University of Virginia scientists wield statistics to prep for coronavirus’ next moves

When the coronavirus reached Virginia, public health officials worried there would be so many patients, they would need to start building field hospitals right away.

Across the country, there were jarring images of states filling parking garages and sports arenas with extra beds, reminiscent of war zone triage.

But a team of University of Virginia scientists, part of the Biocomplexity Institute, told the state to wait. The governor’s stay-at-home order, quarantines and social distancing that began in March 2020 could slow the disease’s spread.

Based on the institute’s projections, Virginia had three or four months to get cases under control. If it didn’t work, the team advised that health workers set up small COVID-19 units that could be moved wherever needed.

A couple months later, Virginia’s hospitals weren’t overwhelmed, rendering auxiliary COVID-19 wards unnecessary. Meanwhile, other states’ surge hospitals sat mostly empty.

“It turned out to be a good decision,” said Madhav Marathe, director of the institute’s network systems science and advanced computing division. “A facility like the one that was established in New York City, the Javits Center, would have cost us $300 to $400 million or thereabouts.”

That was the first policy question state officials posed to the institute’s scientists some 16 months ago. Since then, Virginia’s health and emergency management leaders have relied on the team’s expertise to help navigate pandemic policy.

Marathe’s division, supported by 70 people, including students, is the quiet force behind the scenes, monitoring the disease’s transmission, medical supply levels, vaccinations and the prevalence of more contagious variants.

Even public attitudes and behaviors are factored into their statistical models.

By April 2020, it was clear COVID-19 was going to be a major life-altering event. The state struck a deal with the scientists, who had previously assisted with seasonal flu forecasts. Under the contract, the state has paid the institute $576,000.

Each week, the team gives Virginia projections on the numbers of infected and hospitalized people. In the beginning, public health officials and hospitals feared they’d run out of ventilators, so the institute’s reports included ICU patient estimates, drilled down to the local level.

If you’ve followed the institute’s work, you may have noticed the team warned about a month in advance that the delta variant, a more infectious mutation of the coronavirus, was a looming threat for another surge.

In a prescient interview in April, while the more contagious strain raged in India, Bryan Lewis, one of the U.Va. computational epidemiologists, said the belief that the crisis was over was a dangerous misconception.

There weren’t nearly enough vaccinated Virginians to stop another wave. All one had to do was look at India to know what could happen, he said.

“There’s still a lot of people sitting on the sideline, and those are the folks that can generate those sizable peaks (in cases), especially if we go forward with the attitude that you know it’s over, and that we can start mixing at the level that we were doing pre-pandemic,” Lewis said. “We can’t think of herd immunity as a 70% line, and then we can party again. Thirty percent is still enough folks that you can get a reasonable amount of cases generated.”

The researchers have been two steps ahead of the virus, studying how human behaviors — travel, trade and socializing — affect its spread and evolution. Vast increases in available data, coupled with advances in computer technology and data analytics, make it possible for the team to make quicker disease-related forecasts.

For example, they use networks, such as Facebook, Twitter and Google, as graphic representations of human interactions to study problems.

“Ten years back, we had to explain what networks were, but now networks are ubiquitous,” Marathe said. “In epidemiology, diseases spread when people interact with each other or a person interacts with a pathogen or a vector. The same things happen when we talk about social contagions, with memes and innovations.”

Often their work has been misunderstood as ”predictions.” The scientists are quick to point out they don’t hold a crystal ball. Their models can’t show what will happen but what is likely to happen in certain scenarios.

That means different interventions — masks, social-distancing, policy changes and vaccines — can change the course of the pandemic and avert some of the worst outcomes.

The collaboration among the scientists began in Los Alamos, New Mexico, more than 20 years ago. Marathe and Stephen Eubank were postdoctoral fellows working with Chris Barrett, now executive director of the institute. There, they studied urban transportation to understand how traffic forms and policy could affect its outcomes.

Through transportation modeling, the group had a handle on estimating how people in an urban region come into contact with each other — similar to the way infectious diseases spread.

After 9/11, some wondered whether people should be vaccinated against smallpox, despite that it had been eradicated decades earlier. It was a biodefense question: Should we vaccinate against an extinct disease in case an enemy tries to grow and spread it?

As a result of their contribution to that debate, the group became one of the founding members of a National Institutes of Health consortium, Models of Infectious Disease Agents Study.

In the first MIDAS project, they started tackling pandemic influenza. Many of the issues debated academically then are now playing out with COVID-19, Lewis said. One of the questions was whether schools would even shut down if the government told them to because school boards have jurisdiction.

That has perhaps come full circle with renewed conflicts between the Northam administration and school districts in August over mask-wearing rules. State Health Commissioner Norm Oliver signed an order mandating that schools enforce mask requirements, forcing the hands of school systems that did not want to impose them.

In 2004, Barrett moved the lab to Virginia Tech, where it would become part of the Bioinformatics Institute, later renamed the Biocomplexity Institute. Three years ago, he and dozens of research faculty moved the program to U.Va.’s campus.

Over the years, the scientists have been involved in modeling most major epidemics, including swine flu, Ebola, Zika and MERS. Their research has been funded by the U.S. Centers for Disease Control and Prevention, the National Science Foundation and NIH. They’ve even dabbled with private clients, such as Accuweather, providing flu forecasts.

The institute is also providing weekly pandemic projections for the Department of Defense — national versions of the information given to Virginia. But not all states have gone the route of the commonwealth, consistently using statistical models to inform its public health decisions.

“They were enlightened enough to see the value of models, and I say this because this is not the case universally,” Marathe said. “Our state feels the value of science and evidence-based policymaking, and they take that very seriously.”

Despite working through various scenarios to help Virginia leaders make choices, the group says it stops short of getting political, even if they personally believe the findings support different strategies.

Lewis, whose background is in public health, says it’s important to remember health departments have limited authority. And getting cooperation with mitigation measures — even ones that bear the weight of a mandate — is a challenge. Whether people comply with mask-wearing, vaccinations and social distancing affects their impact.

“The reality is it’s a dance between the public and the government, as we’ve seen to varying degrees of success during COVID here,” he said.

Today, the institute tries to build statistical models to analyze targeted interventions. Rather than broadly answering whether herd immunity would help a community, for instance, they’re trying to get to the bottom of more complex questions, like:

If you vaccinate these specific critical workers, what will happen?

How much benefit comes from convincing this specific demographic group to wear masks?

Are there early warnings or triggers that should be monitored to help inform policy?

Statistical modeling has evolved over the past year and a half. Initially, when there were no vaccines, all the scenarios were based on behavior changes, such as people staying home, closing down airlines and shutting down systems.

The second phase looked at different amounts of vaccinated people in a community and how that worked as a control on transmission.

Srini Venkatramanan, one of the scientists who has developed models for the state’s forecasts, said the questions have evolved.

“When will we hit the vaccine hesitancy wall in different counties, that’s a question that was on the horizon, but like far-off horizon, way back in December, when vaccines were just getting emergency use authorization,” he said. “The model evolution is driven by both what is known, and what do you want to know, and models are supposed to connect those two.”

The third phase — where we are now — centers on mutations. It’s not known whether a new strain could breach existing vaccines, but that’s something the group will continue to monitor. The vaccine has already proven to be less effective against the delta variant than it has been for previous strains, though it is still good at preventing serious illnesses and death.

Researchers at the institute also continue to study the pandemic to find ways to improve public health approaches in the future.

In one paper, a team made a hypothetical case for prioritizing vaccines to individuals who could be so-called super spreaders, rather than focusing on people of specific age groups and occupations. The researchers found that by distributing to people with the greatest number of social contacts and time spent with others, whether a grocery clerk or a young social butterfly, they could significantly reduce casualties.

Preliminary findings estimated the strategy, compared to the age-based approach, could have yielded 3 million to 6 million fewer infections, 180,000 to 306,000 fewer hospitalizations and 51,000 to 62,000 fewer deaths in the United States.

In another project, the institute looked at the potential medical costs of keeping the U.S. economy open in the first wave of the pandemic. That paper used a dozen scenarios to estimate the number of people who would likely get infected and require medical attention and ventilators.

The study found over 96% of the total medical costs could be eliminated if 90% of people complied with lockdowns for 45 days or more, schools were closed and people with symptoms stayed home for two weeks. Without any mitigation, the medical costs would have been over $1 trillion, but with social-distancing alone, that loss could be reduced to $35 billion, according to the analysis.

Despite their scientific approach, the institute’s researchers know some laymen believe a hidden agenda lurks beneath their work, especially if the findings show favorable results for prolonged mask-wearing and vaccinations. Those two have become sore subjects, particularly among Americans who believe they infringe on personal freedoms.

Eubank, a computational epidemiologist, said he wishes people would engage in a debate on public health rather than default to their partisan bases. He compares pandemic mitigation measures to when seatbelt laws were made into a requirement for states to receive highway funding.

“What does it mean to say, ‘I’m not willing to endure any minor discomfort in order that other people don’t get sick from a possible lethal disease? How does that square with other moral values that I hold?’ I think it would really help to get beyond just, ‘Oh, this party says this, and that party says that,’” he said.

“That’s more a question of the media than modelers in general, but I’ve been missing that debate over the last year.”

Elisha Sauers, 757-839-4754, elisha.sauers@pilotonline.com