Skip to Main Content

A version of this story appeared in STAT’s Health Tech newsletter. Click here to subscribe.

Artificial intelligence is not going to stop the new coronavirus or replace the role of expert epidemiologists. But for the first time in a global outbreak, it is becoming a useful tool in efforts to monitor and respond to the crisis, according to health data specialists.

In prior outbreaks, AI offered limited value, because of a shortage of data needed to provide updates quickly. But in recent days, millions of posts about coronavirus on social media and news sites are allowing algorithms to generate near-real-time information for public health officials tracking its spread.


“The field has evolved dramatically,” said John Brownstein, a computational epidemiologist at Boston Children’s Hospital who operates a public health surveillance site called that uses AI to analyze data from government reports, social media, news sites, and other sources.

“During SARS, there was not a huge amount of information coming out of China,” he said, referring to a 2003 outbreak of an earlier coronavirus that emerged from China, infecting more than 8,000 people and killing nearly 800. “Now, we’re constantly mining news and social media.”


Brownstein stressed that his AI is not meant to replace the information-gathering work of public health leaders, but to supplement their efforts by compiling and filtering information to help them make decisions in rapidly changing situations.

“We use machine learning to scrape all the information, classify it, tag it, and filter it — and then that information gets pushed to our colleagues at WHO that are looking at this information all day and making assessments,” Brownstein said. “There is still the challenge of parsing whether some of that information is meaningful or not.”

These AI surveillance tools have been available in public health for more than a decade, but the recent advances in machine learning, combined with greater data availability, are making them much more powerful. They are also enabling uses that stretch beyond baseline surveillance, to help officials more accurately predict how far and how fast outbreaks will spread, and which types of people are most likely to be affected.

“Machine learning is very good at identifying patterns in the data, such as risk factors that might identify zip codes or cohorts of people that are connected to the virus,” said Don Woodlock, a vice president at InterSystems, a global vendor of electronic health records that is helping providers in China analyze data on coronavirus patients.

As different treatments are tried out, he added, “we can also use machine learning to identify what might be working for the virus.”

It’s too early in the outbreak to do those kinds of analyses, but AI tools may help to accelerate that research once more data becomes available. The true impact of AI in responding to the coronavirus outbreak will probably not be known for several years.

Brownstein said efforts to harness the power of AI to predict the course of disease — and the size of the impact — are unfolding at a breakneck pace. “Groups across the country are developing models for the spread (of coronavirus) domestically and internationally,” he said, adding that is working with a Boston startup called Buoy Health to build a symptom-checker to assess the symptoms of coronavirus that distinguish it from seasonal flu.

That promises to be a huge challenge for public health officials in the months ahead, as they work to allocate resources to contain the virus and manage a possible influx of cases to emergency rooms. “The more we are focusing our intervention efforts, identifying cases as soon as possible and quarantining those cases, the better chance we have of limiting the global impact of this virus,” Brownstein said.