I put down ONS weekly all cause registered death data for 2010/w1 – 2023/w52 and pick up daily in-hospital death data for 2017/w1 – 2021/w36 for an undisclosed NHS Trust once more
Indeed so, though only in terms of mention of acute respiratory prior to death. Coming up will be an extensive analysis (several months worth) that looks at folk who came to hospital but didn't die, and we are going to see a dirty great bump in 2019. Bugger - I've gone and spoiled the ending of what should be a cliff-hanging series!
That will be interesting - and I don't blame you for the spoiler, as its really hard to keep some secrets! . I was kinda hoping something fruity might turn up, given that I am totally convinced we all had here had this bug in late Dec 2019. The $1Trillion dollar question is whether in fact Covid-19 was even primarily a respiratory bug at all, or a generalised bug that simply went for any immune system vulnerability it could find, and for that I'd be especially interested in circulatory and cardiac issues like strokes or coronaries, given that they tend to be more immediately obvious than cancers.
PS, is it just me being a bit more conscious of it, or is cancer on a roll at the moment?
Yes, absolutely. And if cancer is indeed proven to be on the march, then we need to somehow figure out if it is Covid, the Vaccines, or both, or something else, that is driving any increase. Here's where international comparisons might be crucial, due to the variables of timing of the infections and the fact that vaccination rates vary so greatly on a global basis.
Its a horrible shame that we never thought to introduce a voluntary 'control group' on the unvaxxed and follow them up regularly - my family and I would have been delighted to assist.
I've seen similar thoughts about breast cancer etc.
Now, this affects our discussions about whether lockdown increased late presentation of common cancers. This was always - entirely reasonably - assumed to be 'a bad thing'. And of course it muddies the waters - if we can blame lockdown for an increase in cancer, then that lets vaccines off the hook.
However the new revelations about Prostate cancer make for very murky waters.
A reduction in excessive screening and early interventions might have two possible outcomes, and they might vary for each cancer. Late screening and treatment might actually improve outcomes, or it might not.
Is the somewhat non-independent nature of the samples as indicated by AR1 due in part or in whole to the T4253H filter? (The effect that today’s sample reflects yesterday’s sample, to some degree.) A question that may elucidate this is whether the the AR1 value is sensitive to the use of different bandwidth settings when filtering the daily data (or, how sensitive, to avoid a yes/no formulation). As you mentioned, maybe not something that leads to the need for a correction, but, curious minds like nooks and crannies, especially when filled with glorious butter.
P.S. Depending on the nature of the filter - I’m not familiar with T4253H flour sifter - it may also introduce a lag in its output, wrt its input, related to its bandwidth. I’m thinking of moving average and first order Butterworth as examples of that.
Yes indeed, filtering of any kind is going to produce distortions and there is an argument for averaging the raw data. But... we could come at this from the other end and use a hypothesised Gompertz-like function. I suspect the differences wouldn't sway matters much.
A finely churned question! Yes, any form of smoothing is going to inflate estimates of rho (AR1). If I run autoregression on the entire daily series maximum likelihood methods yield ρ = 0.510 (p<0.001) for the raw series and ρ = 0.981 (p<0.001) for the smoothed series. Despite such inflation there's still a hefty amount of serial correlation, as we might expect.
Pulling back from nooks and crannies, were there any definitional changes or procedural changes, that might affect how events were classified and counted, in Dec/19 timeframe preceding the plunge to the Mar/20 nadir?
There almost certainly would have been and we can only hazard a guess as to how Trusts (and teams within Trusts) responded. Not just responded neither, for I suspect a great deal of anticipation, and that pre-pandemic dip smells of preparedness more than anything. Analysis of the data is like trying to eat soup with a fork.
At least it doesn't look like covid was prevalent in 2019.
Indeed so, though only in terms of mention of acute respiratory prior to death. Coming up will be an extensive analysis (several months worth) that looks at folk who came to hospital but didn't die, and we are going to see a dirty great bump in 2019. Bugger - I've gone and spoiled the ending of what should be a cliff-hanging series!
That will be interesting - and I don't blame you for the spoiler, as its really hard to keep some secrets! . I was kinda hoping something fruity might turn up, given that I am totally convinced we all had here had this bug in late Dec 2019. The $1Trillion dollar question is whether in fact Covid-19 was even primarily a respiratory bug at all, or a generalised bug that simply went for any immune system vulnerability it could find, and for that I'd be especially interested in circulatory and cardiac issues like strokes or coronaries, given that they tend to be more immediately obvious than cancers.
PS, is it just me being a bit more conscious of it, or is cancer on a roll at the moment?
There's a big pile of goodies to come. Yep, it looks like cancer is going crazy but I always like to get my hands on the data.
Yes, absolutely. And if cancer is indeed proven to be on the march, then we need to somehow figure out if it is Covid, the Vaccines, or both, or something else, that is driving any increase. Here's where international comparisons might be crucial, due to the variables of timing of the infections and the fact that vaccination rates vary so greatly on a global basis.
Its a horrible shame that we never thought to introduce a voluntary 'control group' on the unvaxxed and follow them up regularly - my family and I would have been delighted to assist.
One further thought. I don't know if you saw the recent articles about early detection and treatment of Prostate cancer : https://www.telegraph.co.uk/news/2024/04/06/prostate-cancer-screening-may-do-more-harm-than-good/
I've seen similar thoughts about breast cancer etc.
Now, this affects our discussions about whether lockdown increased late presentation of common cancers. This was always - entirely reasonably - assumed to be 'a bad thing'. And of course it muddies the waters - if we can blame lockdown for an increase in cancer, then that lets vaccines off the hook.
However the new revelations about Prostate cancer make for very murky waters.
A reduction in excessive screening and early interventions might have two possible outcomes, and they might vary for each cancer. Late screening and treatment might actually improve outcomes, or it might not.
Why do you say that?
Because no spike in excess acute respiratory death in 2019.
But you're assuming SARS-CoV-2 is deadly, yes?
Is the somewhat non-independent nature of the samples as indicated by AR1 due in part or in whole to the T4253H filter? (The effect that today’s sample reflects yesterday’s sample, to some degree.) A question that may elucidate this is whether the the AR1 value is sensitive to the use of different bandwidth settings when filtering the daily data (or, how sensitive, to avoid a yes/no formulation). As you mentioned, maybe not something that leads to the need for a correction, but, curious minds like nooks and crannies, especially when filled with glorious butter.
P.S. Depending on the nature of the filter - I’m not familiar with T4253H flour sifter - it may also introduce a lag in its output, wrt its input, related to its bandwidth. I’m thinking of moving average and first order Butterworth as examples of that.
Yes indeed, filtering of any kind is going to produce distortions and there is an argument for averaging the raw data. But... we could come at this from the other end and use a hypothesised Gompertz-like function. I suspect the differences wouldn't sway matters much.
A finely churned question! Yes, any form of smoothing is going to inflate estimates of rho (AR1). If I run autoregression on the entire daily series maximum likelihood methods yield ρ = 0.510 (p<0.001) for the raw series and ρ = 0.981 (p<0.001) for the smoothed series. Despite such inflation there's still a hefty amount of serial correlation, as we might expect.
Pulling back from nooks and crannies, were there any definitional changes or procedural changes, that might affect how events were classified and counted, in Dec/19 timeframe preceding the plunge to the Mar/20 nadir?
There almost certainly would have been and we can only hazard a guess as to how Trusts (and teams within Trusts) responded. Not just responded neither, for I suspect a great deal of anticipation, and that pre-pandemic dip smells of preparedness more than anything. Analysis of the data is like trying to eat soup with a fork.
So much for wide stationary and ergodic
Good work
What you have termed “catastrophic health collapse,”I characterize more strongly - and with reference to intent - as “sinking the damaged ships”
https://www.woodhouse76.com/p/the-allegory-of-the-damaged-ship
Graphs 5 and 6 are problematic for the Narrative.
On the upside, at least the hospital(s) in the undisclosed trust didn't do whatever they did to the level and extent of New York City.
https://substack.com/profile/32813354-jessica-hockett/note/c-56452613?utm_source=notes-share-action&r=jjay2