How to Report Attrition Without Hiding the Failure Rate

Most reports on a connection programme describe the people who finished it. Few describe the people who started and did not. That second group is usually larger, and it is usually where the honest story of the programme lives.

This matters more for social connection work than for most other interventions, because the outcome being measured — whether someone feels less lonely, whether they show up to the next session, whether they keep the relationship going after the programme ends — is also the thing that predicts whether they stick around to be measured at all. A person who drops out at week three because the group felt awkward and unwelcoming is not a data gap. They are the finding.

Here is a sequence for handling attrition data so it holds up when someone asks the second question.

1. Decide what counts as attrition before you collect a single form

Write down, in advance, what counts as a completer. Is it someone who attended a set number of sessions? Filled in an exit survey? Reached a fixed end date regardless of attendance? If this definition is set after you’ve seen the data, it will drift toward whatever makes the numbers look best. That drift is invisible to the person reading the final report and it is the single easiest place for optimism to enter a self-assessment.

2. Report the denominator on page one, not in an appendix

State how many people enrolled. State how many completed. Do this before any outcome figures, not after. A report that says “participants reported a 40% increase in feeling connected” and only later reveals that 61 of 140 enrolees are represented in that figure is not lying, but it is arranging the truth to be missed. Systematic reviews of social prescribing have repeatedly flagged this kind of gap: the underlying evidence base is thin and heterogeneous partly because so many evaluations report on completers only, which makes it hard to compare programmes or trust the size of an effect.

3. Split your reasons for leaving into categories you can defend

“Dropped out” is doing three different jobs when you leave it as one line. Separate at minimum:

Category	What it tells you
Did not return after session 1–2	Programme did not land on first contact — often a facilitation or fit problem
Left mid-programme, no reason given	Genuine unknown — flag it as such, don’t guess
Left mid-programme, stated reason (health, moved, scheduling)	Structural, not necessarily a programme failure
Completed early by design (e.g., matched, referred onward)	Not attrition at all — recode it

Lumping these together produces a single attrition percentage that means nothing, because a 35% figure driven by scheduling conflicts and a 35% figure driven by people quietly deciding the group wasn’t for them require completely different responses.

4. Check whether completers differ from non-completers at baseline

If the people who finished were less lonely, more socially connected, or in better health to begin with, your end-of-programme outcomes describe a healthier subgroup, not the programme’s effect on the population you set out to serve. This is precisely the distinction randomised designs are built to protect against — the Lancet Healthy Longevity trial of volunteering among lonely older adults in Hong Kong is worth citing here specifically because it is one of the few loneliness interventions tested this way, rather than reported as a before-and-after on whoever stayed. Most programme evaluations cannot randomise. That’s a real limitation, not a reason to skip the comparison — run it on whatever baseline data you did collect and report what you find, including “no meaningful difference,” which is itself useful.

5. Say what you are not claiming

If your outcome data only covers completers, say so explicitly next to the headline number, not in a methods footnote. “These results describe the 61 participants who completed the programme, not the 140 who enrolled” is one sentence. It costs you a little shine and buys you credibility with anyone who has read more than one evaluation before.

What this means in practice: if your completion rate is under roughly 60%, lead your report with the attrition breakdown, not the outcome scores. A funder who finds the dropout rate buried on page nine will not trust the number on page one, however real it is.

The vanishing denominator

Name the failure mode so your team recognises it happening in real time: the vanishing denominator is what occurs when the enrolment number quietly disappears between the first slide of a report and the last. It’s rarely deliberate. It happens because completers are the people still in the room when someone finally sits down to write the evaluation, and non-completers, by definition, are not there to be counted. Qualitative work on social prescribing has found that what participants value most is structured, purposeful engagement — which means the people who left because the programme wasn’t that are telling you something a satisfaction score from stayers cannot.

What this does not solve

Honest attrition reporting tells you the programme reached the people who stuck with it and roughly who fell away and possibly why. It does not tell you about the larger group who never enrolled at all — the people social prescribing and community programmes structurally struggle to reach in the first place. That is a recruitment problem, not a reporting one, and no amount of careful denominator work fixes it.