Accuracy of U.S. coronavirus data thrown into question as decline in testing skews drop in new cases

For the first time in months, the daily growth of new coronavirus cases in the U.S. has steadily fallen over the past two weeks, giving some hope to U.S. officials who proclaimed there were “signs of progress” in Southern states that were hit particularly hard.

“No one’s declaring victory,” Adm. Brett Giroir, an assistant secretary at HHS, told reporters on a July 30 conference call. “We continue to see signs of progress across the Sun Belt and diffusely throughout the country.”

But testing shortages in key states and other gaps in Covid-19 data call into question the accuracy of those numbers and whether the outbreak in the U.S. is really improving or whether cases are simply going undiagnosed, epidemiologists say.

The country recorded an average of 52,875 new cases every day over the last seven days, down 19% from an average of 65,285 new cases per day on July 28, according to a CNBC analysis of data compiled by Johns Hopkins University. However, Covid-19 testing has declined as well, falling from a seven-day average of about 814,000 tests per day two weeks ago to about 716,000, a 12% decline, over the same two-week period, according to data compiled by the Covid Tracking Project, a volunteer project founded by journalists at The Atlantic magazine.

The decline in testing is particularly acute in some of the hardest hit states with the worst outbreaks, which further skews the overall case numbers across the U.S. 

In Texas, for instance, new cases have fallen by 10% to an average of 7,381 a day from 8,203 two weeks ago, based on a seven-day moving average. Testing, however, is down by 53% over the same time frame. Meanwhile, the percent of positive tests has doubled over the last two weeks to about 24%, according to Johns Hopkins University. That compares with a so-called positivity rate of less than 1% in New York state, which was once considered the epicenter of the outbreak in the U.S.

“I really have come to believe we have entered a real, new, emerging crisis with testing and it is making it hard to know where the pandemic is slowing down and where it’s not,” Dr. Ashish Jha, director of the Harvard Global Health Institute, said in an interview with CNBC. The Texas data, he said, is “very concerning.”

When asked about the drop in testing, the Texas Department of State Health Services said it’s looking into the situation and has reached out to clinical labs and statewide health-care associations.

“We are looking into why the number of lab tests reported to DSHS has been trending lower over the past few days,” Lyndsey Rosales, a spokesperson, said in a statement to CNBC. “We are closely analyzing our data to uncover any anomalies.”

Texas isn’t the only state that’s seen a drop in testing in recent weeks, according to the Covid Tracking Project. Testing has fallen in other states, including Florida, North Carolina and Tennessee, which are home to some of the country’s largest Covid-19 outbreaks. The national drop in testing is driven largely by southern states, the group said.

Why would anyone get a Covid-19 test that takes 10 days to process, Jha asked. Regardless of what’s driving the decrease in test results in many states, he said epidemiologists and public health specialists are unable to determine how bad the outbreak truly is.

“How pathetic are we as a nation that six months into this pandemic, we can’t get this stuff right. We don’t have enough tests. Tests are taking two weeks,” he said. “We can’t figure out where the outbreaks are getting better or worse because our numbers are so messed up that we’re having to squint at the data.”

The U.S. Department of Health and Human Services, which took over collecting U.S. Covid-19 data from the Centers for Disease Control and Prevention last month, didn’t immediately have a comment for this article.

The drop in testing is especially worrisome in Texas, Jha said. Texas was testing an average of 66,400 people a day at its peak on July 23, based on a seven-day average. As of Aug. 11, that number has fallen by more than half to 29,145. Average daily new cases have declined by 23% over the same period.

After factoring in the drop in testing, the decline in cases doesn’t look so good, Jha said.

“I’m not at all convinced it’s getting better. It may in fact be getting worse,” he said.

Catherine Troisi, an epidemiologist with The University of Texas Health Science Center at Houston, said that while testing varies across different regions of the state, overall, it has dropped off substantially at a critical time in the outbreak. The increasing portion of tests coming back positive indicates that the state should be testing more, not less, especially as school districts prepare to reopen in the coming weeks, which she expects to complicate the outbreak even more.

“We know we’re missing a lot of people. Basically, we don’t know what’s happening,” she said. “We’re not casting a wide enough net, which we know because testing is going down.”

Florida, another state that’s been driving the recent slowdown in new U.S. cases, has also seen testing drop. However, that was likely due to Hurricane Isaias, which hit Florida and other states on the East Coast earlier this month, Cindy Prins, an epidemiologist at the University of Florida, said. 

The seven-day average of daily new cases has dropped by 37% compared with two weeks ago, according to Hopkins data, but testing has declined as well. The state was running roughly 54,000 per day two weeks ago, but that has dropped by about 30% to just below 38,000 reported tests as of Aug. 11.

Following the storm, Prins said she expected to see “a really huge jump” in the number of tests processed every day, but instead that number has struggled to return even to the same levels seen in late July, according to the Covid Tracking Project’s data.

“When your testing goes down you find fewer cases,” said Dr. Karen Smith, the former director of the California Department of Public Health.

President Donald Trump said essentially the same thing at a July 14 press conference. “Think of this, if we didn’t do testing, instead of testing over 40 million people, if we did half the testing we would have half the cases,” he said at the time. 

Smith, however, said the cases don’t actually go away. Public health officials just don’t know who’s infected.

“That’s why you cannot say that a drop in positive cases is a significant change in the actual situation in a community, because you can’t make that claim that there are fewer cases unless there’s a stable amount of testing going on,” she said. 

Availability of testing aside, Smith, Jha and others said the U.S. has been using imperfect data to respond to the outbreak from the outset of the pandemic. Smith cited outdated public health data collection methods in many, especially poorer, more rural states, as one reason for scant data.

Even in states with robust health systems, like Smith’s home state of California, technical reporting issues have plagued the country’s response. A glitch in the state’s CalREDIE data reporting system recently failed to process up to 300,000 coronavirus records. Following the mistakes that led to the glitch, the state’s top health official resigned.

California Gov. Gavin Newsom said earlier this week that the backlog of data was cleared over the weekend, and Smith said the problem is unlikely to impact the integrity of the data now that it’s been corrected. Researchers at Hopkins said they anticipate “erratic fluctuations” in its own popular Covid-19 dashboard “that may result in very large spikes at the county and state level” as California reconciles its unreported cases. 

While the error in California will likely lead to a surge in cases over the next few days as the state clears its backlog, similar problems have plagued other state health departments, Smith said.

When it comes to Covid-19 data collection and sharing, she said the U.S. has become “the wild, wild west.”