is new england covid hospitalization data being retroactively adjusted down?

because the data reported by NYT has sure changed a lot in the last week.

about a week ago i published this piece showing a rise in northeast covid hospitalizations among the highly boosted/vaccinated.

a funny thing happened on the way to herd immunity

i was going back to look at some of these series to see what had happened since, and noticed something interesting:

the trends were all up, but the numbers were, in many cases down, including numbers from the past. this series can have incomplete data, but that makes the recent numbers read low, not high. i am not seeing any obvious reason that they would drop over time. that’s not generally how a series like this works. they backfill, they do not back drain.

and it looks to me like the hospital counts from past days are being adjusted down materially in many cases.

example:

here was connecticut from april 19th (reporting through the 18th)

and here it is now:

as can be seen, 4/23 was a new high for the 70+ group. but the number itself is only 8.1 which is actually lower than what it read on the 18th when reported on the 19th. the data from the 18th in the new plot has been adjusted down significantly. it had read 8.2. it now reads 6.7, 18% lower.

so where did these hospitalizations go?

this does not appear confined to CT. rhode island sees a similar effect and their sudden surge has been “disappeared.”

on the 19th, the data read like this:

it now reads like this:

this is a profound change in the shape of the 70+ curve and a large drop in scale.

on the 19th, it read 7.0 for 4/18. that date is now reported as 5.1, 27% lower. the current day (4/23) is also reported as 5.1.

the aggressive uptick is no more.

massachusetts saw a smaller drop from 9.5 to 9.2 for 4/18

of further interest, new york is NOT seeing these issues. their count for 4/18 rose from 7.1 to 7.2 which seems pretty typical for backfill. (and now reads 8.5 as counts keep rising)

there has been talk of changing the definition of “hospitalized for covid” and i know that new hampshire did this. (i have not seen definitive reports on the others doing so, but given the data shifts, i suspect they may have)

“In an announcement Tuesday, the Department of Health and Human Services said it had redefined COVID-19 hospitalization to include only patients being treated with remdesivir or dexamethasone, drugs used for hospitalized patients with moderate to severe illness. Those hospitalized with milder symptoms or primarily for another cause are no longer included — even if they continue to take up a hospital bed because they are too ill to be discharged.”

unsurprisingly, it has a big effect on the shape of their curves and stopped the rapid rise in 70+ in its tracks and seems to be reversing it.

but this is clearly a data splice and not a trend change that can be trusted as an apples to apples assessment.

call me my cynical paws (again) but it’s hard not to wonder if this change was put in place to avoid having to report something like this from the next door neighbors? (which is what NH was starting to look like prior to the methodological alteration)

and one can certainly see why they’d like to avoid showing this outcome in a 95%+ vaxxed and highly boosted group.

it looks to me like a lot of this data is going or is about to go wonky in the northeast as they shift the goalposts to prevent comparisons and hide what seems to be an out of season spike in the very highly vaxxed.

this current hotspot is looking awkward for the public health narrative of vaccines saving the day and the same gang that has used definitional skullduggery to inflate counts, profits, and fear will now use them to hide outcomes making it difficult to track what is occurring relative to past outcomes.

somehow, i doubt this is being done in the pursuit of accuracy as opposed to politics…