“our list of suspects includes: COVID, long-COVID, lockdown and vaccines. We’ve seen that COVID plays a minor role and that vaccines play a major role in explaining that surge.”
If long-Covid was playing a bigger role than Covid (~5%) in predicting excess death, then would your analysis have picked it up as a better fitting longer lag from the CDR variable?
Putting that question a different way: does your analysis put an upper bound on what contribution a long-Covid effect could be delivering? (since long-Covid could be considered as lagged-Covid, and as there is no lag that exceeds 2 weeks in predictive value, and that contributes just ~5%).
Or does the analysis not exclude the possibility that long Covid has a long and smeared out effect that is not detectable by the optimal lag analysis, but still could contribute more than the 5% of deaths predicted by 2-week lagged CDR?
Finally, if the earlier interpretation is valid, then is it then fair for the jury to conclude that the remaining ~60% of observed excess deaths, which still need an explanation, can be attributed to the remaining suspect in the dock, i.e. Lockdown?
Why thank you kind sir! Ideally I need to find a reliable time series for incidence of long COVID but, yes, this will be a subset of what I've crunched. The wrinkle here is if long-COVID possesses a different profile over time to short-COVID - it most likely does, so I have modelled a hybrid with smearing as you say. The net effect of this is likely to reduce predictive power of all things viral, so we may declare these results as a start but we're only half-way there! I've got some ideas to try but need a spell of cogitation. Yes indeed, that remaining 60% is where my thinking is. If I can improve the model with estimates for long-COVID we might get a handle on lockdown deaths, these including removal of NHS services.
One final thing to bear in mind is all these ideas will make better sense if we flip to modelling all cause excess mortality and not non-COVID excess. It's worth asking why non-COVID excess death is linking to COVID in the first instance and my money is on certification error.
Even with no EPR, you sure you can't find a dashboard that at least reports weekly deaths from one likely candidate - say cancer or CVD? Surely those 5-month deaths must be driven by something more specific that would fit even better.
Agreed. I'm trying to get a fix on long-COVID for starters so that the virus is better represented in the model but the enemy here is time - I'm trying to cap my analytical work at 49 hours per week. I'll be grubbing for cancer and have a nurse grubbing for CVD using FOI. What we have in essence is a promising methodology under development. I'm guessing that as we progress the impact of vaccines will shrink.
And if a contribution from a very squashed sombrero long-Covid effect can’t be excluded, could a similar analysis of the Australian data (if accessible) help quantitate it, as they haven’t had so much time for a long-Covid effect to work its magic?
Dynamite! A question ... If the injections are resulting in increased numbers of both test-+ve and symptomatic CoViD cases (lots of clinical as well as statistical evidence to suggest this), then you would expect CDR (and potentially Long CoViD also) to be correlated with 'vaccination' rates. Have you run inter-variable correlation tests as well as looking at relationships with excess all-cause mortality?
A fabulous suggestion and one that I scribbled down on my bucket list this morning after a spot of cogitation. As you will note in comments above I'm trying to get a better handle on long-COVID so that the virus gets a better footing in the model - at present we have a smearing of short and long COVID. What is most peculiar is that the analysis made is for non-COVID death and not all cause, so there are issues with certification accuracy.
My mistake! Reading non-CoViD and typing all-cause because I was thinking in that direction ... ie. now you've so beautifully established the relationship with non-CoViD deaths and if there are statistically significant inter-variable correlations between combined dose rates and CoViD deaths/CDR/Long CoViD (with or without lags), then it becomes valid to look at combined dose rates vs all-cause, surely? This also gets around the misclassification problem ...
But then in highly vaxxed populations (given shedding and all) at this distance out from the naive infection phase of the pandemic, it's all looking like it's becoming a bit of a feedback loop and teasing out variable impacts is going to be nigh on impossible as we go further down the road?
I guess the only option then is to make comparisons with populations with very low vax uptake.
From one perspective you could say the differentiation between wild virus mortality and self-made is moot in any case. Spike protein is spike protein is spike protein, whichever way you get it (albeit with reported variations). Which brings the argument down to simply establishing which is the more lethal route.
No worries - I had to keep reminding myself that I was looking at non-COVID! The idea here was to flush out certification issues and to minimise the impact of variants but they sneaked in anyways. I shall indeed be flipping back to all cause using the same modelling strategy but it will be rather like eating a plate of spaghetti like you say, and I'm bound to get sauce on my shirt!
Whilst waiting for my tea to brew I quickly whipped this slide up for you. Substack doesn't permit images in comments so here it is on my Google drive...
...what you've got here is a cross-correlation plot of weekly vaccine doses with CDR, subject to first order differencing. There's a curious negative correlation at lag zero and a positive correlation at lag +2 weeks. Thus CDR is rising 2 weeks after vaccine dosing rises. That's quite a finding and I'll be writing this up in a future newsletter.
John, great work! With the finding of the highly significant 23 week lag between vaccination and excess death rates, it would be fascinating if you were to look at the forthcoming autumn Covid booster campaign and predict the excess deaths 5 months later. Kinda like Ferguson's model but one that doesn't input and output garbage!
Nailed it, John! One observation on my part. I think you are alluding that 5% of the non-COVID excess deaths might in fact be COVID deaths, given the correlation with the CDR? Correct me if I'm wrong! But, if so, it might also be the case that disruption to healthcare provision as a result of COVID-related HCW absences might account for this relationship? I can't remember how many months ago it is now but we did try and highlight this (Christmas 2000 I think it might have been) that the NHS quarantine rules, i.e. healthy staff isolating due to potential contact) was more likely to result in harm than any spurious protection afforded due to lack of exposure to a presumed asymptomatic carrier?
In fact, data IS available for NHS staff absences. Perhaps you could introduce that as a variable in your model (instead of the CDR) and see how it fits?
I love it when there's synergy like this - it sure helps me settle on a strategy when I've a long bucket list. I'll be looking at both non-COVID and COVID absences... but you already knew that :-)
Why thank you kind sir! Yep, a bit of alluding going on there and, yep, there are other cards we may play including this one that popped out after lunch..
....here we have a positive cross-correlation between CDR and dosing that pops up in the CDR record 2 weeks after dosing, which would suggest vaccine surges are driving short-term COVID surges (or maybe just positive test results). This may explain that 2 week lag effect for CDR.
I'll flip back to modelling all cause excess mortality at some point soon, which will be a juicier series to delve into. Lots of policy aftermath in that for sure. Nurses confiding in me all confirm total shit treatment (lack thereof) of those identified and wheeled away, so I'd certainly go along with elevated risk of death for asymptomatic COVID. Ironically, then, it is the identification process and withdrawal of care that is driving death rather than anything genuinely viral. This may partly explain the wacko results we are seeing.
Wasn’t it the misuse of PCR Testing that was responsible for huge increases in NHS absences 19/03/2020 - 02/04/2020. Some 73,316 people became absent in 13 days, at the critical period. During this period a&g beds were at the lowest levels for at least 12 years!
It was the low levels of occupancy that alerted me to all of this. With the confusion around ventilating patients (some doctors suggested it caused death) I focused on G&A, ejecting 32,000 in the first quarter of 2020, we then see over 24,000 excess care home deaths in April and May 2020.
Hi, sorry, I'm unable to ask this on your climate post as I don't seem to be a paid subscriber on that. Which leads me to my question: I pay for the vaccine/covid posts; do I need to pay twice so I can read Climate Corner? Thanks 🌞
A very interesting analysis. I am impressed by the fit of the model to the data and the 33% associate between jabs and deaths.
I have asked this question in part 3, but I am still a little confused....The nc xs deaths are -ve over a substantial number of weeks within the modelled data. Surely, if jabs are related to nc xs deaths, we would expect large +ve values of the xs?
Looking at it another way, 5 months after jabs were at a min rate, the nc xs death rate was -2000. So if there were no jabs, would we expect to see nc xs deaths running at -2000 or less? If so, then would we need to explain why fewer people are dying than normal?
Yes indeed, there shouldn't be a negative excess if jabs are the jobby so we are missing something (though this is essentially work in progress). I'm going to try a different approach to calculating excess and I'm going to model raw counts to see how things stack up then. There's also the issue of admin delays causing havoc so I'm going to cough up the cash needed to obtain deaths by date of death for England since they offer Vx by date of Vx. I've got NHS absence data for COVID and non-COVID to look at as an alternative to CDR and I'm on the trail of cancer and other leading causes of death in case this is all lockdown aftermath that correlates with jabs. A lot to do as yet, that will take another half dozen newsletters!
Another fantastic analysis.
“our list of suspects includes: COVID, long-COVID, lockdown and vaccines. We’ve seen that COVID plays a minor role and that vaccines play a major role in explaining that surge.”
If long-Covid was playing a bigger role than Covid (~5%) in predicting excess death, then would your analysis have picked it up as a better fitting longer lag from the CDR variable?
Putting that question a different way: does your analysis put an upper bound on what contribution a long-Covid effect could be delivering? (since long-Covid could be considered as lagged-Covid, and as there is no lag that exceeds 2 weeks in predictive value, and that contributes just ~5%).
Or does the analysis not exclude the possibility that long Covid has a long and smeared out effect that is not detectable by the optimal lag analysis, but still could contribute more than the 5% of deaths predicted by 2-week lagged CDR?
Finally, if the earlier interpretation is valid, then is it then fair for the jury to conclude that the remaining ~60% of observed excess deaths, which still need an explanation, can be attributed to the remaining suspect in the dock, i.e. Lockdown?
Why thank you kind sir! Ideally I need to find a reliable time series for incidence of long COVID but, yes, this will be a subset of what I've crunched. The wrinkle here is if long-COVID possesses a different profile over time to short-COVID - it most likely does, so I have modelled a hybrid with smearing as you say. The net effect of this is likely to reduce predictive power of all things viral, so we may declare these results as a start but we're only half-way there! I've got some ideas to try but need a spell of cogitation. Yes indeed, that remaining 60% is where my thinking is. If I can improve the model with estimates for long-COVID we might get a handle on lockdown deaths, these including removal of NHS services.
One final thing to bear in mind is all these ideas will make better sense if we flip to modelling all cause excess mortality and not non-COVID excess. It's worth asking why non-COVID excess death is linking to COVID in the first instance and my money is on certification error.
Even with no EPR, you sure you can't find a dashboard that at least reports weekly deaths from one likely candidate - say cancer or CVD? Surely those 5-month deaths must be driven by something more specific that would fit even better.
Agreed. I'm trying to get a fix on long-COVID for starters so that the virus is better represented in the model but the enemy here is time - I'm trying to cap my analytical work at 49 hours per week. I'll be grubbing for cancer and have a nurse grubbing for CVD using FOI. What we have in essence is a promising methodology under development. I'm guessing that as we progress the impact of vaccines will shrink.
And if a contribution from a very squashed sombrero long-Covid effect can’t be excluded, could a similar analysis of the Australian data (if accessible) help quantitate it, as they haven’t had so much time for a long-Covid effect to work its magic?
A great idea but my time is fully committed with just the UK data!
Dynamite! A question ... If the injections are resulting in increased numbers of both test-+ve and symptomatic CoViD cases (lots of clinical as well as statistical evidence to suggest this), then you would expect CDR (and potentially Long CoViD also) to be correlated with 'vaccination' rates. Have you run inter-variable correlation tests as well as looking at relationships with excess all-cause mortality?
A fabulous suggestion and one that I scribbled down on my bucket list this morning after a spot of cogitation. As you will note in comments above I'm trying to get a better handle on long-COVID so that the virus gets a better footing in the model - at present we have a smearing of short and long COVID. What is most peculiar is that the analysis made is for non-COVID death and not all cause, so there are issues with certification accuracy.
My mistake! Reading non-CoViD and typing all-cause because I was thinking in that direction ... ie. now you've so beautifully established the relationship with non-CoViD deaths and if there are statistically significant inter-variable correlations between combined dose rates and CoViD deaths/CDR/Long CoViD (with or without lags), then it becomes valid to look at combined dose rates vs all-cause, surely? This also gets around the misclassification problem ...
But then in highly vaxxed populations (given shedding and all) at this distance out from the naive infection phase of the pandemic, it's all looking like it's becoming a bit of a feedback loop and teasing out variable impacts is going to be nigh on impossible as we go further down the road?
I guess the only option then is to make comparisons with populations with very low vax uptake.
From one perspective you could say the differentiation between wild virus mortality and self-made is moot in any case. Spike protein is spike protein is spike protein, whichever way you get it (albeit with reported variations). Which brings the argument down to simply establishing which is the more lethal route.
No worries - I had to keep reminding myself that I was looking at non-COVID! The idea here was to flush out certification issues and to minimise the impact of variants but they sneaked in anyways. I shall indeed be flipping back to all cause using the same modelling strategy but it will be rather like eating a plate of spaghetti like you say, and I'm bound to get sauce on my shirt!
Whilst waiting for my tea to brew I quickly whipped this slide up for you. Substack doesn't permit images in comments so here it is on my Google drive...
https://drive.google.com/file/d/1quG0t3JmhVt_GNY6tFWQTOuC8c8xC2IK/view?usp=sharing
...what you've got here is a cross-correlation plot of weekly vaccine doses with CDR, subject to first order differencing. There's a curious negative correlation at lag zero and a positive correlation at lag +2 weeks. Thus CDR is rising 2 weeks after vaccine dosing rises. That's quite a finding and I'll be writing this up in a future newsletter.
I somehow thought you might find something like this ... It ties in quite nicely with anecdotal reports.
John, great work! With the finding of the highly significant 23 week lag between vaccination and excess death rates, it would be fascinating if you were to look at the forthcoming autumn Covid booster campaign and predict the excess deaths 5 months later. Kinda like Ferguson's model but one that doesn't input and output garbage!
I've certainly been thinking about this. I've a few tweaks to make to models for all cause and non-COVID death - what you see is just the start!
Nailed it, John! One observation on my part. I think you are alluding that 5% of the non-COVID excess deaths might in fact be COVID deaths, given the correlation with the CDR? Correct me if I'm wrong! But, if so, it might also be the case that disruption to healthcare provision as a result of COVID-related HCW absences might account for this relationship? I can't remember how many months ago it is now but we did try and highlight this (Christmas 2000 I think it might have been) that the NHS quarantine rules, i.e. healthy staff isolating due to potential contact) was more likely to result in harm than any spurious protection afforded due to lack of exposure to a presumed asymptomatic carrier?
In fact, data IS available for NHS staff absences. Perhaps you could introduce that as a variable in your model (instead of the CDR) and see how it fits?
You've got ESP - that data is already prepped and waiting for my next modelling phase!
Great minds, eh?! Hee hee.
I love it when there's synergy like this - it sure helps me settle on a strategy when I've a long bucket list. I'll be looking at both non-COVID and COVID absences... but you already knew that :-)
Why thank you kind sir! Yep, a bit of alluding going on there and, yep, there are other cards we may play including this one that popped out after lunch..
.
https://drive.google.com/file/d/1quG0t3JmhVt_GNY6tFWQTOuC8c8xC2IK/view?usp=sharing
....here we have a positive cross-correlation between CDR and dosing that pops up in the CDR record 2 weeks after dosing, which would suggest vaccine surges are driving short-term COVID surges (or maybe just positive test results). This may explain that 2 week lag effect for CDR.
I'll flip back to modelling all cause excess mortality at some point soon, which will be a juicier series to delve into. Lots of policy aftermath in that for sure. Nurses confiding in me all confirm total shit treatment (lack thereof) of those identified and wheeled away, so I'd certainly go along with elevated risk of death for asymptomatic COVID. Ironically, then, it is the identification process and withdrawal of care that is driving death rather than anything genuinely viral. This may partly explain the wacko results we are seeing.
Wasn’t it the misuse of PCR Testing that was responsible for huge increases in NHS absences 19/03/2020 - 02/04/2020. Some 73,316 people became absent in 13 days, at the critical period. During this period a&g beds were at the lowest levels for at least 12 years!
https://www.england.nhs.uk/statistics/statistical-work-areas/covid-19-hospital-activity/
Oh yes indeed! Another story to tease out perhaps - I have bed use data.
It was the low levels of occupancy that alerted me to all of this. With the confusion around ventilating patients (some doctors suggested it caused death) I focused on G&A, ejecting 32,000 in the first quarter of 2020, we then see over 24,000 excess care home deaths in April and May 2020.
Absolutely. A criminal act.
You inspired me to do this: https://metatron.substack.com/p/evidence-of-the-mrna-experiment-causing
Oh boy, that is utterly fabulous! I'll start my next newsletter with this.
Hi, sorry, I'm unable to ask this on your climate post as I don't seem to be a paid subscriber on that. Which leads me to my question: I pay for the vaccine/covid posts; do I need to pay twice so I can read Climate Corner? Thanks 🌞
Hang on - I'll issue you with a complimentary subscription for the climate corner... we can't have you paying twice!
Sorted - Substack should now have issued you with a year comp for my climate corner - thanks for taking an interest in my rantings!
Thanks so much! BTW I wasn't necessarily expecting two subscriptions for the price of one 😳 I know a lot of work goes into your NOT rantings!
My pleasure. Yep, it sure keeps me busy!
A very interesting analysis. I am impressed by the fit of the model to the data and the 33% associate between jabs and deaths.
I have asked this question in part 3, but I am still a little confused....The nc xs deaths are -ve over a substantial number of weeks within the modelled data. Surely, if jabs are related to nc xs deaths, we would expect large +ve values of the xs?
Looking at it another way, 5 months after jabs were at a min rate, the nc xs death rate was -2000. So if there were no jabs, would we expect to see nc xs deaths running at -2000 or less? If so, then would we need to explain why fewer people are dying than normal?
Yes indeed, there shouldn't be a negative excess if jabs are the jobby so we are missing something (though this is essentially work in progress). I'm going to try a different approach to calculating excess and I'm going to model raw counts to see how things stack up then. There's also the issue of admin delays causing havoc so I'm going to cough up the cash needed to obtain deaths by date of death for England since they offer Vx by date of Vx. I've got NHS absence data for COVID and non-COVID to look at as an alternative to CDR and I'm on the trail of cancer and other leading causes of death in case this is all lockdown aftermath that correlates with jabs. A lot to do as yet, that will take another half dozen newsletters!
Thank you John - very much appreciated.