Wearables like the Apple Watch measure your heart rate, and when you visit the doctor, she can review that data with you to accurately monitor your progress with a heart condition. However, when you test positive for COVID-19, your doctor and local health department have no data about who else you may have infected. What if you could bring high quality location data to your doctor's visit? Your doctor can pass data to the city health department so they can execute a contact trace - in other words, identify the people you've likely infected while incubating the virus. Contact tracing today is primarily based off of patients' incomplete memory of who they were with and what they were doing during the last 14 days. Needless to say, patients' memory are imperfect. What if a patient could give their doctor and local health department an extremely high quality copy of location data from their phone? The city can determine your infectious paths and anonymously reach out to anyone you made contact with to prevent spread. For example, if you were in an Uber Pool last Tuesday, and the city estimates you were incubating a virus during that time, then the city can contact the driver and let them know they likely were in contact with someone with the virus. The city is able to determine who the driver was, because the user also has a receipt of that uber pool ride in their email. Epi-Collect can cross-reference your location data and Uber receipts to identify this class of potential super-spreaders, and massively improve the epidemiological outcome for cities with fewer than 500 cases.

What it does

Epi-Collect ( ingests location data from Google using Google Takeout. We are able to retrieve a copy of this data thanks to the GDPR. This data quality is so high quality that, not only do we know exactly when and where an incubating user was, but also we have an estimate of whether that person was in a car, in the public transit system, riding a bike, or doing some other activity. This is called activity detection, and it is attached to every data point from Google Takeout. Additionally, because we use historical rather than realtime data to do contact tracing, users do not need to install a privacy-invading location tracking app to save lives.

How we built it

We built a generic web app that can be used by doctors, city health departments, and academics to ingest Google Takeout location data. It uses React, Python, Postgres, and PostGIS. We released the code under an MIT license so that any city in the world can run their own contact tracing program. If they make improvements to it, then all other users can benefit too. We also built a data access program for researchers and a  privacy philosophy.

Challenges we ran into

We initially thought this data would be best for academic research purposes. Many contact tracing apps are in development, so we also thought we were making a mistake by pursuing a non-realtime project. As we worked through some of the use cases, we quickly realized that historical contact tracing is actually more valuable than realtime tracing. Why do you need realtime tracing if everyone should be staying home in the first place?

What's next for Epi-Collect

We have a meeting with an employee from the Los Angeles Department of Public  Health to discuss potential next steps. We would like to collaborate with city governments and academic researchers. We also would like to finish our front end, which is in dire need of polishing. We also would like to translate the app into as many languages as possible, so that it can be deployed in places where it's needed most.

Try It out



amazon-ebs, amazon-rds-relational-database-service, amazon-web-services, postgis, postgresql, python, react, typescript

Devpost Software Identifier