Anonymized cell phone data is helping Virginia’s local health officials select mobile vaccine sites

In early spring, Virginia health officials launched large community vaccination clinics with the highest of hopes. The sites, set up to administer thousands of doses of coronavirus vaccine, were selected with equity in mind — all placed in communities with diverse populations and high rates of infection. At the time, state Health Commissioner Dr. Norman Oliver said Virginia was doing a “very good job” at expanding vaccine access to underserved communities.

Then, something unexpected happened. Attendance at the sites slowed to a trickle. Some, like a community vaccination center in Danville, were overrun with more out-of-town visitors than locals. Others administered far fewer shots than their capacity allowed. Visits slowed so much in Danville that state officials considered closing the site early, said Dr. Scott Spillmann, director of both the Danville-Pittsylvania and Southside Health Districts.

“There was an initial surge and then the numbers really dropped off,” he added. “And the state was discouraging people from driving long distances to get their vaccinations.”

The experience spurred Spillmann to reach out with a more novel request — data that could help inform the placement of mobile vaccine units. The big sites were good at reaching Virginians who were eager and willing to get their shots, he said. But they weren’t attracting residents in more remote parts of his districts, who weren’t always close to Danville or ready to get vaccinated. 

“We were really trying to ascertain, ‘Where were people living and working?’” he said. “‘Where was the right place to go.’”

The questions spurred a statewide approach that’s still being used today. Spillmann reached out to researchers at the University of Virginia’s Biocomplexity Institute, which developed a data dashboard for state and local health officials soon after Virginia reported its first cases. The reports include information on mobility, drawn from anonymized cell phone data collected by a company called SafeGraph, showing where and when Virginians were traveling to help understand the impact of safety restrictions. 

“They were looking for new sites and thought this particular data offered something new and fresh,” said Madhav Marathe, director of the institute’s Network Systems Science and Advanced Computing Division. It was staff at the Virginia Department of Health, he said, who had the idea of using mobility information to select the best locations to give out vaccines.

The data has become increasingly important as mobile clinics become the preferred method of outreach for many local health departments. By late April, less than a month after vaccines were made widely available to all adults, demand in Virginia was dropping. It plummeted through much of the summer and is back in decline after a brief bump driven by the delta variant, according to VDH data.

The emergence of an even more highly transmissible form of COVID-19 underscored the importance of vaccinating as many people as possible, as quickly as possible, said Bryan Lewis, a computational epidemiologist with the Biocomplexity Institute. But deciding where to set up mobile sites was often a challenge — especially in large and rural health districts. Through much of the spring, it was largely a guessing game, with some clinics failing to administer any vaccines despite the best marketing efforts of local health officials.

Mobility data, though, can offer a clearer picture of where people actually travel day-to-day. The project relies on anonymous cell phone information and specific census data to inform researchers — with a fairly high level of specificity — where people from certain neighborhoods are doing their grocery shopping, for example, or going out to eat. Every week, the Biocomplexity Institute sends VDH a list of the top 25 locations in each health district that could make the best targets for mobile clinics. 

Sites are typically around three-quarters of an acre in area, which gives local health officials more flexibility on where to set up. “We’re not saying, ‘Okay, this dollar store is the right place,’” Marathe said. “We’re saying, ‘This shopping center is the right place,’ because we learned that sometimes they can’t place a site right in front of a particular store.’”

Frequently, there’s wide variability in where different demographic groups are traveling. VDH has focused specifically on Black and Latino Virginians — who have historically received less public health outreach, said Justin Crow, director of the department’s division of social epidemiology — as well as young adults and the unvaccinated population as a whole. Without data, though, it can be difficult to figure out the most popular locations for a diverse range of groups.

“What the analysis revealed is that in areas where there’s a lot of Latinx population or areas where there’s a lot of African American population or areas where there’s a lot of white population, the overlap of their top 25 locations isn’t perfect,” Lewis said. “There’s often quite a disparity between Latinx and Black Virginians, for example, and those are two groups that are a bit more underserved in areas like vaccination or health access.”

The goal is for mobility data, then, is to expand access to the vaccine by meeting people where it’s most convenient for them. Crow said smaller community sites often feel more approachable than large-scale clinics, which — when they were operating — sometimes had long lines or weren’t located in locations that were familiar to many residents.

And the data did help pinpoint more specific locations for mobile clinics Spillman said. At one point during the summer, the report identified a specific Hispanic grocery where his department was able to vaccinate at least 15 people. Churches and volunteer fire departments have also emerged as sometimes unexpected community hubs. 

But as access to the vaccine has steadily improved — even in far-flung communities — many health experts say hesitancy remains as the most persistent barrier. Even the most well-placed clinics often don’t administer more than a handful of shots.

“It’s been more of a trickle, to be honest,” said Chris Garrett, the local health emergency coordinator for the Pittsylvania-Danville Health District. 

In some respects, Virginia metrics have improved by leaps and bounds since the spring. A little more than 61 percent of the state’s total population is fully vaccinated, and rates are even higher for adults, at just over 73 percent. After months of wide disparity, the state has also narrowed the gap in vaccination rates for Black and Latino residents. 

By May, Latinos had become the second-most vaccinated group in the state and are now slightly overrepresented among people with at least one dose compared to their total share of the state’s population. Black Virginians, who account for roughly 20 percent of the state’s population, now make up 17 percent of people with at least one dose. At this point, White Virginians are now the most undervaccinated group compared to their total population share — just over 61 percent of all Virginians and 57.9 percent of people with at least one dose.

Even with those gains, health officials say it’s a slow process of reaching out to the nearly one in five Virginia adults who haven’t been vaccinated yet and sometimes don’t want to be. And right now, there isn’t a clear strategy for overcoming that barrier. 

The unvaccinated population isn’t a monolith, and experts emphasize that people often have good reasons for not wanting the shots. While “trust the science” has become a familiar refrain during the pandemic, medical experts sometimes aren’t a trusted resource for Black patients, said Dr. Katherine Tossas, an assistant professor in the Department of Health Behavior and Policy at VCU School of Medicine. Black mothers, for example, are still far more likely than other groups to report mistreatment based on race or cultural background in health care settings.

“In a sense, we fail to recognize that for you to trust me, I have to come across as trustworthy,” she said. Misinformation has also plagued the vaccine rollout — partially fueled, Tossas said, by initial stories over how fast the vaccine was developed (even though mRNA technology dates back more than two decades). 

She compared it to the rollout of the HPV vaccine, which initially faltered in the United States after being more heavily marketed for its protection against STIs than its protection against cancer. The COVID-19 vaccine also needs a rebrand, she said, and getting more people comfortable with the shot will likely have a significant impact on the future of the pandemic.

While cases in Virginia are once again declining, the state is still reporting an average of 2,220 new infections a day (in late June, the average was less than 200). And though a growing number of residents now have immunity to the virus, either through vaccination or previous infections, Lewis said there are at least a million Virginians who haven’t been immunized against the disease but also haven’t contracted it yet.

“Because these populations are in pockets, it gets hard to predict what will happen,” he said. “But at the moment, there’s still enough fuel out there that this fire could burn for a while longer.” Experts are worried that the holidays could bring another surge among unvaccinated Virginians. And since natural immunity may not be as effective against new variants as vaccinations, there’s continual worry that an even more transmissible version will emerge, leaving unvaccinated people vulnerable to new infections.

Vaccine outreach has been an expensive endeavor for the state, but many officials say it’s reached the point where individual efforts may make more headway than statewide public health campaigns. 

Garrett pointed to a one-on-one conversation he had with an EMS worker just a few weeks ago, emphasizing that the man’s job put him at higher risk of exposure and bringing the virus home to his family. Both the worker and his wife decided to get vaccinated just a few days later.

“I think it is those individual interactions that are making a difference, but it’s going to be a tough slog,” Lewis said. “It may not be the mobile clinic sitting outside the store that’s going to convince you.”