Maryland Smith Students Assist in Contact Tracing Efforts

Delicately balancing data privacy and public health concerns amid the coronavirus pandemic have created roadblocks for accurate contact tracing around the world. But two recent Maryland Smith graduates are hoping to be part of the solution.

When tasked with a project on leveraging big data, master’s students Laura Klett, MFin ’20, and Gabriel Castro, MFin ’20, realized they shared a mutual interest in improving current contact tracing methods. They teamed up to find a way to positively impact their local communities.

“Some people think privacy is more important, while others value health safety more,” says Klett. “We understand how important it is to protect people’s data, but at the same time, we’re seeing that we can combine ideas that are already out there to contribute to people’s overall health.”

Many have tried their hand at developing contact tracing programs, with most using smartphone and Bluetooth technology to log people’s locations and cross-reference it with the known locations of sick individuals.

Google and Apple announced a partnership in search of a solution. Singapore launched its own app called TraceTogether, which senses a person’s whereabouts and locally records data so people can be reached when they are either positively diagnosed or when they have come into contact with someone who has tested positive.

However, these solutions still feature their own flaws, Castro says, including privacy concerns that might make people reluctant to use the apps.

“A lot of contact tracing methods have been executed manually, which includes phone calls and requiring patients to recall details of their previous interactions, creating a huge margin of error. We believe that if we can automate that process, we can reduce the margin of error and better map the virus’ transmission,” says Castro.

With a goal of optimizing the various contact tracing applications and programs already in existence, Klett and Castro decided to create the framework of an application that primarily uses Bluetooth technology. “The important thing here is that Bluetooth is low-strength,” says Castro. “It’s a major advantage because the virus is transmitted across short distances, which fits right into its range.”

After downloading the app, users’ phones will track interactions with people within the phone’s bandwidth and record the data locally. The key, Castro says, is that the data are anonymous. Users are given a push token upon download, an identifying serial number.

When people are positively diagnosed, they can voluntarily submit the information through the app, which is then verified through hospital networks using Blockchain technology. After verification, the user’s consent is required in order for this information to be stored on big data servers with programs like MongoDB and Neo4j that help store, compute and visualize the data.

This visualization component of their project is critical, Klett says, for better understanding how the virus has mutated, where the virus was transmitted from and whether communities are doing an adequate job at containing the threat.

“It takes this very small community analysis of contact tracing and helps us be specific in our analysis in terms of pinpointing cases,” says Klett. “Adding the visualization component lets us understand the origins of a virus and helps us understand the full story.”

Even with all of this information being anonymous, Castro says it is still possible to determine whether a person has a common network of individuals that they interact with frequently.

“We would be able to see their network, whether it’s a sports team, congregation or school, at a visual level without knowing who they are or why they are together,” says Castro.

Castro and Klett, who graduated in May, hope the framework they created might contribute to other contact tracing solutions and help officials make decisions.

“Governors are trying to understand which areas are the highest affected – it’s a matter of probabilities,” says Castro. “We believe that by understanding these complex networks of individuals, we can almost assign probability values for outbreak risks within communities.”