Unfortunately for 8/12 the reported deaths in Israel as 17. This unfortunately agrees with the crude prediction model I created on 8/7. Blue dots to the right of the yellow line indicate new data observed since the prediction was made.
We're going to see if holds through to the end (8/20) - if it does I'll regenerate the prediction 14 days forward and update one post every day - but not send these short updates anymore..
A hit piece came out yesterday making many erroneous claims about one of my previous articles that I have to issue a response to.
I guess we all try to internalise this is some way. Here's my current thinking. 1). https://gabgoh.github.io/COVID/ is a great tool for modelling covid variants with different R0. eg Delta at 5+ , vs 1.5 or so for earlier versions. Essentially everyone will get delta, and it is nearly impossible to put in restrictions effective enough to get R0 < 0 .Maybe welding doors shut. 2.) drrollergator has shown that vaccines are effective in reducing symptomatic infection (say 90%) but not much else : eg death rate same , better or worse depending on age group. 3). This prediction is reconfirming original thesis .. in that only the infection rate really changed.. but once you get to 'a case' , outcomes havent changed that much 4. If that's the case then combining all then we can assume all of Israel will be exposed to this high R0 disease. With their very high vax rate then lets assume that the infection rate is reduced to 10% of total 7m pop (roughly). Death rate of covid is roughly 1% of cases.. so 7M x 10% * 1% = is about 7000. The epidemic tools says it will all be over in 100 days so an average of 70 / day. Delta seems like a first major variant.. ie infectivity or R0. May God help us if the CFR jumps on the next variant.
I just check worldometer.. and Israel has had about 6000 covid deaths so far.. so even with 89-90% vax rates, the arrival of delta means the same magnitude.. if my logic is correct. It is a monster.
There are many cracks in this.. Dr John Campbell (from youtube) ponts out that Israel (and US) give 2nd Pfizer shot at 3 week vs 8+ in UK. This seems to be linked to longer term effectiveness of the drug. It will be interesting to see if death / case rate holds in UK also.... cases do seem to be not spiking as they are in Israel...
OK, then if I may get by with just a quick answer for now - beta-binomial posterior predictive distribution with Haldane's uninformative improper prior | alpha = sum deaths, beta = sum (cases-deaths) from start of pandemic to jan 24th 2021. Use the assumption deaths are 14 days delayed from cases, use the beta-binomial with the alpha and beta parameters, and the n you're using is cases for x-14 (days) back to generate the distribution of possible k death day x
it's intentionally naïve and unfortunately accurate so far.
*deleted previous version just to add teensy extra clarity
The case fatality rate in a population 90% vaccinated is at the moment indistinguishable from the observations of case fatality rates before vaccination rollout. You could use the unvaccinated population's history to predict the vaccinated population's deaths.
There are many variables at play that are hard to isolate. It could be that the benefits offered by the vaccine were overcome by changes in virus behavior due to a variant.
But in a 90% vaccinated population cases are currently growing exponentially and we are observing the same deaths we would have expected from the unvaccinated population before the vaccination rollout.
Why that observation is roughly the same is not answerable with this prediction model.
I would advise against using these results to make any particular inferences about why it is accurate. It is purposely ignorant of most underlying details about variants, age groups, comorbidities, etc.
It is certainly possible that Delta, which appears to be the dominant strain, is just "deadlier" and therefore offsets, coincidentally, exactly the reduction in death you would have otherwise had.
Well here is the line of thought, based off what is known evolutionarily (per Bret Weinstein and others) it's expected that as the virus mutates it will mutate to a more transmissible but less deadly (per case). This is is how all viruses work because essentially the virus wants to maximize the likelihood they will survive and the best way to do so is by infecting as many hosts as possible. So the trade off being they can mutate in the direction of increasing transmissibility while inversely becoming less deadly. Now this only applies to if the virus would have came purely from nature, so if we find out that it's better at transmitting AND more deadly then that makes the lab leak theory (of gain of function) as certain as you can get without absolute direct evidence (and then were in uncharted waters). This is all to say if we see it becoming more deadly the question will be then either ADE or it's all but certain to be a GOF virus that doesn't follow the natural evolutionary pathways. So even if you exclude the vaccine and it's effects you absolutely should see it become less deadly regardless of any drugs/vax/lockdown etc simply because of the natural evolutionary forces. If we see the deadliness remain constant or increase, then ADE, GOF virus are the options on the table that would need to be parsed through.
One way we might be able to discriminate ADE vs GOF signals to look out for is how the deadliness changes through time in the unvaxed and vaxd camps, the logic being if the deadliness is remaining the same in the unvaxd then that could suggest GOF (and if it goes down then it's a likely naturally evolved virus or at least the GOF experiments could have then the thing that started the fire but didn't have a major physical effect), BUT if the vaxd population is now spreading the virus at a higher rate than unvaxd/dying at higher rates than unvaxd then this will give all the signals that ADE is occurring. If this is true then we have a whole other problem on our hands A) Vaxd are spreading the virus more than uvaxd B) The virus is now more deadly to vaxd than it is to unvaxd then this a major red flag at which point vaccination should be halted.
We are have two parts of population which are producing different variants, and those parts are mixed. Total amount of unvaccinated roughly equal total amount of vaccinated. Variants made by vaccinated does not have evolution pressure to be less fatal for unvaccinated and vice versa. What could emerge if those populations are interact. No one knows. We have two strains recombination with different evolutionary pressure, so most likely it just keep ability to spread, and because of fatality rate of this virus is too low for both groups it will not make any difference in virus evolution. In fact it's all only cause it will be able to spread and infect and in some cases cause decease in both groups. For unvaccinated one it already had year more of evolution so it's probably less virulent just because of it. Most pressure from evolution now it just escape vaccine induced immunity.
In fact it is impossible to make any reliable conclusion from any data available. Only one conclusion could be made --- vaccination is not preventing death since March (in US at least). Risk groups was already vaccinated till that point. Because of death rate falls exponentially from age it had no sense to continue. But vaccines itself are deadly and risk benefit for <50 population is on risk side. Even not counting breakthrough or increase of probability to get infected in between doses and etc.
Doc…. any chance you might post your code to Github, or in-line? No worries if not, but the model looks good and would be great to see how you’re calculating the trend.
The model is purposely very naïve. Its ability to predict the values with accuracy is not because it was designed to do so - it only looked at the history of deaths 14 days after cases, using data from before the vaccination rollout. With that, it takes cases now and uses standard probability estimates to generate predictions about the 14-day-out deaths.
The fact that it's accurate simply says that the case fatality rate looks pretty much the same now as it did before.
Because we have so many new variables, like variants, behaviors, etc, it doesn't necessarily mean nothing has changed. Most things have changed. But because the results wind up being roughly identical, that reality needs to be addressed instead of swept under the rug, or obfuscated by finding the few unvaccinated left and saying "they got infected at a higher rate"
But in a future article I could perhaps go into how to construct this naïve model.
The model is specifically not sophisticated - it does not look at age groups; it does not look at variants; it does not look at comorbidities; etc.
That's not to say the math generating the prediction is sketchy. The math fine. It's that the assumptions built in the math are purposely ignorant of everything but what cases and deaths looked like before vaccination. And then the math estimates what they would look like today if the rates were unchanged.
The fact that the predictions are accurate is just another way of saying it looks darn near identical to before vaccination.
I would personally use the term "looks" in this case. Because of how little is being considered in the prediction model.
This model cannot be used to say that Delta is more or less lethal, for example - it is possible that the vaccines protect greatly against death if you become infected but it just so happens Delta is exactly that much deadlier, and they cancel out. Who knows.
So the safest non-mathematical way to phrase the model is "if this were 2020, and cases were this high, we would expect this many deaths." And what we are unfortunately seeing is that deaths this year look the same per case as they did last year (for now), even though Israel is 90% vaccinated.
Source: Our World in Data via 91-divoc.com
Yes, thank you - this is available in previous posts but I neglected to include it here. Apologies
Thanks for doing this. Were you planning on keeping this up much longer?
We're going to see if holds through to the end (8/20) - if it does I'll regenerate the prediction 14 days forward and update one post every day - but not send these short updates anymore..
A hit piece came out yesterday making many erroneous claims about one of my previous articles that I have to issue a response to.
We can't have any of this unauthorised science going on.
Good luck, I look forward to reading it.
I guess we all try to internalise this is some way. Here's my current thinking. 1). https://gabgoh.github.io/COVID/ is a great tool for modelling covid variants with different R0. eg Delta at 5+ , vs 1.5 or so for earlier versions. Essentially everyone will get delta, and it is nearly impossible to put in restrictions effective enough to get R0 < 0 .Maybe welding doors shut. 2.) drrollergator has shown that vaccines are effective in reducing symptomatic infection (say 90%) but not much else : eg death rate same , better or worse depending on age group. 3). This prediction is reconfirming original thesis .. in that only the infection rate really changed.. but once you get to 'a case' , outcomes havent changed that much 4. If that's the case then combining all then we can assume all of Israel will be exposed to this high R0 disease. With their very high vax rate then lets assume that the infection rate is reduced to 10% of total 7m pop (roughly). Death rate of covid is roughly 1% of cases.. so 7M x 10% * 1% = is about 7000. The epidemic tools says it will all be over in 100 days so an average of 70 / day. Delta seems like a first major variant.. ie infectivity or R0. May God help us if the CFR jumps on the next variant.
And what's the chance of that?.. why not look up Dr Robert Malone (@RWMaloneMD) on twitter, and / or ADE from mRNA vaccines on SARs-COV experiments.
I just check worldometer.. and Israel has had about 6000 covid deaths so far.. so even with 89-90% vax rates, the arrival of delta means the same magnitude.. if my logic is correct. It is a monster.
There are many cracks in this.. Dr John Campbell (from youtube) ponts out that Israel (and US) give 2nd Pfizer shot at 3 week vs 8+ in UK. This seems to be linked to longer term effectiveness of the drug. It will be interesting to see if death / case rate holds in UK also.... cases do seem to be not spiking as they are in Israel...
It would be very nice to have the details of model.
Please consider reviewing the initial article - https://drrollergator.substack.com/p/are-cases-decoupled-from-deaths
I have already read the article, but there is no math detail.
My apologies. I can try to correct for this in the future. Could you let me know what level you are asking.
I wouldn't want to "dumb things down" and insult you or provide you with things that looked like gibberish.
Advanced. I know well the stuff.
OK, then if I may get by with just a quick answer for now - beta-binomial posterior predictive distribution with Haldane's uninformative improper prior | alpha = sum deaths, beta = sum (cases-deaths) from start of pandemic to jan 24th 2021. Use the assumption deaths are 14 days delayed from cases, use the beta-binomial with the alpha and beta parameters, and the n you're using is cases for x-14 (days) back to generate the distribution of possible k death day x
it's intentionally naïve and unfortunately accurate so far.
*deleted previous version just to add teensy extra clarity
Thanks
So this is showing that vaccination is not improving odds of dying?
The case fatality rate in a population 90% vaccinated is at the moment indistinguishable from the observations of case fatality rates before vaccination rollout. You could use the unvaccinated population's history to predict the vaccinated population's deaths.
There are many variables at play that are hard to isolate. It could be that the benefits offered by the vaccine were overcome by changes in virus behavior due to a variant.
But in a 90% vaccinated population cases are currently growing exponentially and we are observing the same deaths we would have expected from the unvaccinated population before the vaccination rollout.
Why that observation is roughly the same is not answerable with this prediction model.
This might be a reasonable way to tell whether or not ADE is occurring ahead of time if it begins to outpace the model (just a guess)
Problem with this idea of mine above is that is assuming the variants aren't becoming more deadly than the previous ones I believe
I would advise against using these results to make any particular inferences about why it is accurate. It is purposely ignorant of most underlying details about variants, age groups, comorbidities, etc.
It is certainly possible that Delta, which appears to be the dominant strain, is just "deadlier" and therefore offsets, coincidentally, exactly the reduction in death you would have otherwise had.
Well here is the line of thought, based off what is known evolutionarily (per Bret Weinstein and others) it's expected that as the virus mutates it will mutate to a more transmissible but less deadly (per case). This is is how all viruses work because essentially the virus wants to maximize the likelihood they will survive and the best way to do so is by infecting as many hosts as possible. So the trade off being they can mutate in the direction of increasing transmissibility while inversely becoming less deadly. Now this only applies to if the virus would have came purely from nature, so if we find out that it's better at transmitting AND more deadly then that makes the lab leak theory (of gain of function) as certain as you can get without absolute direct evidence (and then were in uncharted waters). This is all to say if we see it becoming more deadly the question will be then either ADE or it's all but certain to be a GOF virus that doesn't follow the natural evolutionary pathways. So even if you exclude the vaccine and it's effects you absolutely should see it become less deadly regardless of any drugs/vax/lockdown etc simply because of the natural evolutionary forces. If we see the deadliness remain constant or increase, then ADE, GOF virus are the options on the table that would need to be parsed through.
One way we might be able to discriminate ADE vs GOF signals to look out for is how the deadliness changes through time in the unvaxed and vaxd camps, the logic being if the deadliness is remaining the same in the unvaxd then that could suggest GOF (and if it goes down then it's a likely naturally evolved virus or at least the GOF experiments could have then the thing that started the fire but didn't have a major physical effect), BUT if the vaxd population is now spreading the virus at a higher rate than unvaxd/dying at higher rates than unvaxd then this will give all the signals that ADE is occurring. If this is true then we have a whole other problem on our hands A) Vaxd are spreading the virus more than uvaxd B) The virus is now more deadly to vaxd than it is to unvaxd then this a major red flag at which point vaccination should be halted.
They claim ADE now. It's became finally. https://www.journalofinfection.com/article/S0163-4453(21)00392-3/fulltext#2021
We are have two parts of population which are producing different variants, and those parts are mixed. Total amount of unvaccinated roughly equal total amount of vaccinated. Variants made by vaccinated does not have evolution pressure to be less fatal for unvaccinated and vice versa. What could emerge if those populations are interact. No one knows. We have two strains recombination with different evolutionary pressure, so most likely it just keep ability to spread, and because of fatality rate of this virus is too low for both groups it will not make any difference in virus evolution. In fact it's all only cause it will be able to spread and infect and in some cases cause decease in both groups. For unvaccinated one it already had year more of evolution so it's probably less virulent just because of it. Most pressure from evolution now it just escape vaccine induced immunity.
In fact it is impossible to make any reliable conclusion from any data available. Only one conclusion could be made --- vaccination is not preventing death since March (in US at least). Risk groups was already vaccinated till that point. Because of death rate falls exponentially from age it had no sense to continue. But vaccines itself are deadly and risk benefit for <50 population is on risk side. Even not counting breakthrough or increase of probability to get infected in between doses and etc.
Doc…. any chance you might post your code to Github, or in-line? No worries if not, but the model looks good and would be great to see how you’re calculating the trend.
The model is purposely very naïve. Its ability to predict the values with accuracy is not because it was designed to do so - it only looked at the history of deaths 14 days after cases, using data from before the vaccination rollout. With that, it takes cases now and uses standard probability estimates to generate predictions about the 14-day-out deaths.
The fact that it's accurate simply says that the case fatality rate looks pretty much the same now as it did before.
Because we have so many new variables, like variants, behaviors, etc, it doesn't necessarily mean nothing has changed. Most things have changed. But because the results wind up being roughly identical, that reality needs to be addressed instead of swept under the rug, or obfuscated by finding the few unvaccinated left and saying "they got infected at a higher rate"
But in a future article I could perhaps go into how to construct this naïve model.
What do you mean by naive in this context? Simple? Comparing before the vax with after?
Simple is an ok synonym, yes.
The model is specifically not sophisticated - it does not look at age groups; it does not look at variants; it does not look at comorbidities; etc.
That's not to say the math generating the prediction is sketchy. The math fine. It's that the assumptions built in the math are purposely ignorant of everything but what cases and deaths looked like before vaccination. And then the math estimates what they would look like today if the rates were unchanged.
The fact that the predictions are accurate is just another way of saying it looks darn near identical to before vaccination.
I would personally use the term "looks" in this case. Because of how little is being considered in the prediction model.
This model cannot be used to say that Delta is more or less lethal, for example - it is possible that the vaccines protect greatly against death if you become infected but it just so happens Delta is exactly that much deadlier, and they cancel out. Who knows.
So the safest non-mathematical way to phrase the model is "if this were 2020, and cases were this high, we would expect this many deaths." And what we are unfortunately seeing is that deaths this year look the same per case as they did last year (for now), even though Israel is 90% vaccinated.