can covid vaccines cause more covid deaths while looking like they prevent them?

yes, they can. running through some math that people are not going to like.

this is a thought experiment. we’d struggle a bit to fully verify these numbers and viral attenuation is clearly playing a big role so it’s difficult to be overly precise, but i’ll try to keep the key figures within the realm of sanity based on observed outcomes. the point is not to be precise so much as to lay out the manner in which some of the math here can get a little perverse and wind up making an intervention that caused a massive rise in covid deaths look like vaccine efficacy.

let’s lay out a couple of assumptions for our experiment:

1. each person has a 20% chance of getting covid (made up)

2. CFR is 10% (massively too high, but makes the math easier to see and does not affect conclusions)

3. covid vaccines reduce the risk of death from covid (CFR) by 50% if you contract the disease (made up for convenience, but in line with what used to seem to be the case)

4. covid vaccines increase the risk ratio of contracting covid by 4X. (in line with UK data)

5. we have a homogeneous population all with equal covid risk and CFR

imagine case 1: no vaccination

we have 100 people.

20 get covid.

CFR = 10%

2 die.

case 2: 50% vaccination

we have 50 unvaxxed

10 get covid

CFR = 10%

1 dies

we have 50 vaxxed

40 get covid

CFR = 5%

2 die.

we can now compare the two cases:

as can be readily seen, the vaxxed case is worse in total cases and in total deaths. the entire pandemic has been elevated into a much higher valance and the lower overall CFR is swamped by higher case counts.

so you get more deaths overall while making it look like the vaccines are reducing deaths because relative VE is 50%. but absolute VE on societal scale is deeply negative. and the more you vaxx, the worse it will get.

at 100% vaxxed you get 80 cases and 4 deaths, 4X the case counts and double the deadliness of doing nothing.

assessing VE on a post infection basis without accounting for possible effects on likelihood of contracting disease is a common practice, but that’s because we’ve never had a vaccine that makes you more likely to get sick before. and now we do. so incidence adjustment becomes critical.

this gets compounded in complex ways because the higher prevalence in the vaxxed likely leads to higher prevalence in the unvaxxed as they have more opportunities for exposure. (but i am leaving this out as it would require more complicated math where variables are functions of one another and this is not needed to make the point.)

this is why looking ONLY at relative severity effects is a very dangerous and potentially misleading idea.

and this gets MUCH harder to see if we add another parameter.

let’s call case 1 case 1 again.

but now let’s say case 2 is a year later. it’s exactly the same as before with one new twist:

covid has attenuated. its infectiousness is still the same, but it is only 30% as deadly. CFR drops to 3% unvaxxed and 1.5% vaxxed. (so VE is still 50%) (CFR is still 10% in case 1)

so case 1 retains 20 cases, 2 deaths

but now case 2 gets 50 cases as before but 0.5 deaths in the unvaxxed and 0.6 deaths in the vaxxed.

covid cases are well up, but this is easy to blame on a virus despite it having nothing to do with the virus and everything to do with OAS.

but, on deaths and CFR, now it looks like everyone won big.

deaths are down 45%. CFR drops 80%. VE looks strong. vaccines look like they saved the day, like they protected you.

but they didn’t. this is entirely an artifact of comparing a more virulent strain to a less virulent one.

if instead of vaccinating, the society in question had done nothing we would have a “case 1” of no vaxxed, 3% CFR, but only 20 cases.

so we get this:

what makes this one hard is that the data in yellow is a counter-factual. it’s what might have been but never was. we cannot see it, prove it, or provide data.

what people see is the comparison above that looks like a 45% reduction in overall deaths. but that reduction could have been far greater. that 1.1 is 83% higher than what could have been. 55% of those deaths were directly caused by the vaxx driving prevalence higher (but looking like it worked) but they remain unseen, an unrealized potential.

we reside in a higher valance that we would have but do not realize it. an overly simplistic look at the data seems to suggest the opposite.

it takes some sleuthing to see this.

and THAT is a serious problem. the potential to misread overall absolute harm as relative safety and presume that that means that the more people who jab, the lower death goes is significant even with the best of intentions. it’s not easy to separate out all these cross currents at societal scale. with the sorts of silly buggers being played in the US around undercounting the unvaxxed, misattributing cases and deaths, and rigging definitions to hide outcomes, it’s essentially impossible.

to be clear, this has all been a thought experiment. it cannot prove anything about what actually happened. but it CAN prove the potential for some things to have happened and lay out why realizing that they did could be hard to see.

they key parameters here that really make this model move are:

1. assumptions about VE

2. assumptions about greater contagion

3. assumptions about viral attenuation

i think the first one is the most made up. it’s a real jump ball right now (especially post omicron) and while i suspect that figures like 40-50% may have pertained a ways back, the current figures are almost certainly lower and whether they are even still positive before adding in the incidence adjustment is anyone’s guess. the UK data that seemed like the best insight here has been discontinued, as has the israeli. the US data is a mess.

the second one is likely not terribly far off and the third one was likely conservative.

my goal here is not to prove this case but to raise its possibility.

there are some lines here that look fertile for exploration and i want to run down a few in the next week or so and hope others may be inspired to do likewise.

the mathematics here may be trying to tell us something.

and those who fail to learn from math are doomed to repeat this public health morass.