CDC Seroprevalence Estimates Round 2: Convergence with My Algo

Yesterday’s post reported the CDC’s seroprevalence estimates for 4 states, comparing their results with my prevalence estimation algorithm. I was in the same large ballpark, but for three of the four states my estimates were quite a bit lower than the CDC’s.

That was for the first round of data collection, with most of the data collected during the April peak. The CDC also reported a second round, using data collected during the last week of May, by which time there had been more total infections in the population. Here are the CDC’s round 2 estimates compared with mine:

  • Louisiana:  CDC estimate = [results pending]; my estimate = 10.0%
  • Missouri:  CDC estimate = 2.8% (range 1.7-4.1); my estimate = 2.2%
  • Connecticut:  CDC estimate = 5.2% (range 3.8-6.6); my estimate = 18.5%
  • Utah:  CDC estimate = 1.1% (range 0.6-2.1); my estimate = 0.8%

Round 2 shows considerably greater convergence between the two sets of prevalence estimates. As I read the numbers, it seems that the CDC estimates have moved in my direction, indicating that a smaller percentage of people are being infected than they’d originally estimated.

Missouri is illustrative. In late May the CDC estimated that 2.7% of the state population had been infected; in late June the estimate remained nearly the same at 2.8%. In contrast, my Missouri estimates went up from 0.9% in late April to 2.2% by late June. That’s because my estimate uses death rates as a lagging indicator, and by late June the total covid death count had more than doubled from late May. For what it’s worth, total Missouri diagnostic case counts too nearly doubled during that one-month interval.

For Utah the CDC’s prevalence estimate actually went down, from 2.2% to 1.1%. Obviously the total number of infections didn’t decrease; using my algo the total infection rate for Utah nearly tripled. Now though the CDC’s number has dropped to a level that’s very close to mine. Clearly it’s a matter of the CDC changing either its algorithm or its measurement methods between the two rounds.

The wide discrepancy on Connecticut persists from round 1 to round 2. As I speculated yesterday, those early Northeastern states that got hit hard and fast may have experienced unusually high death rates as they grappled with treating a new disease that had overwhelmed the healthcare system.


3 thoughts on “CDC Seroprevalence Estimates Round 2: Convergence with My Algo

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s