How to Design a Loneliness Programme Evaluation You Can Actually Run

Most social connection programme evaluations fail for a boring reason: they were designed to answer the question a funder might one day ask, not the question the programme can actually generate data for. The result is a logic model with an outcome column nobody measures, a pre/post survey with a 40% completion rate, and a final report that says “participants reported feeling more connected” with no denominator attached.

This is a procedure for avoiding that. It assumes you have a small team, no research staff, and a budget that will not stretch to a control group.

Step 1: Decide what you are actually claiming

Before any instrument gets chosen, write down the single sentence you want to be able to say at the end. Not “the programme improves wellbeing” — that is not a claim, it is a mood. Something closer to: “Participants who attended at least four sessions showed a measurable drop in loneliness scores between intake and week 12.”

This matters because the systematic review evidence on social prescribing is consistent on one point: most existing studies report positive outcomes but cannot say much about mechanism or durability, because they weren’t designed to. A 2021 systematic review in the International Journal of Environmental Research and Public Health found consistent gains in self-esteem and confidence but flagged limited trial evidence and wide heterogeneity across programmes. A separate 2021 review in Perspectives in Public Health found all nine included studies reported positive impacts — which should make you suspicious of publication bias, not reassured. If you don’t fix your claim in advance, you will unconsciously shape your measurement toward whatever result looks good, and so will everyone who reads your report generously.

Step 2: Separate isolation from loneliness, and pick one

These are different constructs with different causes, and conflating them is the single most common design error. Isolation is structural — the number and frequency of social contacts a person has. Loneliness is subjective — the gap between the connection someone has and the connection they want. A programme that increases contact frequency can leave loneliness unchanged if the contact is low quality. A programme aimed at loneliness needs a validated instrument like the UCLA Loneliness Scale, which is what the strongest recent RCTs use; a programme aimed at isolation needs a social network inventory, not a mood scale.

Pick the one your intervention actually targets. A weekly drop-in group targets isolation directly and loneliness only indirectly. A structured befriending or behavioural activation programme targets loneliness directly. Measuring the wrong construct produces a null result that looks like programme failure but is actually a measurement mismatch.

Step 3: Build in a comparison, even a weak one

You do not need a randomised controlled trial. You need something better than a single pre/post measure on the same group, because pre/post alone cannot distinguish programme effect from regression to the mean, seasonal mood shifts, or simple maturation. Options in order of feasibility:

1. Waitlist comparison. People who signed up but haven’t started yet act as a rough comparison group. Cheap, imperfect, still better than nothing.
2. Dose comparison. Compare high-attenders against low-attenders within your own programme. This doesn’t prove causation but it’s a defensible signal, and it’s the comparison social prescribing evaluations use most often given the practical constraints.
3. Historical comparison. Compare this cohort’s change scores against a previous cohort’s, if you have that data.

None of these are publishable in a journal. All of them are more honest than the single-group pre/post that dominates programme evaluation reports.

Step 4: Choose your outcome window deliberately

This is where evaluation design collides with the evidence on what actually works. The HEAL-HOA trial on volunteering among lonely older adults in Hong Kong and its follow-up trial on telephone-delivered behavioural activation both used multi-month follow-up, not just an end-of-programme survey — because loneliness reductions that look real at week eight can fade by month six. A 2025 randomised trial of befriending in aged care found measurable drops in loneliness at 8 and 16 weeks, but that same evidence base shows befriending losing head-to-head to structured psychological approaches at 12 months. If your evaluation only measures at programme end, you cannot tell whether you built something durable or something people liked while it was happening.

What this means in practice: measure at intake, at programme end, and at a follow-up point three to six months later — even a short phone check-in. If you can only afford two measurement points, choose intake and the delayed follow-up, not intake and end-of-programme. The delayed point is the one that tells a funder something real.

Step 5: Name your failure mode in advance

Call it the completion-rate mirage: a programme where only committed, already-improving participants fill out the exit survey, producing an average that flatters the programme and says nothing about the people who dropped out at week three. Fix it by tracking attrition explicitly and reporting it alongside outcomes, not as a footnote.

Evidence status of common claims

Claim	Evidence status
Social prescribing improves self-esteem and confidence	Reported consistently, but from heterogeneous, mostly uncontrolled studies
Befriending reduces loneliness scores	Supported by at least one RCT, effect size modest
Structured psychological interventions outperform befriending	Supported by direct RCT comparison
Social prescribing reduces service use (GP, ED visits)	Reported in some studies, not consistently measured across the field
Social prescribing for older adults works overall	Genuinely unclear — only one peer-reviewed RCT exists in this specific area

What this does not solve

An evaluation designed this way will tell you whether your programme moved a defensible outcome for the people who stayed in it. It will not tell you why most social connection programmes reach people who were already motivated enough to sign up, nor will it capture the people who never enrolled at all — often the most isolated group, and the one hardest to measure by definition.