Plans to adapt calculation of exposure risk

[r-r-liu]

Are there plans to adapt the exposure risk calculation to more infectious variants? It seems that they might require less than 15 minutes throughout the day to infect.

[pdehaye]

I have my own (informed) doubts on the utility of Bluetooth, but this adjustment has been explicitly done in New Zealand. Although it is not the main topic of the exchange (rather a comparison of QR and Bluetooth) you can read some of that here, in the last two emails. https://medium.com/geekculture/location-based-tracing-some-insights-from-new-zealand-17a2e9295957

…

On Fri, Jul 2, 2021, 12:39 r-r-liu ***@***.***> wrote: Are there plans to adapt the exposure risk calculation to more infectious variants? It seems that they might require less than 15 minutes throughout the day to infect. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#8>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAY7MXYPEJB5Y7B6FXOM5VDTVWJMRANCNFSM47WO57DA> .

[r-r-liu]

Thanks. As you say, BTE alone is not ideal. But now that NotifyMe functionality has been integrated, I am still interested in whether the exposure risk calculation will be adapted to take into account the variants that seem to spread ever more quickly.

…

On 2 Jul 2021, at 14:24 , pdehaye ***@***.***> wrote: I have my own (informed) doubts on the utility of Bluetooth, but this adjustment has been explicitly done in New Zealand. Although it is not the main topic of the exchange (rather a comparison of QR and Bluetooth) you can read some of that here, in the last two emails. https://medium.com/geekculture/location-based-tracing-some-insights-from-new-zealand-17a2e9295957 On Fri, Jul 2, 2021, 12:39 r-r-liu ***@***.***> wrote: > Are there plans to adapt the exposure risk calculation to more infectious > variants? It seems that they might require less than 15 minutes throughout > the day to infect. > > — > You are receiving this because you are subscribed to this thread. > Reply to this email directly, view it on GitHub > <#8>, or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AAY7MXYPEJB5Y7B6FXOM5VDTVWJMRANCNFSM47WO57DA> > . > — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[r-r-liu]

By the way, one thing is clear from your communication with the NZ health authorities: The NZ app is regarded as a bonafide tool by their contact tracers. Here in Switzerland, SwissCovid seems to have been rejected by cantonal contact tracing groups, who evidently feel that their job consists in sticking people in quarantine, because they can't be trusted to act like responsible adults and quarantine themselves voluntarily. Mind you, I know of know of no reliable data one way or another, only anecdotal evidence that they can't. And the Bundesrat seems complicit in ignoring the original SwissCovid, and now the enhanced one, to death. Even at the height of the second wave, while they were pleading with people to minimize their contacts, wear masks, etc., they never once mentioned that using the app the way it was intended could help. And now, when it would do no harm, as the current incidence is low but that of delta is rising steadily, to recommend that people use the enhanced app and cantonal contact tracers shift their focus from exciting "cops and robbers" detective work to the less glamorous role of supporting people who voluntarily quarantine, the Bundesrat instead chooses to make a "Reisli" before disappearing for the summer.

I have my doubts about integrating NotifyMe functionality into the existing app. To my knowledge, there's nothing that NotifyMe requires that even an iPhone on which SwissCovid doesn't work couldn't meet. But would the enhanced SwissCovid even install on such an iPhone, just so people could have the "check-in" functionality? And why is it called "check in" in the app. This will only confuse those who think, "Finally, a single app with which I can check in anywhere where there's a QR code to scan." That is certainly not the case, and will only be a further reason to reject an app that doesn't meet unjustified expectations.

Summer is the time to pilot the app as a tool for the cantonal contact tracing ... and to gather data on self-responsibility in the populace. A press conference devoted to it and its ideal integration into the contact tracing accompanied by the explicit expectation that it will be used responsibly, or we all risk shutdowns, lockdowns and increase mask wearing as soon as, say, Re of delta reaches a certain value ... I would have wished for all that.

[pdehaye]

When the first reintroduction of COVID into NZ happened, it turns out it was with someone who had "followed all the rules". She had done quarantine, tested multiple times negative, and then was "released" into the general population. I don't remember if she had got infected at the quarantine hotel, or if symptoms just took a very long time to appear. But in any case she had been a very very good citizen (many people insisted on this) and scanned all the QR-codes for the period between her quarantine and her symptoms. The consequence was that within 30 min of a positive test, the NZ contact tracing authorities had the exact list of all the places she had been to, with little chance of forgetting any of them. So, yes, if I was the cantonal health authorities I would indeed consider this a bona fide tool.

Note that this diary does not use much: no magical Bluetooth for which unreasonable claims of precision are made based on completely fallacious experiments](https://medium.com/personaldata-io/inferring-distance-from-bluetooth-signal-strength-a-deep-dive-fe7badc2bb6d), no complicated cryptography that can only generate mistrust. Just "you could take notes on paper of where you have been, but instead here is a QR-code scanner and businesses wanting to act as good citizens will display them". From there the system evolved: from QR-codes to Google/Apple Exposure Notification Bluetooth (with, it seems, some leniency from Google/Apple that other countries didn't benefit from).

What happened in Switzerland was the reverse. The federal layer (EPFL, FOPH) built a solution based on their understanding of "federal science" that was completely divorced from the actual proper science. They implemented a tool from the perspective of the federal level that from Salathe's own admission failed to acknowledge the reality of cantonal contact tracing (their "clients" after all). They put an enormous amount of resources and appeared very busy with what could be directly foreseen would only help in 5% of the cases (1/5^2) and would certainly come at a cost (staff, training, communications). Meanwhile the cantons had to deal with the remaining 95% of cases, while they had serious needs for technical tools to help them and interoperability between cantons' tools. This insistence by the federal level that Bluetooth was the way to go etc is not without consequences still now: it made Google/Apple enter into this, and indeed, as you highlight, also be implicated into the QR-code tracing while it actually is definitely not a necessity.

This 95/5 split is what led me to join the Geneva contact tracing team. I cannot comment on my time there, but can make a few non controversial remarks:

• the cantons are not all the same;
• my colleagues are certainly extremely dedicated to what they are doing (I have met them before working with them, and this has convinced me of joining);
• as you know the population is certainly not homogeneous: some are irresponsible, others are extremely worried or prosocial, and all shades in between

As for the self-responsibility in the populace, I think some data has already been gathered, but that it is somewhat inconvenient for "federal science": if you want to prove your tool works, certainly ascribe any good behavior of app users to the app rather than self-selection bias.

[r-r-liu]

Thanks. I take all your points. I evidently don’t need to convince you, nor you me. Can we agree that we both drop the polemic jabs at the system? I happen to think that apps like the original SwissCovid have a place on controlling the spread of COVID-19. For example, I don’t believe that people don’t use public transportation when they are infectious, or, when they do, that their close contacts have a seriously reduced risk of being infected by them. An app like SwissCovid, with all the limitations that the technology and privacy preservation place on it, is probably the only hope to alert close contacts. I don’t imagine that the SBB would be thrilled to have to put unique QR codes on every wagon in its fleet, although that might be better than what SwissCovid has provided up to now. As I have said else where, the key to coping with increasingly more rapidly spreading mutations seems to me to be, getting people to go into quarantine as soon as possible, and having them identify themselves to contact tracing in the course of acquiring advice on the accuracy of the app’s warning, the recommended duration of quarantine, what to do in the event of symptoms, etc. You mention paper diaries. I have been keeping a daily contact diary since March 13 … 2020. It was obvious to me even at that time that belonging to the risk group meant having more important things to do than produce a memory dump for contact tracing should I become infected. This basically tells me where I was from when to when, and if I was there with others, their names, addresses, emails and telephone numbers, to the extent that I know them. I pay particular attention to public transportation, because I suspect that contact tracers probably never will find any close contacts or sources of infection there, precisely because they don’t feel empowered to do anything about it if they did. Having worked in contact tracing yourself, perhaps you can tell me why we don’t see any statistics on the proportion of new cases that were already in quarantine when they tested positive? This would seem to be a rather evident indicator of the success of contact tracing in breaking infection chains, a good performance indicator not only for the public, but especially for the cantonal contact tracing units themselves. But perhaps I am oversimplifying and contact tracers use less intuitive but for their purposes better measures of performance?

…

On 3 Jul 2021, at 16:19 , pdehaye ***@***.***> wrote: When the first reintroduction of COVID into NZ happened, it turns out it was with someone who had "followed all the rules". She had done quarantine, tested multiple times negative, and then was "released" into the general population. I don't remember if she had got infected at the quarantine hotel, or if symptoms just took a very long time to appear. But in any case she had been a very very good citizen (many people insisted on this) and scanned all the QR-codes for the period between her quarantine and her symptoms. The consequence was that within 30 min of a positive test, the cantonal authorities had the exact list of all the places she had been to, with little chance of forgetting any of them. So, yes, if I was the cantonal health authorities I would indeed consider this a bona fide tool. Note that this diary does not use much: no magical Bluetooth for which unreasonable claims of precision are made based on completely fallacious experiments](https://medium.com/personaldata-io/inferring-distance-from-bluetooth-signal-strength-a-deep-dive-fe7badc2bb6d), no complicated cryptography that can only generate mistrust. Just "you could take notes on paper of where you have been, but instead here is a QR-code scanner and businesses wanting to act as good citizens will display them". From there the system evolved: from QR-codes to Google/Apple Exposure Notification Bluetooth (with, it seems, some leniency from Google/Apple that other countries didn't benefit from). What happened in Switzerland was the reverse. The federal layer (EPFL, FOPH) built a solution based on their understanding of "federal science" that was completely divorced from the actual proper science. They put an enormous amount of resources and appeared very busy with what could be directly foreseen would only help in 5% of the cases (1/5^2). Meanwhile the cantons had to deal with the remaining 95% of cases, while they had serious needs for technical tools to help them and interoperability between cantons' tools. This insistence by the federal level that Bluetooth was the way to go etc is not without consequences: it made Google/Apple enter into this, and indeed, as you highlight, also be implicated into the QR-code tracing while it actually is absolutely not necessary. This 95/5 split is what led me to join the Geneva contact tracing team. I cannot comment on my time there, but can make a few non controversial remarks: • the cantons are not all the same; • my colleagues are certainly extremely dedicated to what they are doing (I have met them before working with them, and this has convinced me of joining); • as you know the population is certainly not homogeneous: some are irresponsible, others are extremely worried or prosocial, and all shades in between As for the self-responsibility in the populace, I think some data has already been gathered, but that it is somewhat inconvenient for "federal science": if you want to prove your tool works, certainly ascribe any good behavior of app users to the app rather than self-selection bias. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.

[pdehaye]

The SBB-Inclusive app listens to the Bluetooth beacons already installed in the trains. It would be trivial to add this functionality to the SBB app itself. What to do with that information is indeed another question.

To me the key to deal with increasingly more rapidly spreading mutations is educating the population to what matters when doing contact tracing: not tracing the virus (who is infected) but tracing where the virus hopped from one person to the next (infection). It looks like it is the same, but due to overdispersion and privacy concerns it is not: it is both more effective and less intrusive (you tie the tracing to places/events rather than people, and you focus on big ones as a priority).

That I have worked in contact tracing doesn't mean I was taking decisions on what to make public. In actual fact, I have filed several Freedom of Information requests (LTrans) with both the FOPH, cantonal authorities and those in charge of approving apps (SwissMedic, GastroVaud) to try to force them to release more information about exactly the things you say.

I wholeheartedly support your last paragraph. Here is what I know and can say publicly:

• The Scientific Task Force has published a paper discussing - alongside scalability - metrics to use to evaluate the effectiveness of contact tracing. Make your own mind about how useful that paper might be.
• Geneva canton did publish the information you mention, at least during the second wave. See here, for instance "Week 42", p. 18. It is presented however in a horribly confusing way, with two statistics alongside each other in the same graph. Trick questions: do we want the yellow curve to go up or down? What about the purple one?
• Geneva still publishes in the more recent reports qualitative information about the clusters - which they already did in July-August 2020 (and which convinced me I could be of help). It would be better if it was quantitative, agreed.
• As a theoretical problem, it is more subtle than what you describe: people might already be in quarantine when they test positive, but maybe they got into quarantine right before, which hasn't prevented their infectiousness pre-test. Also, that number would be highly dependent on whether asymptomatic people in quarantine are tested even if without symptoms (this did happen when they actually implemented this policy, from memory the number of "useful quarantine days" was multiplied by 2.5). As a practical problem, it might be better to go with the simple metric like you did, and if it becomes problematic (because teams end up gaming the metric), to modify it. As a strategic problem, it is very complex. There are two modes of contact tracing, as you probably know: prospective and retrospective. The second is much more costly short term but effective long term. There is a real risk a poorly designed metric would incentivize teams to work with the wrong strategy ("I keep on running right behind the virus, so I can put lots of people in quarantine who will later test positive" rather than "When there is a higher risk of infection, I put a large number of people in quarantine but do so early - their quarantine will last the same length, but if they did get infected I cover the whole period of infectiousness").
• There are a ton of problems in evaluating efficacy with cross border situations (even cross cantonal).

I agree with you that the effectiveness of contact tracing should be assessed much more quantitatively. If the population had that number to track instead of the number of cases, it would focus their attention where they can make the most difference. When the federal level started the cantonal system of lockdown, it was based on four criterias. One for instance was that Re had to be less than 1. That number is calculated with a delay of 14 days, so it is not great as a basis. Another criterion was that "cantonal contact tracing teams were not overwhelmed". Of course it was in the interest of the cantons to say they had everything under control. A quantitative metric would have helped then, and the only ones to have the moral authority to suggest it were in the Scientific Task Force. I am of the opinion that it would have been possible to calculate an estimation of the entropy in the distribution of the virus in the population, and its exponential growth exponent, and that this number would have been much more dynamic than the Re. Essentially it would measure for the contact of each index case "if that contact person now turns out to be positive, how many days of infectiousness would I have missed?". This would come with full "cost" for people who just turn out positive, without knowing who infected them. Or people who are put in quarantine very late.

I will drop the polemic jabs with you, but I might have more evidentiary basis than you to formulate my dejection in that way.