The deaths were put into the daily totals in clumps as investigators tracked them down, which had the effect of boosting the numbers that came out each of those days, said Lilian Peake, Virginia’s state epidemiologist.
“The timing of the problem that we had with the data was very unfortunate,” Peake said, because the backlog coincided with an increase in cases over the holidays.
It’s the latest sign of the challenge public officials across the country face in tracking the disease, relying on systems that in many cases are archaic and in most cases were never designed for an emergency of this magnitude.
Last month, the state of Ohio blamed human error after discovering that more than 4,000 coronavirus-related deaths had been left out of totals for November and December, and an audit in Indiana turned up more than 1,500 deaths mistakenly omitted from coronavirus totals since the pandemic began.
A recent audit in Los Angeles County turned up more than 800 additional coronavirus deaths from the holiday season, and last summer a server outage in California created a backlog of nearly 300,000 cases of covid-19 that had gone unreported statewide.
“The systems, as we’re finding, are fairly antiquated,” said Jennifer Nuzzo of the Coronavirus Resource Center at Johns Hopkins University.
The public is watching the numbers on a daily basis, she said, and any discrepancy fuels anxiety about the true nature of the crisis. While some people fear the pandemic is being underplayed — such as questions about the scope of nursing home deaths in New York — others looking to make a political statement can use the numbers to push the unfounded claim that the coronavirus is not a serious threat.
“Just the level of mistrust that people have in all of this has been eye-opening for me, and also worrisome,” Nuzzo said, adding that she believes audits will eventually show deaths from the coronavirus are being undercounted nationwide.
In Virginia, Peake cautioned that the data system was not designed to meet public expectations for up-to-the-minute information about the spread of the disease.
“It’s human nature to say, ‘I want to see what was the number yesterday compared to what was the number today,’ ” she said. “But we haven’t tracked any disease real-time like this — this is the first time.”
Virginia tracks coronavirus deaths in two ways. The first is through a disease surveillance process that’s mainly set up to chart the spread of infections, not the ultimate outcome of each case, Peake said.
When a lab notifies the state health department that someone has tested positive for the coronavirus, investigators will call that person in a few days to check on their status. But there’s no built-in way to follow up weeks later and see if that person got better, wound up in the hospital or worse.
“There are not enough epidemiologists to track those cases the whole way through,” Peake said.
More deaths are reported through the second method of tracking, in which a computer system scours death certificates for signs that indicate covid-19 as a possible cause of death. Each certificate it identifies is examined by an epidemiologist to make a final determination.
That system had the error earlier this year, Peake said.
Because of a programming flaw, the system stopped recognizing anything with a 2021 date, she said — just as everyone feared would happen with computer systems worldwide two decades ago with the Y2K phenomenon, when software programmers had to deal with the date change from 1999 to 2000.
Once the problem was identified and fixed, officials had to go back and examine each death certificate in the backlog while still processing new cases as they came in.
Peake said the situation points out an issue that’s sometimes overlooked in the rush to parse the daily virus statistics: Deaths should be viewed in terms of when they occurred, not when they are reported.
The recent backlog shows that deaths from the disease climbed to a peak around the third week in January — about two weeks after a spike in infections from the holiday season, when families got together and mask-wearing may have fallen off.
In terms of timing, “that’s exactly what you’d expect to see,” Peake said. She added that she’s confident the data problem has been fixed, but said she couldn’t guarantee other issues won’t arise.
The state has added new computer servers and hired personnel to try to keep up, but Peake said she hopes the situation nationally will lead the federal government to invest more in public health data systems.
“I can’t emphasize enough, our surveillance systems in public health were not created for what the public would like right now,” she said.