Before You Port a Programme to a New Population, Check This

A men’s shed model gets pitched for young fathers. A befriending scheme built for isolated older adults gets rolled out for isolated new mothers. A peer-support group designed for care leavers gets adapted for veterans. Something built and evaluated for one population keeps getting proposed for another, usually because it worked, and working things are scarce enough that nobody wants to start from zero.

Sometimes this is a reasonable shortcut. Often it is how a programme quietly stops working while everyone involved keeps citing the evidence for the original version.

The mechanism, not the demographic, is what transfers

The mistake is treating “this worked for isolated older adults” as if it means “this format works.” It means the specific mechanism — structured recurring contact, low-cost joining, activity as the reason for showing up rather than talk about feelings — worked for a group with a particular set of constraints: retirement removing a daily structure, bereavement or reduced mobility shrinking a social network, and enough free time in the week to attend something at 2pm on a Tuesday. The National Academies’ 2020 consensus report frames isolation in older adults largely through this lens of network shrinkage following role loss.

Move the same weekly-social-club format to, say, new fathers, and the constraint set is different: not too little time-structure but too little discretionary time at all, a social identity (“dad who goes to a club”) that doesn’t yet feel normal to inhabit, and a life stage where the isolating factor is competing demands rather than an empty calendar. The format that worked for one group solves the wrong problem for the other.

The AEI survey on American friendship gives a related warning about assuming demographic similarity means need similarity: the collapse in close friendships since 1990 has been steepest among men, with those reporting no close friends rising fivefold, but the mechanisms behind male friendship loss (work absorbing social contact, discomfort with vulnerability-based bonding) are not simply gender-flipped versions of what drives isolation in women. Porting a female-designed peer-support model to men by changing the flyer photography does not address why men disengage.

A four-question adaptation check

Before adapting a model, answer these about both the original and the new population:

1. What was the isolating mechanism the original model addressed? Loss of role, geographic disconnection, stigma, time scarcity, network shrinkage — name it specifically.
2. Does the new population share that mechanism, or a different one that happens to look similar from outside? Loneliness among remote workers and loneliness among retirees can both present as “not enough contact,” but Gallup’s 2024 workplace data shows fully remote employees report loneliness at 25% against 16% for fully on-site staff — a structural, work-design mechanism, not a network-size one. A model built for the retiree mechanism will not touch the remote-worker one.
3. What was low-cost for the original population to access, and is it still low-cost for the new one? A drop-in daytime group works for retirees with flexible schedules. The same format asks a shift worker or a single parent to give up something they don’t have spare.
4. What was the original success metric measuring, and does it map onto what matters for the new group? Cigna’s 2020 workplace data found lonely workers miss work roughly twice as often for illness and five times as often for stress — a metric relevant to an employer commissioning a workplace programme, and close to useless for judging a community-based befriending scheme for older adults, where the outcome that matters is more likely reduction in emergency contact use or GP visits.

What usually needs to change, and what usually doesn’t

Element	Tends to transfer	Tends not to transfer
Structured, recurring contact (same time, same place, low re-decision cost)	Yes — this is closer to a general principle	—
Activity-first framing rather than “support group” framing	Often	Populations for whom stigma is the primary barrier may need it named differently in each context
Referral pathway (who refers, how)	—	Almost never. A GP-led social prescribing referral works differently from a manager-led workplace referral or a self-referral flyer
Session length and time of day	—	No. This is the single most commonly copied detail that should not be copied
Success metric	—	No. Tie it to the new population’s actual costs, not the old evaluation’s

What this means in practice: don’t ask “did this model work?” Ask “what mechanism did it act on, and does my new population have that same mechanism?” If the answer is a different mechanism, you are not adapting a programme. You are building a new one and borrowing a name.

The named failure mode: costume-change adaptation

This is what happens when a programme changes its marketing to fit a new population without changing its design — new photography, new name, same Tuesday 2pm slot, same referral form, same six-week structure calibrated to a different constraint set. Attendance is lower than projected, it gets blamed on “engagement” or “stigma in this population,” and the underlying mismatch — wrong mechanism, wrong access cost, wrong metric — never gets diagnosed because nobody asked whether the format made sense for this group in the first place.

The qualitative evidence on social prescribing gives a clue to what avoids this: programmes that worked well were the ones where the activity itself carried a sense of purpose and participation for that specific group, not just contact for its own sake. That purpose has to be re-derived for each population, not assumed to carry over.

What this does not solve

This check tells you whether a model’s mechanism fits a new population. It does not tell you whether the mechanism itself is well evidenced — most social prescribing evaluations remain small, heterogeneous, and short-term, per the pooled outcomes literature, and Holt-Lunstad’s mortality-risk work establishes that social disconnection matters, not that any given intervention format reliably reduces it. And it says nothing about reach: adapting a programme carefully for a new population still only serves the people who find their way into it, which in most designs remains people already somewhat willing to seek help.