The UT COVID-19 Modeling Consortium has developed a forecasting model that uses city hospital and cell phone mobility data to predict the progression of COVID-19 transmission, assisting the Austin metropolis area in maintaining the lowest COVID-19 mortality in the state, said the consortium’s associate director Spencer Fox. 

The model can inform when an infection peak will occur, the peak’s magnitude and when hospitals will reach capacity. The model evaluates healthcare needs by giving estimates for city hospital admissions over the next three weeks, including total intensive care unit bed usage, infection reproduction estimates and overall pandemic trends, Fox said.

Like the city of Austin’s five stage alert system, the dashboard uses hospital admission data to guide city health officials, businesses and the general public to estimate Austin’s situation in the coming weeks, Fox said. 

“We found that there’s a strong correlation between city efforts to reduce transmission that ultimately shows up in the data as reducing citywide transmission,” Fox said. “State policies that relax restrictions have been associated with increases in transmission over the course of the pandemic.”

SafeGraph, a company that tracks population mobility through cell phones, provides anonymized data that tells the algorithm how much time the average person spends at home and points of interest in the area like schools, restaurants or businesses, Fox said. 

“Those metrics about mobility give us an idea of how people’s behavior is changing over the course of the pandemic,” Fox said. “People are finding safer ways to actually be mobile in the community over the course of the pandemic, so the same level of mobility today leads to a lot less transmission than it did, in say, February of 2020. Those types of things have actually changed the relationship between mobility and transmission over time.”

The model was formed as a part of the COVID-19 task force in the city, but has been expanded to provide similar projects for the 22 Trauma Service Areas in Texas, Fox said.

“The pandemic has shown how important forecasting models really are for guiding public health policies,” Fox said. “It allows decision-makers to prepare for an impending surge.”

The forecast model would be more accurate with a larger and representative sample, said Rémy Pasco, a graduate student and co-author of the study. 

“If we knew actually how many people are wearing masks in different places, that people are changing their behavior or exactly who has been vaccinated and if people will express the accuracy, that would be nice,” Pasco said. “The issue is that the same people who are more likely to be infected are also the ones less likely to be able to contribute to public health interventions in general.”

Pasco said individuals should be patient and wait a few more weeks until the data indicates the city can go down to stage one, two or three.  

“The dashboards provide really actionable advice for individuals that can help them protect themselves,” Fox said. “You can look at the evolving risks and make decisions about your behaviors based on whether … we’re in the midst of a surge … versus if we’re seeing a decline in the pandemic and cases in hospitalization.”