The Modeling Data Behind Duke's COVID Response Plan

The Modeling Data Behind Duke's COVID Response Plan

Midway through the fall semester, Duke has seen zero COVID transmissions traced back to the classroom. This is a result of a COVID response plan that in several ways distinguished itself from those at other universities: Masks in the classroom, full vaccination of faculty, staff and students, and heavy surveillance testing and quick turnaround of results.

A month of declining and low numbers of infections shows that the response is working; Last week at the Academic Council a team of faculty experts and senior administrators showed how it is working. The speakers presented the modeling data behind their recommendations and how the numbers shaped a plan to safely bring back students to campus and in-person classroom learning.

With a forthcoming article in a leading medical journal, Duke’s effort is gaining national attention. However, the speakers Thursday weren’t declaring victory: High rates of unvaccinated people in the surrounding community and the risk of new COVID variants present continuing concerns as winter approaches.

There was, however, optimism. “We’re not seeing classroom transmission. We’re also not seeing on-field transmission in athletics. We’re not seeing a lot of transmission in daily activities,” said Dr. Cameron Wolfe, an infectious disease specialist at Duke Health. “In clinical settings, we’re finding masks have had a significant effect and offer protection.”

The models used by the response team relied in part on a full year of data collected in 2020-21 while there were fewer unvaccinated students and staff on campus. The success of the testing protocols, contact tracing and social distancing gave the modeling team confidence that the university could safely move to full in-person learning. But other aspects of the response plan were in question.

Surveillance Testing Is Needed

In June, the models were based on the assumption that the campus community would be fully vaccinated and vaccine effectiveness would be above 90 percent against COVID The data then suggested that heavy surveillance testing wouldn’t provide any additional reduction in the spread of COVID. 

“The questions I wanted answered were ‘How do I deliver a great course while keeping people safe?’ and ‘How do I optimize the student experience while keeping people safe?’”
-- Steve Haase

“What we saw in the numbers was what herd immunity looks like,” said Steve Haase, professor of biology and medicine. “Then the Delta variant came along.”

While vaccines continued to be effective against Delta, and provided strong protection against serious illness, the effectiveness in reducing infection against Delta dropped to around 70 percent.  The model showed that campus infections would significantly increase unless heavy surveillance testing was put in place. All residential undergraduates are now tested twice a week.

In the 2020-21 school year before vaccines were available, the use of surveillance testing, along with rapid turnaround of test results, was credited with keeping infections low.

“This is why we committed in June 2021 to continuing surveillance testing of students and the university community,” Haase said.  “At the time we were the only university to do so. Now many are following suit.”

 

But Quarantine Isn’t Necessary

Last academic year, quarantine of students in close contact with others who tested positive was also a central element of the university COVID response. In a semester of mostly in-person learning, continuing heavy use of quarantine would provide a significant disruption for many students.

But the same model that recommended surveillance testing showed that quarantining of close contacts was no more effective in reducing infections than increased testing scheduling of these contacts, Haase said. One reason for this is because the quick turnaround of results allowed students to be quickly isolated if they tested positive.

 

Provost Sally Kornbluth talks with Duke biologists Steve Haase and Grey Wray about the data-driven COVID response plan.

How is COVID Changing on Campus

One valuable aspect of the Duke COVID response is that the team is closely following the changing genetic makeup of the infection. Several variants, predominantly the Alpha or California variant, were first seen as early as January 2021. 

Delta arrived on campus in week 47 of the testing program, in July 2021 and immediately took over. Since week 48, nearly 100 percent of the infections have involved the Delta variant.

“That was an extremely rapid takeover,” Haase said. “What we are looking at now is following the sublineages of Delta.” Indeed, in recent weeks, while infections have declined, as much as 40 percent of the infections have involved Delta sublineages, signal of an expected evolution of the strain.

 

Duke and Community Health

Duke’s response has not just benefited the university community but also assisted the public health of Durham and the wider community. Haase noted that the university’s weekly positive rate of 0.09 percent is lower than the 0.2 percent rate in Durham and 0.3 percent in Wake. Both county numbers probably involve underreporting of cases because neither have significant testing of asymptomatic individuals.

As a faculty member teaching in-person classes this year, Haase said the model was devised to guide faculty and students in having as full an academic and social experience on campus as possible.  “The questions I wanted answered were ‘How do I deliver a great course while keeping people safe?’ and ‘How do I optimize the student experience while keeping people safe?’”

One question asked at the council meeting can’t be answered yet: Will fans be able to pack Cameron Indoor Stadium for basketball season this winter? Wolfe said the team is collecting information about ventilation systems and gathering information from a semester of events on campus, but one important factor will be community trends of infection during the season.

“At the end of the day, the game is played in a closed, packed system. What worked for students last year when they played, what kept them safe, is that they were masked and spaced,” Wolfe said.

Haase added that the best models predicting infection trends work only for two weeks out. “Beyond that there’s no way to make a good prediction,” Haase said.