������Ƶ

When infection rates rise or — in extreme cases — a pandemic hits, it’s impossible to test everyone. Time, money and resources don’t allow it. So what are the best strategies to track and curb disease outbreaks?

������Ƶ Assistant Professor Melissa Zeynep Ertem has led . The research, published in the journal Communications Medicine, used machine learning to examine data from more than 650,000 students attending over 2,100 Massachusetts elementary and secondary schools during the 2021-22 academic year.

Collaborators on the study are Anseh Danesharasteh, PhD ’24, as well as Westyn Branch-Elliman (Boston University), Richard E. Nelson (University of Utah), David Berlin (Weill Cornell Medical College), Dr. Lloyd Fisher (University of Massachusetts) and Elissa M. Schechter-Perkins (Boston University School of Medicine).

“This kind of research is important for understanding any future disease that we can test for,” said Ertem, a faculty member at the Thomas J. Watson College of Engineering and Applied Science’s School of Systems Science and Industrial Engineering. “We learned during the 2009 flu pandemic that school closures helped a lot. Our decisions for COVID-19 were based on past diseases, and this data will improve our knowledge for future airborne illnesses.”

During that academic year, Massachusetts schools used three methods to track COVID:

• Surveillance testing chose individual students randomly from the population.
• Pool testing brought together five to 10 students for one test; if that came back positive, each would be tested individually.
• “Test to stay” required students with symptoms or infected close contacts to be tested; a negative test would allow them to stay in school, while a positive test meant the student was quarantined.

“There are three types of testing because doing everyone in every school would be very costly,” Ertem said. “We could test everybody every morning for a better idea about what is happening, but do we have enough resources for that? Also, would families give consent to test every day or every other day? We need to find the best strategy if or when a pandemic happens again.”

The 2021-22 academic year included students returning to in-person learning, the COVID vaccine became more readily available, the rise of the omicron variant and the loosening of mask restrictions. The study considered many factors, including location, race, ethnicity, socioeconomic status and how incident rates in schools affected the wider community.

One of the biggest challenges that Ertem and Danesharasteh faced was parsing through all the supplied data and creating an algorithm that would offer the best analysis.

“If we just look at the numbers without any model behind it, we will see an increase or a decrease, but we won’t know the cause of that decrease or increase,” Ertem said. “Our methodology tells us whether vaccination and natural immunity were helping. How they affected the overall disease spread involves several complex systems.”

The research concluded that community immunity gained through prior infection or vaccination combined with the various testing strategies were safe and effective for allowing in-person learning. Test-to-stay programs proved the most effective way to curb the spread of infection, with a 3% to 22% decrease in positive test results. Surveillance and pool testing led to smaller decreases between 1% and 4%.

Ertem appreciates the need to have students in school during disease outbreaks and pandemics, but safely needs to be the top priority, and her research can help to guide policymakers’ decisions.