How to Report Attrition Without Burying It or Weaponising It

You have a referral list of 140 people, an end-of-programme attendance list of 61, and a funder meeting in ten days. The question is not whether to mention the other 79. It is how.

Most programme reports handle this one of two ways, and both are dishonest in different directions. The first buries attrition in a methods appendix, reports outcomes only for programme completers, and lets the headline numbers imply the whole referral group improved. The second panics, leads with the dropout rate, and lets a funder conclude the programme doesn’t work when the truth is closer to: it works for the people who stay in it, and you don’t yet know why the others left.

Neither serves you. The first will eventually get caught — a commissioner who cross-references referral numbers against completion numbers, or a second-year evaluation that asks why last year’s cohort size doesn’t match this year’s baseline. The second cedes ground you haven’t lost. Social prescribing’s own evidence base is honest about the fact that outcome data mostly comes from people who engaged; that’s a limitation of the field, not a flaw unique to your project.

Why attrition gets hidden

Programme reports are usually written by the people who ran the programme, for an audience deciding whether to fund it again. That’s a structural incentive to report the version that looks best. Nobody sets out to mislead; they set out to make the strongest honest case, and somewhere in drafting, “honest” gets quietly dropped from the phrase.

The systematic reviews of social prescribing are candid about this at the field level. A 2021 review found that the evidence base does show benefits in self-esteem and confidence, but flagged heterogeneity and thin trial evidence across the studies it drew on. A separate review of social prescribing’s effect specifically on loneliness found that all nine included studies reported positive impacts for the people in them — a result that reads less impressively once you notice these are, almost without exception, evaluations of people who completed the programme, not of everyone referred to it. A 2025 protocol assessing social prescribing for isolation and loneliness in older adults goes further and notes that, despite wide adoption, only one peer-reviewed randomised controlled trial exists in the entire area. If the underlying research mostly measures completers, your local report inherits that bias unless you correct for it.

The completer-bias failure mode

Call it what it is: completer bias. It happens when a report calculates outcome improvement only across the group that finished the programme, then presents that figure as the programme’s effect — without stating, in the same sentence, what share of referrals that completer group represents.

The number is not fabricated. The framing is. “Participants who completed the twelve-week course showed a two-point improvement on the loneliness scale” is a true and useful sentence. “The programme reduced loneliness by two points” is the same finding wearing a bigger claim than it earns, because it silently drops the people who left before week four, week eight, or after one session and never came back.

A structured, purposeful group activity — which is roughly what most social prescribing referrals turn into — appears to produce more benefit than contact alone, according to qualitative synthesis work on what participants themselves report valuing. That’s a genuine finding. But it describes what happens to people who stay engaged with structured activity long enough to experience it. It tells you nothing about the people who never got that far, and a completer-only report implicitly claims it does.

What an honest attrition section looks like

1. Report the full funnel, not just the endpoint. Referred, contacted, attended once, attended the threshold number of sessions your model considers “dosed,” completed. Five numbers, not one.
2. State outcomes against the group they actually describe. “Among the 61 who completed” is not the same claim as “among the 140 referred,” and the report should never let a reader conflate them.
3. Say what you don’t know about the dropouts. If you didn’t collect exit reasons, say so plainly rather than speculating. If you did, report the categories even when they’re unflattering — transport, timing, the group not being what they expected.
4. Compare your attrition rate to something. A raw dropout figure means little without context. If you can’t find a directly comparable published attrition rate for a similar programme, say that too, rather than implying your number sits within some established normal range it may not sit within.
5. Separate isolation-reduction claims from loneliness-reduction claims. These are different constructs with different completer patterns, and the difference matters more than most local reports treat it as mattering.

What this means in practice: if your report states an outcome number, the sentence next to it should state the denominator. Not in a footnote — in the same sentence, or the one immediately after. A funder who has to hunt for your completion rate will assume you were hoping they wouldn’t.

The trials that exist show why this matters

The scarcity of controlled trials in this area makes the few that exist worth reading closely, because they show what happens when someone actually tracks a comparison group rather than reporting on completers alone. A dual randomised controlled trial of volunteering as a loneliness intervention among older adults in Hong Kong is one of the few controlled tests in the literature, precisely because most of what exists is uncontrolled programme evaluation. A 2025 trial of befriending in residential aged care found real, measurable reductions in loneliness scores at eight and sixteen weeks against a control group — a result that holds up because it was measured against people who didn’t get the intervention, not just against the people who received it and stuck around.

That contrast is the whole argument of this note. Controlled comparison tells you what the intervention does. Completer-only reporting tells you what happens to the subset of people for whom it worked well enough to keep showing up. Both numbers can be true. Only one of them answers the question a funder is actually asking.

Evidence-status table for common attrition claims

Claim	Evidence status
“Social prescribing improves self-esteem and confidence in participants who engage with it”	Supported by systematic review, but evidence is heterogeneous and mostly non-controlled
“Social prescribing reduces loneliness”	Positive in all included studies of one systematic review, but those studies measure engaged participants, not full referral cohorts
“Structured activity works better than informal contact”	Supported by qualitative synthesis of participant accounts, not by controlled outcome trials
“Social prescribing reduces isolation in older adults specifically”	Unclear — a 2025 protocol notes only one peer-reviewed RCT exists in this area
“Befriending reduces loneliness in aged care”	Supported by a 2025 randomised controlled trial with a measured effect size at two time points
“Volunteering-based programmes reduce loneliness in older adults”	Supported by a randomised controlled design, one of few in the field

What this does not solve

None of this fixes attrition. Reporting it honestly doesn’t make the 79 who dropped out come back, and it won’t make a thin evidence base thicker. What it does is protect the credibility of the report that says the programme worked for the people it reached — a real and defensible claim, but a narrower one than most programme reports currently make. It also does nothing for the reach problem underneath all of this: social prescribing, like most connection interventions, mostly succeeds with people who were already willing to walk through the door. The honest version of your report should say that plainly, rather than let a completion-rate footnote quietly do the work of hiding it.