Between May and early July, the Galveston County Health District underreported COVID-19 cases in Galveston County.
Then, as a state backlog of cases began to reach local health officials, the district began over-reporting the daily number of COVID-19 cases.
The numbers of recent tests were mingled with results of older tests — giving the appearance that the rate of COVID-19 infection in the county was higher than officials really believed it was.
Recent changes to the way the Galveston County Health District reports COVID-19 tests and test results reveal many gaps and flaws in what officials knew and reported about local spread of the virus earlier this year.
Health officials now say a new system, which debuted last month, provides a clearer understanding of what COVID-19 did in Galveston County in 2020, although they acknowledge bugs remain to work out of the system.
The health district rolled out a new daily COVID-19 reporting format in September. The report, which is generated from state and local data, provides information about the number of diagnosed cases in the county, as well as active cases, deaths and recoveries.
Crucially, the data is sorted by the date COVID-19 tests were collected and by the date the results were ultimately reported to the health district.
Comparing the two data sets reveals how many positive tests were reported late to the county.
From May 17 to July 4, the health district underreported the number of new positive cases by more than 1,300, according to a Daily News analysis of the new data.
For example, during the week of June 21, 1,472 people tested positive for the virus, according to the health district’s new data based on when the tests were collected. At the time, however, the district reported only 1,028 new cases, an undercount of 444.
The situation flipped in July, when the health district began receiving backlogged test information. Between July 5 and Sept. 12, the district reported 1,168 backlogged tests collected at least a week before they were reported.
During the week of Sept. 6, for example, at least 316 tests reported were a week or more old.
That week illustrated the deeper problem created by reporting the spike of backlogged cases. Including the old tests, which were not identified as being old, the positivity rate in Galveston County was 8.9 percent.
Accounting for the backlog, however, the positivity rate was only 3 percent..
“If you look at the reported date, we had a little mini surge in September,” said Phillip Keiser, Galveston County’s local health authority. “If you were looking at it and thinking about it and sweating — ‘Is this it? Is this our next surge?’ — it turns out, if you’re looking at it on a day-by-day basis, it’s not.”
The new reporting method and graphical reporting tools the district is using now are more accurate but aren’t perfect, officials said.
Health district officials have acknowledged problems with the underlying data used to generate the charts and graphs, which make some parts of the graphics incorrect. For instance, a graphic showing cases over time among Hispanic residents shows more than a dozen people died before March 16.
In fact, the first local COVID-19 death didn’t occur until April 3, officials said. The graphic was incorrect because underlying data about when tests were collected and reported was missing, and the system defaulted to counting those deaths as early as possible, officials said.
The graphic was updated this week with a note explaining the discrepancy.
The district’s new reporting system includes specifics about what information has been corrected or updated on a day-to-day basis. It isn’t immediately clear on a given day where a backlogged case is recorded, however.
“Unfortunately, case data doesn’t always come to the health district complete, or clean,” spokeswoman Ashley Tompkins said. “Some fields are left blank. Through case investigations, staff is able to fill in many of those blanks and make sure data being reported on the dashboard is correct.”
The change also had consequences outside of the district’s control. A graphic The Daily News had published daily since April was created using data provided by the health district. It was based on the district’s reported cases and repeated information now known to be inaccurate.
The Daily News submitted a public records request to the health district for a copy of the database now being used to generate the district’s graphic, including information about when tests were collected.
The health district estimated that providing the data would cost the newspaper up to $3,750 because of the work needed to remove private information from the database before release.
If the scrubbed data were released it might not include information, such as dates, useful for charting cases over time and identifying trends, health district officials said.
The Daily News hasn’t yet challenged the district’s stance on the data and is in discussions with the agency about what information it would release publicly and whether that information would be useful in illustrating trends.