Skip to content Skip to navigation

Research & Ideas

Search this site

Can contact tracing work at COVID scale?

Data gathered through a nationwide operation to trace people exposed to coronavirus can help us learn more precisely who needs to be quarantined and for how long.

A digital illustration of people wearing masks, social distancing, walking, and being connected by dotted lines

If wearing masks is controversial, how will Americans respond if COVID-19 tests lead to requests for voluntary quarantines? | iStock/elenabs

What happens when millions of healthy, symptom-free Americans are told to voluntarily lock themselves in a two-week quarantine for the public good, based only on a possible exposure to someone with COVID-19?

We may soon find out.

Contact tracing — the process of identifying and quarantining people who may have been exposed to a person with COVID-19 — is so important to breaking chains of transmission and containing community outbreaks that states plan to hire tens of thousands of contact tracers in the coming months. Tech companies are racing to build apps and technologies to help. However, without building an infrastructure to measure and calibrate these efforts, the push to identify more contacts faster may result in the quarantine of so many people that the whole endeavor becomes untenable.

Identifying contacts of new cases by itself doesn’t slow down the disease; those contacts must also take a transmission-busting action. For COVID-19, that action is an immediate self-quarantine lasting 14 days from exposure. This applies to all contacts — even those with no symptoms or a negative test. Most of these contacts won’t go on to get COVID-19 by the end of that period. These are regular people who will be asked to put their lives on hold for two weeks for the public good. That means staying home from work (and possibly losing income as a result), arranging for groceries and other necessities, and delegating childcare responsibilities. It is not an easy ask, especially for those in socioeconomically challenging situations.

The U.S. is currently reporting upward of 20,000 new COVID-19 cases per day, and each case can average more than 30 identifiable close contacts. If these numbers hold, successful contact tracing could mean that, every day, hundreds of thousands of people would be asked to enter voluntary 14-day quarantines. If this is the case, the number of people under rolling quarantine on any given day in the U.S. could easily run into the millions. In a country where quarantines are voluntary, the sheer number of people under quarantine combined with the onerous length of the quarantine period may encourage many to simply disregard quarantine recommendations, undermining contact tracing efforts.

The need for a data-driven approach to quarantine

As we build our nation’s tracing operations, we need to ensure that they are effective at identifying contacts while attempting to quarantine as few people as possible, for as short a duration as possible. To ensure contact tracing remains viable at scale, we must develop data-driven metrics to evaluate and adapt our contact tracing efforts. Historically, successful contact tracing has been measured by its sensitivity: how many contacts are identified (more is better), how many are contacted (more is better), and how many are quarantined (more is better). However, at scale “more is better” breaks down. We must have corresponding metrics for specificity, to ensure we are also working hard to exclude from quarantine those people who have not themselves become carriers of the virus.

Regularly testing a sample of contacts under quarantine to see how many go on to get COVID-19 can serve as a scorecard to calibrate our tracing efforts. If a large fraction of quarantined contacts test positive, then we know our efforts are bearing fruit, and perhaps we should even be more aggressive. If the percentage were too low, we would know that we are largely quarantining false positives, and we should back off or retarget our approach. Dissecting this data could help us understand risk profiles of different contacts and exposure types. Which types of exposures are more likely to transmit disease, and which are less likely? What age and demographic features place individuals at increased risk during an exposure event? This knowledge would allow us to more precisely target exactly who needs to be quarantined and who can be spared. Establishing and operationalizing sensitivity and specificity metrics rooted in testing can power an adaptive contact tracing effort, one that is continuously calibrated with data.

Learning from contact tracing data

The data from contact tracing generates an invaluable resource — a group of people with known dates of exposure who we can follow over time. Studying what happens to those who go on to get COVID-19 can help us understand key features about the disease itself. What symptoms appear first, and how long after exposure? What trajectories portend worrisome outcomes, and which will be OK?

Consider one possible application: using this data to understand, for those who go on to get the disease, how many days after exposure they typically test positive. Using this knowledge, we could develop a strategy to time a COVID-19 test that could “clear” an individual and end the quarantine short of 14 days. People could return to work sooner. Travel quarantines (international and domestic) could be shortened. The social and economic burden of fighting the disease could be reduced.

Digital exposure notification technologies to augment contact tracing, such as those announced by Apple and Google, can help study and understand the disease, even if they are not adopted broadly enough to become the backbone of entire contact tracing operations. Despite their potential — that smartphone technology can be used in a privacy-preserving way to instantly alert people of recent exposure to someone with COVID-19 — it does not appear that apps based on this technology will be adopted in the U.S. at the levels required to shoulder a meaningful load of contact tracing activities.

Even if they are unevenly used, however, these apps can be useful. In the same way that app-based research allowed Apple and Stanford University to recruit hundreds of thousands of participants for a study on heart conditions, contact tracing apps can have the reach and speed to direct exposed individuals to participate in studies of benefit to science or public health. Such studies could use digital measures of exposure time to better understand which types of interactions ultimately lead to disease transmission or help answer more basic questions about the natural history of the disease itself. But these use cases are not built into the technology; they must be conceived, coordinated, and supported by researchers and public health officials.

The United States is embarking on a monumental operational endeavor — creating, within a few short months, a nationwide contact tracing operation employing tens of thousands of people to chase the novel coronavirus and stop its spread. As we build out the tracing infrastructure to reach and guide individuals exposed to COVID-19, we must also remember to build the data infrastructure to learn from them.

This article originally appeared on the Health Affairs Blog, July 8, 2020. Copyright © 2020 Health Affairs by Project HOPE — The People-to-People Health Foundation, Inc., and reprinted as allowed for the authors’ noncommercial use. The authors have received grant support from the Wallace H. Coulter Foundation, Stanford Bioengineering Department and Stanford Biodesign Program to investigate digital contact tracing for COVID-19.