Coronavirus cases fell especially sharply in U.S. provinces where people stopped going to offices and workplaces, new cell phone data suggests.
Researchers found that infections were about 30 percent lower in the counties where most people stopped going to offices.
In provinces with more activity in workplaces, the number of cases only decreased by about 10 percent after two weeks.
In addition, counties with the most cell phone activity at home had an almost 20 percent lower growth rate in 15-day post-stay-at-home orders compared to counties in the lowest quartile.
The team, from the University of Pennsylvania School of Medicine, says these patterns could be used to estimate COVID-19 growth rates and inform policymakers making decisions about its closure and reopening.
The number of coronavirus cases was about 30% lower in the counties where the largest number of people stopped attending offices, but only 10% lower in the counties with more activity at work (above)
Counties with the most cell phone activity at home had an almost 20% slower growth rate in cases 15 days after stay-at-home orders (above)
“ It is our hope that the counties would be able to include this publicly available cell phone data to help guide reopening policies at various stages of the pandemic, ” said senior author Dr. Joshua Baker, an assistant. professor of medicine and epidemiology at the Medical Faculty.
“Furthermore, this analysis supports the inclusion of anonymized cell phone location data in modeling strategies to predict at-risk counties in the US before outbreaks become too large.”
For the study, published in JAMA Internal Medicine, the team used cell phone location data made publicly available by Google.
Between the beginning of January and the beginning of May 2020, activity data was available for 2,740 counties in the US.
Researchers grouped locations where activity was reported into categories such as homes, workplaces, shops, supermarkets, and transfer stations.
This data was then split into two periods: the first period was January and February – before the outbreak in the US – and the second period was from mid-February to early May when the cases increased sharply.
Unsurprisingly, the team saw marked changes in the cell’s activity around the same time that several states began issuing lockdown orders with an increase in the time they spent at home.
Meanwhile, the number of visits to workplaces decreased significantly, as did the number of visits to shops, restaurants and transfer stations.
While there was some decline in most counties, the level at which it occurred varied.
In the fourth quartile – countries where activity levels were highest everywhere – the average workplace level fell by 25 percent.
However, in the first quartile where activity was lowest, workplace activity decreased by 51 percent
The range in the decrease in activity at the transit level was greatest, with the top quartile seeing an average decrease in activity of only 6.5 percent and the lowest quartile a decrease of 58.5 percent
Researchers found that the counties where workplace activity decreased the most had the lowest rates of new COVID-19 cases in the days that followed.
They even corrected for delay times of five, 10, and 15 days to account for an incubation period, but the results persisted.
For example, in counties in the lowest quartile of activity at work, the number of cases was nearly 30 percent lower after 15 days than in the highest quartile.
Counties with the most activity at home had a 19 percent slower growth rate at 15 days compared to counties with the least activity at home.
Baker says he would like to see in the future whether cell phone data could be used to predict COVID-19 hotspots.
“It will be important to confirm that cell phone data is useful at other stages of the pandemic beyond initial control,” he said.
“For example, is it useful to monitor this data during the reopening stages of the pandemic or during an outbreak?”
“They have the potential to help us better understand patterns of behavior that could help future researchers predict the course of future epidemics or perhaps monitor the impact of different public health measures on people’s behavior.”