How serious is the rise in Covid-19 infections?

I've been bombarded with emails and messages from data scientists who firmly believe that the trend to Covid-19 infections, based on when a specimen was taken, is flattening or even falling.

On the basis of that analysis, they are convinced the government is overreacting by threatening to impose new social distancing measures.

And if you look at the government's Covid-19 dashboard, you will be struck that the seven-day average for positive results is sharply on the rise, whereas there is a modest fall in the seven-day average of results by the date the specimen was taken.

Now to be clear, those who are bossing this show - the prime minister, the health secretary, the chief medical officer and the chief scientific adviser, advised by the Joint Biosecurity Centre - are all convinced that Covid-19 infections are rising in an alarming and dangerous way.

In fact, one senior source, with a close knowledge of the data used by the Joint Biosecurity Centre, said:

"All the data sources point to a rise in the infection rate. [That is true of ] NHS Test and Trace, ONS, ZOE".

The source added that "positivity is rising in care home asymptotic testing (which is unaffected by a rise in demand for tests) and in all age groups. There is no escaping the fact that the second wave is beginning and we need to change the way we behave very swiftly if we are to avoid a national lockdown".

Yuck.

Just to translate a bit of this: my source pointed out that the now routine testing in care homes is showing that infections are on the rise in those settings, which is deeply troubling.

And presumably what it all means is that the Joint Bio Security Centre will soon raise the national Covid-19 alert level from the current level three back up to four - since the government believes that the latest data indicates transmission of the illness is high again and rising exponentially.

It also suggests that in a matter of just a couple of days or so, the kind of new restrictions on our social lives that I've been discussing with you - pubs and restaurants being forced to shut at 10pm, or perhaps being temporarily shut altogether - will be ordered by the PM.

So that is the grim outlook: we've got to get better at social distancing voluntarily and pronto, or our basic freedoms will be stripped from us again.

But to revert to where I started, why does the curve seem to be flattening based on when specimens were taken? Well I am told it is because that line in data almost always falls for the latest four days, because of the lag between when specimens are taken and when they are processed.

And that lag has probably got worse (though the government denies it) because demand for tests has surged so far ahead of capacity.

In other words, it is the data on test results and on hospital admissions - which show a doubling in numbers infected every eight days or so - which has set off klaxons in Downing Street.

Which presumably means the prime minister will be addressing us all via the medium of television sooner than he or we would ideally choose.

Update: My wording about the flattening curve based on the specimen test data was sloppy.

Looking closely at it, the rate of increase declines from about 7 September onwards - so the "normal" understatement of the latest four days doesn't apply, though an increase in the backlog of unprocessed cases might apply. 

More relevantly, this is what a data analyst used by the government says: "Using testing data as a proxy for the epidemic is inherently unreliable as it depends on who is tested.

"You have different delays in any testing system and have to be careful when you see a drop that it isn't an artefact. 

"Data sources like the ONS survey, hospital admissions, ICU admissions is more reliable and all are showing an increase".