AI Predicts Highest Risk U.S. Counties During Covid-19 Surge

On Thursday, the daily Covid-19 case count in the United States soared to more than 185,000, a record, and a total of more than 253,000 people have died since the pandemic began, according to data compiled by researchers at Johns Hopkins University.

The United States is also seeing a record number of hospitalizations in 23 states, even as the full brunt of the current cases will not be felt for about one to two weeks. Nationally, more than 81,000 patients are currently hospitalized with Covid-19.

“Hospitalizations are on the rise in all states simultaneously – with the worst areas being the ones we already know about. But this idea that the entire country is a hotspot right now is borne out clearly by the hospitalization data,” said Andrew Schroeder, vice president of research and analysis at Direct Relief.

Schroeder and his team have been using Facebook-provided data, data from other sources, and AI to predict, visualize, and analyze the spread of Covid-19 down the county level.  They have also been using the data to track mobility — or how much people are moving around — and hospitals in order to fill out a more complete view of the pandemic. Facebook’s methodology, which is based on anonymized data sourced from users who have opted into the program on their mobile devices, can be viewed here.

These insights can be useful for public health and other officials as they determine how to best deploy their resources, set policies, and communicate with the public on how best to mitigate transmission of the virus.

Most At-Risk Counties

If current trends hold, the top five most at-risk counties in the U.S. are Cook (Ill.), Los Angeles (Calif.), Maricopa (Ariz.), El Paso (Texas), and Harris (Texas). The model predicts 2.5 million new cases nationwide by the end of November. Using this figure and existing death rates, more than 400,000 deaths would be expected by January 2021.

“The geography of the current wave has not changed much – areas that were previously concerning are now more concerning – with the exception perhaps of what appears to be a regional acceleration in northern Ohio and western Pennsylvania,” Schroeder said.

Regarding the death rate, Schroeder pointed out that, while treatment outcomes have improved since the onset of the pandemic, there have been few decreases since July, with an approximately 1.8% mortality rate.

Covid-19 hospitalizations and ICU occupancy by state. National curves on hospitalizations, ICU admissions, and ventilator usage are tracked on a tab below the main map. (Direct Relief)

Mobility Data and Virus Transmission

Schroeder believes that mobility rates have also played a significant role in influencing the path of the current wave. He shared his findings related to Wisconsin, which has seen some of the worst trends in this current wave:

“One month ago this was what their rate of mobility looked like based upon the Facebook movement range metric.”

“By Election Day, things had arguably gotten considerably worse, especially in more rural areas, with large portions of the state continuing to move around at rates well above baseline.”

“All the way up to November 8, while discussion of Wisconsin’s dire Covid-19 situation was very public, we can see this situation persist.”

“As of November 17, things slowed back down again in Wisconsin, as statewide targeted social distancing measures have begun to take effect once more.”

“However, we can expect to see the transmissions which occurred during the period of elevated mobility continue to grow for several days more at least, with hospitalization rates lagging cases by about 12 days, and then deaths by roughly another 10 days more than that,” Schroeder said.

More information about Direct Relief’s response to the Covid-19 pandemic be accessed here. 

Share:

More Posts

Send Us A Message