Report the Dropout, Not Just the Completers

A link worker refers 100 people to a walking group over a quarter. Sixty-two attend a first session. Thirty-one are still coming at week eight. The end-of-year report says: “Participants in the walking group programme showed a 40% improvement in loneliness scores.” That sentence is true only of the 31. It is silent about the other 69, and it is exactly the sentence a funder will remember.

This is not fraud. It is the default output of most outcome tracking software, which measures whoever fills in the second questionnaire. But it is a decision, not a fact of nature, and it changes what the number means. If you report it without saying so, you are not describing your programme. You are describing the subset of it that worked well enough for people to stick around and answer a survey.

Why attrition data gets buried

Nobody sits down and decides to hide dropout. It disappears for three ordinary reasons.

First, follow-up surveys only reach people who are still engaged. Someone who stopped attending after two sessions rarely completes an exit questionnaire, so their outcome — good, bad, or neutral — is simply missing from the dataset, not recorded as zero or negative.

Second, commissioning templates often ask for “outcomes achieved,” a phrase that invites you to report on achievers. Nobody asks you to report on people who didn’t achieve anything, so that number never gets built into the spreadsheet in the first place.

Third, and most human: a report full of caveats is harder to write and harder to fund from. There is real pressure, especially near a renewal decision, to lead with the number that looks like proof of value.

The systematic review evidence on social prescribing gives you cover for none of this. The 2021 review in the International Journal of Environmental Research and Public Health found real gains in self-esteem and confidence across the literature, but flagged limited trial evidence and heavy heterogeneity between programmes — meaning the underlying studies themselves are not a clean, comparable body of proof. A separate systematic review in Perspectives in Public Health found all nine included studies reported positive individual impacts, with three showing reduced service use. That is a genuinely encouraging signal. It is also nine studies, not ninety, and a 2025 review protocol notes that the effectiveness of social prescribing for older adults remains unclear, with only one peer-reviewed randomised controlled trial in the area. The base you’re building your local claims on is thinner than the confident tone of most annual reports suggests.

The completers-only report

Call this failure mode what it is: the completers-only report. It has three symptoms, and you can check your own last report against them in about five minutes.

1. The denominator in your headline outcome statistic is the number who completed a follow-up measure, not the number referred or the number who attended a first session.
2. There is no attrition figure anywhere in the document — no funnel from referral to attendance to completion.
3. Non-completion is treated as data loss rather than as a finding. Nobody asked why 69 of 100 people didn’t finish, because the report doesn’t acknowledge they existed.

The fix is not complicated. It is uncomfortable, because it puts a worse-looking number on the same page as your best one.

What an honest attrition section looks like

Report the whole funnel, not the peak.

Stage	Count	Notes
Referred	100
Attended first session	62	38% never attended
Still engaged at week 8	31	50% of attenders dropped by week 8
Completed follow-up measure	28	Outcome data available for this group only

Then state the outcome finding against the group it actually describes: “Among the 28 participants who completed both baseline and follow-up measures, loneliness scores improved by X.” Not “participants.” Not “40%.” The 28 who completed, out of the 100 who were referred.

Where you can, say something about the 72 who didn’t. Even a partial breakdown — how many declined at referral, how many attended once and stopped, how many were still notionally enrolled but unreachable — turns a missing-data problem into a finding about where the programme loses people. A qualitative meta-synthesis on social prescribing found that participants describe the benefit of these programmes as going beyond contact itself, toward restored purpose and meaningful participation — which suggests dropout early on may often mean the activity didn’t supply that, not that the person was simply unreliable. That is a design question, not a data-cleaning inconvenience.

What this means in practice: never publish an outcome percentage without its denominator sitting next to it, and never let “completers” silently stand in for “participants.” If your report has one number a funder will remember, make sure that number carries its own asterisk in the same sentence, not in a footnote.

Where the clinical literature is stricter than social prescribing reporting

Health systems already have a version of this discipline, and it is worth borrowing. The clinician-facing commentary on the National Academies’ 2020 report on isolation in older adults argues for routine, structured assessment rather than ad hoc measurement precisely because informal tracking lets exactly this kind of selection bias creep in unnoticed. A clinical trial that lost 70% of its sample to follow-up would be reported as having lost 70% to follow-up, and reviewers would treat the remaining result with real caution. A social prescribing programme that loses 70% of referrals rarely reports the loss at all, let alone treats its remaining result with equivalent caution.

You don’t need a trial’s rigour to borrow its honesty. Three habits get you most of the way there:

• Report the denominator every time you report an outcome, in the same sentence, not a table two pages later.
• Track attrition by stage (referral, first attendance, mid-programme, completion) so you can say where people leave, not just that they leave.
• Distinguish people who left because the programme didn’t fit their situation from people who left because it didn’t work for them. Those are different problems and need different fixes.

What this does not solve

Honest attrition reporting tells a funder what your numbers actually cover. It does not make the underlying evidence base for social prescribing stronger than it is — the review literature is still thin, heterogeneous, and short on randomised comparisons, and no amount of careful denominators changes that. It also does not reach the people your attrition data is about: the ones who never engaged at all, or who left in week two. Reporting on them accurately is not the same as reaching them. That is a programme design problem, and a report, however honestly written, cannot fix it from the outside.