0 Speaker List


It’s Time for a National Pandemic Prediction Agency

Caitlin Rivers, PhD, MPH

THE BIG IDEA that might save the world from the next catastrophic pandemic isn’t totally buried in the Biden administration’s Covid-19 strategy, but it isn’t exactly above the fold, either. After a flick in the Executive Summary, you’ll have to scroll quite a ways down—to page 115 of the 200-page plan—to find it: “To improve the United States’ preparedness, the Administration will work to secure funding and Congressional support to establish an integrated, National Center for Epidemic Forecasting and Outbreak Analytics to modernize global early warning and trigger systems to prevent, detect, and respond to biological threats.”

That’s it—federal PreCrime for pandemics. Precognitive epidemiology. Make up whatever sci-fi words for it you want; the fact is, one thing the Covid-19 pandemic proved is that pandemics can happen, and certainly will again. Building a place to develop the sophisticated models and simulations that can give a hint of when and where an outbreak will hit, and give guidance on how to stop it … well, that sounds like a pretty good idea.

That notion has been kicking around in wonk circles since the years after the anthrax attacks of 2001, and it comes back up with every big disease outbreak. Two longtime advocates, epidemiologist Caitlin Rivers of the Johns Hopkins Center for Health Security and Dylan George, a vice president at the intelligence agency-affiliated venture capital firm In-Q-Tel, laid it out most recently and in more detail in an article in Foreign Affairs. Think of it, they say, like a National Weather Service, but for predicting and studying pandemics and disease outbreaks rather than hurricanes and tornadoes. It’d combine data gathering capabilities with a centralized approach to the kinds of epidemiological and statistical models that featured so heavily in the first months of the Covid-19 pandemic.

The US climate and weather infrastructure combines data from buoys on the ocean, readings from barometers and thermometers everywhere, and satellite images, using predictive engines to generate analyses and simulations on everything from how climate change is making hurricanes worse to where cargo ships should go to whether you should carry an umbrella. So, similarly, an outbreak analysis center might combine genomic surveillance and public health data with, say, notes on mosquito and bat populations, to point to where the next outbreaks might break out. “We have public health emergencies all the time, even more than people realize,” Rivers tells me. Before Covid-19, there was Zika, Ebola, H1N1, H5N1, SARS, anthrax—not to mention seasonal influenza, or longstanding global threats like tuberculosis. “These crises just feel like they’re continuous, and every time, there’s a need for this analytics capability. But it’s usually just modelers working in academia who volunteer,” she continues.

That’s no way to run a country in an emergency—especially when resources to deal with public health crises come from the federal government but the policies and on-the-ground deployments happen at the state and local levels. “When you are trying to incorporate people with a range of different skills or perspectives in the middle of a crisis, who may not have experience working at the speed of an outbreak or sitting with decision-makers, it’s hard to cobble that together,” Rivers says.

To be clear, she’s been saying that. About a decade ago, she and George were on a task force set up at the Office of Science and Technology Policy to study pandemic predictive capabilities. It looked like one of the problems with the government’s handling of the H1N1 pandemic had been a push-pull in the advice that epidemiologists were giving to responders—dueling models. George says that, at the time, the Centers for Disease Control and Prevention, the heart of the US federal public health infrastructure, didn’t really have the capability to evaluate which models were the right ones at the right moment. And there wasn’t enough of a standing capability to build the best models from scratch. “When a hurricane comes barreling onto the East Coast, we don’t randomly ask modelers at academic institutions in the US, ‘Hey, could you drop what you’re doing and model where this hurricane is going to hit?’ There’s been this progressive investment in people, models, systems, and data to improve forecasting skill,” he says. “We are in the early stages of infectious disease and pandemic forecasting. I’m confident we can get much better at it if we do a similar investment.”

This isn’t just, you should excuse the phrase, academic.

Remember trying to figure out, in early 2020, which Covid model was getting things right? Initially, the influential Institute for Health Metrics and Evaluation—respected, well-funded, and with the imprimatur of the Gates Foundation—was using a “curve-fitting” approach to projecting numbers of potential deaths, rather than a more classical epidemiological approach that looked at the numbers of susceptible, exposed, infected, and recovered people. The IHME model and projections influenced a lot of policy decisions, and got a lot of numbers wrong. Some epidemiologists complained, and by summer the models were more sophisticated. “The IHME models got better and better over time, but they should have started off where they were in the summer, and that didn’t happen because a lot of that learning had to occur in real time during the pandemic,” says Sam Scarpino, director of the Emergent Epidemics Lab at Northeastern University. “A lot of the data systems had to be stood up in real time during the pandemic. The National Weather Service doesn’t learn to forecast a thunderstorm as they see clouds gathering over the plains of Kansas.”

The center would also become a central place to gather all that data, via public health surveillance and lab work—the equivalent of ocean buoys and satellites—and for dissemination of that information to local public health workers. Right now, lots of the most important data is siloed among different researchers and labs. The result is, disease modelers have to wheel and deal to get access to data, and production of models responsive to emergent problems is ad hoc. Meanwhile, if you want to know whether you need to double up your masks today, there’s no National Epidemic Center web page where you can check the forecast.

Leaders and public health officials don’t have a centralized place to go for help, either. That’d be another area where a center could really help—a scientist who’s great at modeling an outbreak lickety-split might utterly suck at communicating those findings in a way that’s useful to a governor, mayor, or FEMA administrator. “Modelers will refer to this as ‘peacetime’ and ‘wartime.’ Wartime, there’s a big effort. But between outbreaks there’s a chance to learn from what happened last time and improve the modeling methods, and improve links with decision makers and other stakeholders,” says Jean-Paul Chretien, a program manager at Darpa who ran the Pandemic Prediction and Forecasting Science and Technology working group at OSTP that Rivers and George were part of, and then worked on pandemic preparedness for the Defense Intelligence Agency.

Working with those stakeholders is its own kind of special skill, one that not all scientists have. A new center could help foster that, too. “When I was embedded in Health and Human Services during the 2009 H1N1 pandemic, and we were developing models to help the national response, many times when I was interacting with decision makers, I only had a limited amount of time to communicate what our model was showing. I didn’t have an hour to give a PowerPoint presentation,” says Bruce Y. Lee, executive director of public health informatics, computational, and operational research at the CUNY Graduate School of Public Health and Health Policy. “You have to figure out: How do I actually communicate this information in a very succinct way, but in a way that’s not misleading?”

As early as the mid-2000s, scientists at what’s now the CDC’s (solid and respected) modeling unit looked at the possibility of creating a government-wide “fusion center” that’d hire new workers and combine their work with that of modeling groups cranking away in various corners of the federal bureaucracy. The bureaucracy resisted the impulse. “CDC was very happy with their capabilities at the time, and Washington didn’t feel strongly that they needed a projective modeling capability that would utilize the state-of-the-art science,” says Nathaniel Hupert, codirector of the Cornell Institute for Disease and Disaster Preparedness and Policy Lead at the Covid-19 International Modeling Consortium. Hupert was the inaugural director of the CDC’s preparedness modeling unit, and hoped to build that federal hub.

That post-H1N1 task force at OSTP favored the idea of a national center, too. But its eventual white paper went to the same place most white papers go. “This is always the problem with public health preparedness,” Rivers says. “There’s usually a period of reflection after a crisis that precipitates new funding and programs, but over time those dwindle.”

Still, every once in a while a disaster is so freaking disastrous that things really do change. The 9/11 attacks resulted in the creation of the Department of Homeland Security, and the anthrax attacks shortly thereafter prompted a massive effort to deal with bioweapons. Covid-19 looks like it could be another one of those singular moments. Rivers wrote a new report on the idea of a pandemic prediction office back in March; then-President Trump’s budget priority document had some of the same ideas. And OSTP convened another group of experts to talk about it in November. (Rivers thinks it might have helped that the last head of OSTP, Kelvin Droegemeier, was a meteorologist who totally got the National Weather Service thing.) “In the same way that a lot of good preparedness ideas don’t get implemented, it didn’t have the same urgency until Covid hit,” Rivers says.

The Biden administration’s plans call for OSTP to look at new ways to collect and share public health data, and ask the National Security Council to evaluate the feasibility and budget for an outbreak modeling center. Nobody knows how much it’ll cost (though if it could avert another Covid, it’d be worth it). And nobody knows whether the Biden administration’s plans will turn into legislation, or get funding. But the need is there. “It certainly makes sense that public health preparedness modeling in this country should be more than a handful of well-recognized stars having the ear of the people making the decisions,” Hupert says.

Those aren’t even the only obstacles. Would this center actually be a place, some bunker in the mountains somewhere, with modelers sitting alongside tactical epidemiology teams waiting to deploy at the moment some bat coughs on a hiker? Or would an existing unit, like the modeling team at the CDC, or IHME at the University of Washington, get deputized? Maybe it’d just happen over Zoom. Either way, it might be cool for the new breed of computational epidemiologists to have somewhere to use their PhDs outside the academy or private industry—and to sit with sociologists, anthropologists, and engineers who can make sure their strategies turn into tactics and action. “We’re probably going to spend the next few weeks trying to come up with how many people it would be, what its function would be, where it would live. I think it would be most effective if it was a federal office in the Department of Health and Human Services,” Rivers says. (Though even that might be controversial, given the longstanding rivalry between CDC, in Atlanta, and the infrastructure of Health and Human Services in Washington, DC.)

Of course if such a center did its job, most of us wouldn’t know—or care—where its headquarters was, or even that it existed. “One thing that’s tough is: If we do our job right, nothing happens,” Rivers says. “But this pandemic has cost trillions of dollars in direct costs and more in indirect costs. These events happen way more often than anyone has internalized. It’s almost every year—it is every year, if you consider influenza. So it is not a once-in-a-hundred-year problem.” Maybe it’d be smart to build a national center whose job was to turn pandemics into something people only had to worry about once a century.

Get A Quote For: