The Week Four Problem in Social Prescribing

If you run a social prescribing group, a walking scheme, or a referral-based befriending programme, you already know the shape of the attendance curve. Session one is full. Session two loses a few people to logistics. Session three is fine. Somewhere around session four, attendance drops hard, and the people who don’t come back mostly don’t come back at all.

This is not a gradual decline. It is a threshold. And because it happens at roughly the same point across very different programme types — walking groups, arts-on-referral, men’s sheds, exercise referral — it is worth treating as a design problem rather than a motivation problem.

Why the drop is not random

Three things converge around the fourth session.

First, novelty runs out. The first two or three sessions carry their own interest — new people, a new space, the mild curiosity of finding out what the thing actually is. By session four, the group has settled into its actual content, which is often less interesting than the pitch.

Second, the referral relationship goes quiet. Social prescribing link workers typically front-load contact: an initial conversation, a warm handover, maybe a check-in call after the first session. By week four that support has usually tapered off, and the participant is now attending, or not attending, on their own initiative. The National Academies’ 2020 consensus report on isolation and loneliness in older adults is explicit that health systems need to treat isolation as an ongoing thing to monitor, not a referral to close out — and a companion clinical commentary on that report pushes further, arguing that routine reassessment, not a one-off screen, is what clinical settings would need to build for isolation to be taken seriously as a modifiable risk factor. Most programmes are not built that way. The support curve and the attendance curve are shaped the same way, and that is not a coincidence.

Third, there is no reason yet to have made a friend. Group cohesion — the thing that actually keeps people coming — usually needs more like six to eight shared sessions to form. At session four, most participants have not yet gained the thing the programme is meant to produce, but they have already spent the energy of showing up three times. That is a bad trade, and people respond to bad trades by stopping.

What the evidence actually supports here

Be careful about what is established and what is plausible extrapolation.

Claim	Evidence status
Social connection is a legitimate target for preventive health, comparable to other modifiable risk factors	Fairly well established — Holt-Lunstad’s 2021 review makes this case directly
Social prescribing improves self-reported wellbeing and confidence for people who complete a programme	Supported by a systematic review of social prescribing outcomes, though the review notes limited trial quality and heterogeneous outcome measures
Social prescribing reduces downstream service use (GP visits, A&E, inpatient care)	Weakly supported — a 2021 systematic review found three of nine included studies reported this; not a settled finding
Attendance collapses at a specific, predictable session number rather than declining smoothly	Not directly measured in the published literature; this is a pattern practitioners report, not a quantified finding
Network size and diversity, not just frequency of contact, predict lower loneliness	Supported — AARP’s 2018 national survey of adults 45 and older found network size and diversity were the strongest predictors, ahead of physical isolation alone

That last row matters for what you build, not just when you check in. A programme that produces one weekly contact point but no broader network is treating the wrong variable.

What this means in practice: don’t schedule your check-in calls around your own capacity. Schedule the one that matters — a call, a text, or a co-attendance offer — to land just before session four, not after it. By the time attendance has already dropped, the intervention is too late; you are now trying to re-recruit rather than retain.

Designing around the threshold, not against it

1. Move the support taper to week five, not week two. If your model currently has a link worker step back after the first session, extend meaningful contact — even brief — through the fourth. This is the single highest-leverage change available to most programmes, and it costs staff time rather than money.

2. Build a reason to return that isn’t the activity itself. A walking group that is only about walking loses people once walking stops being novel. A walking group where three participants have started arranging to meet fifteen minutes early for coffee has built something that survives the activity being ordinary. Eric Klinenberg’s argument about social infrastructure is relevant here: physical spaces that people return to repeatedly, not single events, are what generate durable contact. Pick venues — a café, a library room, a park with a regular bench — that support informal extension of the session, not just the session itself.

3. Tell participants explicitly that week four is normal. Naming the dip removes some of its power. A participant who has been told “most people feel like this is pointless around now, and it usually isn’t” is more likely to push through than one who assumes their own boredom means the programme has failed them specifically.

4. Measure attendance by session number, not by week or month. If your monitoring only tracks overall retention rate at three months, you cannot see the cliff — you only see the average after it. Track session-by-session drop-off and you will find your own threshold, which may not be exactly session four but will be a specific, identifiable point.

5. Treat the group’s self-sustaining capacity as the actual outcome, not a bonus. A programme that has produced a group capable of continuing without staff support after twelve weeks has done something durable. A programme that keeps people attending only as long as staff keep prompting them has produced attendance, not connection.

What this does not solve

None of the above addresses who never gets referred in the first place, which is a bigger problem than retention and a different one. Social prescribing and similar programmes reach people who are already in contact with a referring service — a GP, a link worker, a community organisation. People who are isolated enough to have no such contact point are invisible to this entire design conversation. The systematic reviews on social prescribing outcomes are themselves built on studies of people who engaged enough to be measured; they say very little about the people who were referred and never showed up once. A programme that solves the week four problem perfectly will still only be solving it for the subset of lonely people who made it to week one.