AUSTIN (KXAN) — It’s predictions about the pandemic that have guided public policy, warned hospital systems of surges, and been the subject of numerous reports to KXAN as a way to let the public know what’s to come. A new study published in the Proceedings of the National Academy of Sciences shows that these predictions were among the most accurate in the nation during turbulent surges and declines, and likely saved lives.
In 2020, the University of Texas COVID-19 Modeling Consortium was tasked with helping local leaders predict the outcome of the COVID-19 pandemic with enough lead that hospital and community surges intensive care does not become overwhelming. By predicting surges in a few weeks, local leaders could plead with the public with enough evidence to change behavior.
“When COVID-19 was emerging, we were already thinking about how best to predict the dynamics of COVID-19 and track the pandemic,” said Dr. Spencer Fox, associate director of the UT COVID-19 Modeling Consortium. “So the first thing we did was figure out what data we’re going to feed into the model.”
Fox said the consortium looked to two sources of data to create reliable predictions: epidemiological data and behavioral data.
The most reliable form of epidemiological data, Fox said, was hospital admissions. This is a dataset that the City of Austin and Travis County have also highlighted as a measure for which risk-based guidelines fall within the zone.
But in March 2020, how to measure behavioral data was less obvious – tests weren’t widely available and best practices were just emerging for much of the public. That’s why researchers turned to what you carry in your pocket.
The UT modeling consortium was one of the first to include public movement data in its models, according to UT, even before the famous Institute for Health Metrics and Evaluation modeled by the University of Washington.
The researchers incorporated SafeGraph’s anonymous cellphone data into their projections, and it showed how much time people spent in their own homes compared to school, work, restaurants and other public spaces. It filled a data gap that many other models could not fill.
“This cell phone mobility behavioral data has allowed us to understand how behaviors are changing in our community with the goal of providing a one to two week turnaround and what will happen in terms of healthcare needs. “, said Fox. “So what we found in the study is that overall our predictions were very reliable up to three, even four weeks in some cases.”
But simply collecting the data would not be enough to warn the public of what was to come. For these predictions to be intuitive to the community, they needed to be presented in a way that everyone in the home could understand. That’s why UT researchers turned to an already well-established public forecasting model – spaghetti lines from hurricane forecasting, paired with easy-to-read graphs.
These predictions and data were published on a publicly accessible dashboard, which you’ve probably heard of if you’ve been following KXAN’s pandemic coverage. We regularly use this dashboard.
“I think this report that we just released really gives, you can think of, a blueprint of how to actually respond to a pandemic,” Fox said. “It explains how to collaborate with local public health officials, elected officials, and health systems, what data sources are useful for actually making these kinds of predictions, and it gives a model structure and framework for doing so.”
However, when it comes to this pandemic, other parts of the country won’t have to create their own forecast dashboards from scratch. UT researchers are working to deploy a dashboard similar to the one we use in Austin for every metropolitan area in the country.
“Over the next few weeks, a few months, we should have a dashboard that will actually let you see what our projections are for Albuquerque,” Fox said. “You will be able to select any metropolitan area in the country.”