Earlier today the @UKStatsAuth gave @MattHancock a rap on the knuckles over the testing statistics produced each day by the @DHSCgovuk. In an effort to try to get my head round the numbers I took a deep dive into the spreadsheets. I ended up with a bit of a headache (thread):
This is a mess. A real mess. For all sorts of obvious reasons, testing is incredibly important when it comes to saving lives, understanding the spread of the disease, stopping a second wave, track & trace etc. Yet our data on this is not just poor: it's actively misleading.
Let's start with the overall numbers the government produces each day. You've prob seen charts like this. Look! We're WAY past the government's 100k target. Great news, right! Esp if you care about arbitrary targets! 
But the overall fig includes many types of tests, some of which are clinically led, some research, some never even happened. Start with "pillar 1" - mostly tests on hospital patients/health workers. These are "gold standard" tests & they've been stable since early May
Then you've got pillar 2: this is clinical tests for everyone else: drive-in tests, tests for key workers. Combine the two and you're still not up to 100k, meaningless as that target may or may not be.
Then you've got tests that are mailed out to people under pillar 2. There are lots of these. But no-one in government seems to have a clue how many of them have actually been carried out and returned. Either way, this is the main bulk of the increase in recent weeks
Then (still following?) on top of that you've got another couple of pillars: pillar 3 which is those serology antibody tests which may or may not be reliable, and pillar 4 which includes other testing for academic research, the ONS etc. Note this last bit isn't clinically led
The problem with this is that it's a very broad church: every test either carried out or mailed out. Some of those tests never even happened. Sometimes patients needed to be tested more than once. Yet these charts give us no indication of the number who actually have the disease
We do know how many people have been tested under those pillar 1 "gold standard" tests in hospitals. These peaked in early May and have been in around the same ballpark since then. But here is where things get fishy...
Up until a couple of weeks ago @DHSCgovuk was publishing figs that suggested that far more people had been tested than that. Then they abruptly stopped producing these other figs and have removed them from their website (I had to trawl through the archives to produce this chart)
This is v weird and, in the wonky world of statistics, v irregular. Here's the official @DHSCgovuk response. Basically: the numbers were dodgy so they've been withdrawn. But it's now been two weeks and we don't have a clue abt one of the most important datapoints in this pandemic
Now in fairness, govt is up against it. And we're not the only country who haven't got decent figs on the no of people tested. But some are. Look at Singapore or Hong Kong or Belgium where there's far more data on the disease, its spread, its incidence epistat.wiv-isp.be/covid/
This stuff matters. Where did people catch the disease? At home? At hospital? In the community? At work? Unlike many other countries we have no statistics publically available on this. NHS/PHE may have such stats but if so they don't publish them. Unlike in, eg, Belgium or HK
There's a paradox here. Deaths statistics published each week by the @ONS are world class. But the other stuff we get from @DHSCgovuk is closer to banana republic standard. That we don't know how many people have been tested months after this epidemic began is pretty astounding
Here's a vid of me trying to get my head (and, hopefully, yours) round what on earth is going on with the UK's testing data. Bit of an experiment so excuse rough edges. Let me know if it's helpful. If so I'll try to do other such deep dives in future
