News came out on Wednesday, May 20, that Georgia DPH has been including antibody tests in their total testing numbers, but not including antibody positives in the confirmed case numbers. They are working to fix this reporting issue on the web site, but I wanted to explain a little more about what the problem is and why it matters.
UPDATE: As of June 2, Georgia is now sharing the number of antibody tests and positive antibody test results, as well the number of PCR tests and number of positive PCR test results. However, we learned that these testing numbers are only for tests reported via their electronic reporting system. About 10% of cases are reported outside of the electronic reporting system, and we don’t have testing numbers for these cases. As a result, our test numbers are actually under reported. Learn more about the GA DPH COVID web site.
- Viral and antibody tests have been intermixed in the Georgia test totals, but only positive viral tests are listed as cases.
- Georgia DPH started providing serology (antibody) testing totals on their web site as of May 27. At that time, there were 77,835 antibody tests, or 15% of the total.
- Many states have been criticized for mixing these test and/or case numbers, and even the CDC has been combining these numbers for some reason.
Why is this a problem?
There are two different types of tests for COVID-19. They are performed differently, and they tell you different things. They also have different rates of false positive and negatives.
- A viral (also called PCR or molecular) test tells you if you have a current infection. This is the nose and/or throat swab type of test that any Georgia resident can get free of charge.
- An antibody (serology) test tells you if you had a prior infection. These are blood tests that come at a cost to an individual or their employer, or may be covered by health insurance.
As a result, it’s important to keep these two types of tests separate, so that we know how many tests for active infections are actually being performed, and how many of those tests are positive. The percent positive is an important tool to ensure testing is adequate and interpret case numbers, so we need the right data to calculate this metric.
Antibody testing is interesting in order to determine how many Georgians may have recovered from a COVID infection, but it doesn’t identify who is currently sick, so it has a very different function in tracking the spread of coronavirus.
How big the problem?
Antibody tests make up about 15% of total tests as of May 27, so while the problem does affect a large number of tests, it is a small percentage of our total tests, so we know we have still seen a massive increase in viral testing, and our percent positive is still fairly low. It’s just not as low as we thought, because our viral test numbers are lower that the numbers we’ve been using, but the case numbers are unaffected.
What about the historic testing data?
When Georgia DPH starting publishing antibody testing numbers, they did not provide historical numbers. This meant our testing data was very skewed on sites that rely on published testing data. To deal with this, I have estimated the impact of increasing numbers of antibody tests over the past 5 weeks and removed these estimates from total testing numbers over time, so that I can graph a more accurate % positive over time. Learn more about my estimates here.
How was this discovered?
On Sunday, May 17, the total number of cases was revised down slightly, and many of us who follow the data closely had questions for GA DPH as to why, since this is a cumulative number and shouldn’t go down. I noted this in my Key Points when I discussed Sunday’s results.
On Monday, May 18, Georgia DPH posted on Twitter:
This led to a lot of questions about how Georgia was tracking antibody tests and their results. Journalists reached out to DPH, and Nick Wooten at the Columbus Ledger-Inquirer asked the right questions to get the answer we needed.
What other states do this?
Virginia was the first state to be exposed widely for this practice of mixing antibody tests with viral tests, but other states have done the same, and have already fixed their reporting to separate out the two types of tests. Other states including Texas and Pennsylvania have also come under fire for reporting this way, and then The Atlantic discovered that even the CDC themselves combine these two numbers in their reporting of tests and cases.
So while it’s frustrating that Georgia has been doing this, it’s hardly unprecedented. I suspect that due to the publicity over this issue now, more states will admit to this practice and start separating out numbers going forward. The CDC has also said they are working to separate out the numbers on their web site.
What about all the other problems that have been reported with the DPH site?
You’ve no doubt heard in the news or seen the AJC articles about “all the problems” with DPH data. Other than this testing issue, the problems have been display/graphing issues, not incorrect data. In my opinion, the DPH is working overtime and doing their best with a whole new level of data sharing that they’ve never had to do before, and mistakes happen. I do think they are trying to get a better handle on the reporting to regain the public’s trust.
I wrote two other articles to address previous issues that you may have heard about with the DPH site if you’d like to learn more:
- Did Georgia graph cases with the dates out of order?
This addresses the issue with a single graph that briefly had a non-chronological X-axis. This was widely reported on local and national news. I also address a couple of other minor errors that were identified and corrected with the GA DPH site.
- What’s the problem with the DPH graphs?
This addresses the “backdating” or “lag” that has been discussed with the way DPH graphs cases and deaths over time. It’s not really a problem, per se, in that there is nothing wrong with the way they are graphing the data. However, it has been the subject of intense criticism, particularly by the AJC, which mischaracterized it as a change in the way Georgia was counting or reporting cases.