VDH’s Data Update Problem

by Carol J. Bova

Governor Ralph Northam and Virginia’s public health officials say they want to “follow the science” and “follow the data” when managing the COVID-19 lockdown. Unfortunately, the data keeps changing.

Last week the Virginia Department of Health made 1,021 changes to the dataset of regional COVID-19 cases by onset date between March and October — adding 1,361 cases to the total. Forty-five percent of the dataset’s 2,258 regional entries from February through October were changed. The VDH dashboard has no footnotes explaining why the changes were made or the source of the new data.

Onset date should show increasing or decreasing trends. The VDH website says, “For cases, the date of symptom onset is the closest date to when a person’s symptoms began. This measure shows increasing or decreasing trends in the number of COVID-19 cases and deaths in Virginia.” VDH also says, “Please note that data are considered preliminary and subject to change.”

Citizens must ask what validity or usefulness these changes offer one to eight months after they’ve already been considered and incorporated into the Virginia COVID-19 models and forecasts? The R0 (reproductive number) that suggests the likely spread of infections is based on recent case numbers. Were the missing numbers enough to skew the results?

Those model results and forecasts were used to evaluate the results of social behavior on virus spread and impose or relax restrictions. How long has this backfilling of months-old data been going on?

I was able to find the changes because I noticed changes in some onset dates that I had recorded. I downloaded the dataset covering dates through November 15 and so I could use that as a point of comparison with the November 23 set. Previous datasets are not available online, only the most current.

Why was months-old data entered last week? Were the changes the result of recent entries of old information to the Virginia Electronic Disease Surveillance System (VEDSS)? VDH says, “These sources include laboratory reports of COVID-19 test results, case investigation interviews conducted by the health department, monitoring of close contacts, and syndromic surveillance for coronavirus-like illness which is a system used to receive surveillance data from these various sources and report that data to CDC.”

There are questions that VDH needs to answer:

  • Why would old data show up now? Who dropped the ball?
  • Was contract tracing ever done on the additional cases?
  • Has VDH come up with a plan to ensure timely entry of data in the future?

Some citizens might take the questioning a step further and wonder if VDH has “stockpiled” cases to guarantee a November surge to justify any actions the Governor may choose to take. Without earlier symptom onset datasets, it’s impossible to know when the backfilling started and to what degree it occurred. Without the actual data, I have to admit the thought has crossed my mind.

Carol J. Bova is a writer living in Mathews County.

There are currently no comments highlighted.

14 responses to “VDH’s Data Update Problem

  1. 1,0221 changes? Typo I assume. Please put the comma where it should be. 🙂 If 10,221,that would be pretty amazing.

    Hey, I suspected they were going to come out with a pile of stockpiled cases in time to impact the election, but it didn’t happen. I’m not sure what the motivation would be now. I too have had my confidence in this data source shaken, but suspect the breakdowns may be just more pandemic fatigue.

  2. Well, better latent than never, as Freud would say.

  3. Today’s dataset has 290 new changes (up and down) with a net increase of 186 from April through October.

  4. Cases appear to have more than doubled since the beginning of October, but hospitalizations haven’t quite doubled, and deaths are actually down.
    (Click the buttons for cases, hospitalizations, & deaths)
    Are more cases that aren’t serious enough to require hospitalization being found? Are treatments improving? Does anybody know?

  5. Paul,
    Earlier in the pandemic, testing was directed only to the sickest patients, so only a small fraction of mild or asymptomatic cases were identified. There is more testing by local health depts now where numbers are rising, so it’s likely more mild cases in the communities are being identified.

    There are almost 24,000 nursing home residents in Virginia, plus staff . VDH advises that nursing homes, during an outbreak, to repeat testing for individuals testing negative every 3 to 7 days until no new cases are identified for 14 consecutive days. In November, the VHHA has reported from 350 to 1100 pending tests each day for patients and staff together.

    Routine testing of nursing home staff is based on community COVID-19 activity:
    low activity (less than 5% county positivity rate) = testing at least monthly;
    medium activity (5 – 10% county positivity rate) testing at least weekly; and
    high activity (greater than 10% county positivity rate) testing twice weekly.

    Hospitals test all patients upon admission now.
    So all that adds up to more cases identified.

    In September, The Lancet said, “greater patient survival suggests that treatment of severe disease has improved. How much of this improvement is due to better supportive care and how much to pharmaceuticals is a matter of debate.” But the fact is, more patients are surviving. There aren’t enough controlled studies to know for sure which treatment is best at what stage and severity of the infection, so there are conflicting reports. Drug A may only work early on, while Drug B is best for severe cases. But the medical community is reporting their findings and sharing them openly.

  6. The term “case” seems to be used to describe positive tests whether the subject is ill or not.

    • Yes. For epidemiological purposes, a positive test result = a case, even if a person has no symptoms.

    • Probably no more than one-third to one-half of infections are showing up on reports, if you work the numbers backwards from the CDC infection fatality rate. Dr. Gottlieb said 15% of U.S. population probably already had it heading into this surge, and another 15% would be infected during the winter spike.

      • I’d be surprised if it’s that high a detection rate. Every time the test positivity rate is above 5% in a community, it indicates there probably wasn’t enough testing to identify the extent of the spread if you go by statements from RAND Corp in their weekly reports.

        Last notes I recall on it, the UVA Model used 15% for confirmed case detection based on NYC’s 12% detection rate based on 15,000 random antibody tests in May.

  7. Is there any clarity about the “real time” that the daily statistic updates represent? Are results posted at 5 p.m. yesterday from tests done 3to 5 days earlier or are they just dumped into the data when they are received?

    • Everything entered into the system (VEDSS) by 5 p.m. is supposed to be posted on the dashboard the following morning. Test results could be from less than a half hour to 7 days or more and are added when received if entered into the electronic system. If lab results are faxed, they get input by someone in the health department. Deaths are recorded when the death certificate is processed and reviewed. This can take a few days to a few weeks.

      VDH hospitalization numbers only reflect those cases reported to VDH about a person in the hospital. They do not track how many are reported after a test and later are hospitalized. The VHHA report is based on information sent to them directly by the hospitals.

  8. Update Note: As of 11-28, since 11-15, VDH has made a total of 1510 changes in case numbers, for March 5 to October 31, adding 1604 cases.

Leave a Reply