higher vaccine rates associated with greater rise in hospitalization

if vaccines work to stop severe outcomes, then why aren't they working?

for this analysis i used the NYT covid hospitalization data. it can be accessed HERE. vaxx rate data from OWID HERE.

my goal was to look at hospitalization rates in relation to vaccine adoption. to do this (and because i could not find a way to automate it) i took the 5 least vaccinated states and compared them to the 5 most vaccinated states. it seems a reasonable variance as the most vaxxed average 80% vaxxed and the least average 56%. because of vagaries in the data, namely an annoying tendency to use “<1” instead of an actual number, i had to throw out VT and RI. i replaced them with HI and NY, the two next most vaxxed. this almost certainly weakened the case against the high vaxx states as VT and RI look to have much greater increases in 70+ (400-600%) but as will be seen, this did not eliminate the signal.

different states report hospitalization so differently and express such different seasonality that it can be difficult to compare them to one another. but they can be compared to themselves, so i took the hospitalization rates for the most vaxxed group (70+) and for “all” and compared them to a year ago to establish % change from july 10th 2021 to july 10th 2022. by using this methodology, they act as their own controls as reporting, age, demographics, and who knows what other imponderables remain constant.

(i would have liked to stratify this by all age groups, but the data was too incomplete and tainted by the non-numerical “<1” result, especially in 2001 data.)

this data was quite striking from the start, esp as the vast majority of states are showing higher hospitalization rates in nearly every category.

this alone (in conjunction with milder variants) gives some real pause on overall outcomes in over 70’s.

(high vaxx blue, low vaxx green)

and even more so in “all” (IN would be green if there were a visible bar)

in both cases, 4 of 5 best performers are low vaxx and 4 of 5 worst are high vaxx.

the extent of the variance becomes clear when we take an average and a median. as can be seen the two measures align well.

in both series, the % rise in low vaxx states was lower than in high vaxx and this was more prevalent in “all” than in “70+” which is a VERY provocative outcome as the old tended to be high vaxx everywhere and the bigger divergences in jab rate came in the young.

it also shows that the old (who are most highly vaxxed and boosted) are seeing greater rise in incidence of hospitalization than the young who are less so. combine that with the high vaxx/low vaxx divergence and the signal here really starts to pop.

we can see something interesting in the scatterplots as well:

the R2 on 70+ is lowish (but noticable), but is mostly suppressed by outliers (ME, TN) and the rest of the series has strong divergence. if you remove those 2, the remaining 8 now fail to overlap and show extremely high R2 for a series this noisy.

one can argue about whether this is fair or amounts to egregious p-hacking (and i’ll probably side with p hacking), but if nothing else, it seems to provide some useful insight, especially as we start to look at the “all” series where we would expect much greater divergence in vaxx rate.

here, we’re seeing quite strong R2 for such a noisy series even before we remove outliers (ME, TN). (shown in red and counted in R2)

the lack of overlap is quite striking. only one low vaxx state exceeds a 67% rise and only one high vaxx state is less than a 136% rise. that’s a very provocative grouping.

and if we remove them just for visualization sake, this becomes more clear and we start getting into extreme alignment. (though again, this starts to look like p hacking and i would urge against to drawing convulsions from it, especially when the full data is still so interesting. this is just to illustrate a point.)

the bottom line is we’re seeing exactly what we would expect if variants were now homing in preferentially on the vaccinated and either becoming more severe for them or (more likely in my view) still providing some protection level of zero or greater but more than swamping it by making contracting covid more likely.

the fact that variance by vaxx level is so much greater in overall than in 70+ bolsters this.

we’re seeing strong association with more vaxx, greater rise in rates of hospitalization for covid, and more oddly, it’s well out of season in most of the high vaxx states.

this sudden unseasonable surges pattern is aberrant relative to past covid outcomes which adds further cause for suspicion of new external drivers. (i suspect one could advance an argument about immune exhaustion from repeated boosting here as well)

new york provides a typical example:

july this year vs july last year does not look good and they have already had one big surge that far exceeded last summer. another seems to be beginning both early and from a worryingly high base. this starts to look more an more like a virus “jumping the banks” due to herd immune fixation and/or a people less and less able to resist it. one would always expect to see that signal most in the oldest whose generalized immune systems are less capable and thus cannot pick up the slack from inapt antibody response.

i realize i’m becoming a bit of a one note flute here, but this signal just keeps popping up everywhere in completely independent data series.

i’m not sure any one can rise to the standard of “proof” but as a mosaic, especially given what we know about the underlying biology, this starts to get awfully compelling to ignore.

the simple fact that this vaccine program is not working on society scale is getting far to obvious to ignore.

if we could get any health agencies interested in actually stepping up and assessing this, then we could all go grab a beer and put our feet up, but as it appears they are disinclined to do so, the pulling of threads is left to us.

and so we shall.