During the 2014 Ebola outbreak in Guinea, global health researcher Anne Liu struggled to convince public health officials that apps and other technologies could help manage the spread of disease. To beat back the outbreak, officials had to locate every person an Ebola patient may have interacted with while they were infectious, in a process called contact tracing. Liu and her colleagues wanted investigators to use apps to compile information, rather than pen and paper. At the time, it was a hard sell.
“The fight was more, is technology going to be useful at all,” says Liu, now the senior technical advisor at the Clinton Health Access Initiative. “I don’t think that’s the battle anymore.”
Now, during the ongoing novel coronavirus pandemic, some experts are turning to technology to take over the contact tracing process entirely. Countries like Singapore and China are using cellphone-based tools to identify and monitor people who might have been exposed to someone with COVID-19. The United Kingdom is building a contact tracing app, and in the United States, Google and Apple partnered to build a Bluetooth-based tracking system that can automatically log people’s interactions.
Whether this type of tech could help halt a disease outbreak is still unclear; it’s never been studied before. Some experts are optimistic that automating contact tracing could scale up the COVID-19 response in the US. But creating systems that can do that work is only the start of the conversation.
“This type of technology is not a silver bullet. It has to be part of a comprehensive public health strategy,” Liu says. “Technology is usually the easy part.” These new systems won’t be useful without a dramatic increase in the amount of testing done in the United States, clear messaging, and strong integration with public health policies.
Shifting the burden
The goal of contact tracing apps is fairly simple. They’d log every phone within a certain range of a person, and if that person later tested positive for COVID-19, they could send an alert to each phone that had been nearby. Google and Apple turned to Bluetooth, which can monitor the other phones in your area without tracking your specific location. There are limitations to this approach: Bluetooth casts a wide net and may struggle to tell if two phones were actually close enough for their users to transmit a virus between them. The systems also wouldn’t be able to monitor the contacts of people who didn’t agree to use it or of people who don’t have smartphones.
That’s why app-based tracking will not be a full replacement for manual contact tracing, and public health agencies still need to vastly expand the contact tracing workforce in the United States. Apps could, though, take on some of the work and make the process more efficient.
“Generally what’s done is a huge amount of manual work. We have to increase the public health workforce, and this new technology could ease a lot of that burden,” says John Brownstein, an infectious disease epidemiologist and chief innovation officer at Boston Children’s Hospital.
An automatic system can do things that a manual contact tracer can’t. It’s hard for people to remember everyone they interact with each day, let alone over a week or two-week period, and an app would take some of the guesswork out of the equation. It would also be able to flag people who someone may not know they interacted with, like a stranger they stood next to in a subway car.
“Let’s say you’re an infected grocery store worker,” says Ranu Dhillon, a researcher in the division of global health equity at Brigham and Women’s Hospital. “There’s no way to track who you may have exposed.” An automated system could comprehensively identify those previously unknown contacts at a huge scale.
An automated system is also fast. “The biggest advantage, I think, is speed,” Dhillon says. It can take a few days for a contact tracer to manually track down everyone on their list for each case — and in that time, infected people could be passing the virus on to others. “Instantaneous notification can make a big difference,” he says.
What happens next?
Once those notifications are made through the app, the second wave of work begins. It’s still not clear if the system will give public health agencies any information about the identified contacts of a known positive case.
“Typically, in contact tracing you want a health official to have some ability to follow up,” Liu says. Manual check-ins or follow-ups may not be as important for COVID-19 contacts as they were for Ebola contacts — for Ebola, tracers circle back to contacts every single day for a few weeks. People exposed to COVID-19 are usually just asked to monitor themselves for two weeks.
Liu still thinks that it would be helpful for officials to know who the contacts of each case were, just like they would with manual contact tracing. She acknowledges that it might be a difficult ask given that people may feel differently if that information is processed through an app using 4G or software. “I can understand that in settings where people are concerned about privacy, that might be a challenge,” she says. At a minimum, contacts should still be given some sort of public health information by the app so that they can make a decision about what steps to take after finding out they’d been exposed.
The recommendations that each contact would get depends a lot on the infrastructure and policy in place for the disease response. Ideally, Dhillon says, every contact could be tested for the novel coronavirus, and then potentially tested again around a week later (to make sure a first test, if negative, wasn’t incorrect).
Right now, in the United States, there isn’t nearly enough testing available for that to be possible. Instead, it’s more likely that a notification about a possible exposure to a person with COVID-19 would trigger a recommendation to quarantine for two weeks. That’s still useful, Dhillon says. “I’d rather isolate all contacts than isolate everyone. Contacts are still far fewer people than the whole population.”
Without testing, though, the potential burden of a notification would be higher. “You’d be asking a lot of people to potentially quarantine themselves based on a potential contact,” Brownstein says. If the Bluetooth system wasn’t specific enough, and it flagged people as contacts who weren’t actually at risk, those false positives could be stressful and overwhelming. Too many perceived false alarms may also make people less likely to follow instructions after an alert.
If people are regularly alerted to a possible exposure, though, it may have more to do with the caseload in their community than with the reliability of Bluetooth — particularly if the tech can be refined. “If people are flagged multiple times, it probably indicates that people in their social networks are the ones coming back positive,” Dhillon says. “It shouldn’t be as much an issue of false alarms as of a cluster of cases.”
If people are getting lots of alerts, it could also mean that the number of people sick with COVID-19 is so high that contact tracing is no longer the best strategy — and that everyone should be isolating anyway. “That’s when you’re in a situation, such as New York City is in right now, when you just have widespread community transmission. You have to make an assumption that everyone’s been somewhat exposed,” Liu says.
Telling someone that they’ve been around someone with a case of COVID-19 isn’t as useful when the disease is everywhere. Contact tracing isn’t done manually when case counts are high, so it might not be worth having automated systems turned on, either. Decisions around when and where to send notifications have to take that context into consideration, Liu says. “That way you help avoid this issue of notification fatigue.”
Communication is key
Along with ensuring that a system works well and fits into the public health ecosystem, the officials working with automated contact tracing have to push out clear communications to users. “A big part of it is the messaging itself,” Liu says. “Regardless of the technology, you have to make sure the messaging itself is clear.”
Messaging from public health officials about the COVID-19 pandemic has been inconsistent and often misleading, which doesn’t bode well for clear communications around any automated testing system. Still, it’s something that experts say could make or break the proposed systems.
People deciding if they want to use an app or opt into a tracking system need to understand exactly what the technology is for and how it’s being used. They also need to know what a notification from the system would mean and what they should do with that information. An alert wouldn’t necessarily mean a person is in imminent danger, but it could mean that they’re at a certain level of risk of developing COVID-19 — and they have to know what those risks are. “We need to have a lot of public messaging where people understand what the notification is, and can put it into context,” Dhillon says.
Clear communication would also increase the odds that people would choose to use the automated system, which needs a high percentage of participation in order to work. In Singapore, for example, around 12 percent of the population downloaded a contact tracing app — but statistically, that translates to only around a 1 percent chance of two people both using it. With such a low rate of use, it’s not going to catch many random contacts. It’s hard to say how many people in the United States would need to use an app to give it value.
“What would be ideal from an epidemic control perspective is not necessarily what is ideal at a social and policy level, which is just as important,” Dhillon says. “We need enough people to opt-in to make a dent.”
The pandemic is moving at unprecedented speed, and public health experts are sprinting to build the tools they think might help bring it under control. “It’s a little bit of flying the plane while still building it,” Dhillon says. Any automated contact tracing program would have to be carefully monitored to see how well it helps contain COVID-19, how people are interacting with it, and if it’s flagging more people than actually would be at risk from an exposure.
Whatever the systems eventually end up looking like, they have to be introduced alongside public health infrastructure to ensure they have as big an impact as possible. “The tools can’t be used in isolation,” Liu says. “You have to make sure you have the policies in place to support them.”