Updated: Jul 16, 2020
COVID-19 has affected billions of lives worldwide. Across many countries, governments have to decide between maintaining stringent restrictions on movement of people and opening up the economy. A few weeks ago, Jay Inslee, the governor of Washington State — the state where I live and where Microsoft and Amazon have their headquarters, announced a new contact tracing initiative and tweeted that “Until a treatment or vaccine is developed, contact tracing is the fastest and safest path to recovery.”
So, what really is contact tracing? Why can it help? Can we use pervasive technology to empower the contact-tracing initiative? I try to answer these questions in this post.
The core challenge for tracking the spread of COVID-19 is that infected people can spread the disease even before any symptoms appear. So, tracking and isolating people who are COVID-positive is not sufficient. In fact, this is one of the key factors that makes the disease so dangerous. We need to identify and quarantine COVID-positive patients when they get infected, not when the symptoms appear. One possibility is to achieve this by testing everyone frequently. Given the limited testing infrastructure, this is an uphill task. So, how can you identify likely COVID-positive patients without doing tests and without waiting for symptoms?
This is where contact tracing comes in. Since infections spread through contact, we can trace recent contacts of individuals who get tested positive. A contact is usually defined as an interaction where you were less than six feet away for a period of ten minutes or more. If Alice gets tested positive, we can ask them to list the people they met within the last seven days, mark these people as high risk and ask them to self-quarantine (or test them). This allows us to quarantine high-risk individuals early and helps us significantly weaken the spread. This is what we call contact tracing.
1,371 volunteers have been trained to do contact-tracing in Washington state alone. These volunteers will interview COVID-19 patients and identify and inform their likely contacts. In practice, this approach of manually interviewing every patient has many challenges. First, Alice might not remember everyone they met — how do you remember who was shopping next to you at the supermarket? Second, this approach of asking each patient and trying to retrace their steps is not scalable, particularly in areas that are severely hit and have a shortage of trained individuals to do contact-tracing. As a result, it is natural to ask if we can enlist technology to empower the contact-tracers and help them be more efficient and accurate?
Digital Contact Tracing using Bluetooth
Many technology companies (like Apple and Google) and researchers have pushed for using Bluetooth for digital contact tracing. The idea is fairly simple. Many of us carry our smartphones with us when we go out and interact with others. These smartphones have Bluetooth capability — on a high level, this means that they are capable of transmitting and receiving wireless messages. If you’ve ever paired your phone with wireless earphones, or your smartphone with your FitBit/AppleWatch, Bluetooth is what you use to make this connection. Can you use the same technology that helps you connect your AppleWatch to your iPhone to identify if someone was standing next to you at the supermarket?
The idea is to have the Bluetooth devices shout-out some messages and record all the messages they hear. As you go around shopping in a restaurant or a shopping mall, your phone shouts out a random identifier, say Microwave, and listens for what everyone else is shouting. When you are done shopping, you not only have some chips, hummus, and salsa, but you also have a list of random identifiers on your phone: Kangaroo, Croissants, Sofa, Apricots, etc. Over the next few days, if someone who was shouting Croissants was found to be positive, the healthcare provider could list Croissants on a public database. Once you find this word on a public database, you know that you have heard this before and are therefore at a risk of an infectious-contact.
Why does your phone shout some random words instead of your real name? This is done to protect your privacy. You may not want the world to know that you have an infection, but you might be okay with listing Microwave on a public list.
Using Signal Strength to Improve Digital Contact Tracing
There is one problem with this plan: Bluetooth can work at a long range up to 100 ft. So, this is likely to throw a lot of false positives, i.e. people who were not in contact might be labelled so. The fact that you heard Croissants doesn’t necessarily mean that Croissants was standing next to you. They might be two aisles away and it is just a false alarm in that case. If these false alarms happen too often, the system quickly becomes unusable and may even lead to unnecessary panic.
One way to avoid false alarms is to not just record what your phone heard, but to also record how loud it was. If a person is close, the signal is going to be strong (or loud). If the person is far away, the signal becomes weaker. So, you need to really worry if you heard loud Croissants for long enough. Otherwise, you are fine. This idea has become central to many many contact tracing apps being built worldwide.
In fact, your phone measures how loud every message is through a number called signal strength. If the signal strength number is high, your phone is likely to be close to the other phone. If not, you are further off. But how loud is loud enough or how high should a signal strength be for you to be concerned? This is where things get murky. The exact number depends on the phone you have, how you carry it, and if there are any obstacles in the surrounding. If you carry your phone in your pocket, the observed signal strength is lower than when the phone is in your hand. This makes the technique challenging. I will go into the details of this approach and talk about alternatives and the trade-offs in my next post.
You might be wondering — why can’t I just use the GPS on my phone? After all, I can use it to figure out where I am. This is a good idea, but has two practical problems. GPS tracking is not accurate enough. It can have errors of 20–30 feet, especially indoors, which is where a lot of personal contacts happen. Second, you will have to share your location with a central entity, which can leak (or actively misuse) your location data.
Digital contact tracing is not perfect, though. There are several problems that still need to be solved:
Not everyone owns a smartphone: If few people use these apps, the gaps in contact tracing become large. Some estimates suggest that if less than 60 percent of the population uses digital contact tracing, it has little value. This is particularly problematic in vulnerable populations like prison inmates, poverty-stricken neighborhoods, etc. where the smartphone density is low. If digital contact-tracing alone is used for allocating healthcare resources and become the center of our strategies, we will end up completely missing out on large groups of people.
Privacy: Tracing contacts in itself, irrespective of digital or manual contact tracing, may reveal information that you don’t want revealed. Specifically, it will reveal who you associate with. If you go to a bar frequented majorly by homosexual individuals or meet people who are part of an anti-government coalition, these connections get revealed during contact tracing, and may present risk to you and others.
Accuracy: Even with signal strength, digital contact-tracing is not very accurate. A phone that is 2 ft away might appear to be 20 ft away if you put it in your backpack under a bottle of water. When these signals get muffled by obstacles, their ‘loudness’ isn’t accurate anymore.
Digital contact tracing is a tool that, even with its limitations, can help reduce the burden on our healthcare systems. However, we must be wary of seeing technology as a savior that will magically make all problems go away. Technology is a tool, like a hammer. You can use it to build amazing furniture, but you can also use it to harm others. Digital contact tracing is yet another tool and it must be used to empower trained health professionals and volunteers to do their best, but it’s not a replacement for them.
These systems require a larger conversation around privacy and consent. While it is easy to embed consent into the apps, i.e. you share your data only with your explicit permission, that in itself is not enough in the absence of policy frameworks around it. If your government makes it compulsory for you to share the data or sends you to prison for not sharing the data, the consent button on the app is worthless.
Finally, interpreting data collected by such apps requires a lot of caution. If the apps alone decide where the focus of the healthcare system goes, it is going to be a huge injustice to the billions of folks who don’t have digital access. If anything, more human resources should be dedicated to these areas without digital access, given that they are particularly vulnerable. I would end this with a shout-out to my employer, Microsoft, which has embraced this philosophy of using technology to empower every person and organization on this planet to achieve more. This philosophy should underline, now more than ever, how we use technology to fight one of the biggest challenges of our lifetimes — technology not as a replacement of human endeavor but as an enabler of human skill and effort.
Originally published at https://www.linkedin.com.
Watch this space for the next part.