Before You Reuse a Loneliness Measure on a New Population, Check This

Most validated loneliness instruments were built and tested on older adults. If your programme serves teenagers, new parents, shift workers, or refugees, and you have picked up the UCLA Loneliness Scale or a version of the De Jong Gierveld scale because it is the one your funder recognises, you have already made an adaptation decision — whether or not you noticed making it.

That is not automatically a problem. It is a problem when nobody checks whether the instrument still measures the thing it claims to measure in the new group.

Why the source population matters

The National Academies’ 2020 consensus report on social isolation and loneliness in older adults is one of the most cited pieces of evidence in this field, and for good reason: it is rigorous, and it drove a lot of the routine-assessment guidance now used in clinical settings. But it is a report about people over 65, whose social lives are shaped by retirement, widowhood, mobility loss, and shrinking household size. A commentary on the report published later that year pushed clinicians to build isolation screening into practice — again, for that population.

Contrast that with the American Enterprise Institute’s 2021 survey on American friendship, which found 15% of men now report having no close friends, up from 3% in 1990 — a fivefold increase concentrated in working-age adults, not retirees. Or the Cigna 2020 workplace report, where 73% of workers aged 18 to 22 reported loneliness, driven by things a geriatric instrument was never built to capture: job insecurity, digital-only teams, the absence of a stable peer cohort. Or the Harvard Making Caring Common survey, which found 61% of young adults aged 18–25 reported serious loneliness, alongside a specific and different marker — about half said no one had spent more than a few minutes recently asking how they were doing in a way that felt genuine.

These are not the same phenomenon wearing different clothes. An older adult’s loneliness item about “having someone to call in an emergency” measures something different from a 19-year-old’s experience of not being asked a real question in weeks. The AARP’s 2018 survey of adults 45 and older used the same 20-item UCLA scale as the academic literature, which is exactly why it is comparable — and exactly why borrowing it wholesale for a group with a different social structure is not automatically safe.

The three checks before you reuse an instrument

1. Does the item content match the population’s actual social structure? Items about spouses, adult children, or retirement communities do not translate cleanly to a programme serving single parents or recent graduates. Read every item and ask what social role it assumes.
2. Has anyone validated this version in a comparable group? Not the same country, not the same age band roughly — the same group. If the answer is no, you are piloting an instrument, not deploying a validated one, and your evaluation plan should say so.
3. Are you measuring loneliness, isolation, or something else entirely? These get used interchangeably and they are not the same. Isolation is structural — network size, contact frequency. Loneliness is subjective — the gap between wanted and actual connection. Holt-Lunstad’s 2015 meta-analysis treated these as distinct risk factors with different odds ratios; a 2023 review in BMC Public Health lists inconsistent measurement as one of the field’s central methodological problems. If your instrument conflates the two, your outcome data will conflate them too.

The borrowed-instrument problem

Programme teams reach for an existing scale because building one is expensive and validating one is slower still. The failure mode is not the borrowing — it is not reporting the borrowing. A funder reading your outcomes report assumes “loneliness reduced by X points on the UCLA scale” means what it meant in the original validation studies. If you have dropped three items, reworded two for reading level, or administered it to a population the scale was never tested on, that assumption is false, and the funder has no way to know it.

What this means in practice: if you adapt an instrument, document exactly what you changed — items dropped, wording altered, population tested — and report that alongside your results. A funder can work with “we adapted a validated scale, here’s how” far more easily than with results that look clean but rest on an instrument doing something it was never shown to do.

Evidence status

Claim	Evidence status
Loneliness and isolation are measurably distinct risk factors	Well supported — Holt-Lunstad’s 2015 meta-analysis, replicated across the literature
Existing scales generalise cleanly across age groups	Not supported — most validation work is in older adults; younger-population items differ substantially, per AEI and Harvard survey findings
Adapting an instrument without revalidation preserves its meaning	Unsupported assumption, frequently made anyway
Inconsistent measurement is a field-wide problem, not just a programme-level one	Well supported — flagged explicitly in the 2023 BMC Public Health review

The Surgeon General’s 2023 advisory treats loneliness as a population-level health issue requiring routine measurement across settings, which is likely to push more programmes toward reusing whatever instrument is closest at hand, simply because it exists and has a name attached. That pressure will not slow down. It makes the documentation step above more urgent, not less.

What this does not solve

None of this tells you which instrument to use for a population that has never been studied — there often isn’t a validated option, and building one is beyond most programme budgets. It also does not address the reach problem underneath all social connection measurement: whatever instrument you choose, it will only ever describe the people who showed up to be measured, not the ones who never entered the programme at all.