For the last two weeks, the Iowa Department of Public Health has been altering the dates of some COVID-19 tests on its website, a practice that the department says is intended to eliminate multiple tests of a single person, but a top infectious disease expert warns will give a false picture of the state’s coronavirus history.
Two weeks ago, state officials said they had fixed a “glitch” on the coronavirus data reporting website. For months, the Iowa Disease Surveillance System (IDSS) had been erroneously backdating positive and negative test results. The issue occurred when a person was tested multiple times. If their first COVID-19 test was negative and a second test was positive, the positive result was defaulting to the date of the first test. The glitch skewed positivity rates by assigning positive and negative test results to months-old dates.
Dr. Caitlin Pedati, the state’s medical director, said she became aware of the backdating issue in late July, and Gov. Kim Reynolds said she learned of it the second week of August. Both blamed the issue on “antiquated” software that was not built to handle the amount of coronavirus data. The glitch was fixed on Aug. 19, changing daily positive and negative case counts, and the number of tests that were reported to have been collected each day.
But a Times Citizen analysis of the data on the state’s coronavirus website shows that in the two weeks since the glitch was fixed, the number of COVID-19 tests being reported per day on past dates has been changing. Between Aug. 19, when the state fixed the glitch and new daily testing numbers replaced old numbers, and Wednesday this week, daily testing totals of Hardin County residents were altered for 67 out of 161 total days. For more than three-quarters of those days the fluctuations were just one or two tests, but other dates’ totals have changed by four or five tests.
The changes are altering cumulative test totals - and positivity rates - for past days and weeks. For example, earlier this summer - on June 8 - the state reported that it had tested 728 Hardin County residents to date. The numbers that were on the website on Aug. 19 showed that the June 8 cumulative test total was 666. And on Wednesday this week, the cumulative testing total for June 8, according to the new numbers on the website, was 645.
Amy McCoy, a spokesperson for the Iowa Department of Public Health, replied to Times Citizen questions about the changes this week. She said that when an Iowan is tested more than once, the latest result is used and the older result is removed.
“For a person with multiple positive tests, we count the second positive only if the positive occurs greater than 90 days from the previous positive test (per CDC guidelines for an individual who tests positive multiple times),” McCoy wrote in an email on Tuesday. She has not replied to follow-up questions about the implications of that policy, or where guidance for the practice came from.
The issue of past testing numbers changing was reported on Wednesday by the Bleeding Heartland blog.
While the changes may seem to pose less of an issue that backdating test results – they’re not skewing data for today or even the past week – experts warn the practice could hinder the state’s ability to fully understand the history of the disease in Iowa.
Eli Perencevich, an infectious disease doctor at the University of Iowa, said it’s standard epidemiological practice to leave data unaltered.
“You can’t go back and exclude people in the past based on new information now,” he said on Wednesday. “If they were tested then, you have to keep that information.
“It’s a new glitch,” he said of the data alterations.
Whereas the problem that resulted in the backdating of COVID-19 test results artificially lowered current infection rates by assigning them to dates in the past, Perencevich said this new practice could affect response to new outbreaks.
“For historical purposes, we need to understand where we were in the pandemic,” he said. “We’re trying to understand what works. There were things we’ve done – we had pretty significant social distancing happening early in the outbreak and then we removed restrictions in 77 counties and then 22 counties. We can go back and look at the data and look at how those interventions by the state impacted COVID cases. But if they’re removing the cases, we can no longer study the impact. We can’t determine what worked in the past to determine what we should do in the future.
“There are several of these things that IDPH has done that are just kind of no-nos in simple epidemiology,” he said.
At a press conference on Wednesday, Reynolds, when was asked whether Iowans can trust the data being published by the state, said “They can have full confidence in the data because the overall data hasn’t changed.”
She again referenced the limitations of the state's disease tracking system.
“We’re continuing to work with an antiquated system with IDSS,” Reynolds said. “There continues to be just issues in trying to take a system that was never designed to deal with the amount of data that we’re feeding through it and the scope at which we’re collecting data from hospitals and clinics across the state. It’s very complicated, and that causes issues.”
Perencevich pointed to an offer from Iowa medical experts this spring to work with the state to build models to track the disease and project future trends. The Iowa Department of Public Health rejected those offers.
“There are a lot of state employees – from the University of Iowa, Iowa State, some from Northern Iowa – folks that would want to help, but we’re not being pulled in. And I get that they’re underfunded and overworked and this could be because they’re stretched thin, especially now that things are exploding, but part of this is their own doing.”