For every 10,000 people who get an appointment with their GP, 167 fail to secure one and then seek treatment at A&E instead.
That's what researchers at Imperial College London calculated from answers to questions in the GP Patient Survey.
It sounds like, and it is, a tiny proportion of patients. But because there are more than 345 million GP appointments a year, the researchers conclude that annually 5.7 million people attend A&E after failing to get an appointment.
And that's 26% of all unplanned attendances at A&E. Which is a huge proportion, if they're unnecessary visits.
But that is a big "if".
That's because surveys of people who actually have attended A&E paint a dififerent picture. Research by the College of Emergency Medicine, which examined the case notes of attendees, concluded only 15% of people should have gone to their GP.
That's 2.1 million patients - still a huge number to deal with. But a lot less than 5.7 million.
Why the discrepancy? Different statistical estimation methods, for starters. One has worked back from the A&E end of the question, one has projected forward from the GP surgery experience.
But the interesting idea it raises in my mind is.... what if it actually suggests that people don't flock to A&E without good reason.
(I'm going to play very fast and loose with the figures here and subtract one set of findings from the other.)
You could argue it shows 3.6 million very sick people who needed hospital treatment attempted to see their GP first. When they failed to get an appointment they went to A&E, and then discovered it was the place they needed to be anyway.
That doesn't fit the current "narrative" that our A&Es are swamped by minor problems. What it would suggest is that many people are behaving very cautiously, trying to see their GP even when they're quite sick.
As I say, I've played fast and loose with the stats. But if you've got a better explanation for the discrepancy, get in touch.