NHS COVID-19 launch: Risk-scoring algorithm criticised, the downloads, plus public told to 'upgrade their phones'
- Reference: 1600962689
- News link: https://www.theregister.co.uk/2020/09/24/uk_covid19_app/
- Source link:
Intended for users in England and Wales*, version two-point-oh addresses the [3]problems that scuppered the original version, and uses a privacy-centric decentralised model built upon the [4]contact-tracing Exposure Notification API , a joint Apple and Google project incorporated into both iOS and Android.
You can find the source code for the iOS app [5]here , Android [6]here , and the server backend [7]here .
This is a landmark moment for Britain's fight against coronavirus. In the absence of a viable vaccine, epidemiologists argue that the only way for life to return to something resembling normality is an aggressive programme of testing and tracing.
First, testing: identify who has COVID-19 and isolate them so they can't spread it to anyone else. Secondly, tracing: figure out who they've been in close proximity with, and warn them to get tested and isolate until it's confirmed they aren't carrying the virus.
Winter is coming, and with it the UK's COVID-19 contact-tracing app – though health minister says it's not a priority [8]READ MORE
Smartphone apps are the best way to provide that tracing need. Their ubiquity means healthcare providers can cover more of the population, while simultaneously scaling tracing beyond what would otherwise be possible manually.
Separately, it addresses problems caused by the fallible human mind. People can forget interactions, or they might simply not know whom they interacted with. An app handles that by automatically creating a record of any device within a certain proximity.
How low can you algo?
While the app uses Apple and Google's hastily made APIs, at its heart is a [9]custom risk-scoring algorithm designed to rank the risk of exposure. Alerts are sent based on the amount of time spent in close proximity with a confirmed coronavirus carrier.
Being within two metres (about 6.5ft) of a sufferer for five minutes is ranked at 300 points. If you're in medium proximity, defined as between two and four metres (6.5-13ft), the amount of points accumulated is halved to 150. In order for the app to issue an alert, the user needs to cross the threshold of 900 points.
In practice, that means that you only need to be in close proximity with an infected person for 15 minutes to receive a warning. For those slightly further away, that's 30 minutes.
It's not hard to imagine how you might cross that 900-point threshold, even at a medium distance. You could be on the same bus for half an hour. You could share a workplace – even one with social distancing in place, and desks carefully spaced out. While you can understand why the app's algorithm would conceivably take a more aggressive approach in isolating those potentially infected, it's inevitable that false positives will be an ongoing element of this app.
In just the first few hours of its launch, the NHS COVID-19 app has accumulated a steady flow of downloads. At the time of writing, the Android version had been downloaded more than half a million times, according to its page on the Google Play Store.
Although Apple doesn't offer any public figures, Boston-based analytics firm Apptopia estimated it has received over 170,000 downloads since its launch. It currently ranks supreme on the list of free apps.
Not all users who downloaded the app have a UK-registered phone. Although over 99 per cent of iOS downloads came from UK users, it's an entirely different story on the Google Play Store, where nearly 40 per cent of all downloads came from devices using international versions of the app store.
Around 6.4 per cent of all Android downloads are attributed to devices from New Zealand. Why? The UK was estimated to have nearly 60,000 New Zealand-born residents in 2015. It's plausible that many of those are recent arrivals – perhaps doing their " [10]big OE " – who brought their phones with them, or Brits who lived in New Zealand for some time and have since returned home.
Other nations that appear in the "top 10" include close neighbours, such as Ireland, Norway, the Czech Republic, and Hungary.
App doesn't work with iPhone 6 models or older
Another criticism raised against the app is its incompatibility with older yet still fairly common devices such as the iPhone 6. NHS COVID-19 also refuses to run on devices using Android 5.1 Lollipop and earlier. These accounted for nearly 15 per cent of all Android devices in April of this year.
When [11]confronted with this problem on BBC Breakfast , UK Health Secretary Matt Hancock artfully evaded the question, instead celebrating the public's interest in the app. "I'm delighted to hear so many people are trying to download it – that's excellent," he said.
He later admitted that some may have to upgrade their devices in order to use it.
"Only a small proportion of devices don't have the latest iOS software that was released in April. The reason for that is it needs an accurate version of Bluetooth," he said. "Of course, there are technical requirements – in the same way that if you don't have a phone you can't download it. The best thing that they can do if they want to get the app is to upgrade."
A critic that we have not named said of the app: "It's also utterly useless for contact tracing, hardly ties in with the manual T&T service. It's certainly the flagship of gov.uk's response to covid-19 in that it's too late, too little and generally bollocks." ®
* Scotland and Northern Ireland both have their own homegrown apps.
Get our [12]Tech Resources
[1] https://www.theregister.com/2020/09/24/uk_contact_tracing_app_goes_live/
[2] https://www.theregister.com/2020/06/18/uk_test_trace_app_winter/
[3] https://www.theregister.com/2020/06/18/uk_ditches_central_database_model/
[4] https://www.theregister.com/2020/05/21/apple_google_coronavirus_api/
[5] https://github.com/nhsx/covid-19-app-ios-ag-public
[6] https://github.com/nhsx/covid-19-app-android-ag-public
[7] https://github.com/nhsx/covid19-app-system-public
[8] https://www.theregister.com/2020/06/18/uk_test_trace_app_winter/
[9] https://covid19.nhs.uk/risk-scoring-algorithm.html
[10] https://en.wikipedia.org/wiki/Overseas_experience
[11] https://twitter.com/BBCBreakfast/status/1309027404829143041?
[12] https://whitepapers.theregister.com/
Re: At this time, what's the point?
So what's the outcome of a notification, 14 days quarantine?
And a whopping fine if you refuse. Not much of an incentive to use the app.
Re: At this time, what's the point?
selfish: adjective
caring only about what you want or need without any thought for the needs or wishes of other people
Re: At this time, what's the point?
selfish
You've missed the point by a long way. The purpose of track and trace is to prevent spreading the virus unknowingly, but it only works with comprehensive testing. If you are notified, you have to be able to get a test and either rule yourself in or out. The majority of track and trace notifications on the app will be false positives.
With the governments failure to get testing working properly and in the necessary volumes, there is no testing with track and trace . With no testing of track and trace alertees, the number of people under lockdown will be immense, but not compared to the number of people unknowingly still spreading the virus.
We'd be better off with a stricter, shorter nationwide lock down to let the currently infectious recover whilst minimising community spread. Fewer people-weeks would be under lock down (and fewer people will be infected) if we lock down now for a short sharp period, than will be if we interminably lock down healthy people for 2 weeks for the next 6 months.
Re: At this time, what's the point?
We'd be better off with a stricter, shorter nationwide lock down
That's been the theory for months. The UK has had an extended lockdown and it hasn't led to the desired result.
Infection rates are a misleading indicator of risk. Hospitalisation, ICU and death rates are far more important. The more the mortality rate starts to look like that of the flu – yes, I know in the UK it's more than this – the lest justified repressive measures are and at some point the economy will simply no longer be able to pay for them.
Re: At this time, what's the point?
selfish: adjective
What has that got to do with anything?
Install this app, and when you get a notification that you've been in proximity to someone else you have to isolate for 14 days. Even if you know you haven't left the house in over a week, the app says you did. Choose to ignore it? The Dept of Health says you can be fined £1,000 or more but Hancock says it's advisory, so who do you believe, and can you afford to argue? At least manual contact tracing can apply common sense to decide between low risk or close (high risk) contacts
Re: At this time, what's the point?
Hancock was at pains to emphasise this morning that following an app notification quarantine is voluntary (in contrast to a call from Track and Trace). So given that those not following the rules are more likely to become infected, and then more likely to disregard the quarantine, this looks like being really effective!
Re: At this time, what's the point?
Oh you can get them. Just not easily if you need them. But tests are also being carried out on a broad spectrum of the population, including those at risk who are effectively in quarantine. Got a relative who's been tested four times over the last couple of months even though she's got virtually no contact with other people and this despite the evidence* that symptonless testing is not very useful.
* The main reason why it's being phased out in Germany for people returning from places with relative high infection rates.
Covid-19 side-effect
2 million >3 year-old phones dumped.
Having taken a look at the app, I can't see how it took this long to develop, considering the core functionality is provided by Google/Apple.
It's a few pages with a menu, and a symptom quiz.
"Smartphone apps are the best way to provide that tracing need"
Several experts have gone on the record in the last few days stating that apps are only a secondary means of tracing, and that human intervention is the most effective method. Unfortunately, these days if you want to do anything at all you must have an "app". But demanding users of an app upgrade their phones to use it (particularly when it's supposedly such an important app) is just another example of the all too prevalent arrogance of web devs - " it works fro me on my hot kit, so you must upgrade ". It should be possible to write an app of this apparently rather simple kind that doesn't need a bleeding edge phone to run.
But in any case a rather obvious point has already been made publicly, that the most potentially vulnerable sectors of the community to COVID (the very old and the very poor) are less likely to [a] have an "upgraded" smart phone and [b] be moving around meeting strangers that much.
Re: "Smartphone apps are the best way to provide that tracing need"
Indeed and the numbers from Germany would seem to back this up: 18 million downloads, 5,000 notifications in 3 months in a country with a case load of around 10000 per week. A great success according to the Health Minister, but what else would he say?
BTW. you could, of course, use the same app to map the spread of other dangerous illnesses*. But, having been developed by SAP, this would obviously be an add-on for another € 20 million.
* Apparently, there are indeed other dangerous airborne respiratory diseases! As if SARS-COV2 isn't enough there is something called influenza (or "latin flu") which can also kill people! Run for the hills!
Botch 2.0?
A lot of potential false positives unleashed at a Test&Trace system without sufficient testing capability is a recipe for failure.
In my country, the contact tracing app is out for quite a while, but I am not aware of any published figures about the alerts handed out by the app, the number of false positives, etc.
It remains to be seen if the Corona-apps are any good.
Small houses
I live in a terraced house 5 metres wide. What happens if my neighbour is infected on the other side of the wall? Will I get tagged too?
Or I am sitting on my sofa by the window and the infected builder is parked outside in his van doing his paperwork. Do I get a call?
Can this app see walls? Or only distance?
Re: Small houses
Can this app see walls? Or only distance?
No, only signal strength and location. You can do some calculations to try and mitigate propagation and absorption issues but false positives are not considered a problem by the app. And, to be fair, by epidemiologists. All track and trace systems can and should yield false positives, this is fine as long as you also pick up the true positives. The app is an attempt to automate a part of the process and thus increase its scale as well as encourage users to take some of the precautions seriously. I remain unconvinced of the efficacy and I'm sure I'm not alone.
I do find it interesting that the Android version supports down as far as version 6.0 (released in May 2015), yet the iOS one requires 13.5 (released in May 2020).
To be fair to Apple - iOS 13.5 will run on an old 2015 iPhone 6s. Apple still support their older hardware with up to date OS versions, but most of the Android phone manufacturers abandon their hardware updates after a couple of years.
(Weird - I am defending Apple, but own Android, and use a PC)
At this time, what's the point?
You can't get a test anyway - the government are currently saying no test unless you have symptoms. So what's the outcome of a notification, 14 days quarantine? Might as well just lock us all down again for 3 weeks.