that roughly a quarter of adults aged 65 and older are socially isolated. That is a structural measure and belongs in a different column entirely.

And the Survey Center on American Life’s 2021 friendship survey counted named close friends: 12% of Americans reported none, up from 3% in 1990, with the share of men reporting at least six close friends falling from 55% to 27%. A network-size count is closer to what Putnam was tracking in 2000 through club membership and informal socialising than to anything the loneliness scales capture.

Five figures. One is a validated scale score, one a frequency threshold, one a composite severity category, one a structural classification, one a friendship count. The 12-to-61 spread is mostly instrument.

The UK bought comparability; Japan bought a ministry

The 2018 strategy A Connected Society was the first national loneliness strategy published by any government, and its most consequential technical decision was administrative rather than programmatic: it embedded loneliness measurement in the Office for National Statistics. A standing set of questions, asked repeatedly by a statistical agency, produces something no US figure currently does — a trend line where the instrument is held constant and the population is the only thing moving.

Japan created the second dedicated ministerial post in February 2021, with Tetsushi Sakamoto appointed under the Cabinet Office, and the two governments held the first bilateral loneliness ministers’ meeting in June of that year. The joint message established loneliness as a formal cross-government policy area. What it did not establish, and what the published record here does not settle, is whether the two countries are counting the same thing. Ministerial coordination has moved faster than measurement harmonisation.

The pooling problem runs into the intervention gap

The AHA statement names the absence of intervention evidence as the central research gap. Heterogeneous exposure definitions make that gap harder to close from either end.

Consider social prescribing, the intervention most directly funded by the UK strategy. The 2021 systematic review in Perspectives in Public Health found all nine included studies reporting positive individual impacts, three of them reductions in GP, emergency, social worker, or inpatient use. A parallel review in the same year reported gains in self-esteem and self-confidence while noting limited trial evidence and substantial heterogeneity across programmes. Positive across nine studies is encouraging; it is not an effect size, because the studies were not measuring a common outcome on a common scale.

Holt-Lunstad’s 2021 argument that social connection belongs alongside diet, exercise, and smoking in preventive frameworks depends on this being fixable. Blood pressure is comparable across countries because the cuff is the same. Loneliness prevalence is not comparable across countries, and often not across surveys within one country, because the questions are not.

What a better dataset would look like

A harmonised core module, fielded in the same calendar year in each participating country, reporting three quantities separately rather than as a composite: a short validated loneliness scale, a network-contact count, and household composition. Publication should state the fielding date, not only the release date — the gap between the two is where most misreading of these figures begins.

Until that exists, the honest reading of the AHA’s 30% is that it describes a real and clinically consequential association whose exposure variable is a family of related measures rather than a single one. The direction is settled. The magnitude is inherited from whichever instruments the constituent studies happened to use.