Source: Calculated from Virginia Department of Health data, available at this URL.
Virginia’s COVID-19 case count includes both confirmed and probable cases. And that’s a good thing, because the technology of testing has changed over the course of the pandemic.
Confirmed cases are those who test positive with a PCR test for COVID. These tests have to be processed at laboratories that can perform the DNA sequencing needed to identify COVID-19 DNA. These typically require a nasal swab, but there are also DNA-based “spit tests”.
Probable cases are of two sorts, per the Virginia Department of Health usage. One, they can either test positive using an antigen test, which reacts to certain proteins on the surface of COVID-19. (These are, in general, faster but less sensitive than PCR (DNA) tests. They still require some sort of nasal or saliva specimen.) Two, they can have some symptom of COVID and have a known exposure to a COVID-19 infected individual.
(Note: Antigen test are not the same as blood-based antibody tests. Antigen tests flag the presence of an active COVID infection by detecting the virus on nasal membranes or in saliva. Blood antibody tests flag individuals who have recovered from some past COVID infection. They flag the long-term byproducts of the body’s immune system response to COVID.)
It’s tough to say for sure, but as I read the numbers, the vast majority of our “probable” cases are from positive antigen tests. That’s based on the count of positive antigen tests on the Virginia Department of Health website. And inferred less directly from the relatively modest share of new cases for which Virginia obtains usable contact-tracing information. (As I do the math, I think that Virginia is only able to get information about any close contacts in about 12% of all new COVID cases. That’s a combination of about 40% of new cases interviewed, and about 30% of those providing their contact information.)
And so, as the technology for testing has changed, Virginia’s numbers have kept pace. To the extent that the rapid-reading antigen tests are displacing PCR (DNA) tests, the probable cases need to be included in the totals in order to keep track of all positive test results.
Other states made one-point-in-time decisions to include those antigen-test “probable” cases in their case counts. That can lead to large one-point-in-time reporting anomalies, as documented in Post #912. Based on the experience of North Dakota and other states, at least some state consortia have made that the standard.
(There is a nuance here, based on exactly how CDC defines probable cases, versus how the states are defining it. Nobody but the most nerdly will care, but CDC requires something in addition to an antigen test in order to classify a case as probable. Basically, you have to have symptoms or have been exposed, before CDC will accept a “spit test” as evidence of probably COVID-19. By contrast, based on the plain language, Virginia (and ND) count anyone with a positive antigen test as a probable COVID-19 case.)
Finally, let me get to the point: Adjacent jurisdictions may or may not include those probable cases in their counts.
Maryland does not include those cases. The Maryland COVID-19 dashboard counts confirmed cases only and explicitly mentions PCR tests only.
The District of Columbia COVID-19 dashboard is unclear about whether or not they include those cases. But because they require a confirmatory PCR test for those who have a positive antigen test, my guess is that DC probably does not include those in their counts.
Today, the Washington Post shows the Virginia new case count at 33, and the Maryland new case count at 18 (seven-day moving average, new COVID-19 cases / 100,000 / day). That looks like a fairly large discrepancy, for two states that were more-or-less in lockstep for most of the pandemic.
But if I net those “probable” cases out of the Virginia data, the apples-to-apples comparison is 23 in Virginia versus 18 for Maryland. Which is much more in line with their historical relationship.
My first conclusion at this time is that antigen testing has become so commonplace that states really need to include those in their counts. It’s not crystal clear, but to the extent that these are displacing PCR (DNA) tests, then I think the case for including them is compelling.
My second conclusion is that, beyond a handful of states, I have no clue which states do and don’t include them. And that the difference in the count — 30% in the case of Virginia — is material. In this case, it made it look as if Virginia’s recent experience was much worse than that of Maryland. When, in fact, accounting for the missing “probable” cases, the new case counts are roughly comparable.
