News

Astronomers Preview Giant Magellan Telescope, Discuss Extraterrestrial Life at DRCLAS Event

News

Drug May Reduce COVID-19 Mortality Rates for Diabetic Patients, Study by HMS Professor Finds

News

Computer Science Courses Pilot New Teaching Fellow Inclusion Training

News

Harvard Economics Professor Isaiah Andrews Wins MacArthur Grant

News

Harvard Medical Society Renamed in Honor of First Black Tenured Professor, Physician-Scientist Hinton

HMS Researchers Develop New Tool for Early Detection of Local-Level COVID-19 Outbreaks

Researchers at Harvard Medical School have developed a new machine learning-based online tool for the early detection of COVID-19 outbreaks.
Researchers at Harvard Medical School have developed a new machine learning-based online tool for the early detection of COVID-19 outbreaks. By Ryan N. Gajarawala
By Meera S. Nair, Crimson Staff Writer

Researchers at Harvard Medical School have developed a new machine learning-based online tool for the early detection of COVID-19 outbreaks within individual U.S. counties.

The COVID-19 Outbreak Detection Tool — which was developed in partnership with researchers at Massachusetts General Hospital, Georgia Tech, and Boston Medical Center — includes an interactive map dashboard that color codes counties by predicted COVID-19 case count doubling time. The tool also includes a “data explorer” table which can sort counties by a variety of relevant parameters, such as 14-day new case trends or average daily cases in the past week.

To predict the doubling time of COVID-19 cases in each country, the tool takes into consideration various reported parameters, including COVID-19 cases and deaths, face mask mandates, social distancing policies, changes in tests performed, and rates of positive tests. The tool also accounts for the CDC’s Social Vulnerability Index, which quantifies the health-related resilience of individual communities when confronted with natural disasters or disease outbreaks.

Zhaowei She, a Ph.D. student at Georgia Tech and member of the research team, explained how the COVID-19 Outbreak Detection Tool has already been successful in identifying “regional outbreaks that are often underreported in nationwide media,” describing an ongoing outbreak in the North Slope Borough of Alaska which the tool was able to predict and monitor.

“People in the rest of the U.S. might not be able to notice this outbreak if this tool was not there,” she said.

In a MGH press release, Turgay Ayer, an associate professor of industrial and systems engineering at Georgia Tech, underscored the need for outbreaks to be detected early, noting that one major challenge public health officials are facing with coronavirus is that “it may take days or even weeks for humans to manually detect an outbreak.”

“Our data-driven machine learning–based solution significantly speeds up and automates that process,” Ayer added.

In an emailed statement, Assistant Professor of Radiology Jagpreet Chhatwal wrote that the tool was able to detect multiple local outbreaks in advance, citing examples in Johnson County, Iowa; Craig County, Oklahoma; and Monongalia County, West Virginia.

Chhatwal wrote that the new outbreak detection tool — which is updated at least once a week — could help contain the pandemic writ large by breaking the transmission chain between individual counties.

“In order to contain the spread of COVID-19 and minimize the damage, we need to break the transmission chain as soon as possible at the community level,” Chhatwal said. “The COVID-19 Outbreak Detection Tool is designed to detect such outbreaks early on, so that appropriate actions – such as closure of schools, restaurants, etc. – can be taken right away. By taking action at the local level, we can also avoid lockdown of the entire state.”

—Staff writer Meera S. Nair can be reached at meera.nair@thecrimson.com.

Want to keep up with breaking news? Subscribe to our email newsletter.

Tags
ResearchHarvard Medical SchoolScience NewsCoronavirus