3 ms·

The UK "hack" for the bluetooth was quite clever from what I saw - in essence they implemented the "background discovery of an iPhone" feature. One concern I d

by g_p 6y ago

The UK "hack" for the bluetooth was quite clever from what I saw - in essence they implemented the "background discovery of an iPhone" feature. One concern I do have with this approach is that the decentralised system inherently can only model the "risk-to" an individual from an epidemiological perspective. It can't model the "risk-from" side of things. To avoid Sybil type attacks, you need to use centrally issued (and thus verifiable) test results to notify others. Unless tests are administered very promptly, with results immediately available, it seems difficult to make this work in the decentralised approach - with a mean and median "time to symptoms" of around 5.1 days [1], and someone at their most infectious for the 1 to 2 days prior to experiencing symptoms, it seems decentralised apps will struggle. If A is infected on day 0, then on day 3 they may infect B and C. On day 5, A experiences symptoms, goes online, and requests a test, saying they have symptoms. They realistically would get a test the following day. and the result the day after that. That would be day 7, and at that point, B and C might be notified by the app. The issue is that B and C are now on day 4 of their own infections, the second day of their most infectious period, walking around unaware. They find out the day before they would themselves (on average) experience symptoms. Clearly the virus is not as deterministic in behaviour as in this simple model, but the above seems to suggest to me that to get any real value out of this (i.e. preventing the next generation of infection), you need to either be able to respond to symptoms reports, or significantly expedite the testing process. In the above situation, a same-day test-and-result would enable B and C to be notified on day 2 or early on day 3 of their own infection, potentially having some real impact. But to achieve this, the latency of testing would need to drop hugely. [1] https://hub.jhu.edu/2020/03/09/coronavirus-incubation-period/ https://hub.jhu.edu/2020/03/09/coronavirus-incubation-period...

ppf 6y ago
Exactly what I was thinking - I don't see how this approach will work in practice with the time it takes to find and positively identify a Covid-19 case, as the chain of contacts will become enormous.
- g_p 6y ago
  If the clinical results of the 20-minute lab-free test the UK is trialling [1] are positive, I could see this working. But that's a risk. To my mind, this is a simple matter of reducing the time the "next generation" of infected people are in the community before being notified. If that can be below 3 days, I think the disease can be managed based on the law of averages. If not, it seems it will continue to spread. That's why, absent a national "Amazon Prime Now" style <= 2 hour delivery window for tests (and some way to verify the results of these tests for input to a decentralised app), I don't see this working. The app will only really be needed/helpful in cases where you don't know people you are in close or prolonged proximity to, i.e. shops and public transport. In these cases, it makes sense to notify as early as possible, even at the risk of some false positives for those not actually infected, but who were close. The question is how many FPs there will be, but this also happens with human-based contact tracing too. [1] https://www.theguardian.com/world/2020/may/21/uk-coronavirus-test-with-20-minute-wait-being-trialled https://www.theguardian.com/world/2020/may/21/uk-coronavirus...
dignick 6y ago
> The UK "hack" for the bluetooth was quite clever from what I saw - in essence they implemented the "background discovery of an iPhone" feature. I don't think the government deserves the credit they have received for this. The "hack" was well understood and documented before this App was created, and it has been implemented poorly [1]. [1] https://github.com/nhsx/COVID-19-app-Android-BETA/issues/10 https://github.com/nhsx/COVID-19-app-Android-BETA/issues/10
coob 6y ago
> The UK "hack" for the bluetooth was quite clever from what I saw - in essence they implemented the "background discovery of an iPhone" feature. It wasn't that clever, it didn't work: https://github.com/nhsx/COVID-19-app-iOS-BETA/issues/2 https://github.com/nhsx/COVID-19-app-iOS-BETA/issues/2
s_dev 6y ago
>The UK "hack" for the bluetooth was quite clever from what I saw Irish researchers precluded this approach from being effective though -- Bluetooth as a technology is just not reliable enough for this specific use case. https://www.irishtimes.com/news/ireland/irish-news/covid-19-precision-of-tracing-apps-in-doubt-after-tcd-study-1.4247865 https://www.irishtimes.com/news/ireland/irish-news/covid-19-...

by g_p 6y ago

ppf 6y ago

Exactly what I was thinking - I don't see how this approach will work in practice with the time it takes to find and positively identify a Covid-19 case, as the chain of contacts will become enormous.

g_p 6y ago
If the clinical results of the 20-minute lab-free test the UK is trialling [1] are positive, I could see this working. But that's a risk. To my mind, this is a simple matter of reducing the time the "next generation" of infected people are in the community before being notified. If that can be below 3 days, I think the disease can be managed based on the law of averages. If not, it seems it will continue to spread. That's why, absent a national "Amazon Prime Now" style <= 2 hour delivery window for tests (and some way to verify the results of these tests for input to a decentralised app), I don't see this working. The app will only really be needed/helpful in cases where you don't know people you are in close or prolonged proximity to, i.e. shops and public transport. In these cases, it makes sense to notify as early as possible, even at the risk of some false positives for those not actually infected, but who were close. The question is how many FPs there will be, but this also happens with human-based contact tracing too. [1] https://www.theguardian.com/world/2020/may/21/uk-coronavirus-test-with-20-minute-wait-being-trialled https://www.theguardian.com/world/2020/may/21/uk-coronavirus...

dignick 6y ago

> The UK "hack" for the bluetooth was quite clever from what I saw - in essence they implemented the "background discovery of an iPhone" feature. I don't think the government deserves the credit they have received for this. The "hack" was well understood and documented before this App was created, and it has been implemented poorly [1]. [1] https://github.com/nhsx/COVID-19-app-Android-BETA/issues/10 https://github.com/nhsx/COVID-19-app-Android-BETA/issues/10

coob 6y ago

> The UK "hack" for the bluetooth was quite clever from what I saw - in essence they implemented the "background discovery of an iPhone" feature. It wasn't that clever, it didn't work: https://github.com/nhsx/COVID-19-app-iOS-BETA/issues/2 https://github.com/nhsx/COVID-19-app-iOS-BETA/issues/2

s_dev 6y ago

>The UK "hack" for the bluetooth was quite clever from what I saw Irish researchers precluded this approach from being effective though -- Bluetooth as a technology is just not reliable enough for this specific use case. https://www.irishtimes.com/news/ireland/irish-news/covid-19-precision-of-tracing-apps-in-doubt-after-tcd-study-1.4247865 https://www.irishtimes.com/news/ireland/irish-news/covid-19-...