University of Massachusetts Amherst biostatistician Nicholas Reich has been awarded grants from the U.S. Centers for Disease Control and Prevention (CDC) and the National Institutes of Health (NIH) to support innovation and expansion in his groundbreaking pandemic forecasting work.
Reich received a $350,000 award from the CDC to advance the COVID-19 Forecast Hub, which he and his team developed in April. They use the same ensemble approach, unifying multiple models from top forecasters and institutions around the world, for COVID-19 projections as they do for the CDC-designated UMass Influenza Forecasting Center of Excellence.
Each week, the COVID-19 hub is updated with four-week-out projections for COVID-19 deaths in the U.S. and by state. Its centralized, open-science data repository is relied on by the CDC, data journalists and the public for accurate and up-to-date forecasts of the current trajectory of the pandemic in the U.S.
Reich notes that the new investments from the CDC will support a shift toward creating better pandemic forecasting in the U.S. and globally. COVID-19 "represents an unprecedented challenge for the infectious disease modeling community," Reich says. "This award enables us to track, evaluate and synthesize forecasts from dozens of research groups around the world. Our work continues to underscore the importance of looking at multiple different infectious disease models, just as weather forecasters do with hurricane projections, if we want to have a good sense of what is coming next with COVID-19."
In addition to compiling the weekly COVID-19 forecasts from multiple models, Reich's team will work with partners to obtain new relevant data streams, develop new adaptive ensemble methodologies for synthesizing the hub's forecasts and create new forecasting methodologies.
In the other award, Reich and UMass Amherst colleagues Dan Sheldon, an artificial intelligence (AI) researcher; Andrew Lover, an infectious disease epidemiologist; and biostatistics Ph.D. student Casey Gibson received a $300,000 grant from the NIH to develop new statistical methods for their individual mechanistic Bayesian forecasting model. It's one of the models featured on the COVID-19 Forecast Hub, the FiveThirtyEight COVID-19 Forecast tracker and the CDC website.
Led by Sheldon, associate professor in the College of Information and Computer Sciences, who addresses large-scale data-scientific challenges using massive data sets, the team will explore machine-learning approaches for pandemic scenarios.
"New methods are needed to leverage the wealth of surveillance data at fine spatial and temporal granularity, together with associated information about policy interventions and environmental conditions over space and time, to reason directly about the mechanisms to forecast and understand the transmission dynamics of SARS-CoV-2 transmission," the researchers say. "These methods must use sound statistical and epidemiological principles while being flexible and computationally efficient to provide real-time forecasts that can guide public health decision-making and response to the highly dynamic aspects of this global crisis."
These new frameworks need to be developed quickly because of "the very real potential" for COVID-19 to become an endemic, and perhaps seasonal, pathogen in the U.S., causing recurrent waves of the disease.
"Forecasting pandemics of emerging pathogens poses a set of new and important challenges," Reich says.