Reporting Attrition Honestly in Social Prescribing Evaluations

A social prescribing programme enrols 120 people. It reports outcome data on 74. The report does not say what happened to the other 46. This is the most common integrity problem in connection-programme evaluation, and it is almost never dishonest in intent — it is just easier to report on the people who finished than to explain the people who did not.

That gap matters more here than in most health interventions, because the thing being measured — loneliness, isolation, social confidence — is also the thing that predicts who drops out. People who are more isolated are less likely to have the stable routines, transport, or social support that make it easy to keep attending. If your endline sample is systematically less isolated than your baseline sample, your outcome numbers are measuring who stayed, not what the programme did.

What counts as attrition, and why the category matters

Not all dropout is the same, and lumping it together hides the story a funder actually needs.

Category	What it looks like	What it implies about the programme
Disengagement	Stops attending, no contact	May reflect poor fit, logistics, or the intervention not working for that person
Resolved need	Stops attending because goal met	A success that looks identical to disengagement in raw numbers
Life disruption	Illness, bereavement, house move, caring responsibility	Unrelated to programme quality, but still lost data
Administrative loss	Contact details lapse, no follow-up capacity	A measurement failure, not a participant one
Death	Applies disproportionately in older cohorts	Should never be silently folded into “lost to follow-up”

The National Academies’ 2020 consensus report on isolation in older adults is a reminder of why the death category needs separating out on its own line: roughly a quarter of adults over 65 are considered socially isolated, and isolation itself is a mortality risk factor. A programme working with an older, isolated cohort will lose participants to death at a rate that has nothing to do with intervention quality. Reporting that figure alongside disengagement, rather than inside it, is not a cosmetic choice.

The week four problem

Across social prescribing evaluations, the drop tends to cluster early — often within the first four to six weeks, before a participant has had enough contact with a group or activity to form the sense of belonging that keeps people coming back. Call this the week four problem: the point at which the people most likely to benefit from sustained connection are also the people most likely to have already left.

A meta-synthesis of qualitative social prescribing research found that participants who stayed described benefits going beyond social contact — restored purpose, meaningful participation, a reason to leave the house — and that structured, purposeful activity mattered more than contact alone. That is a finding about the people who stayed long enough to say so. It tells you almost nothing about the people who left in week two, and it should not be reported as though it does.

What this means in practice: report attrition by week, not just as a single end-of-programme percentage. A flat 38% dropout rate looks manageable. The same 38% concentrated entirely in weeks one to four tells a commissioner the programme has an induction problem, not a general retention problem — and that’s a fixable, specific thing to fund.

How to report it without either hiding it or overclaiming from it

1. State the denominator on every outcome, every time. “Loneliness scores improved among 74 of 120 enrolled participants” is a materially different sentence from “loneliness scores improved.”
2. Break attrition down by category, using something close to the table above, even if some categories are estimates from case notes rather than clean data.
3. Report attrition by week or month, not as a single aggregate figure, so a reader can see where the loss concentrates.
4. Compare baseline characteristics of completers and non-completers where you have any baseline data at all — loneliness score, isolation indicators, age, referral source. If completers were less isolated at baseline, say so plainly.
5. Do not extrapolate outcome effects to non-completers. A common and quietly dishonest move is to write “the programme reduced loneliness” when what the data supports is “the programme was associated with reduced loneliness scores among those who completed at least eight sessions.” The second sentence is less flattering and more true.
6. Name what you don’t know. If administrative loss means you cannot say whether someone disengaged or moved house, write that. A funder trusts a report more, not less, for admitting a gap.

Evidence status of the claims underneath this

Claim	Evidence status
Social prescribing produces self-reported wellbeing gains among participants who engage	Reasonably supported across several systematic reviews, though trial evidence is thin
Structured, purposeful activity outperforms unstructured contact	Supported by qualitative synthesis; not tested experimentally
Isolated people are more likely to disengage early	Plausible and consistent with isolation research generally; rarely measured directly within programme evaluations
Specific dropout percentages generalise across programme types	Not supported — attrition reporting is too inconsistent across the sector to benchmark against

A 2021 systematic review of social prescribing and loneliness found that all nine included studies reported positive individual impacts — a figure worth treating with caution precisely because studies with heavy, unexamined attrition are more likely to publish a positive result among the survivors.

What this does not solve

Honest attrition reporting will not make a weak programme look strong, and it will not by itself fix the week four problem — that requires changing induction, not measurement. It also cannot substitute for baseline data you never collected. If a programme has no record of who dropped out or why, no amount of careful reporting after the fact recovers that information; the fix has to happen at intake, before the first session, not in the write-up.