In several datasets the ratio between unvaccinated and vaccinated COVID mortality is greater in 2021 than 2022. So it might explain why states with a low percentage of vaccinated people would also have elevated mortality in 2021 relative to 2022. By 2022 many unvaccinated people had acquired natural immunity, so relative to a baseline of unvaccinated people, vaccines seem to have provided a greater immunization advantage in 2021 than 2022: sars2.net/rootclaim3.html#Comparison_of_COVID_deaths_in_2022_relative_to_2021_by_Spiro_Pantazatos.
That's a good thought. If that is true, then 2021 vaccination rates should be negatively correlated with 2021 mortality. I just checked, however, and they are uncorrelated (all ages COVID r=-0.16, p=0.26, all ages total deaths r=-0.11, p=0.4) so that doesn't explain it. Also, the same trend was observed from 2022 to 2023, when many unvaccinated people had acquired natural immunity. Which datasets are you referring to, and did they correctly categorize and/or include vaccinated people within 14 days post-injection?
When I compared the percentage of vaccinated people at the end of 2021 against COVID deaths per capita in 2021, I got a Pearson's correlation coefficient of about -0.62 with a p-value of about 2e-6:
I referred to the English ONS data, the Dutch CBS data, and the newer Czech record-level dataset published in November 2024. In all of them I treated people as vaccinated immediately after the first dose and I didn't exclude partially vaccinated people. And I also looked at the CDC dataset for COVID deaths by vaccination status, which only includes unvaccinated and fully vaccinated people so that partially vaccinated people are excluded.
You're right on the correlation, I had a brain blip and omitted a denominator in one of the variables. However, that negative correlation is still there in 2020, 2019 etc. and is explained by pre-existing variation in COVID comorbidities (see my original post on Robinson's paradox) and not ratio of unvax to vax, and the trend continues from 2022 to 2023 after most unvaxed got immunity, and so I don't see how that can explain the findings. Plus the regression using current year counts deaths to predict following year should adequately adjust for baseline differences. I'm not familiar with Dutch CBS or Czech data, but for the ONS data, see this recent article. https://pubmed.ncbi.nlm.nih.gov/40028449/ They report increasing ASMR in vaccinated over time and also a large bias in the data.
When I used the same methodology as in my previous comment, I got a correlation of about -0.11 in 2020, -0.62 in 2021, -0.35 in 2022, and -0.01 in 2023. See the plot I added here: https://sars2.net/rootclaim3.html#Comparison_of_COVID_deaths_in_2022_relative_to_2021_by_Spiro_Pantazatos. So the correlation was not that far from zero yet in 2020. But the correlation was the furthest from zero in 2021, which seems to have been the year when vaccines had the greatest impact on the COVID mortality rate of vaccinated people relative to unvaccinated people.
In the code below I calculated a relative COVID mortality risk in this CDC dataset that extends from October 2021 up to April 2023: https://data.cdc.gov/Public-Health-Surveillance/Rates-of-COVID-19-Cases-or-Deaths-by-Age-Group-and/54ys-qyzm. For each combination of observation week and age group, I got the expected deaths by multiplying the unvaccinated mortality rate by the vaccinated population size, and I divided the total vaccinated deaths on each MMWR year by the total expected deaths on the same year. The resulting mortality risk of vaccinated people relative to unvaccinated people was about 0.06 in 2021, 0.13 in 2022, and 0.22 in 2023:
So since my relative risk was about 1.7 times higher in 2023 than 2022, it's not surprising that by 2023 the gap in COVID mortality between less-vaccinated and more-vaccinated states would've gotten even smaller than in 2022. The CDC dataset only extended up to April 22nd 2023, but if I would've been able to include the rest 2023 in my calculation, then the RR in 2023 might have been even higher.
---
In the ONS data, I think the ASMR of vaccinated people increases over time because vaccinated people are more impacted by the healthy vaccinee effect in 2021 than 2022. The same thing happens in the Czech data, Dutch CBS data, and Barry Young's NZ data.
Because of the temporal/time-varying HVE, there's a reduced number of deaths in the first few months after vaccination, but the ratio between unvaccinated and vaccinated mortality gradually gets lower over time because the HVE wears out and because unvaccinated people acquire natural immunity: https://sars2.net/rootclaim3.html#Comparison_to_English_ONS_data.
Thanks what about correlations with total deaths in 2020 and 2019? I might expect correlations with COVID deaths in 2020 to be closer to zero as COVID deaths were not consistently being defined and the 2020 COVID waves were largely localized to (later) higher vaccinated areas like NY etc., which would have given those areas natural immunity before 2021 as well, which might explain the stronger 2021 negative correlation and impact you describe. I'll take a closer look at the spreadsheet you posted...At first pass it looks like the HVE may help (at least partially) explain those results rather than vax benefit.
The new Italian paper basically just showed that people with n-1 doses had elevated mortality when the nth dose was rolled out. But the same phenomenon can also be seen in the Czech data even though there are no Fentonian cheap tricks involved: https://x.com/henjin256/status/1897277592924983336.
A flaw of the Italian paper was that they only looked at the mortality risk in people with two but not more doses, and not people with two or more doses. Among people with two or more doses, the mortality risk of vaccinated people relative to unvaccinated people remained stable even after the third dose was rolled out. (See my reply to the tweet I linked.)
In order to explain why the ASMR of people with two doses shot up when the third dose was rolled out, Fenton et al. hypothesized that people who died soon after getting the third dose were misclassified under the second dose. But if it's correct then there should also be a big increase in the mortality of people with two or more doses when the third dose is rolled out. But that is not the case, because there is only a big increase in the mortality of people with two but not more doses.
In several datasets the ratio between unvaccinated and vaccinated COVID mortality is greater in 2021 than 2022. So it might explain why states with a low percentage of vaccinated people would also have elevated mortality in 2021 relative to 2022. By 2022 many unvaccinated people had acquired natural immunity, so relative to a baseline of unvaccinated people, vaccines seem to have provided a greater immunization advantage in 2021 than 2022: sars2.net/rootclaim3.html#Comparison_of_COVID_deaths_in_2022_relative_to_2021_by_Spiro_Pantazatos.
That's a good thought. If that is true, then 2021 vaccination rates should be negatively correlated with 2021 mortality. I just checked, however, and they are uncorrelated (all ages COVID r=-0.16, p=0.26, all ages total deaths r=-0.11, p=0.4) so that doesn't explain it. Also, the same trend was observed from 2022 to 2023, when many unvaccinated people had acquired natural immunity. Which datasets are you referring to, and did they correctly categorize and/or include vaccinated people within 14 days post-injection?
When I compared the percentage of vaccinated people at the end of 2021 against COVID deaths per capita in 2021, I got a Pearson's correlation coefficient of about -0.62 with a p-value of about 2e-6:
vax=fread("https://data.cdc.gov/api/views/rh2h-3yt2/rows.csv?accessType=DOWNLOAD")
vax=vax[date_type=="Report"&Date=="12/31/2021",.(state=state.name[match(Location,state.abb)],vax=Administered_Dose1_Pop_Pct)]
t=fread("https://sars2.net/f/wonderstatecovidyearly.csv")
merge(t[year==2021],vax)[,cor.test(dead/pop,vax)]
I got yearly deaths with underlying cause U07.1 by state of residence from CDC WONDER: https://wonder.cdc.gov/mcd.html.
I referred to the English ONS data, the Dutch CBS data, and the newer Czech record-level dataset published in November 2024. In all of them I treated people as vaccinated immediately after the first dose and I didn't exclude partially vaccinated people. And I also looked at the CDC dataset for COVID deaths by vaccination status, which only includes unvaccinated and fully vaccinated people so that partially vaccinated people are excluded.
You're right on the correlation, I had a brain blip and omitted a denominator in one of the variables. However, that negative correlation is still there in 2020, 2019 etc. and is explained by pre-existing variation in COVID comorbidities (see my original post on Robinson's paradox) and not ratio of unvax to vax, and the trend continues from 2022 to 2023 after most unvaxed got immunity, and so I don't see how that can explain the findings. Plus the regression using current year counts deaths to predict following year should adequately adjust for baseline differences. I'm not familiar with Dutch CBS or Czech data, but for the ONS data, see this recent article. https://pubmed.ncbi.nlm.nih.gov/40028449/ They report increasing ASMR in vaccinated over time and also a large bias in the data.
When I used the same methodology as in my previous comment, I got a correlation of about -0.11 in 2020, -0.62 in 2021, -0.35 in 2022, and -0.01 in 2023. See the plot I added here: https://sars2.net/rootclaim3.html#Comparison_of_COVID_deaths_in_2022_relative_to_2021_by_Spiro_Pantazatos. So the correlation was not that far from zero yet in 2020. But the correlation was the furthest from zero in 2021, which seems to have been the year when vaccines had the greatest impact on the COVID mortality rate of vaccinated people relative to unvaccinated people.
In the code below I calculated a relative COVID mortality risk in this CDC dataset that extends from October 2021 up to April 2023: https://data.cdc.gov/Public-Health-Surveillance/Rates-of-COVID-19-Cases-or-Deaths-by-Age-Group-and/54ys-qyzm. For each combination of observation week and age group, I got the expected deaths by multiplying the unvaccinated mortality rate by the vaccinated population size, and I divided the total vaccinated deaths on each MMWR year by the total expected deaths on the same year. The resulting mortality risk of vaccinated people relative to unvaccinated people was about 0.06 in 2021, 0.13 in 2022, and 0.22 in 2023:
> t=fread("Rates_of_COVID-19_Cases_or_Deaths_by_Age_Group_and_Updated__Bivalent__Booster_Status_20241231.csv")
> d=t[outcome=="death"&age_group!="all_ages"&vaccination_status=="vaccinated"]
> d[,.(rr=round(sum(vaccinated_with_outcome)/sum((unvaccinated_with_outcome/unvaccinated_population*vaccinated_population),na.rm=T),2)),.(year=mmwr_week%/%100)]|>print(r=F)
year rr
2021 0.06
2022 0.13
2023 0.22
So since my relative risk was about 1.7 times higher in 2023 than 2022, it's not surprising that by 2023 the gap in COVID mortality between less-vaccinated and more-vaccinated states would've gotten even smaller than in 2022. The CDC dataset only extended up to April 22nd 2023, but if I would've been able to include the rest 2023 in my calculation, then the RR in 2023 might have been even higher.
---
In the ONS data, I think the ASMR of vaccinated people increases over time because vaccinated people are more impacted by the healthy vaccinee effect in 2021 than 2022. The same thing happens in the Czech data, Dutch CBS data, and Barry Young's NZ data.
Because of the temporal/time-varying HVE, there's a reduced number of deaths in the first few months after vaccination, but the ratio between unvaccinated and vaccinated mortality gradually gets lower over time because the HVE wears out and because unvaccinated people acquire natural immunity: https://sars2.net/rootclaim3.html#Comparison_to_English_ONS_data.
Also read these: https://sars2.net/rootclaim.html#Fenton_and_Neils_miscategorization_bias, https://sars2.net/uk.html#Problems_with_Fenton_and_Neils_n_1_dose_misclassification_hypothesis, https://www.researchgate.net/publication/387220055_A_DETAILED_ANALYSIS_OF_CLAIMS_OF_MISCATEGORIZATION_BIAS_IN_STUDIES_OF_COVID-19_VACCINE_EFFECTIVENESS.
Thanks what about correlations with total deaths in 2020 and 2019? I might expect correlations with COVID deaths in 2020 to be closer to zero as COVID deaths were not consistently being defined and the 2020 COVID waves were largely localized to (later) higher vaccinated areas like NY etc., which would have given those areas natural immunity before 2021 as well, which might explain the stronger 2021 negative correlation and impact you describe. I'll take a closer look at the spreadsheet you posted...At first pass it looks like the HVE may help (at least partially) explain those results rather than vax benefit.
The new Italian paper basically just showed that people with n-1 doses had elevated mortality when the nth dose was rolled out. But the same phenomenon can also be seen in the Czech data even though there are no Fentonian cheap tricks involved: https://x.com/henjin256/status/1897277592924983336.
A flaw of the Italian paper was that they only looked at the mortality risk in people with two but not more doses, and not people with two or more doses. Among people with two or more doses, the mortality risk of vaccinated people relative to unvaccinated people remained stable even after the third dose was rolled out. (See my reply to the tweet I linked.)
In order to explain why the ASMR of people with two doses shot up when the third dose was rolled out, Fenton et al. hypothesized that people who died soon after getting the third dose were misclassified under the second dose. But if it's correct then there should also be a big increase in the mortality of people with two or more doses when the third dose is rolled out. But that is not the case, because there is only a big increase in the mortality of people with two but not more doses.