Thread by @ADMBriggs

Stats from week two of the 'NHS' Test and Trace programme also out today, covering 4/6/20 - 10/6/20.

Having been buried in local complex cases during this period, has been interesting to try and make sense of the macro picture (mini-thread)

gov.uk/government/pub…

Says 5,949 cases referred to programme, 4,366 were contacted (73%), (8,096 referred previous week, with 72% contacted)

And given the importance of timeliness, 78% reached within 24hrs of referral compared with 75% week before.

44,895 contacts identified - this is 10.2 contacts per case. Significantly up from 8.9 per case the week before.

Both weeks, 90% of those contacts were reached and asked to isolate.

That's 9.3 contacts reached per case this week
8.1 the week before

(hear me out)

But here's the interesting bit - 3,572 contacted cases were non-complex, those contacted by tier 2 of test and trace.

These cases were responsible for 4,715 contacts that were reached (contacted by tier 3).

That's 1.3 contacts reached per case.

The week before it was 4,392 contacted cases and 5,282 contacts reached.

That's 1.2 contacts reached per case. Basically the same.

Also, is says in the methods document that those contacts not reached ONLY relates to non-complex cases. Meaning that the percentage of contacts reached for complex cases (tier 1) is - by default - 100%.

The percent not reached for tiers 2 and 3 is therefore 47% (4,205 not reached, 4,715 reached).

Down from 48% the week before.

This would be a much more genuine way of presenting the %age contacts not reached data.

Now, just 794 cases were complex - managed by those of us in local PHE teams.

These cases were responsible for 35,975 contacts reached - 45.3 CONTACTS REACHED PER PERSON (I would use bold rather caps given the option).

88% of ALL CONTACTS reached from just 614 cases.

The week before, 1,434 complex cases, reached 41,667 contacts, at 29.1 contacts per person.

So vast majority of contacts are through tier 1 - complex cases. And this explains all of the increase in numbers of contacts from week 1 to week 2.

Why??

Is this a data problem (error/confounder/sampling) or is this real?

Before we all get cynical about tiers 2 and 3 doing any work - they have to follow up every contact individually one by one.

If we phone a care home, then immediately that might be 30 contacts. One call.

Importantly, we've changed to how we record contacts in the past 10 days or so.

Now it's recorded directly onto CTAS - the new PHE computer system for contact tracing whereas before it was recorded on our completely different usual disease surveillance tool.

Furthermore, we may locally record all other residents in care home (for example) as a contact, whether they were actually within 2m or not.

This makes sense from a public health perspective as in such a high risk setting they are potentially all at risk and may need to isolate

But that may not be what public/politicians think when we say 'contact'.

And if there's more than one resident in a care home affected, then their contacts are going to be counted twice (can't see whether that's dealt with in the methodology document).

And there are other considerations - changing numbers of cases in care homes and hospitals, schools opening up, it's busy (though not sure if 'busier'), the issues picked up by @TimHarford and folk on @BBCRadio4 More or Less the other week.

Where I want to get to with this is that we really SHOULDN'T try to draw too many conclusions from just two weeks of 'NHS' (PHE) T&T.

The data are too difficult to interpret without knowing the impact of all the operational changes that are going on behind the scenes.

My take is that T&T is probably getting better, it definitely will get better, and it'll take time.

It's a massive operation. New staff, new computer system, new protocols. Lots of orgs involved (PHE, NHS, local gov, social care etc)

I would definitely NOT use these data to draw conclusions about public engagement with it, success or not of social distancing, and impact on numbers isolating.

Instead, ignore these first 2 wks and probably go from now on when inferring something about COVID policy.

And I would definitely change how the data are reported - particularly the percentage of contacts reached.

That's a particularly useless analysis as it currently stands.

May be helpful @felly500 @tim_esPH @Jeanelleuk @ADPHUK @LGAWellbeing

The numbers were done quickly so may not be bang on but think they're about right.

Oh - and it's also definitely the right thing to be doing. Please, please do isolate if you're asked to.

also @trishgreenhalgh @bengoldacre @benthebray @deb_cohen @ShaunLintern @HPIAndyCowper @HSJEditor perhaps.