Can you trust COVID-19 test data? In these states, maybe not.

Public health officials in some states are accused of bungling coronavirus infection statistics or even using a little sleight of hand to deliberately make things look better than they are.

The risk is that politicians, business owners, and ordinary Americans who are making decisions about lockdowns, reopenings, and other day-to-day matters could be left with the impression that the virus is under more control than it actually is.

In Florida, Gov. Ron DeSantis has said his decision to begin reopening his state has been driven by science, and federal epidemiologists have praised his administration’s daily release of COVID-19 related data as especially granular and user-friendly.

But questions about the integrity of the state's public health data were raised anew when the chief architect of Florida’s coronavirus website was fired this week after a dispute over what information should be made public. Rebekah Jones, an information systems manager with the Florida Department of Health, announced in an email to researchers Friday that she was reassigned from her duties overseeing an online dashboard that provides daily snapshots of Florida's COVID-19 infections, testing, and deaths.

In preparation for reopening the state, Mr. DeSantis has lashed out against early prognostications of gloom and doom, saying in recent weeks that many of the state’s hospital beds are lying empty and some testing sites were closed because there was not much demand.

He's used data from the dashboard – including the relatively low rate of people testing positive for the coronavirus – to build support for reopening the state. Still, he said, reopening will happen in phases. This week, restaurants and retail businesses were allowed to open at 50% capacity.

But the firing has provided new fodder against the Republican governor as he defends his handling of the coronavirus outbreak. Mr. DeSantis has also come under fire for his handling of the state’s unemployment system, which broke down after being inundated by hundreds of thousands of Floridians who suddenly lost their jobs because of the economic downturn caused by the outbreak.

“We were told the reopening Florida was built on studying the data. If that data was wrong or manipulated, that puts countless Floridians at risk for exposure to COVID-19," said state Rep. Tracie Davis, a Democrat from Jacksonville and a member of the House Health Committee.

“We do know our state is being reopened and we now have a question mark about the data," she said.

The timing of her reassignment came a day after the state's May 4 rollout of the first phase of the governor’s reopening plan. It was unclear if Ms. Jones was being asked to make Florida appear more solidly in compliance with the White House's criteria for reopening. At the time, Florida was showing a downward trajectory in new infections and rates of positive tests – as well as meeting the rest of the federal criteria.

During a Tuesday press conference, Mr. DeSantis dismissed the matter as “a nonissue” and praised the dashboard as a national model.

Afterward, the governor's spokeswoman, Helen Ferre, said in an email that Ms. Jones had “exhibited a repeated course of insubordination," asserting that Ms. Jones had made “unilateral decisions to modify the Department’s COVID-19 dashboard without input or approval from the epidemiological team or her supervisors.” Ms. Jones, whose dismissal Monday was first reported by Florida Today, could not be reached for comment.

But it's not just Florida facing criticism for its handling of coronavirus data. 

In Virginia, Texas, and Vermont, for example, officials said they have been combining the results of viral tests, which show an active infection, with antibody tests, which show a past infection. Public health experts say that can make for impressive-looking testing totals but does not give a true picture of how the virus is spreading.

In Georgia, one of the earliest states to ease up on lockdowns and assure the public it was safe to go out again, the Department of Public Health published a graph around May 11 that showed new COVID-19 cases declining over time in the most severely affected counties. The daily entries, however, were not arranged in chronological order but in descending order.

For example, the May 7 totals came right before April 26, which was followed by May 3. A quick look at the graph made it appear as if the decline was smoother than it really was. The graph was taken down within about a day.

Georgia state Rep. Jasmine Clark, a Democrat with a doctorate in microbiology, said the graph was a “prime example of malfeasance.”

"Sadly it feels like there’s been an attempt to make the data fit the narrative, and that’s not how data works,” she said.

Republican Gov. Brian Kemp's office denied there was any attempt to deceive the public. According to CNN, "Georgia Department of Public Health spokeswoman Nancy Nydam said the issue was due to 'incorrect sorting logic' that did not consider the date of the confirmed cases."

Guidelines from the Trump administration say that before states begin reopening, they should see a 14-day downward trend in infections. However, some states have reopened when infections were still climbing or had plateaued. States have also been instructed to expand testing and contact tracing.

The U.S. has recorded 1.5 million confirmed infections and over 90,000 deaths.

Vermont and Virginia said they stopped combining the two types of tests in the past few days. Still, health officials in Virginia, where Democratic Gov. Ralph Northam has eased up on restrictions, said that combining the numbers caused “no difference in overall trends."

In Texas, where health officials said last week that they were including some antibody results in their testing totals and case counts, Republican Gov. Greg Abbott said Monday that the numbers were not being commingled. Health officials did not respond to requests for clarification.

Georgia's Department of Public Health also regularly publishes a graph that shows cases over time, except new infections are not listed on the day they came back positive, which is the practice in many other states. Instead, Georgia lists new cases on the day the patient first reported symptoms.

That practice can shift the timeline of the outbreak and make it appear as if the state is moving past the peak.

Georgia Gov. Kemp spokesperson Candice Broce insisted that the governor’s office is not telling the department what to do and that officials are not trying to dress up the data to make Mr. Kemp look better, saying that “could not be further from the truth.”

As for the May 11 graph, Ms. Broce said public health officials were trying to highlight which days had seen the highest peaks of infections. “It was not intended to mislead,” Ms. Broce said Tuesday. “It was always intended to be helpful.”

Thomas Tsai, a professor at the Harvard Global Health Institute, said the way Georgia reports data makes it harder to understand what the current conditions are, and he worries that other states may also be presenting data in a way that doesn't capture the most up-to-date information.

Jennifer Nuzzo, a senior scholar at the Johns Hopkins Center for Health Security, said a lot of these cases are not necessarily the result of any attempt to fool the public. For example, she said, states may not have updated information systems that allow them to tell the difference between an antibody test and a viral test.

Still, if states are mixing a lot of testing numbers together, “you’re not going to be able to make good decisions about reopening and about what level of disease you have in the community,” Ms. Nuzzo said.

This story was reported by The Associated Press. Ms. Smith reported from Providence, Ms. Long from Washington, and Mr. Amy from Atlanta. AP writers around the world contributed to this report. 

Editor’s note: As a public service, the Monitor has removed the paywall for all our coronavirus coverage. It’s free.