Everyone quotes "37% fewer GP visits". Where does that number actually come from?

#1Marion K.(Joined Jul 2024 · 38 posts)June 15, 2026, 11:24 am

House rule, so here I am asking.

I sit on the committee of a very small charity, we are trying to keep a weekly creative group going, and I have now seen the same set of numbers three times in a fortnight. Local paper: 37 percent fewer GP visits. A slide at a partnership meeting: 37 percent fewer GP appointments and 27 percent fewer hospital admissions. And a template our funder sent round, which cheerfully suggests we say that arts on prescription saves £216 per patient.

Nobody at any of the three could tell me where the numbers came from. The man with the slide said "it's the APPG report" in a tone that closed the subject.

I am not trying to be difficult, I am trying to write a bid I can defend if somebody asks me a question at the interview. Does anybody actually know what study this is?

#2mikeatthepiano(Joined Feb 2025 · 19 posts)June 15, 2026, 2:02 pm

I have heard the £216 twice this year and both times the person saying it thought it was a national figure.

#3Yusuf A.(Joined Aug 2025 · 11 posts)June 18, 2026, 10:47 pm

Right. I have spent an evening on this because I write bids too and I have quoted at least one of these numbers in the past, which I am not proud of.

All three trace to the same place: a cost benefit evaluation of Artlift in Gloucestershire, written by Dr Simon Opher, dated 9 December 2011, covering 2009 to 2012. It is a PDF on Artlift's own website and it is about as long as a decent essay. Everything downstream is quoting it, usually through two or three intermediaries.

Start with what it is, in its own words: "a simple observational study". Before and after, no control group, and the analysis rests on 90 patients out of roughly 500 who were referred.

Now the three numbers, in order.

The 37 percent. The report's consultation figures are 1,014 in the year before, 454 in the first six months after, and 320 in the second six months. Its own sentence is that six months after seeing an artist, consultation rates dropped by 37 percent. So the 37 is the months 7 to 12 figure. Two lines away the report prints its own full year number: "Percentage reduction = 24%". The 37 is real and it is in the document. It is simply not the year, and the year is the thing everybody thinks they are quoting.

The 27 percent. This one is a straightforward substitution and it is the one I would most want your man with the slide to see. Spend was £157,473 before and £115,050 after, which the report describes as a 27 percent reduction in overall spend. That is NHS spend, not admissions. Admissions appear separately, as 54 before and 33 after, with no percentage attached to them and no significance test. "27 percent fewer admissions" does not appear in the report at all. Somebody put the spend percentage next to the admissions row at some point and it has been travelling ever since.

The £216, which is the one I had assumed was a headline finding. It is a subtraction the report never actually prints. Working from its own cells: 500 patients, 2.7 fewer GP consultations each at £39 a consultation, which is £105 per patient of GP time. Then a line reading "Reduction observed in Healthcare spend in 90 patients = £42,423", which divided by 90 gives £471 per patient. £105 plus £471 is £576. Artlift cost £180,000 across 500 patients, so £360 per patient. £576 minus £360 is £216.

The arithmetic is correct. The problem is that the £471 leg is a per patient figure computed on 90 people and the £360 is computed on 500, so the numerator and the denominator describe different populations.

And the sentence I would put at the top of anything quoting any of it, which is Opher's own: the study "does not imply causality".

#4Dr Rhian VaughanMusic therapist moderator(Joined May 2024 · 91 posts)June 22, 2026, 9:36 am

That reading is correct, and I have gone back to the report to check it rather than take it on trust, which is what the house rule is for.

Two things I would add, from the methods side rather than the arithmetic side. I am a music therapist and not a health economist, so treat me as somebody who can read a study design rather than somebody who can price a service.

The first is what an uncontrolled before and after cannot rule out, and there are four candidates here rather than one. Time passing, and a lot of what people are referred for improves on its own. Regression to the mean, which is the big one: people tend to be referred at the point when things are worst, and consultation rates are highest at exactly that point, so the following six months will look better whatever happens in them. Selection, because the 90 who were analysed are not a random slice of the 500. And attrition, since people who dropped out are not in the after column. All four push the same flattering direction. That is why the design cannot support a causal claim, and it is why Opher said so himself.

The second is what happens if you take the 410 people who were not analysed seriously. Nothing in the report tells you what they cost or what they saved. If they saved nothing, the scheme costs more than it saves. Nothing rules that out, and nothing rules it in either.

The wider position, so this does not read as a hatchet job on one small evaluation from 2011. A systematic review published on 9 July 2024, by Jensen and colleagues in Frontiers in Public Health, screened 7,805 records and included 25 studies, and states in terms that no randomised controlled trials were identified in the search. Its meta analysis did find a statistically significant improvement in wellbeing. That result rests entirely on uncontrolled before and after designs, so it should never be described as a trial finding, and it is not one.

What that adds up to is a Tier 3 label, and I want to be careful about what a Tier 3 label means, because getting this wrong does as much harm as overclaiming. It means the saving has not been demonstrated. It does not mean the courses are worthless, and it does not mean the people in them are wrong about what they got. Does arts on prescription save money shows the whole trail rather than just the conclusion, and reading arts in health research covers the distinction between a trial and an evaluation, which is the one doing most of the work here.

Marion, for your bid: what you can defend is that participants value the courses, that the wellbeing signals are consistent across a lot of uncontrolled studies, and that no saving has been demonstrated. That is a weaker sentence than the template's and it is one you can still be standing behind at the interview.

#5Marion K.(Joined Jul 2024 · 38 posts)June 24, 2026, 7:19 am

The admissions one has genuinely rattled me, because I said it out loud in a room of about forty people in March.

#6Yusuf A.(Joined Aug 2025 · 11 posts)July 16, 2026, 1:44 pm

Bid went in last week with the honest sentence in it, more or less as Rhian phrased it, and I took all three numbers out.

I also emailed the man with the slide. Politely. He replied within an hour saying he had inherited the deck from a predecessor and had never opened the source, thanked me, and asked for the link, which is a better outcome than I had braced for.

One last thing for anybody arriving here from a search. Every one of these numbers is in a real document written by a real GP who was open about its limits on the page. Nobody fabricated anything. The distortion happened entirely in the retelling, one report quoting another quoting another, and by the fourth hop the caveats have fallen off and only the percentage is left. That is worth remembering before assuming bad faith about whoever quoted it at you.