Building a Referral Pathway That Survives Contact With Demand

A GP practice or a hospital discharge team decides that loneliness is a clinical problem worth acting on, finds a walking group or a men’s shed nearby, and starts sending people there. Within six weeks the group has forty new attendees, no more chairs, and a volunteer coordinator who was already running the thing on evenings and a small grant. The referrals stop being useful to anyone. The health side calls it a success because referrals went out. The community side calls it the reason they no longer answer the phone to that practice.

This is the standard way these partnerships fail, and it is avoidable. It fails on the community organisation’s side of the relationship, not the clinical side, which is why most guidance aimed at health systems misses it. The National Academies’ 2020 report on isolation in older adults is right that health systems should routinely identify people who are isolated. It says almost nothing about what happens to the organisation that has to absorb the people identified. This piece is about that half of the arrangement.

Step 1: Map capacity before you sign anything

Before a referral pathway exists on paper, find out what the community organisation can actually hold. This is not a survey question — ask for numbers.

• How many new people per month can they onboard without a waiting list forming?
• Is that number set by physical space, volunteer time, or funding, and which of those is the actual constraint?
• What happens to that number if the group loses one key volunteer?

A systematic review of social prescribing found that most included studies were small and short, with little attention paid to what receiving organisations could sustain — heterogeneity across programmes was one of its central caveats. That gap in the evidence base exists because programme design usually starts from the referrer’s list of options, not the receiver’s ceiling. Start from the ceiling.

Step 2: Decide what “referral” actually means

“Referral” covers at least three different things, and health and community partners frequently discover they meant different ones only after the first cohort arrives.

Model	What happens	Who bears the follow-up cost
Directory	Clinician hands over a leaflet or link	Nobody — and this is why directory-only models show the weakest outcomes
Warm handoff	A link worker or care navigator makes the introduction and checks the person attended	The health-side link worker, at least once
Structured programme	The person joins a defined, time-limited activity with a named contact at the community end	Shared, and agreed in advance

A qualitative synthesis of social prescribing found that the benefit people describe goes beyond having made contact — it is restored purpose and meaningful participation, and structured, purposeful activity outperformed unstructured contact on that measure. Directory referral rarely produces either. If the partnership defaults to handing out leaflets because it is the cheapest version to set up, it should not be called a referral pathway. It should be called information provision, and evaluated as that.

Step 3: Build the feedback loop back to the referrer

Most pathways move information in one direction: clinician refers, community organisation receives. Almost none move it back. The clinician who referred someone in March has no way of knowing whether that person ever turned up, stayed, or left after one session — which means the clinician cannot learn, and the record shows a “successful referral” regardless of what actually happened.

Set this up as a fixed, minimal exchange, not an open-ended reporting relationship the community organisation has to staff:

1. Confirmation of first attendance, within two weeks.
2. A single flag if the person stops attending within the first month — no explanation required, just the flag.
3. A quarterly count of total referrals versus total people still engaged at 90 days.

That third figure is the one worth watching. A protocol reviewing social prescribing for older adults noted that despite growing adoption, there is only one peer-reviewed randomised controlled trial in the area and the effectiveness of the model remains genuinely unclear. Retention at 90 days is a rough proxy that at least tells you whether the pathway is producing sustained contact or a single well-intentioned visit.

Step 4: Fund the receiving end, not just the sending end

Health systems commission link workers, care navigators, and referral software. They rarely commission the community organisation’s side of absorbing referrals — the extra chairs, the extra volunteer hours, the extra intake conversation. The UK’s 2018 national loneliness strategy was notable for funding social prescribing infrastructure at the system level, but even there the money tends to sit with the referral mechanism rather than with the groups being referred into.

This matters because the organisations doing the receiving are often exactly the kind of small, informal social infrastructure — a library reading group, a community centre, a faith-based lunch club — that Eric Klinenberg’s account of social infrastructure describes as chronically under-resourced relative to its social function. And that infrastructure is shrinking on its own terms: one 2025 study tracking twelve categories of third place across the United States found closures in every category between 2019 and 2021, concentrated in the census tracts with the highest social vulnerability. Fortune’s 2026 reporting put a sharper number on one slice of that decline — bowling centres down 32% since 2005, physical library visits down by nearly half between 2019 and 2022. A referral pathway built on top of infrastructure that is disappearing needs to either fund that infrastructure directly or expect the pathway to fail when the venue closes.

What this means in practice: if your commissioning conversation only has a line item for the referral mechanism — a link worker’s salary, a piece of software — and nothing for the organisations being referred into, budget for that gap now. A pathway that floods an unfunded group is not a partnership. It is cost-shifting with better paperwork.

Step 5: Agree review points and an exit before you start

Set a date — three months is reasonable — where both sides look at the numbers from Step 3 and decide, in writing, whether to continue, adjust the referral rate, or stop. Put this in the agreement at the start, not as a crisis response once the group is overwhelmed. The organisation that has to say “we can’t take any more” six weeks into an unplanned surge is in a much weaker position than the one that agreed in advance that 15 referrals a month was the ceiling and 20 triggers a joint review.

Where the evidence is solid and where it is thin

Claim	Evidence status
Social connection interventions can improve self-reported wellbeing outcomes	Reasonably supported, across multiple reviews
Structured, purposeful activity outperforms unstructured contact	Supported by qualitative synthesis, not yet by trials
Social prescribing reduces GP or emergency service use	Mixed — some included studies report this, others don’t measure it at all
Warm handoff outperforms directory referral on retention	Plausible and consistent with the isolation literature, but not directly tested head-to-head in the sources available
The specific 90-day retention threshold recommended here	A practical proxy proposed here, not a validated benchmark from the literature

The clinical case for treating isolation as a health issue is not the weak part of this. Loneliness and isolation carry a mortality risk comparable to other well-established risk factors, and the National Academies has already told health systems to screen for it routinely. The weak part is everything downstream of the referral — what the community organisation can absorb, whether anyone checks that the person stayed, and who pays for the extra chair.

What this does not solve

This is a design pattern for the referral mechanism, not a fix for the underlying shortage of places to refer into. If the third places in a given area have already closed — the library branch, the bowling centre, the community hall — no amount of careful handoff design creates somewhere for the referred person to go. It also does nothing for the person who never gets referred at all: pathways built through clinical contact only reach people already inside a health system, which leaves out anyone who avoids GPs, is uninsured, or simply never mentions it in a ten-minute appointment. A well-run pathway helps the people who reach it. It says nothing about the much larger group who don’t.