Mar 24, 2020 - Technology

Tech firms crunch coronavirus data to track disease spread

Illustration of fingerprint patterns with a tracking target on the middle print.

Illustration: Eniola Odetunde/Axios

Tech companies are using artificial intelligence and other tools to comb through coronavirus data to track cases and find transmission hotspots.

Why it matters: Finding patterns in otherwise discrete data points could help make sense of where and how the virus is spreading in the U.S., and could aid in allocating the country's limited testing and treatment capabilities.

Driving the news: IBM, through its Weather Company subsidiary, is planning to launch an incident map as soon as Wednesday that tracks confirmed coronavirus cases and deaths at the county level.

  • The company is reviewing various forms of coronavirus information on county and state websites using AI, including its Watson natural language processing that turns text into machine-friendly data.
  • The project checks the sites for updates every 15 minutes and aggregates them into its map, which can be found on the Weather Channel mobile app or the weather.com website. (IBM owns the digital assets affiliated with the Weather Channel and supplies the network with its weather data.)
  • "We've spent 30 years keeping people safe — that's been our mission through severe weather," Sheri Bachstein, global head of IBM's The Weather Company, told Axios. "And so we feel like we are really uniquely positioned to do this during this type of a crisis. We have the capabilities to provide data at a more local level."

The big picture: The map ties in with a tech industry push, much of it relying on AI, to marshal data to assess where the coronavirus may already be and where it might be headed.

IBM is separately using the data that powers its map alongside other information to create an interactive dashboard for researchers. It will also provide a trends chart showing whether the number of cases is accelerating, flattening or declining.

  • And the company is leading the industry side of a public-private partnership the Trump administration announced Sunday aimed at giving COVID-19 researchers access to supercomputing resources to aid their work. Other companies involved in the project include Google, Amazon and Microsoft.

Google subsidiary Kaggle, an online community of data scientists, opened a competition last week to use data to forecast the number of cases and fatalities that will be confirmed between March 25 and April 22 in a number of regions around the world, with one subset dedicated to California.

  • The primary goal is to identify factors that appear to affect the transmission rate of COVID-19, Kaggle CEO Anthony Goldbloom told Axios.
  • "Does controlling for temperature or humidity improve forecasts? What is the impact of policy actions like school closures, cancelling large gatherings, self isolation policies?" Goldbloom said in an email. "Our hope is that the community can find helpful relationships that are used as an input to healthcare and policy decisions."
  • Separately, Kaggle is hosting an effort coordinated by the White House Office of Science and Technology Policy to make academic literature on COVID-19 and related pathogens available in a machine-readable format, and called on AI experts to use the data to help answer key questions about the virus.

SparkBeyond, a startup that provides AI-driven data analysis tools to businesses, gathered publicly available data on infections and patient routes in Italy to create a heat map predicting the risk of contracting the virus in certain locations, CEO Sagie Davidovich told Axios.

  • The company, which is working on a U.S. map, used AI to draw insights from the data — like proximity to a gas station correlating with viral spread.

Kinsa Health has long made data from its internet-connected thermometers available online, which is proving useful in tracking the coronavirus in real time.

Yes, but: Even with sophisticated analytical tools at the ready, there remain big holes in the data that's available in the U.S., as testing remains very limited and the amount of information that's available to the public varies by state.

Go deeper