A few miles south of Goldsboro, in a county with thousands of acres of sweet potato and tobacco fields and speckled with hog farms, lies a ZIP code with one of the highest COVID-19 infection rates in North Carolina.
In mid-July, as new case counts surged across the state, no other part of Wayne County had a higher infection rate — not even the ZIP code home to Neuse Correctional Institution, where more than 400 inmates have tested positive for the novel coronavirus.
As of this week, more than 300 of the 10,700 people estimated to live within ZIP code 28333 have tested positive since the pandemic began.
Almost a third of the population there — 32% — is Hispanic. Nearly 40% of the population is Black. And situated near the edge of the ZIP code's border, just south of an Air Force base, is a major chicken plant.
Many of the ZIP codes throughout the state with the highest infection rates have similar characteristics.
Wayne County spokesperson Joel Gillie said he knew of no specific event or outbreak that caused the high infection rate in ZIP code 28333, which encompasses the unincorporated rural community of Dudley.
Reporters with the NC Watchdog Reporting Network analyzed months of ZIP-code-level coronavirus data to learn why some areas have seen high infection rates or spikes in case growth. The reporting team used the state's own COVID-19 data, along with demographic data calculated by Carolina Demography, a research group housed at UNC-Chapel Hill.
Interviews with dozens of county health directors pointed to new testing sites, population density or concentrations of cases in prisons and nursing homes as reasons for spikes or high infection rates. But some county health officials say it's unclear why some of their ZIP codes had been hit harder by the coronavirus.
"The virus spreads how it spreads," said Chatham County spokesperson Zachary Horner of the ZIP code with the second-highest infection rate in the county. "This is one of those things where community spread has been in place for months now, and so it's hard to track exactly where it comes from."
The uncertainty highlights the limitations of tracking the disease with data alone. Given the constraints of testing and contact tracing, health experts say sources of the virus are sometimes elusive.
'These Nameless Things'
On May 1, the state Department of Health and Human Services started releasing the number of coronavirus cases by ZIP code.
"We're just trying to share as much information so folks feel like we're being transparent," said DHHS Secretary Dr. Mandy Cohen when asked about the value of releasing data by ZIP code.
About 800 ZIP codes stretch across North Carolina, some crossing borders into other states.
Paul Delamater, a health geographer and professor at UNC-Chapel Hill, said although that number can make the data difficult to work with, it can also provide more fine-grain detail about what's happening closer to home.
"I live in Orange County, and I don't think I would feel like looking at a map of North Carolina counties and looking at Orange County should tell me how I feel about my risk. But ZIP code, it's just more detail," Delamater said. "If you think about the difference between looking at all of North Carolina as one, versus looking at counties, you feel like you gain a little information there. If you look at ZIP codes, it's even more."
But as a way to track hotspots, they're far from perfect.
"ZIP codes are kind of just these nameless things," Delamater said. "There's no real kind of context to go along with them."
Their geographies aren't fixed, for example, since their borders can change when the U.S. Postal Service recalculates delivery routes to serve shifting populations. They also don't match up perfectly with the standard population measurements used as part of the U.S. Census.
ZIP codes can even cross county lines, so people in many rural areas of North Carolina have ZIP codes describing a place name for a post office in an adjacent county, even though they may live closer to a post office in their own county.
"ZIP codes, fundamentally, were created for mail delivery, not for tracking disease or tracking populations," State Epidemiologist Dr. Zack Moore said.
The estimated population counts also vary wildly. Some ZIP codes in the state's data have fewer than five residents, others as many as 80,000. That can create problems with what epidemiologists call "stability."
"Sometimes there can be real challenges when you're dealing with a small population or a sparsely populated ZIP code where you can see huge fluctuations in the number of cases per population that might be attributed to a single outbreak or cluster," Moore said. "It bounces around a lot — the number and rates — in some ZIP codes."
But that doesn't mean health officials on the state and local level haven't put the data to some use.
'Bang For Our Buck'
In mid-June, for example, a team at DHHS set about tackling a thorny logistics problem.
Almost as soon as the pandemic began, signs were clear that the disease was disproportionately affecting Black and Hispanic communities — the same groups that have historically suffered from worse outcomes in the American health care system.
With testing resources expanding, state health officials needed a plan.
"To get to health equity in our testing strategy for COVID, we had to commit a disproportionate amount of resources to historically marginalized populations," said Dr. Shannon Dowler, chief medical officer for North Carolina Medicaid who steers the initiative.
County-level data was too blunt an instrument, the team found. So they turned to case counts for ZIP codes.
"We wanted to dig in a little deeper, particularly with historically marginalized populations, and make sure we were being laser focused on how to get the most impact," Dowler said.
Breaking the target population into three groups — Black, Latino and American Indian — they filtered first for population density. For Black residents, they looked for older populations with higher proportions of chronic conditions.
"We're going to get a lot of bang for our buck testing historically marginalized populations, particularly the African American population, in these areas," Dowler said.
For Latino residents, the team looked for migrant farm camps and areas flush with construction jobs. They looked at the handful of areas in the state where American Indian residents are most concentrated.
Then they removed places where testing sites already existed.
At the end, they had narrowed their list to almost 200 ZIP codes for the CHAMP initiative, launched in July. The program offers no-cost testing through three contractors.
The data, Dowler said, showed they were on the right track.
"What we saw was the ZIP codes we selected were all either in or immediately adjacent to the highest rates," she said.
In addition to contact tracing, testing and working closely with long-term care facilities and schools, Iredell County has also relied on more granular data to tackle COVID-19.
"Data is the most significant driver in guiding our limited resources and efforts," Iredell County Health Department spokesperson Megan Redford said.
Beyond The Data
But as Vickie Bradley points out, data on ZIP codes often doesn't tell the full story.
Bradley serves as the secretary for Public Health and Human Services for the Eastern Band of Cherokee Indians.
In this region in the westernmost part of the state, 12 ZIP codes encompass tribal land. One of them in particular, 28719, stands out with a higher infection rate compared to the rest of Swain County.
"There's no way to look at one ZIP code and make assumptions about COVID," Bradley said.
In March the tribe began rigorous testing of residents and people who work on tribal land, which included asymptomatic people at a time when federal and state guidelines said otherwise. The tribe also initially limited access to tribal lands and shut down businesses – while still paying tribal workers.
Bradley was among those who advised Principal Chief Richard Sneed on measures to take to mitigate the spread. But leaders needed data to know what steps to take. Bradley worked with hospital executives at the Cherokee Indian Hospital Authority to track a wide array of coronavirus metrics to help leaders make decisions to protect the public.
"It took a lot of infrastructure development and effort," Bradley said.
Bradley said they are detecting more cases of COVID-19 because testing is so widespread – and to prevent the spread they need to know who is infected, she said. Paired with testing, the Cherokee Indian Hospital Authority organized several contact tracers who sleuth out where those who are infected have been – and notify people who are most at risk of having contracted the virus.
Altogether, the tribe had tested 7,116 as of early Wednesday – and so far only 2.1% have tested positive – far lower than North Carolina's positivity rate of 7%.
"We believe all things are held in common and if we really believe that, we are responsible to help each other survive," Bradley said.
Following The 'Bread Crumbs'
Individual counties, as well as the state, are using contact tracers to identify the sources of spread. But many local health leaders contacted by the N.C. Watchdog Reporting Network could not attribute spikes in COVID-19 cases to any one source.
In some cases, community spread made the exact route of the virus murky.
Robeson County Assistant Health Director Melissa Packer noted that the virus there has spread from localized hotspots into the general community, where it can be transmitted in areas like grocery stores, gas stations or restaurants. Early in the pandemic, Packer said, the emphasis was on nursing homes and meat processing facilities.
"Now the emphasis is on widespread community spread," she said.
In Hoke County, health officials say transmission through social gatherings with limited social distancing – like family events, church services and funerals – has been the primary route of infection.
For Karen Lachapelle, Edgecombe County health director, and the department's director of nursing, Susan Johnson, there have been few cases for which contact tracers couldn't pinpoint the source. But community spread means those sources are spread across the county.
"There's not one group or one thing that we can attribute a large number of cases to," Johnson said.
Even when a major vector of the virus does emerge — a meatpacking plant, for example — the resulting spike isn't always obvious.
"It's actually a lot more challenging to identify based on a ZIP code or a location to track outbreaks, because we actually track the data from where they live, their residence," Dowler said. "A lot of people live, work and play in different places in this state and sometimes that can make the bread crumb trail a little harder to follow."
But health experts say the state — and the counties — can do more with the data than it's doing now.
Delamater, the professor at UNC-Chapel Hill, has been using the state's data to look at "where things are going in the ZIP code, instead of where they're at right now."
"It helps us understand the trajectory of places," Delamater said. "The trend map is thinking more about, is COVID getting worse or better or the same in my ZIP code? Are things getting better or worse or the same?"
But counties don't always think in terms of ZIP codes, said N.C. State professor Julie Swann, who worked as a science advisor for the H1N1 pandemic response at the Centers for Disease Control and Prevention.
Rather than knowing which ZIP codes are outliers, Swann said, counties might know which neighborhoods or communities are hotspots.
Coronavirus is moving fast, and many counties don't have enough resources and are scrambling to get everything done, Swann said.
"Would it be better if they knew?" Swann said of some counties not knowing their ZIP codes are outliers.
"Absolutely. But there are some real challenges here. There is no normal. … I think the better question to ask is, should counties be looking at data at a more granular level than county? The answer to that is yes."
This story was jointly reported and edited by Kate Martin and Frank Taylor of Carolina Public Press, Ames Alexander and Gavin Off of the Charlotte Observer; Aaron Sánchez-Guerra, Dave Hendrickson, Jordan Schrader, Julian Shen-Berro and Lucille Sherman of The News & Observer; Nick Ochsner of WBTV; Emily Featherston of WECT; Tyler Dukes of WRAL; and Jason deBruyn and Jeff Tiberii of WUNC.