How Apple and Google address one of the most difficult parts of tracking COVID-19 exposures

Here we say that The interface comes out Monday to Thursday and on particularly new Fridays. Well, today was a very new Friday. Here’s your emergency newsletter …

Thursday evening, I wrote about some of the limitations of using the Bluetooth chip in your smartphone to track the spread of COVID-19. Obviously, on Friday morning, Apple and Google announced what could be the most significant collaboration in the history of both companies – a concerted effort to use the Bluetooth chip in your smartphone to track the spread of COVID-19.

Russell Brandom and Adi Robertson had the details in it The edge:

The new system, which is documented in a series of documents and white paper, would use short-range Bluetooth communication to establish a voluntary contact discovery network, keeping comprehensive records of phones that have been close to each other. Official apps from public health authorities can access this data and users who download it can report if they have been diagnosed with COVID-19. The system also warns people who download them if they were in close contact with an infected person.

Apple and Google will be introducing a few iOS and Android APIs in mid-May and ensuring that these health authorities’ apps can implement them. During this phase, users will still need to download an app to participate in contact tracking, which could limit acceptance. But in the months after the API is completed, the companies are working to build tracking functionality into the underlying operating system, as an option immediately available to anyone with an iOS or Android phone.

If you are not yet familiar with how contact tracking helps stop the spread of disease, start with this explanation from my colleague Nicole Westman. Public health authorities have long been sending workers, both on foot and by phone, to contact people who may have been exposed to someone who carries an infectious disease. In light of the COVID-19 pandemic, countries around the world are experimenting with apps that attempt to use the widespread adoption of smartphones and the signals they receive to identify new potential cases. And as I explained yesterday, public health officials with whom I spoke have expressed skepticism about the effectiveness of such efforts.

First let me say that I am happy that large companies are working hard on the COVID-19 response and working together. Fast-paced, bold action can save lives, and it’s fine that not everything that is tried works perfectly – if at all. It is also true that, given the number of questions remaining around the Apple / Google collaboration, it is currently impossible to say how effective it could be. I hope it is very effective!

That said, Bluetooth-based contact tracking approaches have at least three major issues, experts have told me. (Surprisingly, privacy isn’t one of them, at least not to me; the privacy design of the Apple / Google system is pretty smart. Moxie Marlinspike has some bickering thoughThe bigger issues are: It’s hard to get people to download a new app, Bluetooth signals can be unreliable, and a focus on tech solutions can reduce pressure on public health agencies to hire people to track contact, too although there is much more evidence of the effectiveness of those workers than for smartphone apps.

So let’s see how the Apple-Google partnership tries to address some of these points.

The main part of the problem the API project is trying to address is adoption, especially in the second phase of the project. Apple explained it to me like this: once you update your phone to the latest OS version, and When you sign up for the contact tracking API, your phone starts sending Bluetooth signals to nearby phones and recording signals sent by other phones. The best part about this system is that it works retroactively – once you download a public health app associated with this system it will share your “proximity events” from the past 14 days. Adi Robertson describes the process here.

By creating a central API for our two major smartphone operating systems, Apple and Google provide a valuable resource to public health authorities working on contact tracking apps that will work in all jurisdictions around the world, even when people resume their journey. It is hard to imagine anything like this being done in any other way than at the operating system level; only these two companies can make this possible.

An open question is whether you’ll be pinged about exposure if you’ve updated your phone’s operating system but haven’t downloaded a public health app. It seems as the answer is yes, based on what we’ve seen – which would go further in addressing the issue of adoption than any other proposal I’ve seen. If the answer is that you still need to download an app to get the notification, the basic problem has not really disappeared.

We will see.

So what about the reliability of Bluetooth signals? A strong signal has a range of about 30 feet – well beyond the 6-foot distance that authorities have asked the public to maintain. And the signal is binary, not relative – it can only say “these two phones came close” instead of “this phone was 6 feet from that phone”. That raises concerns that many of the nearby events recorded by our phones will be false positives – instances where you were relatively close to someone who reported an infection, but may not be close enough to become infected yourself.

Apple says it’s still investigating this, but notes that public health apps can include the duration of proximity when determining what counts as proximity. (The suggested time I heard today was five minutes.) With a five-minute interval, you would be less likely to evoke false positives from someone jogging on the street.

Which begs the question, what are the circumstances during the pandemic where people (1) are within 30 feet of you, for (2) five minutes or more that (3) you don’t really know? (If you knew them well, you would probably also discover that they had COVID-19.) Some suggestions I heard today: supermarket employees; people who wait in long lines for things (such as access to supermarkets); warehouse workers; and mass transportation. As cities begin to reopen, more use cases may emerge. But it seems like a passive system that informs people in those situations about potential exposures that can at least provide some level of protection. The question is whether the system eventually generates more signal than noise – whether Bluetooth finds more real positives than false ones.

We will see.

After all, does Silicon Valley rely too hard on untested software solutions while a proven manual solution might suffice? That is the argument in a paper this week from Duke University’s Margolis Center for Health Policy. Authors Mark McClellan, Scott Gottlieb, Farzad Mostashari, Caitlin Rivers and Lauren Silvis write:

Ideally, when a new case of COVID-19 is identified, local public health officials will ensure that the affected person is isolated, and that their close contacts are identified and asked to be quarantined. However, existing local public health capacity for such response activities is very limited, and many jurisdictions have abandoned contact tracking in favor of mitigation measures at community level. To allow a return to case-based interventions as the incidence decreases, these capabilities need to be expanded. Enhanced capacity will be most effective when coordinated with healthcare providers, health systems and health plans and supported by timely electronic data exchange. Mobile phone based apps that record proximity events between individuals are unlikely to have enough distinctiveness or acceptance to achieve a public health benefit while introducing serious privacy, security and logistics issues. Instead, timely contact tracking can be achieved through enhanced public health research complemented by technology and community-level partnerships.

But other researchers have argued that COVID-19 simply spreads too easily to detect manual contact to find a viable solution to the problem. Something passive and automated is needed to slow down the transmission speed, according to a paper from March in Science. Luca Ferretti, Chris Wymant, Michelle Kendall, Lele Zhao, Anel Nurtay, Lucie Abeler-Dörner, Michael Parker, David Bonsall and Christophe Fraser write:

Traditional manual contact tracking procedures are not fast enough for SARS-CoV-2. However, a delay from confirming a case to finding their contacts is not inevitable. In particular, this delay can be avoided by using a mobile phone app.

Perhaps the best way to think about Apple / Google’s announcement is that in a world without a coherent federal response to the ongoing disaster, we should rely on a patchwork of partial solutions. In such a world, I have no objection to Apple and Google trying to build a contact trace, even if I’m concerned that people will expect too much from it. As always, I am open to a pleasant surprise.

We will see.

Those good tweets

Talk to us

Send us tips, comments, questions and Bluetooth signals: casey@theverge.com and zoe@theverge.com.