Question: Did you see that VA was being accused of “cooking the books” because they were including antibody test results in their denominators? The real question is why isn’t everyone else? We won’t have an idea of the average severity, the R0, or the death rates if we aren’t trying to find and include as many ill in the denominator as we can.
Answer: Dang, I did not see that. And when I went to the Veteran’s Administration COVID-19 website, I didn’t find evidence that they are including antibody test results in their denominators. But then I realized that you might be referring to Virginia… ah ha! There are indeed lots of stories on this issue, including this prominent article published in The Atlantic. How did I miss them!? Anyway, according the the latest — a CNN report published yesterday — Virginia health officials say they are no longer combining COVID-19 PCR diagnostic test results with antibody results. In my opinion, this change is a good thing.
To elaborate, I agree that we do want and need to be conducting antibody tests to understand how far the virus has spread and have a deeper understanding of its severity. I’d love to see a two-stage cluster design household survey to really get reliable population-level estimates. In this space, CDC did just announce a study that will test samples from blood donors in 25 cities for antibodies to understand population-level prevalence. When it comes to state-level reporting of cases, however, I think it’s best to avoid lumping antibody tests with PCR tests for several reasons:
- Monitoring Incidence: New cases cannot be identified by antibody tests. We need widespread testing that identifies new cases (incidence) so we can track how the virus is spreading in real time.
- Making Re-Opening Decisions: The National Guidelines (described in Q&A of 4/19) that Governors are using to varying degrees to determine re-opening include a metric — proportion of tests that are positive — that can be influenced by the inclusion of antibody tests. The Richmond Times reported last week, “With a higher number of tests in the state’s log due to antibody tests, Virginia’s positive rate was skewed down by a percentage point, from 15% to 14%.” The test positivity rate should be based on diagnostic tests; other data should not be added because they may obscure our understanding of the situation, especially when these data are being used to inform re-opening decisions.
- Monitoring Testing Capacity: Lumping the two types of tests may present a rosier picture of a state’s testing capacity and infrastructure than is the truth. As we focus on a test, trace, isolate approach, it is key to understanding how testing capacity is changing over time, including how many diagnostic tests per capita are being conducted. We need to compare apples with apples.
- Minimizing False Positives: Until the population-level prevalence is much higher, antibody tests will yield far more false positives, which will artificially inflate the number of cases (see our Q&A of 4/15 for a deep dive into the sensitivity/specificity challenge).