Thread by @_johnbye

The govt has quietly admitted that some people tested by private labs were double counted, after days of blaming "technical issues" for a lack of data on number of people tested in the @DHSCgovuk daily statistics.

This seems to come after a slap on the wrists from @UKStatsAuth.

May 19: stats for the number of people tested are missing from the daily update.

"Data for the number of people tested is unavailable for pillar 1 – daily and cumulative figures are therefore not reported today. We expect to resume reporting tomorrow".

web.archive.org/web/2020051919…

May 20: the data returns "following technical issues, and a transfer of cases from pillar 1 to pillar 2". The various corrections suggest the total number of people tested to date may have been over counted by several thousand...

web.archive.org/web/2020052103…

This is the key bit.

Some cases have been transferred FROM pillar 1 (NHS) TO pillar 2 (private) labs

Number of pillar 2 tests done has been adjusted UP by 17,517

BUT

Number of people tested in pillar 2 has been adjusted DOWN by 3,132

They've been over counting people tested!

(By way of explanation, if the daily figure is LOWER than the difference between the running totals for today and yesterday, that means previous figures were adjusted UPWARDS, as more people have been added to the total than the daily number would account for. And vice versa.)

The latest "time series" spreadsheet (which includes previously missing data) says 17,792 people were tested in pillar 1 on the missing day.

But the correction says only 13,955 were added to the pillar 1 total.

This means 3,837 people were moved from pillar 1 to pillar 2.

It gets worse.

934,168 people tested up to the 18th + 16,287 tested on the 20th + 13,955 from the stated correction = 964,410

The total given is actually 968,247

This is what it would be if the 3,837 people that were moved to pillar 2 weren't deducted from pillar 1. 🤔

So it now looks like:

1) 6,969 people (3,132 + 3,837) have been over counted in pillar 1.

AND

2) 3,837 people are now being counted in BOTH pillars!

Either that or they DIDN'T correct the people data, despite saying that they DID, and "only" over counted pillar 2 by 3,132.

May 23: the number of people tested disappears again, for both private and NHS labs this time.

"Due to technical difficulties with data collection we cannot provide people tested figures today".

A week later the full numbers STILL aren't available.

web.archive.org/web/2020052415…

May 24: pillar 1 numbers return but pillar 2 is still missing.

"Due to technical difficulties with pillar 2 data collection we cannot provide people tested figures today. One laboratory has been unable to report their data due to technical issues."

web.archive.org/web/2020052514…

Interestingly, for the first time they also admit "there may be a small percentage of cases (around 2%) where the same person has had more than 1 positive test result for coronavirus. The government will correct any figures if they have subsequently been found to have an error."

May 26: still no pillar 2 data, supposedly due to "technical difficulties", despite the issue from the 24th (IT problems in Milton Keynes) being fixed now. They've also only just added data from a lab in Cambridge which has been running since May 11th!

web.archive.org/web/2020052614…

May 27: @DHSCgovuk confirm some people have "had multiple tests or tested positive more than once [in pillar 2]... We are ensuring that the way these people are counted in a rolling cumulative total is consistent with pillar 1."

web.archive.org/web/2020052800…

May 27: despite the clarification today from @DHSCgovuk, the PM's spokesman says he "hadn't noticed" the data has been missing all week and can't explain it.

Maybe I can help him... 🙄

independent.co.uk/news/uk/politi…

May 27: after a sharply worded letter from the @UKStatsAuth on May 11, @MattHancock finally replies and notes "we are reconciling our approach to reporting across the different testing pillars to ensure consistency". This explains why data was pulled now.

statisticsauthority.gov.uk/correspondence…

May 28: there's a new disclaimer - "Reporting on the number of people tested has been temporarily paused to ensure consistent reporting across all pillars."

In other words some people tested in private labs are STILL double counted. NOT technical issues.

web.archive.org/web/2020052901…

May 29: bizarrely, even though the @DHSCgovuk site explains exactly why data isn't available, @BBCNews don't know and neither does NHS England's medical director when they ask him. He replies "I can't give you a specific answer on the testing numbers". 🤔

https://twitter.com/BBCNews/status/1266410029911085057

May 29: new problem - someone at @DHSCgovuk can't count. Pillar 1 stats say 27,719 people were tested yesterday, but the total only went up by 21,719. One of these numbers is out by 6,000. No mention of this in the notes so it seems to be a genuine error.

web.archive.org/web/2020052921…

So that's the story so far.

There's a problem with the way the government collates data from private labs. About 2% of people have been tested multiple times and counted as a new person each time, unlike for NHS labs.

Also, there seem to be glaring errors in the data published.

All of this raises concerns, as the tracing system relies on fast and accurate testing. If the government can't even tell us how many people have been tested with any degree of confidence, can we rely on this data?

And why did they blame "technical difficulties" until Thursday?

Update: The 21,719 figure seems to be the correct one, based on the data in the spreadsheet. It looks like the website data that says 27,719 is a typo (or wishful thinking). Doesn't exactly inspire confidence.