I’m starting to think I should make HHS hospitalization data a weekly section of this newsletter.
In case you haven’t read my previous two issues, here’s the situation: in mid-July, hospitals stopped reporting their counts of COVID-19 patients to the CDC, and instead began reporting to the HHS. Since then, HHS’s national hospitalization dataset has been unreliable. HHS’s counts of currently hospitalized COVID-19 patients are far higher than the concurrent counts reported by state public health departments, and HHS’s numbers often rise and fall significantly from day to day without clear explanation.
I, along with other COVID Tracking Project (CTP) volunteers, have been monitoring both hospitalization counts daily—the two counts being, HHS’s numbers and state-reported numbers compiled by CTP. Rebecca Glassman (data entry volunteer and resident Florida expert) and I have drafted a blog post for CTP about the biggest discrepancies we’ve seen, which will be published in the next few days.
Here’s a little preview of the issues we’re calling out:
- In six states, HHS’s counts of currently hospitalized COVID-19 patients are, on average, at least 150% higher than the state’s counts. These states include Maine, Arkansas, New York, Connecticut, New Hampshire, and Delaware.
- Both Florida and Nevada saw unexplained spikes in their HHS counts which were not matched by corresponding spikes in state counts.
- The state of Louisiana actually reports more currently hospitalized COVID-19 patients than HHS does, even though the definitions used by both sources suggest that this discrepancy should be the other way around.
- Many states do not have publicly available or easy-to-find definitions for how currently hospitalized COVID-19 patients are classified.
- HHS’s counts on August 6 were very low across the board, with significant drops in the number of hospitals reporting in every state.
If you are a local reporter in any of the states mentioned here and would like to investigate the discrepancies in your area, please reach out to me! I’m happy to share the data underlying this analysis.