Latest News

 

Powering COVID-19 Collaborations and Communicating Risks to Public

Published on December 16, 2020 by Aaron Dubrow



The UT COVID-19 Consortium site has received more than 600,000 visitors. Results of the embedded models are studied by state, local, and national leaders as they handle the various waves of the pandemic.

When will COVID-19 infections peak in our communities? How well do interventions, like school closures or social distancing, limit the spread of the disease? What are the risks of going back to school? Where are more medical resources needed and where should testing sites be located?

These are the types of questions that have weighed on our minds during this pandemic. The answers help decision-makers determine how and when to open up schools, stores, and services, and let the public make informed decisions.

The UT Austin COVID-19 Modeling Consortium at The University of Texas at Austin (UT Austin), led by Lauren Ancel Meyers, has developed some of the leading epidemiological models of how the disease spreads. The models incorporate the latest science on virus transmission and real-time data on social distancing from cell phone data.

One of the first cutting-edge models to be released by the consortium was the national projections for COVID-19 mortality across the U.S., broken down by metropolitan service area. "The model uses geolocation data from cellphones to determine the impact of social distancing within each specific place," said Meyers. "Using more granular cell phone mobility data, UT can project COVID-19 deaths for the next three weeks."

The portal's School Risk dashboard projects the expected number of students or staff who will arrive infected in a given week based on recent estimates for COVID-19 prevalence in each US county and the size of a school or pod.

The models and data are processed using the Frontera supercomputer at UT Austin, the fastest at any university in the world, and have influenced both local and national policy makers.

Starting in April, UT researchers have shared the results of pandemic forecasts with the public through the COVID-19 Portal. In addition to the Austin projections, Texas projections and school opening risks. The COVID-19 Consortium site has received more than 600,000 visits since it launched, and is studied by state, local, and national leaders as they handle the various waves of the pandemic. Last month, the researchers added a new tool to help parents and educators estimate COVID-19 infection numbers at their school. This will expand to a future dashboard to explore the impact of COVID-19 testing effectiveness for schools.

Developed by the Meyers team and researchers at the Texas Advanced Computing Center (TACC), the portal hosts one of a handful of models that the national media and public have used during the pandemic to inform their understanding. Dig deeper and it becomes clear that the portal, and the models it communicates, reflect state-of-the-art parallel computing, software engineering, data visualization, and portal design — gracefully hidden to make the available data appear seamless and easy to investigate.

"At the end of the day, models are impactful," said Maytal Dahan, TACC Director of Advanced Computing Interfaces and a collaborator on the project. "By expanding the model with visualization and interactive web access via a science gateway, we're able to make the most impact for public understanding."

Building on a Long-Standing Relationship

Meyers' team has worked with TACC for years. The COVID-19 Portal builds on efforts that begin in 2010 to create the Texas Pandemic Flu Portal, with support from the state and the CDC. Developed in the wake of the H1N1 flu, the portal integrated Meyers' models and tools at the time.

When the coronavirus began spreading in China, Meyers was among the first U.S. epidemiologists to take note and to use available data to predict its expansion, which she determined was far broader than China was admitting at the time. The explosion of cases worldwide motivated Meyers to re-boot the portal as a site for researchers — her team, collaborators, and in the future the broader community — to run complex analyses, and for the public to get critical information.

As a partner on the project, TACC has assisted in four important areas: offering high performance computing cycles, scaling and improved code performance, visualization, and science gateway portal development.

Re-Tooling Models for High Performance Computing

TACC's computational experts, led by research assistant Kelly Pierce, have helped update the codes Meyers' team uses so they can run efficiently on supercomputers. This allows them to run real-time analyses — based on daily cell phone GPS data — faster than would otherwise be possible, and also to run many more forecasting simulations to better characterize the uncertainty inherent in the predictions.

The projections released by the team, showing the potential that deaths from the pandemic peaked, or will peak in one or two weeks, were among the first in the U.S. to use real-time data, and were far more careful in their projection about the range of uncertainty that other models — a point they highlighted in the report that accompanied their April 17 forecast.

The re-tooling of the modeling code packaged the computations so they can be re-run on other systems, or at a later date. This enhances the transparency and reproducibility of the code and provides confidence that there are no errors in the model.

"Using our skill and expertise to create robust software gives researchers the ability to quickly ramp up models, and easily generate and reproduce results," Dahan said. "We were able to enhance their coding practices by containerizing their software so it can run anywhere, and provide faster turnaround times."

The software development work was supported in part by the Extreme Science and Engineering Discovery Environment (XSEDE)'s Novel and Innovative Projects effort.

Communicating Risk to Diverse Audiences

Visualization experts at TACC assisted with the creation of data representations to communicate the models' results. This effort, led by TACC director of Health Informatics Kelly Gaither, was a departure from the typical visualizations TACC produces, which are highly technical and geared towards researchers.

"What makes this situation special is the risk communication aspect of it," Gaither said. "It's high stakes. We're not just communicating to a scientific audience. The audience for this information extends from the lay public to the researcher to decision-makers, so we have to communicate in a range of ways."

18 TACC staff members have contributed to the UT COVID-19 Consortium, providing expertise in computing, visualization, and portal development to the effort.

Accessibility, clarity, and accuracy are all critical aspects of the visualizations that are being developed.

"Validation and verification are a huge piece of what the Meyers group does," Gaither said. "Making sure when we communicate that we are not misrepresenting anything and are communicating in as clear a manner as we can is essential."

A Platform for Cooperation and Communications

The third aspect of the collaboration was the development of the portal framework itself.

"By being able to offer a user portal for COVID-19 we can effectively communicate science to the wider community," said Dahan, who in addition to her TACC role is also co-director of the National Science Foundation-funded Science Gateways Community Institute. "We not only reach other researchers, but also local, state, and national policy makers and the wider community."

The public-facing aspect of the site is intended for a broad audience and lets the public, policy makers, and government staffers access timely, science-driven predictions, Dahan explained. But the interface also allows researchers to login to the portal, execute simulations on high performance computing systems, do parameter sweeps, visualize results, and compare them to previous model runs.

In the future, the group intends to add the ability to do comparative visualizations using different models, and to produce even more geographically-defined predictions.

"Right now, we're citing risks to states, cities, counties, or metro areas," Gaither said. "But we hope to get to the point where we can analyze what's going on at the school district, zip code or neighborhood level, like having a digital twin of a community."

Further models will use the portal to find optimal solutions to locating testing sites and distributing vaccines once they are available.

TACC is mostly known as the nation's leader in providing advanced computing resources to U.S. academic researchers, but the center employs more than 80 PhD scientists, who are experts in diverse topics. These staffers lead funded research projects and are frequently collaborators on outside research. Even for those staffers who are frequent collaborators, this experience has been different.

"In the 20 years I've been at TACC, this has been the most tightly integrated research project I've worked on," Gaither said. "We have been embedded with the team and assisted in all aspects: from helping the team use our resources and services to bringing value-add capabilities that they didn't have. We've augmented their team and allowed them to springboard to the next level."

"The mortality projections website that the team at TACC put together has helped our COVID-19 modeling work have a public-health impact beyond anything we could have hoped to achieve ourselves," said James Scott, associate professor of Statistics at UT Austin, and one of the creators of the mortality model. "Because of the visibility and user-friendliness the site offers, our projections have garnered nationwide attention — from Anderson Cooper to the CDC, and from the Rhode Island Governor's Office to the Department of Defense.

"It goes to show what you can accomplish when you bring together a team of statisticians and epidemiological modelers with a group of world-class computational scientists like we're blessed to work with at TACC."


Story Highlights

UT Austin COVID-19 Modeling Consortium, led by Lauren Ancel Meyers, has developed some of the leading epidemiological models of how the disease spreads. The models incorporate the latest science on virus transmission and real-time data on social distancing from cell phone data.

Starting in April 2020, UT researchers have shared the results of pandemic forecasts with the public through the COVID-19 Portal.

Developed by the Meyers team and researchers at TACC, the portal hosts models that state, local and national decision-makers, media, and the public have used during the pandemic to inform their understanding of risk.

The portal, and the models it communicates, reflect state-of-the-art parallel computing, software engineering, data visualization, and portal design, enabled in part by staff at TACC.


Contact

Faith Singer-Villalobos

Communications Manager
faith@tacc.utexas.edu | 512-232-5771

Aaron Dubrow

Science And Technology Writer
aarondubrow@tacc.utexas.edu

Jorge Salazar

Technical Writer/Editor
jorge@tacc.utexas.edu | 512-475-9411