How to Structure Roles Inside a Social Prescribing Group So It Doesn't Collapse by Week Four

Most social prescribing groups run on one role: facilitator. One person books the room, greets everyone, keeps the conversation moving, and notices if someone stops coming. That works for the first three sessions. Then the facilitator gets pulled onto another caseload, or the group grows past the size one person can track, and attendance falls off a cliff. Nobody planned for it to fail — nobody planned the roles at all.

This is what practitioners in this field sometimes call the week four problem: attendance holds for the first two or three sessions, driven by referral momentum and curiosity, then drops sharply once the novelty wears off and the group has to sustain itself without external push. A single-facilitator structure has no mechanism for catching that drop. This piece sets out a role structure that does.

Why one role isn’t enough

Social prescribing works, when it works, by giving someone structured, purposeful contact rather than contact alone. A 2022 qualitative meta-synthesis found that participants describe the benefit of social prescribing extending beyond simple social contact to a restored sense of participation and purpose — structured activity outperforms unstructured mixing. But structure has to be maintained by someone, and a systematic review of loneliness-focused social prescribing found that of nine included studies, all reported positive individual impacts, yet the review is candid that programme design and delivery varied enormously, with little standardisation of how groups were actually run day to day. The evidence says structured groups work. It says almost nothing about who inside the group should be doing what.

Absent that guidance, most programmes default to the model borrowed from clinical services: one professional, many clients. That model assumes the professional is doing all the relational labour. It isn’t sustainable, and it isn’t how healthy informal groups actually function. Robert Putnam’s account of associational decline and Ray Oldenburg’s account of the third place both describe robust informal groups as having distributed, often informal roles — someone who always shows up, someone who brings in new people, someone who keeps the physical space running. None of those groups had a facilitator in the clinical sense. They had a structure that didn’t depend on one irreplaceable person.

The three roles

Split the single facilitator role into three, whether or not you can staff all three with paid time.

1. The host. Owns the logistics: the room, the time, the recurring booking, refreshments if there are any. The host’s job is to make sure the group has somewhere to be, reliably, whether or not anyone from the referring service turns up that week. Eric Klinenberg’s account of social infrastructure is useful here — the physical and administrative scaffolding of a space measurably shapes whether contact happens at all, independent of the people inside it. If the host role isn’t assigned, this scaffolding quietly depends on the health worker’s calendar, and health workers’ calendars change.

2. The anchor. A participant, not staff, who is committed to showing up every session regardless of who else does. The anchor is what a group has instead of a facilitator’s continuous presence: a known, stable face that returning members can orient to. Groups with an anchor survive staff turnover; groups without one restart from zero every time the facilitator changes. This is close to what the peer-support and befriending literature is pointing at — a 2025 randomised trial found that structured befriending in aged care reduced loneliness scores measurably over eight and sixteen weeks, and a separate randomised trial of prosocial volunteering among lonely older adults in Hong Kong found similar reductions. Both interventions work by installing a reliable, low-formality human relationship inside the person’s routine, not by adding more clinical oversight. The anchor role is a way of building that same reliability into a group setting rather than a one-to-one one.

3. The connector. The person actively responsible for noticing who’s missing and reaching out — not to case-manage them, just to say the group noticed. This is the role that catches the week four drop before it becomes permanent attrition. It does not need to be the facilitator, and arguably shouldn’t be, because the facilitator is usually the person also running the session in the room and has no spare attention for absence-tracking. It can be a link worker doing a fifteen-minute check between sessions, or a participant willing to send a message.

None of these roles requires clinical training. That is the point: they are load-bearing but not skilled in the sense that requires a professional qualification, which means they can survive staff rotation if you build them into the group’s habits rather than into one person’s job description.

Assigning the roles: a sequence

1. Before the first session, decide who is host. Book the room for the full run, not session by session — a group that has to be rebooked weekly will die the first week someone forgets.
2. By session two, identify a candidate anchor from among participants — someone who has come twice, seems likely to keep coming, and is comfortable being visibly present. Ask them directly, not implicitly. “Would you be willing to be someone new people can look for” is a real ask, not a compliment.
3. By session three, assign the connector role and set the absence rule: if someone misses a session without notice, the connector reaches out within 48 hours, by whatever channel the person prefers. This has to be routine, not exceptional, or it gets skipped the first busy week.
4. From session four onward, the facilitator’s job shrinks to running the content of the session itself. If the facilitator is still doing all three jobs at session four, the structure hasn’t taken.

What this means in practice: if your programme has one paid facilitator and a group of twelve, do not let that one person also be the group’s institutional memory, its safety net, and its only source of continuity. Recruit the anchor and connector roles out of the group itself by session three. If nobody in the group is willing to take either role, that is diagnostic — it usually means the group hasn’t yet built enough trust to sustain itself, and it needs more facilitator-led sessions before it can be handed any autonomy.

Evidence status: what’s actually supported here

Claim	Evidence status
Structured, purposeful group activity outperforms unstructured social contact	Reasonably supported — qualitative meta-synthesis, consistent with wider social prescribing reviews
Social prescribing reduces loneliness in older adults	Weak — reviews report positive findings across nearly all included studies, but a 2025 protocol notes only one peer-reviewed RCT exists in this specific area
Peer-anchor and befriending-style relationships reduce loneliness measurably	Moderate — two randomised trials, still a small evidence base for the field overall
A three-role structure (host/anchor/connector) specifically improves retention	Not tested — this is an operational inference from the literature above, not a finding of any cited study

Be honest with funders about that last row. Nobody has run a trial comparing single-facilitator groups against role-distributed ones. This structure is a reasonable extrapolation from what’s known about social infrastructure, peer relationships, and associational life — not a proven intervention in its own right.

What this does not solve

This structure addresses retention inside a group that has already formed. It does nothing for the much larger problem of who never gets referred, never attends a first session, or attends once and is too anxious to come back — the population social prescribing reviews consistently struggle to reach precisely because programme design starts from people who already showed up. It also assumes you have at least one participant willing to take on visible responsibility, which will not be true in every group, especially early in a programme’s life. And it says nothing about what happens when the anchor themselves stops coming — which is exactly the kind of single-point-of-failure problem this whole structure is meant to avoid, and which no version of it fully eliminates.