Five health experts on the state’s coronavirus data backlog

After state officials this month disclosed hundreds of thousands of coronavirus tests had not been previously reported — a backlog that has distorted metrics used to gauge the toll of the pandemic — Gov. Greg Abbott said Sunday that the current data is “far more accurate than what we had last month.”

The disclosure of the backlog was the latest in a string of data problems that have plagued the state’s public accounting of the pandemic and came as schools and universities prepared to resume instruction for the fall term.

While patients were notified of their test results, the cases were not reported to some local health departments, which meant those departments couldn’t assign contact tracers to determine who may have been exposed to the virus.

The problems left some local officials and lawmakers frustrated.

“This pattern of folks at every level pretending like they have everything handled while behind closed doors they’re so overwhelmed and behind that the data is meaningless… it’s toxic and dishonest,” state Rep. Erin Zwiener, D-Driftwood, said on Twitter.

Collin County took the unusual step of stamping its coronavirus dashboard with a warning that officials lack confidence in the data provided by the state.

“The Commissioners Court is 100% certain that the COVID-19 data being reported for Collin County is inaccurate,” County Judge Chris Hill wrote on Facebook last week. The state has since assigned a dozen investigators to update the county’s data.

Lara Anton, a spokesperson for the Department of State Health Services, said positive tests revealed in the backlog show there might have been more infections in July than state officials previously knew about, but the “additional cases would not have changed the overall trend or the recommendations for the public.”

Several issues caused the delay in reporting more than 850,000 test results, some dating back to March. The first logjam of some 350,000 tests accumulated because the state could not process enough test results each day before an Aug. 1 system upgrade expanded its capacity.

In July, the state identified a separate issue with a private lab’s data: An errant question mark in one data field prevented about 354,000 test results from transferring to the state’s system. Two other laboratories that the state began working with had lengthy onboarding processes and formatting issues that caused delays in uploading their test results, Anton said.

The agency has fast-tracked the onboarding of more than 300 labs to provide testing data as quickly as possible, Anton said. The normal process can take between six and eight months to work out the bugs with uploading results to the electronic system.

Meanwhile, the state’s positivity rate — the rolling average of people who test positive — swung wildly this month as backlogged tests were added, soaring to 24.5% Aug. 11 and plummeting to 10.8% a week later.

Abbott, who frequently cites the metric to explain his decision-making, has said anything above a 10% would raise a “warning flag” while a sustained rate below that could prompt him to loosen restrictions.

The Texas Tribune spoke to epidemiologists and health experts about the backlogged tests. The Tribune also spoke to the same experts about school reopenings.

Question: What’s the effect of the backlog — and the later data dump — from a public health standpoint?

Catherine Troisi, infectious disease epidemiologist at UTHealth School of Public Health in Houston: The problem of course is that you’re just not sure what’s happening in the community… If I had a test two weeks ago and the result came back today and said I was positive, I would be a tick on August 24th instead of when I was tested, and ideally when I had symptoms. So that makes it hard, too. You’re dumping all these cases now that are from a while ago so it … can really skew the information about when cases are occurring.

Dr. Ron Cook, professor at Texas Tech University Health Sciences Center and the city of Lubbock’s public health authority: The biggest problem … is the fact that when we find out a week or two later, several days post facto, in some cases 10 days or so later, that we get a positive test result, then who knows how many thousands of people they’ve spread that to exponentially, right? Ideally you want to get a positive test as soon as possible, call them as soon as possible and get in contact with them to find out who they’ve been in contact with closely.

Dr. James McDeavitt, dean of clinical affairs at Baylor College of Medicine: I want to start with just a word of sympathy for the people running the state reporting system. This is unprecedented.

Just a couple of facts to put in perspective: In terms of mandatory reporting and public health issues, if a lab has a positive [tuberculosis] test, for example, it must be reported. However, there are relatively few positive TB tests, and there’s no expectation that you report negative tests, only the positive ones. Well, with COVID-19, the expectation is you’re reporting all the positives, and you are also reporting all the negatives, so that’s a huge ramp up of reporting requirements for the state. So that’s number one.

Second factor: the poor analysts at the state, they’re getting data feeds from I think 600 different laboratories across the state. Commercial labs, hospital labs, independent labs in all kinds of different electronic formats — and in some cases non electronic formats, there are paper, pencil and fax records that they’ve all got to reconcile — so it’s been a real challenge I think for them to get ramped up. Hopefully a lesson we’ll learn out of this pandemic – locally, regionally and nationally – is we have to have better data and reporting systems in advance of having to deal with something like this.

That note of sympathy aside … we’ve seen the hospital numbers come down really sharply and the community cases staying stubbornly flat or maybe drifting down just a little bit but not as quickly as the hospital numbers are declining. This was a disconnect that made no sense to me three weeks ago: that we’d see such a difference in the rate of decline between these two numbers. Well it turns out, I think what has probably been happening is this backlog of cases is now being loaded into the system and they’re appearing to us as new cases; cases that could have been from a week or two weeks or four weeks ago.

I’m not exactly sure how far back the backlog goes. So the impact of that is we lose a major signpost. Until the backlog is resolved, it is hard to interpret the state reporting numbers. Presumably, once they are caught up the data will have more integrity, although there are some additional methodologic issues they need to work through. For example, is a positive test result reported on the day the sample is collected, the day the test is run or the day the data is entered into the system? How is the state accounting for multiple tests of the same individual?

Angela Clendenin, instructional assistant professor of epidemiology and biostatistics at the Texas A&M School of Public Health: I think you have to understand how we got here to be able to understand what it means moving forward. … Whenever you’re faced with a disease like COVID that we’ve never seen before and we knew little about in the beginning, and early on our ability to test was limited, we went about our process like we do with every other infectious reportable disease. Testing was going to be only at the public health laboratories, and then capacity was quickly exceeded. Then once we got capacity through commercial laboratories, then we’re running out of reagents [chemicals required for testing] and then when we have reagents. It was [a shortage of] swabs and then when we had swabs, now it’s [a shortage of] people to run the test. So that causes a delay of up to seven days in some cases.

Rajesh Nandy, associate professor of biostatistics and epidemiology at the University of North Texas Health Science Center at Fort Worth: Of course, that skews our perception of where the state of COVID was at that time — a month or two months ago when the backlogs are coming from. Obviously that’s not desirable because we need the data to be accurate whenever we are trying to assess the current state of COVID and study the trends, and also when you’re trying to see what lies ahead.

But that said, these numbers do not have much impact in studying the current trend.

One thing I always tell people from the media … the numbers that we see from the county health departments or the state health department … they’re not useless, but they are highly qualified and unreliable in terms of studying the trend. …

There is also the issue of volatility in the number of new reported cases based on the current level of testing. So, when the testing level is high, the numbers tend to go up for obvious reasons. Whereas when some counties are having some shortage of test kits — at that time what happens is the number of reported new cases is low but the positivity rate is high. So that’s another layer of volatility in the data.

Disclosure: Texas Tech University, UTHealth, the University of North Texas, and Texas A&M University have been financial supporters of The Texas Tribune, a nonprofit, nonpartisan news organization that is funded in part by donations from members, foundations and corporate sponsors. Financial supporters play no role in the Tribune’s journalism. Find a complete list of them here.