News →

Virginia Coronavirus Stats Only Capture ‘Tip of the Iceberg’

Technician in lab coat watches samples of COVID-19
A technician at GenetWorx lab in Glen Allen tests samples of COVID-19 (Crixell Matthews/VPM). 

The number of positive COVID-19 cases in Virginia has quadrupled in the last week.

But experts say the daily numbers released by the Department of Health is still an undercount exacerbated by some of the country’s lowest per capita testing.

Alex Krist, an associate professor of family and population medicine at VCU, said the state data represents “just the tip of the iceberg.”

He’s seen an increasing number of patients with respiratory problems at Fairfax Family Practice. In an interview on Thursday, Krist estimated the practice, which has 11 locations, has diagnosed more unofficial coronavirus patients than the official, tested tally for the entire county.

State officials say they have other tools at their disposal. They’re tracking visits to emergency rooms and urgent care clinics, where visits related to respiratory illness have risen sharply.

Researchers in Virginia say the data will improve with time as testing increases and they learn more about the spread. A team at the University of Virginia is already synthesizing data from around the world to understand the virus and predict its spread.

At least one model has already come out. Researchers at the University of Washington project Virginia could suffer over 2,000 deaths, with a peak in mid-May. But like everything else with coronavirus, those estimates may change quickly.

Testing Shortages

Krist says he’s seen a growing number of patients in recent weeks with respiratory problems. In some cases, Krist is confident he’s seeing a COVID-19 infection based on their symptoms and travel history.

But he can’t know for sure without a test -- a tool he’s used only rarely.

“We've mainly reserved testing for people where it's going to change the type of care that we provide,” Krist said, citing guidance from the Virginia Health Department.

Krist is concerned that the Washington DC area may be harder hit than we currently understand. It’s close to New York and has flights across the world.

“It could very well be kind of a next epicenter for cases of coronavirus and we should be testing people and trying to track this to try and curb the spread,” he said.

Virginia ranks in the bottom third of testing per capita, according to a VPM analysis of COVID-19 data. Only six states and the territory of Puerto Rico have lower testing rates.

State officials say the federal government prioritized harder-hit states for testing.

Chart of COVID-19 testing
Source: Virginia Department of Health

But testing has increased in the commonwealth. On Monday, the state reported results from 1400 new tests, up from 360 a week before (a number that was unusually low).

Dr. Laurie Forlano, deputy commissioner for population health for VDH, said the state hopes to broaden its criteria on who could get tested as more became kits become available. And she said it’s not unusual to have incomplete information on the spread of infectious diseases.

“In things like respiratory illness -- flu, for example -- we do the best to capture as much of the case counts as we can and then make estimates,” Forlano said.

The daily coronavirus numbers produced by the state only include people who’ve been tested. But Forlano says the state is also tracking times when doctors diagnose someone with COVID-19 but don’t actually test them.

State officials can then use this data to look for trends, searching for words like ‘cough’ or ‘COVID.’ VDH commonly uses “ syndromic surveillance ” during the flu season to spot clusters of cases or to track injuries after a natural disaster.

“It's a rougher system, but it's pretty remarkable how it almost always tracks the way we would expect it to,” she said.

The department’s data shows about four percent of patients admitted to Virginia’s emergency rooms last week showed symptoms of coronavirus, like fever with a cough. That’s up from 1.5% from the month of February.

Crowdsourcing Pandemic Data

At some point, this so-called syndromic surveillance might be done digitally, according to Madhav Marathe, a professor in biocomplexity at the University of Virginia.

“You can imagine apps on your phones where you essentially press a button every day saying ‘Were you feeling sick or not?’” Marathe said.

Marathe said this sort of crowdsourcing may be critical to understanding pandemics. But the data will never be perfect, with an entire field, called “now-casting,” devoted to getting a grip on real-time data.

“Even trying to tell what the correct number today is a scientific challenge because of measurement errors, because everybody doesn't report it at that time, because of coding errors,” Marathe said.

Still, there’s more data than ever before, and more computer power to analyze it. For example, UVA's Biocomplexity Institute has tracked whether airport closures have helped slow the spread of the disease. They’ve also built two COVID-19 dashboards that synthesize data ranging from the World Health Organization to news stories. 

Screenshot of UVA's COVID-19 dashboard
UVA's COVID-19 Surveillance Dashboard uses a number of open source data sets to try to cross-check data on the spread of the disease.

Marathe says this kind of data-driven pandemic analysis is still relatively new.

“We have made tremendous progress as a society to be able to even talk about it in this manner,” Marathe said.

The Biocomplexity Institute recently won a $10 million grant from the National Science Foundation to extend its pandemic planning work using advances in artificial intelligence and big data.

The Institute is also putting the finishing touches on a model it’s developed alongside health officials. Marathe expects the model will improve day-by-day as researchers tweak its predictions based on the latest numbers. 

They expect to debut it in the coming weeks.