Iowa's COVID-19 website rewrites history every day

If you visit coronavirus.iowa.gov and view the graphs on the “case counts” page, you might expect to learn how many Iowans were tested for COVID-19 on any given day, and how many of those tests came back positive or negative.

You would be wrong.

Every day, records of hundreds or thousands of old tests disappear from the website. Consequently, it is impossible to reconstruct an accurate picture of Iowa’s testing numbers or positivity rates, either statewide or at a county level.

“FIXING” ONE PROBLEM CREATED ANOTHER

The disappearing act began after the Iowa Department of Public Health (IDPH) claimed to correct a backdating problem that had persisted for months. The reporting system attached a single date to each individual. Anytime an Iowan was tested multiple times for COVID-19, the result from the newest test would be filed to the date of that person’s first reported result.

The upshot was that new case numbers and positivity rates did not reflect the prevalence of the virus, and old case and testing numbers were inflated as the system continually added new results.

Nurse practitioner Dana Jones brought the backdating to light. (It later emerged that State Medical Director Dr. Caitlin Pedati and her epidemiologist team had been aware of the problem for weeks.) Two days after Bleeding Heartland, the Associated Press, and the Iowa Falls Times Citizen reported on Jones’ findings, the Iowa Department of Public Health announced that the “glitch” had been fixed.

But outsiders soon spotted more discrepancies in the official numbers after the August 19 “fix.” According to data collected by Dr. HyungSub Shim, a behavioral neurologist in Iowa City, 27,912 old tests went missing from August 19 to August 31.

Scott Foens has been tracking daily statistics since the earliest weeks of the pandemic. He found this week that “There was not one test date from the daily testing total data reported on August 20th that matched the data reported on September 1st for Iowa.” Hundreds of tests have been subtracted from most dates since early June.

After the state began publishing positive and negative results from antigen tests on August 28, Shim observed, “We are still losing tests from previous test dates. Just overnight with the addition of the antigen tests and daily PCR tests, we lost at least 6700 old tests.”

Iowa Falls Times Citizen editor Sara Konrad Baranowski has been tracking Hardin County’s statistics since late July and reported on September 1 that “numbers on the website are being altered months later.”

A Times Citizen analysis of the testing data on the website shows that over the last weekend – between Friday, Aug. 28 and Sunday, Aug. 30 – Hardin County testing numbers as far back as May were being altered. Over the course of those two days, the state removed 20 tests on dates between May 18 and Aug. 3, and added a new positive cases for Hardin County on Aug. 6, more than three weeks ago.

You might guess that the IDPH’s epidemiology team would be working to fix these anomalies. Again, you would be wrong.

NO APPARENT PLANS TO FIX THIS PROBLEM

Before the backdating problem became public knowledge, Rob Ramaekers, the lead epidemiologist for the IDPH Surveillance Unit, told Jones, “We recognize this is a problem and have been working on logic to handle it.”

In contrast, the department has not indicated that anyone is trying to stop old tests from disappearing. IDPH spokesperson Amy McCoy shared the following information with the news media on September 1.

For Iowans who get tested more than once, the latest result is used, the older one is removed.

  • For a person with multiple negative tests, we count only the most recent date of test.
  • 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).
  • Additionally, as a part of a positive case investigation, all information received from the laboratory – name, address, date of birth, phone number – are verified. As an example, the investigation may result in address adjustment (therefore county of residence) due to the incorrect zip code.
  • Why would state officials remove older test data by design?

    Why would they count only the most recent test for someone who tested negative more than once?

    Knowing how many Iowans tested negative as well as how many tested positive on any given day is important for calculating the positivity rate statewide or in a county. We need accurate baseline numbers to know whether community spread is accelerating or slowing down in any part of the state.

    Incidentally, it’s not clear McCoy’s explanation fully accounts for the anomalies. On August 28, the day the IDPH began including antigen test results in published totals, Shim observed that at least 6,700 tests had disappeared from the website overnight. The IDPH “still hasn’t said where those tests are going,” Shim added. “I think that they are removing previous tests on people who have been retested, but we only added 6003 tests yesterday, so that doesn’t account for it all.”

    Governor Kim Reynolds has often boasted about Iowa’s supposed “transparency” in providing information about the pandemic. She is proud of the A+ grade the state website received from The Atlantic’s COVID Tracking Project.

    Rewriting the past on a daily basis is not A+ work, and it’s the opposite of transparent. Iowans should be able to rely on published numbers, without wondering how many tests disappeared from whatever day we are checking.

    A FIX THAT WOULD STICK

    To restore trust, the IDPH needs to heed the pleas of public health and medical professionals and release a full dataset of COVID-19 tests and results, anonymized to protect patient privacy. As Dr. Ryan Flannery, a family physician in Washington County, explained in a recent interview with Bleeding Heartland,

    I have to be on the front lines selling all of this. I’m the one that has to have faith in the information that I’m sharing so that I can actually boost patients’ confidence in what we’re supposed to do.

    Because right now, if people don’t have confidence in the data, then why are they going to listen to anything I recommend or anybody recommends about quarantining and isolating and other treatment options that we have right now to stop the spread of this disease?

    Flannery argued that it’s crucial to fix these problems before cold and flu season.

    We just need to know where the starting point is to know how bad potentially we’re going to be at if things get worse. And if we don’t have the great accurate numbers to start with, it’s going to really hamper our efforts to take care of people the way we need to if things were to get worse. […]

    I just want to have the accurate data. […] That’s all I’m asking.

    I’ll update this post after the governor’s news conference, scheduled for 11:00 am on September 2. Reynolds’ staff don’t allow me to participate, but I hope other reporters will ask the governor and/or Dr. Pedati to justify the practice of removing old results for individuals who are retested for COVID-19.

    UPDATE: Katarina Sostaric of Iowa Public Radio asked about continuing problems with the website after the IDPH claimed to have “fixed” the backdating issue. “What is going on with the state’s data, and how can anyone rely on what they’re seeing?” Reynolds and Pedati answered the question without acknowledging the central problem.

    Reynolds said, “They can have full confidence in the data, because the overall data hasn’t changed.” (Not true: thousands of old tests continue to disappear.) She attributed some problems to the state’s “antiquated system” for disease reporting, which was “never designed to deal with the amount of data that we’re feeding through it.”

    Reynolds also claimed that sometimes the IDPH doesn’t get complete information from clinics, so data may be adjusted later. That does not explain the practice of removing old tests for Iowans who have been tested more than once.

    Pedati stepped up to explain that “all systems” require maintenance and surveillance and monitoring. She was referring to maintenance on Saturday, August 29, which caused more than 200 COVID-19 deaths and tens of thousands of tests to temporarily disappear. But that problem was resolved in hours, whereas the issue discussed in this post is ongoing.

    Pedati said the state was looking to improve the reporting system and claimed it had been a “primary concern” for the department before the pandemic.

    Like Reynolds, she asserted that occasional “adjustments” to the data happened after the state confirmed details with patients or providers. “We’re going to keep making those adjustments, because it’s the right thing to do for Iowans.”

    When another reporter attempted to follow up, asking why this “antiquated system” hadn’t been updated before the pandemic, Reynolds and Pedati walked away.

    I sought comment from Professor Joseph Cavanaugh, who leads the University of Iowa’s Department of Biostatistics, on the practices for reporting of COVID-19 testing on the IDPH website, as described by McCoy. He replied via email on September 2,

    From my perspective, for the sake of transparency and the ease of interpretability, the protocol that the IDPH should use in reporting test results is fairly straightforward. On each day, the number of tests conducted (PCR and antigen) should be reported. When the test result is obtained, it should be recorded for the same day that the test was conducted. Since PCR results are often delayed, recent daily positives may vary, but eventually, they should stabilize. Past test results should not be changed, even if a subject is tested multiple times. Any deviations from this protocol should be clearly articulated on the IDPH website.

    According to Dana Jones, she uncovered the backdating anomalies through simple arithmetic (and careful attention to detail). If the data are reported using this straightforward protocol, the results will stand the test of arithmetic. Otherwise, the data will be inscrutable, and historical data will be in a constant state of flux.

    Dr. Eli Perencevich, an epidemiologist and infectious disease specialist at the University of Iowa, agreed on September 3 that “no tests should be removed.”

    If you were tested in May, that needs to stay in the database no matter what happened in the future. We need this historical data so that we can go back and analyze the impact of policies and interventions and understand the complete story of COVID in the state. It would be like removing football scores from last week because the teams played this week, or something like that. We need to look at football scores for the whole season and not just the latest game.

    LATER UPDATE: The disappearance of tests is not an occasional problem related to “scheduled maintenance.” Dr. Shim posted more graphs on September 10 showing the anomalies.

    Dana Jones continues to track this problem as well.

    • What a mess...

      I cannot know, as I have no direct knowledge. But I can imagine that as this health crisis developed so quickly, there was never a team of smarter-than-the-average-bear leaders with enough competence to anticipate, plan, and articulate to the IT team a devils-in-the-details level of data and reporting requirements. Then add to that, system testing probably gets short-changed because we’re in a crisis right now. I actually can feel the pain of the IT people in the trenches trying to do what is right when in many cases in this mess there has been no leader giving adequate guidance or definition as to “what is right” or what are the specific requirements.

      This episode reminds me of when I kept thinking over and over in the weeks/months before the Democratic caucus, where is the public evidence that this new phone application has been adequately tested? It’s going to be a mess, probably. And it was a mess. The world of IT can be really, really fun, but it can also be a crying shame embarrasing mess when a project lacks enough really smart people (from top to bottom) and adequate time to properly implement. And then when the public facing leaders are either dishonest or simply don’t know what they’re talking about, it just adds to the frustration of really smart people on the outside trying to make sense of it all.

    • I just tried to get info from IDPH covid website...

      …and it won’t show me any information at all. I check that website every other day (though I no longer trust what I read), and I don’t think the latest problem is on my end.

      To me, the biggest arrogance on the part of the governor’s office is not even bothering to respond to local-government requests for permission to enact mask mandates. If I were bungling a major responsibility as badly and obviously as Reynolds is bungling covid, I wouldn’t be so quick to unnecessarily disrespect local government officials. I think they will remember in two years.

    • Thank you

      Thank you so much for continuing to dig for the truth and inform Iowans of what our government is doing.

    You need to signin or signup to post a comment.