Part 1: How it works and what’s happening now
Everyone’s waiting for the silver bullet — a miraculous treatment, prophylactic, or vaccine that will put a definitive end to the pandemic. In the interim, we know what it will take to reopen the broader economy and mitigate risk: Testing, tracing, and isolation are the best available tools for reducing the number of new infections.
Now that testing is finally ramping up across most of the United States, we need to make sure we can use test results to stop the spread. While new cases are inevitable, contact tracing is critical to curtailing transmission. States like New York are including tracing capacity in their key metrics for reopening, enabling precise regional monitoring and avoiding blunt-force shutdowns.
Of course, tracing is much easier said than done, especially with the new coronavirus: Effective tracing requires widely available testing and rapid results; any lag time increases the risk of presymptomatic and asymptomatic transmission. Our inability to scale testing efforts has made it even harder to scale contact tracing. Outbreaks are happening in heavily populated areas where the virus was spreading before tests were available and where interactions are difficult to track; as social distancing measures relax, contact tracing will become even more difficult.
In the first of a two-part series, we’re looking at the use of contact tracing in previous epidemics, current approaches to coronavirus contact tracing, and what it will take to deploy a large-scale contact tracing program in the United States.
How contact tracing works
John Snow’s 19th-century maps of cholera infections in London set the foundations for modern epidemiology. By tracing cases and interviewing victims’ families, he learned that the disease was spread through water, not the air, and shutting down a contaminated water pump could end an outbreak. In modern times, contact tracing is a proven tool in containing outbreaks: It was used during the Ebola virus and SARS outbreaks and is used to track sexually transmitted infections.
Contact tracing is part public health work, part detective work: Tracers work backward from infected cases to identify people who may have been exposed so that they can be tested, isolated, and treated; they also monitor the condition of contacts and connect them to support they need while quarantined.
Contact tracing is different for each disease because what constitutes a “contact” depends on how the disease is transmitted — measles can be transmitted by being in the same room as someone, whereas HIV requires sexual contact or needle sharing. What contacts are advised to do will also differ depending on the disease.
When a patient gets a positive coronavirus test, the lab reports the results back to both the patient’s doctor and the local health department. A contact tracer is then assigned to the confirmed case and will call the person to ask about symptoms, find out who they have been in contact with, and create a plan for isolation, including how to get groceries, medicines, and other essentials delivered.
Recreating the map of a patient’s movements can be challenging if the patient is unwell or if they’ve had sustained contact with people they don’t know, such as at a store or while using public transit. Tracers will ask about the people they met, the errands they ran, and the places they visited — they’ll ask “what did you have for breakfast?” or “where did you go for lunch?” to help jolt a patient’s memories.
That patient’s contacts will be told when they were exposed but not the identity of the person who exposed them. Tracers will then regularly check in with isolated patients and quarantined contacts to monitor their health and help determine when they may be able to safely leave home again.
Current CDC guidance recommends patients who test positive isolate themselves until several criteria are met, including that they have had three days without a fever and that other symptoms are also improving. For close contacts who have been exposed, the CDC recommends a 14-day quarantine after the last date of exposure. “Isolation” is the term used for people who have tested positive for the virus, and “quarantine” is used for people who may have been exposed. In practice, both terms mean the same thing: Stay away from others.
What’s happening now
Our understanding of what constitutes a contact for a COVID case is evolving; currently, a contact is someone who was within six feet of an infected person for at least 15 minutes, starting two days before the infected person began experiencing symptoms. This means a contact is unlikely to be a retail employee or a delivery person; most contacts are between friends or family members, or in service settings like salons or care facilities — and the number of sustained interactions will likely increase as social distancing measures are eased. One of the big challenges for the new coronavirus is “invisible” transmission: Patients appear to be most infectious before they experience symptoms and many show only mild symptoms or none at all.
The success of contact tracing will depend on a number of factors, including the size of a regional tracing team, the number of new cases in a community, the availability and speed of testing, and how readily potential contacts respond.
This month, an NPR survey found that 44 states and the District of Columbia are planning to expand their contact tracing teams, collectively increasing them from about 11,000 to more than 66,000 over the coming weeks. New York State is currently recruiting up to 17,000 contact tracers toward its goal of 30 tracers per 100,000 people. New York City plans to hire an additional 5,000; as of May 17, the city had hired 50, pending background checks.
These numbers, large as they are, may not be enough. In April, a Johns Hopkins paper recommended 100,000 contact tracers in the United States. Based on an hourly rate of $17, the average pay for a community health worker, this would cost approximately $3.6 billion for one year — with the added benefit of creating new jobs at a time of record unemployment.
Not unlike any other healthy supply chain, the tracing workforce must surge and flex based on evolving needs; hotspots will need more tracers than communities with lower rates of transmission. Wuhan, China, employed about 81 contact tracers per 100,000 residents, whereas New Zealand has about four per 100,000.
To make an impact in any location, regional tracing teams need to detect a significant portion of new symptomatic cases, reach their contacts, and encourage testing and quarantine — and they need to do it as fast as possible. While social distancing measures are in effect, tracers will generally only need to identify and test two or three contacts per case. But as regions relax these measures, the number of contacts per case could rise closer to 20.
Several countries — including Taiwan, South Korea, Iceland, and New Zealand — have demonstrated that aggressive contact tracing can be an effective measure in helping to control the spread of COVID-19. These countries have dedicated substantial resources to building both the staff capacity and technological tools to support these efforts. The tech-heavy methods used by Taiwan, Singapore, and South Korea may be difficult to replicate in the U.S., but New Zealand and Iceland’s approaches could be achievable if we can establish a large enough contact tracing workforce.
In short order, America will need to quickly recruit and train people who can build rapport, ask smart questions, and follow leads to reach every possible contact and convince them to self-isolate.
There are no standard education requirements for contact tracers, but they need technical skills — such as an understanding of medical concepts and the ability to use CRM software like Salesforce — as well as strong interpersonal, cultural sensitivity, and interviewing skills. In New York, tracers are required to have a high school diploma or equivalent; team supervisors need a bachelor’s degree, with a nursing license preferred. Community support specialists, who support the needs of contacts who are isolating, need a bachelor’s degree and experience in social services.
Even with the requisite education and experience, workers who are new to contact tracing will need training. Johns Hopkins University is offering a free online course, and the CDC has published guidance, resources, and a sample training plan. Could unemployed journalists, call center employees, or Peace Corps volunteers quickly reskill and apply their talents to contact tracing?
And all of these newly minted contact tracers will need more than training to make a dent in transmission rates. The “testing, tracing, isolation” strategy starts with testing to identify new cases; without extensive testing, tracing won’t deliver results. A recent Harvard report suggests the U.S. will need to deliver 5 million daily tests by early June and 20 million by late July to support “massive-scale testing, tracing, and supported isolation as the path to pandemic resilience for a free society.” Covid Exit Strategy tracks daily, state-by-state progress toward two goals: reaching 500,000 tests per day and 4 million tests per day.
The efficacy of tracing also depends on patients and contacts adhering to isolation and quarantine protocols. Once contacts are identified, tracers must build trust and offer connections to support and services they’ll need while isolated. Contacts need to truly isolate or quarantine themselves, which can be hard for people who live with others. Some localities are using hotels, dorms, or other facilities to offer patients and contacts free accommodation if needed.
Traditional contact tracing methods work to curb the rate of transmission, and we know how to do it effectively. Every expert we contacted emphasized the importance of expanding existing tracing programs; some also saw the potential for an additional “layer” of technology to fill gaps and help the U.S. scale its efforts. Next week, we’ll look at how technology solutions might augment time-tested, human-powered tracing to bring coronavirus contact tracing to scale — and what it will take to make tech tools useful.
Part 2: The role of technology
Traditional contact tracing methods work to curb the rate of transmission, and we know how to do it effectively. But manual contact tracing is a labor intensive process of interviews and detective work. Given the speed and scale of coronavirus transmission, some experts have suggested that relying solely on manual tracing is infeasible and that tech-based tools are needed to improve tracing by helping people recall movements, track contacts, and notify those who may have been exposed.
Every expert we contacted emphasized the importance of expanding existing tracing programs; some also saw the potential for an additional “layer” of technology to fill gaps and help the U.S. scale its efforts. What role should technology play in augmenting human-powered tracing efforts?
Tracing depends on data: where patients have been, who they were near, and how long they were interacting with contacts. Anyone who has tried to recall their past two weeks of activities and conversations will tell you how difficult it is to remember. One answer might lie in the proliferation of mobile devices — but apps have faced a number of hurdles, from privacy concerns to low adoption rates.
Our relationship with health data, on an individual level, often depends on our personal tolerance for risk and what we stand to gain. This risk-benefit ratio is a main point of contention for contact tracing technology — as with any health intervention. In extreme circumstances, people may be more willing to share personal data, but many technology solutions prompt questions about the trade-offs between individual rights and public health. In the United States, the unique dynamics between federal, state, and city governments — as well as the private sector — further complicate contact tracing technology initiatives.
In the second of a two-part series on contact tracing, we look at existing apps, what works and what doesn’t, and nontraditional data sources that can fill gaps in COVID-19 data.
The current state of COVID contact tracing apps
Different approaches to COVID digital contact tracing are being proposed and implemented around the world. Each reflects differing medical, technical, and social factors — including testing and manual tracing capabilities, the availability of smartphones, and societal expectations around privacy. Some contact tracing methods rely on GPS and other location data, some use Bluetooth, and some use a combination of different data sources.
Countries such as China, South Korea, and Israel are collecting and sharing a range of personal data, including closed-circuit TV footage, credit card transactions, and mobile phone location data. Other apps — such as Iceland’s Rakning C-19 and North and South Dakota’s Care19 — are collecting users’ location data. If a user tests positive, they will be contacted by health authorities and asked to share their information.
But this data isn’t perfect: GPS can be inexact in crowded places, such as apartment buildings, and the precision of mobile tower data varies. This type of tracking has also prompted significant privacy concerns. To address this, many — including Singapore, Australia, and the United Kingdom — are using Bluetooth technology instead. Bluetooth technology tracks when other phones are within a certain distance for a set period of time — usually within six feet for 15 minutes. When a user reports a positive COVID-19 diagnosis, the app can notify the people who may have been exposed and, in some countries, local health authorities.
Apple and Google’s “exposure notification” technology, launched earlier this month, also uses Bluetooth. Switzerland just released the first app based on this API; several countries, as well as three U.S. states, are currently developing tools using this technology. Unlike other tools, these can run continuously in the background without depleting battery life or requiring users to unlock their phones.
New Zealand has taken a different approach by creating a voluntary “digital diary” app, which allows users to scan QR codes at different locations to track their movements. It is intended to supplement manual tracing efforts.
The United States has not announced any plans for a national contact tracing tool. Currently, different states and cities are developing their own tools, leading to concerns about interoperability and security. In the U.S., there’s also a possibility that a non-government app could become most widely adopted; if it’s free to use, then user data would likely be harvested for commercial purposes.
The MIT Technology Review launched Covid Tracing Tracker to review “the deluge of apps” against several criteria, including transparency and whether the app collects any unnecessary information. Regardless of which solutions emerge as “winners,” the success of these tools will rely on manual tracing efforts and widespread testing. Assuming notification systems do work, contacts will need access to testing and support during isolation or quarantine.
There’s work to be done
Every COVID contact tracing app is new — and like any digital disruption, this moment presents an opportunity to reconsider how to build a secure, accessible user experience that preserves privacy and encourages adoption.
Most public health experts acknowledge it will be a challenge to get enough people to use voluntary tech tools, and attitudes toward adoption often reflect cultural and societal norms. A recent poll found almost 60% of Americans couldn’t or wouldn’t use the system Apple and Google have developed. In the same poll, 40% said they would be uncomfortable using the app to anonymously notify their contacts if they tested positive for COVID-19. Similarly, in a study by the Brookings Institution, less than one-third of Americans indicated they would likely download and use a mobile contact tracing app. This is much lower than polls in Europe, where polling has found 50% of British, 70% of Swiss, and 80% of French people would use these tools.
Uptake of Singapore’s TracingTogether has only reached 20% since launching in March. In Iceland, 38% of the population has downloaded the country’s app — the highest penetration of a voluntary tracing app so far. This higher uptake is attributed to the app’s clear data retention rules, but authorities say the tool has only had a minor impact compared with manual tracing efforts. Australia’s COVIDSafe tool has been downloaded by 6 million people — about 1.5 million short of its 40% target — and the number of downloads has been slowing as case numbers decline across the country.
What’s the minimum level of adoption required? Estimates vary; some studies suggest 40% of the population would need to use a tool for it to be broadly effective, others suggest 60-80%. To attract and retain voluntary users, apps will need to prove their value — to individuals and to the public at large.
Collecting any type of personal data will raise questions about privacy and security; with health data, those questions multiply. And the concerns are not unfounded: North and South Dakota’s app has shared data with Foursquare and Google’s advertising system, India’s contact tracing app made it possible to locate users who had tested positive, and Qatar’s system leaked users’ personal data.
Even Bluetooth solutions, which aim to preserve privacy, take two different approaches. Centralized systems gather anonymized data and upload it to a remote server; decentralized systems, including Google and Apple’s API, store data on a user’s phone. While the centralized model can give authorities more insight into the spread of the virus, the decentralized approach is thought to offer a higher degree of privacy. Due to these concerns, some countries have announced a shift to a decentralized system. University of Glasgow researchers have proposed a blockchain-enabled app called BeepTrace, designed to provide greater privacy, security, and user control over data permissions.
These tools mean people don’t have to rely solely on their memories of where they were or who they came into contact with. But they exclude anyone who doesn’t own a smartphone — often among the most vulnerable to COVID-19, including older people and migrant workers.
Even well-designed, accessible apps are likely to both overreport and underreport possible contacts. Bluetooth and GPS signals don’t recognize walls, face masks, and other barriers that might mitigate transmission. And self-reporting doesn’t always result in trustworthy data: Health experts know that people don’t always tell the truth about their behavior. People could neglect to record important details or inaccurately report self-diagnosed cases; even worse, trolls could report fake cases.
Successfully deploying any technology product requires user trust and confidence. Introducing an effective coronavirus contact tracing app requires public trust on a massive scale. We’ve already seen fraud that capitalizes on communications technology used in manual tracing efforts: Scammers have sent text messages with malware links or phishing ploys. In addition to balancing user concerns over privacy with public health’s need for accuracy, apps will need to provide a high level of security and protections against abuse.
Approaching the problem from a new angle
If contact tracing efforts need data, and apps aren’t widely adopted, what are the other options for getting useful information? Anonymous, population-level trend data can help us assess adherence to social distancing measures and figure out where to focus efforts.
Health authorities in Italy have used aggregate mobile location data provided by major cellphone carriers to gauge compliance with lockdown policies. And in Taiwan, authorities have used location-tracking mobile apps to monitor and enforce quarantine.
New York Times reporting used aggregate cell phone and mail forwarding data to track people’s movements. A University of Chicago study used cell phone data to identify how crowded different businesses get and how long people typically stay; MIT did a similar study to assess the transmission risk at different locations.
As the broader economy reopens, regions will need the ability to pinpoint outbreaks — perhaps down to the ZIP code level. Could existing devices and systems be adapted to collect data, without the effort and expense of developing and installing new technology? Existing sources of data — transit logs, credit card purchases, traffic cameras, social media, real estate transactions, or smart devices — could be used in nontraditional ways to provide public health insights. But there’s significant potential for algorithmic bias and other ethical pitfalls; any locality that collects nontraditional data for contact tracing will need to mitigate those risks with a plan for its equitable and accountable use.