Thread by @Kit_Yates_Maths

*Number nerd warning ahead*

So last week I posted about this discrepancy between Public Health England's (gov.uk/government/sta…) positivity rates and those given on the covid dashboard (coronavirus.data.gov.uk).

https://twitter.com/Kit_Yates_Maths/status/1355135287270240257

PHE positivity rates seemed to be coming down much faster than COVID dashboard rates at both national and regional level, despite both PHE and the Dashboard purporting to use the same metric (PCR testing) to calculate positivity.

https://twitter.com/Kit_Yates_Maths/status/1355135281805074433

After lots of to-ing and fro-ing with PHE we managed to establish that PHE are now including Lateral flow tests in their positivity figures (despite what it said in the report).

Lateral flow tests have ramped up a lot recently.

Why does this make a difference to positivity?

Well, lateral flows tests are often used to find positive cases in asymptomatic individuals. This means that their rates of positivity are usually a lot lower than those of PCR tests (this is unrelated to the accuracy of the test- although they are also known to be less accurate)

So throwing a bunch of tests with lower positivity rates in to the figures will necessarily lower the positivity rate, even if the true background positivity rate isn't changing.
Really these figures should be reported separately which, fortunately, is what the dashboard does.

The small victory we won is that the PHE report now explicitly states that their positivity rates use a mix of PCR and lateral flow tests.
Even so, with changing proportions of tests this does make their headline figures unhelpful for comparisons of positivity over time.

A small complicating factor is that before 27th January, all people who tested positive with a lateral flow test would have that positive confirmed with a PCR test. Because of the high volume of PCR tests being undertaken, this strategy was abandoned on 27th Jan.

Now England's case figures include those found both from PCR tests and lateral flow tests. There's no reason to believe this is will lead to significant over or under reporting, but it's worth bearing in mind.

Why is this all important?

Well, the positivity data and the case data are used (or at least should be used) to inform important decisions about how and when to ease restrictions.

If positivity appears to be falling faster than it really is, this might give a misleading picture of the situation and lead to incorrect decisions being made.

(I'll discuss this a bit more at the @IndependentSage briefing at 13:30)

/ENDS

Bob Hawkins has been tremendously helpful in chasing all of this up and picking out the difference.
Most of the credit should go to him for his dogged detective work.