Background: An interesting data problem.
With the end of the COVID-19 public health emergency back in May 2023, there’s no longer any tracking of the number of new cases in the U.S. The image above (note date) is what you see today, if you look at the old New York Times map of COVID in Virginia.
But the CDC continues to collect COVID-related information reported by hospitals, licensed clinical labs, and from death certificates.
If (when?) the next wave of infections occurs, or the virus mutates in some meaningful way, you can either wait for that to hit the news — if it does — or you can try to infer it from those still-available data sources.
And that’s a reasonably interesting problem in inference from data. We seem to have at least weekly reporting, by state, of:
- Deaths from COVID and percent of all deaths from COVID.
- New hospitalizations for/with COVID
- Hospital ER/OPD visits with COVID as a diagnosis
- COVID PCA test positivity rate.
Separately, the CDC continues to track the proportion of cases by COVID variant, keeping tabs on the ongoing mutation of the virus. Answer: Yep, still evolving. But no clue if the new variants are materially different from the prior ones.
State-level trends data from the CDC can be accessed on this page.
That said, if you read the footnotes (which, of course, I did), you’ll realize that everything but deaths and hospital admissions is based on some sample-of-convenience maintained by the CDC.
These numbers may be adequate for monitoring national trends, but may be unreliable for state-level data. Separately, any impact on COVID deaths will appear weeks after a wave has started. And, depending on the data reporting, hospitalizations and other data may also reflect a considerable time lag.
Separately, again reading the footnotes, the time lag inherent in the hospital data has increased considerably. All hospitals continue to report, but the reporting has changed from daily submissions to weekly submissions with some time lag. So the hospital data remain relatively accurate, but now reflect a much longer time lag than they did during the COVID-19 public health emergency.
Virginia test positivity week ending June 9.
The biggest caveat for this next piece is that the CDC does not track test positivity at the state level. So there’s something about the data that does not pass CDC’s quality test when taken down to that level. At least, not for every state.
That said, the NY Times shows the following, for Virginia, ending the week of June 9. Out of the blue, there’s this big uptick in the fraction of PCA tests for COVID that are positive. (Separately, looking at other states, this does not appear to be an artifact of the data, e.g., most states show no such big uptick.)
Source: New York Times, annotations mine.
To be clear, in the past, the test positivity data tracked factors such as cases and hospitalizations quite well, at least qualitatively. From the same NY Times page, here’s how test positivity lines up against COVID-19 hospitalizations:
I don’t know what this is. Odd are that it’s just something screwy with the data. Checking the NY Times analysis for Maryland and DC, no such uptick appears in those jurisdictions. And nothing appears to be showing up in any of the other measures.
Plausibly, the source for the test positivity data has changed, and has become less reliable (or less closely linked to upticks in cases and hospitalizations).
Of note, though, the 33% test positively rate shown at the end of the graph was only surpassed during the big Omicron wave of the winter of 2022-23.
So I don’t know if that’s a data anomaly, or a harbinger of something. Odds are, it’s a glitch in the data. Either way, it seems worth watching if you live in Virginia.
