New changes to state’s virus data confuse experts, residents alike

While the state has added to the metrics it publishes, it has stopped running others and shifted its method of counting cases. This has confused ordinary Georgians as they decide whether Gov. Brian Kemp was right to begin reopening the state’s restaurants, movie theaters and hair salons.

On April 27, at 8:25 p.m., Georgia passed a tragic milestone: 1,000 deaths during the novel coronavirus pandemic

Or maybe it didn’t. Within hours, the state Department of Public Health said on its website that the death toll was actually 995.

Or maybe it was 982, based on a county-by-county count DPH also published.

Death counts and other essential figures DPH publishes on COVID-19 have ticked up and down as the state changes the way that it reports them. While the state has added to the metrics it publishes, it has stopped running others and shifted its method of counting cases. This has confused ordinary Georgians as they decide whether Gov. Brian Kemp was right to begin reopening the state's restaurants, movie theaters and hair salons.

In the past week alone, DPH rolled out a new data dashboard that added bar and pie charts. But they deleted a daily count of completed tests, making it difficult for the public to determine if Georgia is meeting a key White House criteria for reopening.

Gone is a county-by-county list of reported deaths by gender and age, which locals were using to verify their own counts. Some figures are written in such small type that they are no longer legible. Others are 10 times as large as they once were. Much of it is now published in a format that makes it difficult for researchers to use.

» RELATED: ‘Confused and scared’: Georgians frustrated over shifting virus data

» RELATED: Amid conflicts over outbreak's path, Kemp weighs Georgia's next steps

Meanwhile, the state Dept. of Public Health has not posted measures that Commissioner Kathleen Toomey said leaders are using to make their decisions. Data on hospitalizations, cases of flu-like illnesses, and share of positive cases of all tests performed were "encouraging," she said in a YouTube video released late Monday, but gave no numbers.

A DPH spokeswoman did not respond to inquiries by The Atlanta Journal-Constitution on why it changed its reporting, but did say the agency is trying to fix problems.

“The website dashboard is new as of yesterday and there are aspects of some data being worked on,” spokeswoman Nancy Nydam said in an email. “The data team is working as quickly as they can to make sure we are providing the most accurate, up-to-date information possible.”

Corrections are to be expected as a state copes with an unprecedented pandemic in real time, but they add to the skepticism of those who are trying to make sense of the disease’s impact on Georgia during a deeply politicized time. So many changes have been done without clear explanation that the state has opened the door for critics to say the numbers can’t be trusted.

“People start saying, ‘Oh you’re trying to hide the numbers’,” said Kennessaw State University professor J.C. Bradbury. The economist is not a public health expert, but he has tracked DPH figures closely. When the state unveiled its new dashboard Monday evening, many of the state’s figures no longer matched what he had recorded.

“I think everyone gets mistakes will be made,” said Aaron Levine, an associate professor at Georgia Tech’s school of public policy who has also been critical about how DPH presents its data.. “This is real time. There’s lots of public attention with messy data sources coming from all sorts county health departments.”

“But just be transparent about it. If there needs to be a change, just acknowledge it and move on. I think that goes a long way towards the trust.”

Bradbury, Levine and others have taken to Twitter to point out inconsistencies or errors. Sometimes, they’re quietly corrected on DPH’s page with no explanation, Levine said.

» DASHBOARD:   Real-time stats and charts tracking coronavirus in Georgia

» MORE:   Map tracks coronavirus globally in real time 

At times, state officials have run the risk of misinforming the public.

Charts and graphs introduced last week on DPH’s reporting website may appear to the casual reader as showing a dramatic decline in new cases, but that’s misleading. A different counting method pushes back the date a case is tallied as “new” by days or weeks before what was originally reported, so those figures will always be artificially low for days as results trickle in.

There are some important improvements. The chart that makes it appear that new cases are rapidly dropping in Georgia now has a better disclaimer, Bradbury and Levine noted.

A “Frequently Asked Questions” page describes how the state tallies much of the information, although some of the descriptions are not clear.

Yet some of the FAQ’s descriptions show the data are not what some may think they are.

Consider “Hospitalizations,” a metric that debuted Monday with the state’s new data dashboard. They were reported at 4,837 as of 5 p.m. Tuesday.

That figure is not the number of people currently hospitalized in Georgia for the novel coronavirus. It’s the number of Georgians who happened to be hospitalized when they were tested.

If someone was hospitalized after they tested positive, they are never included in that count, Levine noted. This is of little use to those trying to determine if the state has enough hospital beds.

“It’s not an indication of resource utilization in any meaningful way,” Levine said. “It’s a quirk of when we’re testing people.”

Georgians are caught in the middle.

Mary Phelps, 41, of Sandy Springs hoped to use DPH figures to decide when it’s safe for her family to return to normal life, but found it confusing when she could not square newly posted data with figures it had posted previously.

As a financial sector professional, she is used to working with data and wonders whether what has been published is true.

“The uncertainty about the data adds to my concern,” Phelps said. “When will I be comfortable if I don’t trust the data?”

Staff writers Pete Corson, Emily Merwin DiRico, and Isaac Sabetai contributed to this report.