Reporting Attrition Honestly When You Write Up a Connection Programme

A programme runs twelve weekly sessions. Forty people start. Eighteen finish. The report says: “participants showed a 34% improvement in loneliness scores.” True, as far as it goes. It is also the single most misleading sentence you can put in front of a commissioner, because it describes the eighteen who stayed and says nothing about the twenty-two who didn’t.

This is not a hypothetical failure. It is the default way attrition gets handled in social connection programme reporting, because the alternative — reporting on everyone who enrolled — usually produces a worse-looking number, and nobody wants to hand a funder a worse-looking number voluntarily.

Why the average hides the story

When you report an outcome only for completers, you are running an unstated selection filter. The people who finish twelve weeks of a group programme are, almost by definition, the people for whom it was already working well enough to keep showing up. Reporting their average change tells you about the programme’s effect on people it was already succeeding with. It tells you nothing about the people it lost.

Systematic reviews of social prescribing consistently flag this as a structural weakness in the evidence base, not a one-off reporting lapse. Reviews of both wellbeing outcomes and loneliness outcomes note heterogeneous designs and thin follow-up data, which is a polite way of saying that most published evaluations are built on whoever was still there at the end.

There’s a specific failure mode worth naming: the survivors’ average. It’s what you get when a report presents pre/post scores for completers only, with the dropout number mentioned once, in passing, in a methods paragraph nobody reads. The number is real. The claim built on it is not.

What honest attrition reporting actually requires

1. Report the denominator at every stage. Enrolled, attended session one, attended the midpoint, completed. Four numbers, not one.
2. State the completion rate as a headline figure, not a footnote. If 45% finished, that goes in the same paragraph as the outcome data, not three pages later.
3. Report outcomes for completers and enrollees separately, and label them as such. “Among those who completed the programme” and “among all who enrolled” are different claims. Say which one you’re making every time you make it.
4. Say what you know about why people left. Even a rough breakdown — moved, scheduling conflict, no longer wanted to attend, unreachable — is more useful than silence. If you didn’t collect this, say that you didn’t, rather than letting the gap disappear.
5. Don’t impute a positive outcome for dropouts. Carrying forward a baseline score, or assuming no change, is a common workaround but it manufactures data you don’t have. State the missing data as missing.
6. If the sample is small, say so in the same sentence as the finding. A 34% improvement among eighteen people is a different kind of claim than 34% among four hundred.

None of this is exotic statistical practice. It is closer to plain disclosure. The reason it doesn’t happen by default is organisational, not technical — attrition looks like programme failure, and programmes are usually reported by the people who ran them.

Why attrition itself is informative, not just an embarrassment

Dropout in a connection programme is rarely random noise to be apologised for. It often tracks the same variables that predict loneliness in the first place. The American Enterprise Institute’s 2021 survey found the share of American men reporting no close friends had risen fivefold since 1990, and network size and diversity are among the strongest predictors of loneliness identified in the broader literature. People with thinner networks and less practice sustaining regular contact are plausibly also more likely to stop attending a weekly group. If that’s true, your dropout group is not a random subtraction from your sample — it is disproportionately the population the programme was aimed at in the first place.

Qualitative work on social prescribing points in a similar direction: participants describe benefit as coming from restored meaningful participation and purpose, not contact alone, and structured, purposeful activity appears to matter more than simply showing up. If a programme isn’t delivering that structure early, quitting is a rational response to a mismatch, not a personal failing on the participant’s part. A report that treats all dropout as identical, unexplained noise misses this entirely.

What this means in practice: before you write the report, pull attendance data by session number and look for a cliff, not a slope. A steady trickle of dropout across twelve weeks is a different problem than losing a third of the group between session three and four. The second pattern tells you something specific went wrong early — content, facilitation, room, timing — and it is fixable in a way that “some people naturally drop off” is not.

The evidence-status table for what you can actually claim

Claim	Evidence status
Social prescribing improves self-esteem and confidence for those who stay engaged	Reasonably supported across multiple reviews
Social prescribing reduces loneliness for completers	Supported by small, heterogeneous study samples
Social prescribing changes outcomes for the full enrolled population, including dropouts	Not established — this is rarely measured
Structured, purposeful activity outperforms unstructured contact	Supported by qualitative synthesis, not by controlled comparison
Dropout patterns can be used to diagnose programme design faults	Plausible and consistent with related survey data, but not directly tested

What this does not solve

Honest attrition reporting will not make a leaky programme look good, and it should not. It also does not fix the underlying reach problem: even a programme reported with full rigour on completers and dropouts alike is still describing the people who enrolled in the first place, not the substantially larger group who never approached a link worker, GP, or community organisation at all. Reporting who left tells a funder something true about programme design. It tells you nothing about who never showed up to leave.