Update to NHS COVID-19 app brings improved warnings, end to 'ghost' notifications
- Reference: 1603998006
- News link: https://www.theregister.co.uk/2020/10/29/nhs_covid_19_app_update/
- Source link:
Like all apps using the GAEN (Google/Apple Exposure Notification) contact tracing APIs, NHS COVID-19 uses Bluetooth handshakes between devices to identify the duration and proximity between contacts. The latest version of the API changes how it calculates interactions by considering an additional data point: timing.
By observing the duration between Bluetooth broadcasts, the app is better placed to disregard random "echoes" produced by signals being bounced off surfaces, in turn creating more accurate readings.
[1]Explaining the update to the API , the Alan Turing Institute's Professor Mark Briers wrote:
"The GAEN API now provides health authorities with time sequenced attenuation values, with the attenuation value being a 'raw' measurement at a particular point in time, rather than being aggregated into histogram format. This means that we have several more pieces of information to use (in a fully privacy preserving manner), and can better approximate transmission risk."
This approach to predicting the values of distance between users uses a Bayesian process called an " [2]Unscented Kalman Smoother , and presupposes that two devices are unlikely to have major fluctuations of distance during a contact.
Version 3.9, which is available to download from today via the App Store and Google Play Store, also addresses the problem of "ghost notifications," which have plagued users of the app, and prompted some to self-isolate where it wasn't medically necessary.
Previously, the app would warn users if there had been a "possible exposure" to someone who has tested positive for the disease, even if it didn't meet the threshold for self-isolation. As previously discussed by this publication, the NHS Covid-19 [3]uses a point-based scoring system to determine whether someone should self-isolate , based on the duration and proximity of the contact.
In response to user complaints, the NHS addressed the problem by sending a follow-up notification to users telling them to disregard the previous message. This was a sticking plaster, however, and the API has now removed these notifications altogether.
Separately, the algorithm developed by the NHS to determine risk has been altered to take into account the improved API, as well as a better understanding of the infectiousness of the disease. In practice, this means the threshold upon which someone is instructed to self-isolate has been reduced.
Being within two metres (roughly 6.5ft) of an infected person is scored by the app at 300 points. Medium proximity, which is defined as between two and four meters (6.5ft to 13ft), is scored at 150 points. Previously, you needed to rack up 900 points in order for the app to issue a self-isolation warning. Now, that's been reduced to 120 points.
In an article published on [4]the DHSC's Technology in the NHS blog , the current chief of Track and Trace, Gaby Appleton, noted this will result in more people confined to their homes for a period of two weeks.
"The lowering of the risk threshold is expected to increase the number of people asked to self-isolate by the app, having been in close contact with someone who has tested positive," she wrote.
"We believe lowering the threshold is necessary to reduce the R rate and break the chain of transmission."
The NHS COVID-19 app had somewhat of a rocky start. Early issues prevented those on low incomes from [5]claiming financial support grants for those self-isolating . As reported earlier this week, localisation issues prevented some those with Android devices using unsupported languages from [6]accessing key functionality of the app .
Another major flaw was the lack of interoperability between contact tracing apps used in other regions of the UK. The DHSC said it's [7]aiming to add compatibility between the apps used in Scotland, Northern Ireland, Jersey, and Gibraltar in the future, and is consulting with the National Cyber Security Centre (NCSC) to ensure this is accomplished in a secure manner. ®
*The app does not cover the entire UK: Northern Ireland's contact-tracing app is called StopCOVID NI, and Scotland has Protect Scotland. Both of the countries' apps are based on the decentralised protocol.
Get our [8]Tech Resources
[1] https://www.turing.ac.uk/blog/updates-algorithm-underlying-nhs-covid-19-app
[2] https://towardsdatascience.com/the-unscented-kalman-filter-anything-ekf-can-do-i-can-do-it-better-ce7c773cf88d
[3] https://www.theregister.com/2020/09/24/uk_covid19_app/
[4] https://healthtech.blog.gov.uk/
[5] https://www.theregister.com/2020/10/23/nhs_covid19_contact_tracing_app/
[6] https://www.theregister.com/2020/10/28/nhs_covid19_contact_tracing_app/
[7] https://www.gov.uk/government/news/nhs-covid-19-app-updates-across-england-and-wales
[8] https://whitepapers.theregister.com/
Re: What?
I think. you are only expected to self-isolate if you have been in close proximity to someone for 15 or more minutes. Hopefully your queue and self-checkout takes less time than that.
Re: What?
The details the make it into the story don't include the time limit. 13ft is almost exactly 4 metres. It's unclear from any of the linked articles what exactly the points system is, and strikingly the 120 point limit is below the amount of points awarded for the lower exposure, so why have the threshold?
No doubt this is all calculated somehow, they may weight the points based on the filter output for example, but have a look at the only article referred to that mentions the points change (I'm being generous, it's a link off a referenced article), https://healthtech.blog.gov.uk/2020/10/29/how-the-nhs-covid-19-app-is-making-the-most-of-cutting-edge-global-technology/
Now, I've not spent 15 minutes in a shop queue for a long time, yet I got the low exposure notification warnings. So where is the guarantee that walking down the street isn't enough to trigger an isolation warning without sight of the scoring system?
Re: What?
Shelving and products attenuate the radio energy. That situation won't produce the same signal as an unobstructed 2m-4m proximity.
Re: What?
But it could read 2m as 4m, which is above the warning threshold. Of course, in that specific circumstance it's unlikely you'd be that close for that long, whatever the time limit is that the app/back-end is basing it on, but there are other circumstances where you may well be close to someone for that long, even though separated by a barrier of some sort. eg sitting 2m away from someone at work, facing in opposite directions or even separated by a perspex screen.. The scoring threshold does seem to be a bit low, and as has been mentioned is not helpful if we don't know for sure what exposure duration is being used.
Good
I posted about these horrible ghost notifications yesterday. Downvoted amazingly?
But glad this is happening. Thanks for covering it.
It's a bit scary to get them.
Re: Good
You were downvoted because the "ghost notifications" story was covered a few weeks ago and was well known about. It even made the national news on BBC, ITV, Sky etc as well as some front page headlines.
Re: Good
But why does that make my personal experience down vote worthy? :(
What?
>Medium proximity, which is defined as between two and four meters (6.5ft to 13ft), is scored at 150 points. Previously, you needed to rack up 900 points in order for the app to issue a self-isolation warning. Now, that's been reduced to 120 points.
So, standing in the next aisle in a supermarket to a person who tests positive despite being separated by a barrier over two meters high will land you under 'house arrest'? Isn't that a rather low risk situation to convert into "thou shalt self-isolate for 14 days'? And they expect people to volunteer to use this.