Among the many pandemic-related decisions community leaders must make during the time of COVID-19 is when and how students of all ages should return to in-person classes.
The decision affects virtually every American student from preschool through graduate study.
Recently, a group of students in Ellen Kuhl’s ME233 class — Data-Driven Modeling of COVID-19 — took decision-making matters into their own hands by learning to create computational models that predict how different reopening strategies would affect the spread of COVID-19 on campus.
What they learned went far beyond the nuts-and-bolts intricacies of epidemiological models. The students quickly realized that no matter what reopening strategies universities employ, college campuses are essentially destined to become superspreader hotspots, not just on campus but for the surrounding communities in which they are located. No strategy is perfectly safe.
Kuhl’s modeling class was structured as a series of foundational projects to introduce the students to the essential components of epidemiological modeling and machine learning culminating in a capstone group project in which teams of students created working models of COVID-19 spread.
One such team comprised graduate students Dhiraj Indana, Hannah Lu, Joseph Pace and junior Cortney Weintz. This month, those four became co-authors, with Kuhl and postdoctoral scholar Kevin Linka, on a peer-reviewed paper published in the journal Computer Methods in Biomechanics and Biomedical Engineering.
The capstone project for all the teams was to model various components of the challenge facing administrators — how many students to allow back, what dormitories to populate and how densely to populate them, and what classes to offer.
“It’s classic diffusion modeling,” says Indana, a doctoral scholar in mechanical engineering, about his interest in the class. “At its heart, COVID-19 is a math problem, but my motivation was to help my family, my friends and my fellow students to be safer and take better precautions.”
Indana’s team followed a group of universities that had tried to reopen in the fall and based their models on that rich source of data. They closely monitored data from 30 U.S. campuses with the greatest number of COVID-19 cases. More than half hit peak rates of 1,000 cases per 100,000 people per week within two weeks of returning to class. For comparison, an incidence of just 50 is generally considered “high risk.” At some schools, as many as one-in-five students became infected within the fall term, and four institutions recorded over 5,000 cases. From there, the group hoped to backtrack — or reverse-engineer — which policies worked best, and which did not work at all.
The model showed that lots of campuses saw a huge uptick in cases within the first week or two of reopening, says Indana, who is originally from India, where case numbers spiked mid-September but have since been steadily declining. “It’s almost unavoidable,” he says, “What’s important, however, is how good the university is at testing and identifying these positive cases quickly to rapidly squeeze the disease.”
Lu, a doctoral scholar in energy sciences who grew up in China, normally models things like the flow of oil or water below ground, but like Indana, found that her previous studies translated well to epidemiology. “Whether you’re talking about groundwater or a disease through a population, it all flows,” she says.
Lu said that as one who hopes to pursue a career in research it was an opportunity to explore how the skills she was amassing could be applied outside her particular area of expertise. Lu enjoyed applying her engineering skills to a real-life problem that affects most every person on the planet right now. “It’s quite exciting, actually,” she says.
Team member Pace, a doctoral candidate from Fresno, California, who studies the mechanical properties of human skin, was struck by the diversity of the team — both geographic, with members from India, China and the U.S., and academic, drawing students from engineering, energy and computer science — that played a part in the ultimate success of their work.
“The main thing that differentiated our group is what topics we were able to look at based on our diverse backgrounds,” he says. “I feel like we ended up using pretty much all our complementary knowledge in the model.”
The lone undergraduate on the team, Weintz, of Menlo Park, California, brought computer science expertise. He says that two findings from the team’s model leapt out at him: First is the value in trying to keep the initial number of cases as low as possible, perhaps by testing students prior to their traveling to campus. And second, once the students arrive on campus, how important frequent testing and strict enforcement of quarantines are. “Even if 19 out of 20 students quarantines effectively, you can still get an outbreak,” Weintz says. “You really have to shoot for near-total compliance and that’s pretty much impossible.”
Witnessing the teamwork of her students, Kuhl found herself learning through the experience. “Even after 15 years at Stanford, I’m still amazed by our students,” she says. “They find each other in diverse and multidisciplinary teams, discuss important problems, identify challenges and come up with creative and innovative solutions to big problems. To me, that’s what makes teaching at Stanford unique, even in the middle of a pandemic.”