Americans have been keeping their distance from one another since late March. But new data suggests that those habits are coming to an end for more than 10 million people.
According to an NBC News analysis of cellphone location data provided by the analytics and marketing company Cuebiq, people in more than 450 counties across the country have started to come near one another more frequently. And as people begin to gather in greater numbers, health officials are watching for a new round of coronavirus spikes.
Cuebiq analyzed detailed location data from more than 15 million mobile devices in the contiguous U.S. from March 1 to June 2 to determine how often people were standing alongside one another and up to 50 feet away from someone else while away from home. The data shows that while contact levels are still low nationally, they are creeping back up, and in some places they are near or back to pre-pandemic levels.
Cuebiq's data does not attempt to tie social distancing behavior to transmission rates of the coronavirus. And at 50 feet, the width of a basketball court, you would be more than eight times the recommended guideline for minimum distancing from the Centers for Disease Control and Prevention. But the data, which can be taken as a measure of how often people come near other people, show the extent to which Americans changed their behavior to slow the coronavirus and where that behavior is changing back.
Most Americans began slowly decreasing the number of times they came near others in early March. In Los Angeles, Cuebiq's data shows, contact between residents had its first big decrease on March 13, the day President Donald Trump declared COVID-19 a national emergency and six days before California issued the country's first stay-at-home order. In Cook County, Illinois, which includes Chicago, contacts showed a drop on March 14.
On March 26, when Colorado and Kentucky issued stay-at-home orders, distancing was at its highest, as Cuebiq's data shows that Americans came in close contact with one another about a quarter as often as they had in February.
Health officials say the dramatic change in behavior was necessary to slow the spread of the coronavirus and prevent hospitals from being overwhelmed.
Lauren Ancel Meyers, a professor at the University of Texas at Austin who leads a department that models coronavirus death patterns based on similar cellphone mobility data, said part of the reason New York City's outbreak became an epidemic was that the region didn't begin distancing until the virus was already actively spreading in the area. Other cities got ahead of their outbreaks, and that saved lives.
"If we had not intervened and done something to slow the transmission of the virus in March, the models suggest, we would have seen very large and devastating epidemics in many U.S. cities," Meyers said.
Download the NBC News app for full coverage and alerts about the coronavirus outbreak
COVID-19, the disease caused by the coronavirus, spreads mainly through close contact, and a study published this month found that social distancing was the most effective way to prevent transmission.
Data from Apple Maps in May showed that Americans in certain cities began driving and walking around again. But Cuebiq's data suggests that even as they're moving about, Americans are coming nearby one another more often. The rate of nearness is less than half of February's levels, and it has gradually increased since March. And in some of America's most populous counties, including Los Angeles, Manhattan (New York County) and Miami-Dade, Florida, nearness is even lower than in the rest of the country.
There are exceptions, like Camden County, Missouri, where rates surged by a third over Memorial Day weekend, the same time that a bar held a crowded pool party at Lake of the Ozarks. Bay County, Florida, and Baldwin County, Alabama, both home to beach cities, have had contact rates above February levels since early May. Bay County began reopening May 4, and reopening in Alabama began May 11.
Health experts anticipate more gatherings as cities and states continue to loosen coronavirus restrictions and as warmer weather draws Americans out. The crowding has them keeping an eye on coronavirus numbers and looking for spikes, which would be a flag that governments need to tighten restrictions.
"A reopening suggests that there will not be a reclosing," said William Hanage, an associate professor of epidemiology at Harvard University's T.H. Chan School of Public Health. "As the virus is given more opportunities to spread, we can be confident it will take them."
Meyers said: "The virus is still the same virus. It can still spread just as fast as it could in February and March. Now we're entering a relaxation period where people are not staying at home and are being asked to take precautions. We don't know if that will be sufficient."
But that's not to suggest that an increase in crowding will automatically lead to a surge in new cases. Meyers said many Americans are still wearing masks and washing their hands more frequently, behaviors that can cut down on transmission.
Experts stress that as people begin to venture out again, cities and states will need to aggressively trace people with whom an infected person came in contact and isolate them before they have the opportunity to spread the virus further.
"We need to shift the thinking from cases to transmission events," Hanage said. "Some of the social distancing may be more important than others. The important thing is that we limit the opportunity of the virus to spread to large numbers of people."
Cuebiq collects location data from a panel of about 15 million devices through mobile apps. Cuebiq collects location data only from opted-in devices whose users have agreed to share their anonymous location data with Cuebiq in apps such as GasBuddy & MyRadar. Cuebiq analyzes anonymous and aggregated data at the county level to determine a device's home area, where a particular device spends a significant amount of time, particularly in the evening. Its assessment of whether two devices come close to each other is based on checking the location data to see whether a device is within 50 feet of any other device within five minutes.