The Week Four Problem: Where Social Prescribing Referrals Actually Fall Away

A link worker can usually tell you the referral number and the first-visit number. Ask for the fourth-visit number and most cannot produce it, because nobody built the spreadsheet to hold that column. That gap is where a programme’s real attrition lives, and it is worth treating as a specific operational failure rather than a vague truth about “engagement.”

Why the first visit is the wrong thing to measure

Referral-to-attendance is the metric commissioners ask for, so it is the metric that gets tracked. It is also the easiest one to inflate. A person who has been referred by a GP or link worker, handed a leaflet, and possibly walked to the door with someone, will often turn up once out of politeness, curiosity, or a sense of obligation to the person who referred them. That first attendance tells you almost nothing about whether the activity is going to do anything for them.

The systematic review of social prescribing and wellbeing found real gains in self-esteem and confidence across the studies it covered, but it also flagged how heterogeneous and thin the underlying trial evidence is — programmes vary wildly in design, and most evaluations were not built to track anyone past an initial snapshot. A related review of social prescribing and loneliness specifically found that all nine included studies reported positive impacts for the people who stayed engaged. Neither review can tell you what proportion of referred people that “stayed engaged” group actually is, because dropout before any meaningful dose of the intervention is not what these evaluations were designed to see.

That is the problem with practice-note-level decisions, as distinct from headline evidence: the headline evidence is about people who kept coming. The people who did not are invisible in the same data.

Where the drop actually happens

Programme-level attendance registers, where they exist and are looked at honestly, tend to show the same shape: a moderate drop between referral and first attendance, a much larger drop between first and second or third attendance, and then a plateau. The group that survives to session four or five tends to keep coming for a while. The people this guidance is for are the ones who show up once and never again, or who come twice and stop.

Two things are happening in that window, and they call for different responses.

The activity was never going to fit. Someone referred for social isolation gets placed in a walking group, discovers on arrival that the pace or the terrain does not suit them, and does not come back. This is a matching failure, not an engagement failure, and no amount of follow-up phone calls fixes it. The qualitative meta-synthesis on social prescribing and loneliness found that participants describe benefit as extending beyond mere social contact — restored participation, purpose, a reason to structure a day. If the specific activity does not offer any route to that, one visit is enough to find out, and the person is not wrong to stop.

The activity fits, but nobody has told the person their absence would be noticed. This is the more recoverable failure, and it is the one programme design can actually act on. A first session is usually facilitated warmly because it is a first session. By the third or fourth, the novelty has worn off for staff too, and a missed week goes unremarked. For someone whose baseline problem is a thin or shrinking social network — the AARP Foundation’s national survey of adults 45 and older found network size and diversity were the strongest predictors of loneliness, well ahead of demographic factors — an unremarked absence confirms exactly what they already suspected: that their presence or absence does not register to anyone.

What to build instead of a follow-up call

A single phone call after a no-show is better than nothing but is not a system. It treats each drop-off as an isolated event rather than a pattern with a known location.

1. Track attendance by session number, not by month. A calendar-based register hides the shape of dropout. A register organised by “session 1, session 2, session 3” for each participant makes the collapse point visible within a few cohorts.
2. Build a deliberate difference into sessions two and three. If the format is identical to session one, there is no new reason to have come back. A small addition — a named role, a returning face who remembers them, a task with continuity — gives session two a purpose distinct from session one.
3. Assign a specific person to notice absence, not a rota. Diffused responsibility for follow-up means nobody does it. One named contact for the group, whose job includes noticing gaps, closes this.
4. Separate the two failure modes in your own records. Note whether a non-return follows a stated dislike of the activity versus a silent disappearance. These need different responses: rematching in the first case, direct outreach in the second.
5. Report fourth-session retention to your commissioner, not just first-attendance rate. If you do not ask for this number to be tracked, it will not exist, and you will be making programme decisions on the metric least connected to actual benefit.

What this means in practice: if your attendance register is organised by calendar date rather than by session number for each individual, you cannot see the collapse point at all. Rebuild the register first. Everything else in this list depends on being able to see where people stop.

Evidence status

Claim	Status
Social prescribing produces gains in self-esteem and confidence among people who engage	Reasonably supported, though trial evidence is thin and heterogeneous
Structured, purposeful activity works better than contact alone	Supported by qualitative synthesis; not yet quantified
Network size and diversity predict loneliness more than demographics	Supported by a large national survey
A specific “week three or four” collapse point in attendance	Plausible and widely reported anecdotally by practitioners; not established in the published evidence base
Named-contact follow-up reduces attrition	Operational inference, not a tested finding

Be honest with yourself about which row that table you are relying on. The last two are practice logic, not research findings, and should be described that way to a funder.

What this does not solve

None of this touches the people who were never referred in the first place — those without a GP relationship, without a link worker in their area, or without the confidence to show up to something unfamiliar even once. A programme that fixes its retention curve perfectly is still only serving the subset of an isolated population that made it through the door. The National Academies’ 2020 consensus report on isolation in older adults made a related point about health systems generally: most tools address people already inside a service relationship, and say nothing about identifying the roughly one in four older adults who are isolated and never surface to any system at all. Retention work is worth doing. It is not reach work, and funders should not be allowed to conflate the two.