[Closed] Managing Covid with data..

I am also a data insights person and you would love to think so.

A few issues are:

The people paying for the analysis don't like the results
The current government are incompetent
The results are not consistent enough to share
Different points of data collection disagree with each other
There haven't been many governments who are willing to share such information
No-one bothered to use a common way to gather data so you have 3,000 different file formats to try and ingest
Oh, look, there is a squirrel. It is a much more public and friendly way to lead you blindly by distraction

so we have thia data on a daily basis therefore can we not create an algorithm that advises People by postcode if they need to lock down/dont visit other postcodes?

I suppose a key word there is 'advises'. The more targeted your measures are, the more punitive they feel if you're the subject of restrictive measures and others down the road can carry on as normal. It would make the people who are subject to those restrictions less willing to cooperate

When there was a broad blanket lockdown it made any actions outside the home feel conspicuous. But there was also both the sense of it being a shared burden and also that there was nothing you were missing out on - everything was shut, nobody was going about their life as normal. If you create small local lockdowns so that life can continue as normal for everyone else then those people only have to go a short distance down the road to be able to look, act and feel normal.... so they will. You'd create a system that drives people from areas of infection into scenarios where you're allowing opportunity for transmission.

We have about 8 months of covid “data” i assume we know by postcode where an infection occurs?

Yes we do.

And good location data about where the infection was contracted…

That's a weak point in the current data. A significant number of people don't know where they got it. Others will be pretty sure, but I don't think we have robust data on this.

so we have thia data on a daily basis therefore can we not create an algorithm

essentially we do - this is the sort of modelling that those pesky experts have been working on - its how we decide on areas that go in to different tiers/levels etc.

that advises People by postcode if they need to lock down/dont visit other postcodes?

how would you like that information conveyed? how would you make people understand which postcodes they were going to? I live pretty close to a local authority border and there's enough confusion here about which council areas different businesses are in without going as granular as postcodes.

I like in one postcode, my office (if I go there) is in another, my kids school is somewhere else, my wife's work is in yet another (and multiple different locations), and all of them have people coming to them from dozens of postcodes.

Obviously this is a gross over simplification… also would this data not allow us to forecast areas/locations that are likely to become an issue???

What it ignores is the public will just ignore rules they think don't apply to them (or don't want to apply to them) and some poor person has to try to enforce them, how do you know if the person walking down the street or to a particular cafe if from a good postcode or a bad one?

I am a data person too and like to see stats before I make decisions, I also think people are more likely to follow rules and guidelines if you show them the proof and treat them like adults.

Not just postcodes but things like XX % of infections happened at the pub, YY % at school and Z % in the garden centre. People would then understand why they are closing pubs for a few weeks for example.

Why that data isn't shared is the interesting thing. Not available, cant be reported on, doesn't show the correct message etc.

I don't think any countries share this data so not just a UK government thing.

"Managing Covid with data.."

You can only prolong or string it out for so long.

Not just postcodes but things like XX % of infections happened at the pub, YY % at school and Z % in the garden centre. People would then understand why they are closing pubs for a few weeks for example.

people have no idea where they got infected. The various lock down strategies we’ve had have been more about generating data than acting on it

obviously locking everything down reduces infection but what we’ve done since is let go of the brakes on various actives - allow some sorts of gathering and not others.  It’s to find out what will happen, it’s not because we know what works.

We have about 8 months of covid “data” i assume we know by postcode where an infection occurs? And good location data about where the infection was contracted

Aw, bless. Your faith in the systems and processes is possibly misguided.

WCA summarised it quite nicely. Ooh look, squirrels...

Bear in mind as well that the disease having a 2-week asymptomatic incubation period massively complicates any sort of testing and data collection. Our first line, the poke-up-the-nose tests most people get are a "do you/don't you have Covid at this moment in time?": They tend to be used if someone is symptomatic or knows they have been exposed, so could very easily be two weeks behind any curve.

The antibody tests, which are "have you had Covid?" aren't as common, and don't give any context or history as far as I know. It's just "you've had it" rather than "you had it three weeks ago" etc. But we don't know yet whether those antibodies are strong enough to prevent re-infection, so it's not actually telling you anything of much value.

If everyone got the same symptoms at the same severity, it would be easier to trace, but a person who has only the mildest symptoms could go through infection without ever feeling the need to get a poke-up-the-nose test and therefore that infection only becomes apparent if they have an antibody test, therefore the data is next to useless to anyone but epidemiologists.

Also bear in mind that the lockdown measures, except in a very few targetted cases, aren't aimed at preventing people getting infected. People are going to get it, there's no avoiding that - the incubation period means the horse has long bolted by the time we start looking at the stable door. The restrictions are simply to slow the rate enough that medical services don't get totally overwhelmed like they threatened to do back at the start, and to stall for time while cleverer people than me develop vaccines.

Note: My background is NHS data analysis, though it's not my role at the moment. The data in this case comes with so many quality warnings it's not even funny....

Just one specific point to address:

also would this data not allow us to forecast areas/locations that are likely to become an issue???

By the time they've been forecast, they're likely already an issue. And they'll be the same areas they are during any epidemic: The ones with the highest populations in the smallest amount of space and the highest levels of deprivation.

We have about 8 months of covid “data” i assume we know by postcode where an infection occurs?

Certainly in Scotland, we don't have that data - it's all by healthboard. So we know by healthboard where an infection occurs. My wife is on the Government level data team.

Part of the problem is data collection is governed by data protection laws. If you have one known infection in one postcode that makes the person who is infected identifiable - which breaks data protection laws (a lot of FOIs about Covid are being rejected on this basis). So even if that data were collected, you'd not be able to publish it. Another problem is that if someone is infected, who's to say that they caught it in their own postcode? It doesn't really deal with solving the problem, where looking at in on a macro scale, i.e. healthboard or local authority level, deals with the problem more effectively.

Not just postcodes but things like XX % of infections happened at the pub, YY % at school and Z % in the garden centre. People would then understand why they are closing pubs for a few weeks for example.

Why that data isn’t shared is the interesting thing. Not available, cant be reported on, doesn’t show the correct message etc.

Some of that data is shared, if you go looking for it. But it comes with HUGE caveats.

e.g. Anne has covid, Bob lives with Anne. Bob gets covid - the assumption is Bob got covid from Anne - but he might have got it from someone down the pub, the guy he works with or perhaps more likely Anne and Bob both got it from their mate Chris who has not been diagnosed.

Dave gets diagnosed with Covid. Last night he was out with Eric, Fred, and Gina in the pub - he decides not to tell them or track and trace, because (a) he believes the data will be used against him - he shouldn't have been in the pub cos he was a close contact of Bob (b) he doesn't want E, F, G to have to isolate and blame him (c) he thinks the whole think is a conspiracy made up by the Chinese and he's not going to play that game.

Anne and Bob's daughter Helen, gets covid, from one of them. She's asymptomatic and before anyone's been told to self isolate she's been to school with her whole class including Iona who sits quite close to her. Iona gets covid, so sounds like she's a definite "school transmission" but Iona was at a party the night before Helen was at school possibly passing the disease on, and she got pretty friendly with a guy called James at the party - she doesn't have his number. She knows the party shouldn't even have been on so she decides not to tell T&T about it.

James' mum Kath got covid. She's no idea where from. She met up an old friend Liz for a coffee before she got symptoms. Liz had covid in March but is feeling much better now. She doesn't think she'll get it again and is quite relaxed about the whole thing - she doesn't wash her hands for 20s. Liz went to the supermarket on the way home and ducked round the line of people waiting to sanitised their hands and grabbed 2 pints of milk before using the touch screen self service machine.

Mike was next to use the supermarket machine. He's really hungry so eats a chocolate bar on the way home. 6 days later Mike gets symptoms. He can't remember where he's been in last week but he had been in the pub collecting take away beer and he heard on facebook that the barman Chris had minor symptoms but didn't bother to get a test. Mike shares a van at work with Nigel, they aren't supposed to share a van any more due to Covid but sometimes its easier, neither of them want to get the sack. Mike tells Nigel (because he knows he lives with his elderly mum) and they agree Nigel should get tested but not tell T&T where he got it.

Nigels's son Oscar never had any symptoms but isolates when Nigel did. Two of his school friends, Paul and Quentin get symptoms. Did they get it from Oscar? Or did they actually pass it to Oscar first and Oscar passed it to Nigel? If Oscan and Paul did transfer it was it at school or when they met up to play football in the park?

Rachael, Steve and Tracy all got covid about the same time and had been together last weekend but aren't sure who their patient zero was. Tracy works in a care home so might have got it at work, and T&T will probably list it coming from there - but Rachael and Steve seemed to get it at the same time. The protection measures in the care home stopped Tracy passing it on to anyone else and she wasn't at work for two weeks. She was feeling fine in the after a few days so met one of her old colleagues Ursulla for coffee. Ursulla now works as a delivery driver because the pay is better than the care home. She doesn't get paid if she doesn't do deliveries so when symptoms start she delays getting tested, unfortunately she speaks to ~ 100 people a day. Track and trace say they don't need to find those people because the contacts were on the doorstep and quite short.

Victor got told to self isolate 10 days ago as an App contact of a total stranger, and had just started developing symptoms. T&T will list him as being "transmitted" by the stranger on the bus, but actually he got covid from victoria when she dropped the parcel off at his door and he stood talking to her for so long because he was bored.

etc etc,

And that's before you put in place anything like lockdowns or local restrictions which skew the stats - right now the transmission rate in pubs in England must be essentially zero, whereas the transmission in private parties is definitely not (although it should be). Therefore I conclude that pubs are not responsible for transmission and should reopen. Also transmission rates in schools are high, but in gyms are low - so close schools but open gyms.

You realise you've just written The Penguin Book of Pandemics 🙂

Certainly in Scotland, we don’t have that data – it’s all by healthboard. So we know by healthboard where an infection occurs. My wife is on the Government level data team.

No its much more granular than by healthboard: https://public.tableau.com/profile/phs.covid.19#!/vizhome/COVID-19DailyDashboard_15960160643010/Overview

Pick your LA from the drop down at the top and it will then show you be area (are these council wards?)

You realise you’ve just written The Penguin Book of Pandemics

Yeah - I'm in a Zoom call from hell, and writing that made it look like I was awake and attentive!

Brilliant post from poly.

FFS @poly what happened to William, Xander, Yolanda and Zoe? Don't you even care?

I’m of the opinion that local data is not as helpful due to stochastic variation. Finding one area that looks good, can be countered by one that looks poor. Hence NHS regions are probably the right level, nation too large and UK unhelpful.

singletrackworld.com/forum/topic/wheres-the-coronavirus-were-all-going-to-die-conspiracy-thread/page/507/#post-11495435

i think TiRed sums it up nicely on the 'other' covid thread.

FFS @poly what happened to William, Xander, Yolanda and Zoe? Don’t you even care?

Xander... bugger I was going to go with Xavier but someone asked me a question on Zoom and I thought I'd better get back to what I was being paid to do! Clearly it would have been Yvonne. There's a Quentin - I don't think it needed a Yolanda too...

FFS @poly what happened to William, Xander, Yolanda and Zoe? Don’t you even care?

Shouldn't it be Biff, Kipper and whoever else? I cant remember, my kids are a bit older now.

Interesting thread. I think a key piece of the puzzle that's been overlooked is the challenges of people who aren't particularly comfortable with data being told to make decisions based on data. Given my job is translating data, it's something I have to work with a lot...

The underlying problem is that BoJo - and many of the current crew - are old-school politicians, where you get what you want through rhetoric, backslapping bonhomie and public school networking. They're not very comfortable with dealing with large amounts of data, or weighing up different datasets to understand which to pay attention to and which to ignore.

You can see it in their overarching approach to this mess. Someone clever has explained to them that the most important number to focus on is new daily cases - possibly after trying to get them to understand more complex concepts. If this number gets big, we're in trouble, so everything should be focused on making it smaller.
So you have this incredibly blunt metric, that doesn't capture the impact of the pandemic, and an executive who only really know enough to turn a dial based on whether the number's getting bigger or smaller.

The more granular targeting and preemptive decision-making based on postcode level modelling is way out of their league; even if someone could explain it to them, they'd never buy into it sufficiently to stake their political reputations on it.