Other parts: rootclaim.html, rootclaim2.html, rootclaim3.html.
Kirsch posted this table of the English ONS data, where he pointed out that in April to May 2021 when there was a low number of COVID cases, the ratio between unvaccinated and vaccinated ASMR was higher than in December 2021 to February 2022 when there was a high number of COVID cases: [https://kirschsubstack.com/p/official-uk-ons-data-shows-the-covid]
However in April to May 2021 there were still many people who had been recently vaccinated, so vaccinated people were still heavily impacted by the temporal healthy vaccinee effect (which is a term coined by Jeffrey Morris for the stronger form of HVE which lasts for the first weeks or months after vaccination, as opposed to the inherent HVE that remains in place even long after vaccination).
From the plot below you can roughly see the magnitude of the HVE from the dashed line which shows non-COVID ASMR. The plot below also shows that the ratio between unvaccinated and vaccinated ASMR was in fact elevated during the Delta peak in August 2021 and the COVID wave that peaked in December 2021 to January 2022. But in October 2021 when there was a lower number of COVID deaths than in the surrounding months, the ratio between unvaccinated and vaccinated ASMR was also lower than in the surrounding months:
t=fread("http://sars2.net/f/ons.csv")
t=rbind(t[ed==7&month<="2021-03"],t[ed==9])[cause!="Deaths involving COVID-19"]
t[,x:=as.Date(paste0(month,"-15"))]
t[,vax:=factor(status!="Unvaccinated",,c("Unvaccinated","Vaccinated"))]
a=t[age!="Total",.(cmr=sum(dead)/sum(pop)),.(age,vax,x,cause)]
p=a[,.(y=cmr[1]/cmr[2]),.(age,x,cause)]
p=rbind(p,t[age=="Total"&status%like%"accin",.(y=asmr[1]/asmr[2],age="All ages"),.(x,cause)])
xstart=as.Date("2021-1-1");xend=as.Date("2023-6-1");xbreak=seq(xstart,xend,"year");yend=max(p$y)
ggplot(p)+
facet_wrap(~age,dir="v",ncol=2)+
geom_hline(yintercept=2:8,color="gray91",linewidth=.4)+
geom_hline(yintercept=1,color="gray60",linewidth=.4)+
geom_vline(xintercept=seq(xstart,xend,"year"),color="gray60",linewidth=.4)+
geom_rect(data=p[,max(y),age],aes(xmin=xstart,xmax=xend,ymin=0,ymax=yend),linewidth=.4,fill=NA,color="black",lineend="square")+
geom_line(aes(x,y,linetype=cause),linewidth=.6)+
geom_text(data=p[,max(y),age],aes(x=as.Date("2022-7-1"),y=yend,label=age),hjust=.5,fontface=2,vjust=1.5,size=3.8)+
labs(x=NULL,y=NULL,title="Ratio between unvaccinated and vaccinated mortality rate\nin English ONS data")+
scale_x_continuous(limits=c(xstart-.5,xend+.5),breaks=xbreak,labels=year(xbreak))+
scale_y_continuous(breaks=seq(1,7,2))+
scale_linetype_manual(values=c("solid","11"))+
scale_alpha_manual(values=c(1,0),guide="none")+
coord_cartesian(clip="off",expand=F)+
guides(color=guide_legend(order=1),linetype=guide_legend(order=2))+
theme(axis.text=element_text(hjust=0,size=11,color="black"),
axis.title=element_text(size=11,face=2),
axis.text.x=element_text(margin=margin(3)),
axis.ticks=element_line(linewidth=.4,color="black"),
axis.ticks.length.x=unit(0,"pt"),
axis.ticks.length.y=unit(4,"pt"),
legend.background=element_blank(),
legend.box.spacing=unit(0,"pt"),
legend.direction="horizontal",
legend.justification="right",
legend.key=element_blank(),
legend.key.height=unit(13,"pt"),
legend.key.width=unit(21,"pt"),
legend.margin=margin(,,5),
legend.position="top",
legend.spacing.x=unit(2,"pt"),
legend.spacing.y=unit(0,"pt"),
legend.text=element_text(size=11,vjust=.5),
legend.title=element_blank(),
panel.background=element_blank(),
panel.spacing=unit(2,"pt"),
plot.margin=margin(5,5,5,5),
plot.title=element_text(size=11,face=2,margin=margin(2,,4)),
strip.background=element_blank(),
strip.text=element_blank())
ggsave("1.png",width=4.8,height=4.1,dpi=300*4)
system(paste0("magick 1.png -trim -resize 25% -bordercolor white -border 24 -colors 256 1.png"))
The next plot shows that there was a period with elevated COVID deaths that lasted from around August 2021 until January 2022. But between January and February 2022 when there was a big drop in COVID deaths, there was also a clear drop in the ratio of all-cause ASMR between unvaccinated and vaccinated people:
library(data.table);library(ggplot2)
t=fread("http://sars2.net/f/ons.csv")
t=rbind(t[ed==7&month<="2021-03"],t[ed==9])
t[,x:=as.Date(paste0(month,"-15"))]
t[,vax:=factor(status!="Unvaccinated",,c("Unvaccinated","Vaccinated"))]
a=t[age!="Total",.(y=sum(dead)/sum(pop)*1e5),.(age,x,vax,cause)]
a=rbind(a,t[age=="Total"&status%like%"accin",.(x,y=asmr,age="All ages",vax,cause)])
p=a[cause!="Non-COVID-19 deaths"]
p[cause=="Deaths involving COVID-19",cause:="COVID deaths"]
xstart=as.Date("2021-1-1");xend=as.Date("2023-6-1");xbreak=seq(xstart,xend,"year")
ggplot(p)+
facet_wrap(~age,dir="v",scales="free_y",ncol=2)+
geom_vline(xintercept=seq(xstart,xend,"year"),color="gray60",linewidth=.4)+
geom_rect(data=p[,max(y),age],aes(xmin=xstart,xmax=xend,ymin=0,ymax=V1),linewidth=.4,fill=NA,color="black",lineend="square")+
geom_line(aes(x,y,linetype=cause,color=vax))+
geom_text(data=p[,max(y),age],aes(x=as.Date("2022-7-1"),y=V1,label=age),hjust=.5,fontface=2,vjust=1.5,size=3.8)+
labs(x=NULL,y=NULL,title="Monthly mortality rates in English ONS data")+
scale_x_continuous(limits=c(xstart-.5,xend+.5),breaks=xbreak,labels=year(xbreak))+
scale_y_continuous(breaks=\(x){y=pretty(x,4);y[y<.85*max(x)]},labels=\(x)ifelse(x<1000,x,paste0(x/1e3,"k")))+
scale_color_manual(values=c("#66f","#f66","black"))+
scale_linetype_manual(values=c("solid","11"))+
scale_alpha_manual(values=1:0)+
coord_cartesian(clip="off",expand=F)+
guides(color=guide_legend(order=1),linetype=guide_legend(order=2))+
theme(axis.text=element_text(hjust=0,size=11,color="black"),
axis.title=element_text(size=11,face=2),
axis.text.x=element_text(margin=margin(3)),
axis.ticks=element_line(linewidth=.4,color="black"),
axis.ticks.length.x=unit(0,"pt"),
axis.ticks.length.y=unit(4,"pt"),
legend.background=element_rect(color="black",linewidth=.4),
legend.box.margin=margin(,,2),
legend.box.spacing=unit(0,"pt"),
legend.direction="horizontal",
legend.justification="right",
legend.key=element_blank(),
legend.key.height=unit(13,"pt"),
legend.key.width=unit(21,"pt"),
legend.margin=margin(3,4,3,3),
legend.position="top",
legend.spacing.x=unit(2,"pt"),
legend.spacing.y=unit(0,"pt"),
legend.text=element_text(size=11,vjust=.5),
legend.title=element_blank(),
panel.background=element_blank(),
panel.spacing=unit(2,"pt"),
plot.margin=margin(5,5,5,5),
plot.title=element_text(hjust=.5,size=11,face=2,margin=margin(2,,4)),
strip.background=element_blank(),
strip.text=element_blank())
ggsave("1.png",width=5.47,height=4.3,dpi=300*4)
sub="The values for individual age groups are crude mortality rates per 100,000 person-years, and the values for all ages are age-standardized mortality rates per 100,000 person-years. The data was taken from Tables 1 and 2 of the ONS dataset \"Deaths involving COVID-19 by vaccination status, England\". Data for the first three months of 2021 was taken from the July 2022 edition and data for other months was taken from the August 2023 edition. People are counted as vaccinated on the day of the first dose. COVID deaths are deaths where COVID was mentioned anywhere on the death certificate."
system(paste0("mogrify -trim 1.png;magick 1.png \\( -size `identify -format %w 1.png`x -font Arial -interline-spacing -3 -pointsize $[43*4] caption:'",gsub("'","'\\\\'",sub),"' -splice x80 \\) -append -resize 25% -bordercolor white -border 24 -colors 256 1.png"))
Someone in Kirsch's Substack comments asked why February 2021 was an outlier in Kirsch's table, because in his table the ratio between unvaccinated and vaccinated ASMR was about 2.0 in January 2021, 5.8 in February, 3.7 in March, and 2.6 in April. I pointed out that in February a large part of vaccinated people had been vaccinated in recent weeks, and the ONS dataset has a very low number of deaths in the first 2 weeks after vaccination: [stat.html#Plot_deaths_by_weeks_after_vaccination_and_age_group]
I also pointed out that there was high COVID mortality in both January and February 2021. In January 2021 COVID ASMR accounted for about half of total ASMR in both unvaccinated and vaccinated people, but in February COVID ASMR now accounted for about 41% of total ASMR in unvaccinated people but only about 20% of total ASMR in vaccinated people, so vaccinated people had a much bigger drop in the percentage between January and February. I think it's because in Janury 2021 many vaccinated people had been vaccinated so recently that there had not yet been enough time for the immune response to build up after vaccination. But in February 2021 a larger part of vaccinated people had been vaccinated at least 2 or 3 weeks ago so the vaccines were now more effective. So the COVID deaths increased the ratio between unvaccinated and vaccinated ASMR more in February than January:
ons=fread("https://sars2.net/f/ons.csv")[ed==7&age=="Total"]
o=ons[status%like%"accin",asmr[cause=="Deaths involving COVID-19"]/asmr[cause=="All causes"]*100,.(month,status)]
o[,.(unvaccinated=V1[1],vaccinated=V1[2]),month][,ratio:=unvaccinated/vaccinated]|>mutate_if(is.double,round,1)|>print(r=F)
# month unvaccinated vaccinated ratio
# 2021-01 47.3 53.7 0.9
# 2021-02 41.3 20.3 2.0
# 2021-03 17.9 6.0 3.0
# 2021-04 6.3 1.8 3.5
# 2021-05 2.6 0.7 3.8
# 2021-06 3.5 0.8 4.6
# 2021-07 13.5 2.6 5.3
# 2021-08 23.6 5.0 4.7
# 2021-09 22.1 6.6 3.3
# 2021-10 19.8 6.4 3.1
# 2021-11 24.7 6.5 3.8
# 2021-12 27.7 5.0 5.5
# 2022-01 32.3 10.1 3.2
# 2022-02 18.7 7.4 2.5
# 2022-03 14.9 7.5 2.0
# 2022-04 17.0 9.6 1.8
# 2022-05 8.9 4.3 2.1
Another reason why the ratio between unvaccinated and vaccinated ASMR was higher in February than January is probably because vulnerable groups of people were priorized for early vaccination.
In the next plot of Dutch CBS data if you look at the left panel which shows people who are not in long-term care, the black line shows that the ratio between unvaccinated and vaccinated ASMR was clearly elevated during the COVID wave in the winter of 2021-2022. But the black line was still higher in May 2021 than in the winter of 2021-2022, because the HVE was stronger in May 2021. You can estimate the magnitude of the HVE based on the gray line which shows the ratio of non-COVID ASMR: [rootclaim3.html#Comparison_to_Dutch_CBS_data]
From the light blue line in the next plot, you can see that in the Czech record-level datasets the ratio between unvaccinated and vaccinated non-COVID ASMR follows a fairly smooth downward curve like in the English and Dutch data. But the bright blue line shows that the ratio between unvaccinated and vaccinated all-cause ASMR is elevated in the Czech Republic in the winter of 2021-2022:
In the Substack I linked previously, Kirsch also wrote:
I did it by 10 year age groups (using a pivot on Table 2) just to make sure and here's what it looks like for 60-69:
As you can see, the ASMR jumps instantly with Omicron at a time when the unvaccinated are decreasing their mortality.
That's hard to explain how the vaccine is helping with such a huge disparity where the vaccinated start dying in droves exactly when Omicron rolls out. Coincidence?
So you'd have to argue that the data is invalid.
The healthy vaccinee effect explains why the ASMR of people with two but not more doses shoots up when the third dose was rolled out, because the healthy vaccinees moved under the third dose and people with elevated mortality risk remained under the second dose:
t=fread("http://sars2.net/f/ons.csv")[cause=="All causes"&age=="60-69"]
t=rbind(t[ed==7&month<="2021-03"],t[ed==9])
p=t[,.(dead=sum(dead),pop=sum(pop)),.(date=as.Date(paste0(month,"-1")),status=sub("(,| or).*","",status))]
p=rbind(p,a[status!="Unvaccinated",.(dead=sum(dead),pop=sum(pop),status="Ever vaccinated"),date])
p[,status:=factor(status,unique(status))]
p[,cmr:=dead/pop*1e5]
ybreak=pretty(p$cmr,6);ystart=0;yend=max(ybreak);xstart=min(p$date);xend=max(p$date)
pal=c(hcl(c(21,11,6,0,30)*10+15,90,70),"black")
frac=p[status!="Ever vaccinated",.(frac=pop/sum(pop),status),date]
ggplot(p,aes(date,cmr,color=status))+
geom_area(data=frac,aes(y=frac*yend*.9999,fill=status),linewidth=.15,show.legend=F,alpha=.3)+
annotate("rect",xmin=xstart,xmax=xend,ymin=ystart,ymax=yend,linewidth=.4,lineend="square",fill=NA,color="black")+
geom_line(linewidth=.4)+
geom_point(size=1.1)+
scale_x_date(breaks=unique(p$date)[c(T,F)],date_labels="%b\n%y")+
scale_y_continuous(limits=c(ystart,yend),breaks=seq(0,yend,1000),labels=\(i)ifelse(i<1000,i,paste0(i/1e3,"k")),sec.axis=sec_axis(trans=~./yend*100,name="Percentage of group"))+
labs(x=NULL,y="Crude mortality rate per 100,000 person-years")+
ggtitle("English ONS data, ages 60-69: All-cause CMR per 100,000 person-years")+
scale_color_manual(values=pal)+
scale_fill_manual(values=pal)+
coord_cartesian(clip="off",expand=F)+
guides(color=guide_legend(ncol=3))+
theme(axis.text=element_text(size=11,color="black"),
axis.ticks=element_line(linewidth=.4),
axis.ticks.length=unit(3,"pt"),
axis.title=element_text(size=11),
legend.background=element_blank(),
legend.box.just="center",
legend.box.spacing=unit(0,"pt"),
legend.direction="vertical",
legend.justification="center",
legend.key=element_rect(fill="white"),
legend.key.height=unit(13,"pt"),
legend.key.width=unit(26,"pt"),
legend.margin=margin(,,5),
legend.position="top",
legend.spacing.x=unit(2,"pt"),
legend.spacing.y=unit(0,"pt"),
legend.text=element_text(size=11),
legend.title=element_blank(),
panel.background=element_blank(),
plot.title=element_text(size=11,hjust=.5,face=2,margin=margin(,,3)))
ggsave("1.png",width=6,height=3.7,dpi=300*4)
system("magick 1.png -trim -resize 25% -bordercolor white -border 24 -dither none -colors 256 1.png")
In the previous plot the "Second dose" category meant people with two but not more doses. But if you look at people with two or more doses instead, their CMR remains stable even after the third dose is rolled out:
t=fread("http://sars2.net/f/ons.csv")[cause=="All causes"&age=="60-69"]
t=rbind(t[ed==7&month<="2021-03"],t[ed==9])
t[,status:=sub("(,| or).*","",status)]
t[,status:=as.integer(factor(status,unique(status)))]
t[,date:=as.Date(paste0(month,"-1"))]
p=do.call(rbind,lapply(1:5,\(i)t[if(i==1)status==1 else status>=i,.(dead=sum(dead),pop=sum(pop),status=i),date]))
p[,status:=factor(status,,c("Unvaccinated",paste(c("One","Two","Three","Four"),"or more doses")))]
p[,cmr:=dead/pop*1e5]
ybreak=pretty(p$cmr,6);ystart=0;yend=max(ybreak);xstart=min(p$date);xend=max(p$date)
pal=c(hcl(c(21,11,6,0,30)*10+15,90,70),"black")
ggplot(p,aes(date,cmr,color=status))+
annotate("rect",xmin=xstart,xmax=xend,ymin=ystart,ymax=yend,linewidth=.4,lineend="square",fill=NA,color="black")+
geom_line(linewidth=.4)+
geom_point(size=1.1)+
scale_x_date(breaks=unique(p$date)[c(T,F)],date_labels="%b\n%y")+
scale_y_continuous(limits=c(ystart,yend),breaks=seq(0,yend,1000),labels=\(i)ifelse(i<1000,i,paste0(i/1e3,"k")))+
labs(x=NULL,y=NULL)+
ggtitle("English ONS data, ages 60-69: All-cause CMR per 100,000 person-years")+
scale_color_manual(values=pal)+
scale_fill_manual(values=pal)+
coord_cartesian(clip="off",expand=F)+
guides(color=guide_legend(ncol=3))+
theme(axis.text=element_text(size=11,color="black"),
axis.ticks=element_line(linewidth=.4),
axis.ticks.length=unit(3,"pt"),
axis.title=element_text(size=11),
legend.background=element_blank(),
legend.box.just="center",
legend.box.spacing=unit(0,"pt"),
legend.direction="vertical",
legend.justification="center",
legend.key=element_rect(fill="white"),
legend.key.height=unit(13,"pt"),
legend.key.width=unit(26,"pt"),
legend.margin=margin(,,5),
legend.position="top",
legend.spacing.x=unit(2,"pt"),
legend.spacing.y=unit(0,"pt"),
legend.text=element_text(size=11),
legend.title=element_blank(),
panel.background=element_blank(),
plot.margin=margin(5,10,5,5),
plot.title=element_text(size=11,hjust=.5,face=2,margin=margin(,,3)))
ggsave("1.png",width=6,height=3.7,dpi=300*4)
system("magick 1.png -trim -resize 25% -bordercolor white -border 24 -dither none -colors 256 1.png")
Fenton et al. have hypothesized that the reason why the mortality rate of people with two but not more doses shot up when the third dose was rolled out was because people who died soon after getting the third dose were misclassified under the second dose. But if that was the case, you'd expect there to also be an increase in the mortality rate of people with two or more doses.
But that's not the case, as you can also see from the next plot where I plotted an excess crude mortality rate in ages 60-69 relative to the general population of England. It shows that in December 2021 people with three doses had about 45% lower crude mortality rate than the general population of England. So conversely people who remained under the second dose had elevated mortality in December 2021, because the total mortality of people with two or more doses remained stable. In October 2021 people with three doses had -60% excess mortality which was even lower than in December 2021, but in October the population size of people with two doses was still so large that people with two doses still had negative excess mortality. But the excess mortality of people with two doses gradually increased over the next three months as more people moved under the third dose, so that the monthly person-years of people with two doses dropped from about 432,000 in October 2021 to about 31,000 in January 2022:
download.file("https://www.ons.gov.uk/file?uri=/peoplepopulationandcommunity/birthsdeathsandmarriages/deaths/adhocs/1343dailydeathsbydateofoccurrence1stjune2014to31stmay2023bysingleyearofageengland/dailydeaths2014to2023england.xlsx","dailydeaths2014to2023england.xlsx")
library(data.table)
sum0=\(...)sum(...,na.rm=T)
kimi=\(x){e=floor(log10(ifelse(x==0,1,abs(x))));e2=pmax(e,0)%/%3+1;p=!is.na(x);x[p]=paste0(sprintf(paste0("%.",ifelse(e[p]%%3==0,1,0),"f"),x[p]/1e3^(e2[p]-1)),c("","k","M","B","T")[e2[p]]);x}
ons=fread("http://sars2.net/f/ons.csv")[cause=="All causes"&age=="60-69"]
ons=rbind(ons[ed==9],ons[ed==7&month<="2021-03"])
p=ons[,.(dead=sum0(dead),pop=sum0(pop)),.(status=sub("( or booster|,).*","",status),month)]
p=rbind(p,ons[status!="Unvaccinated",.(dead=sum0(dead),pop=sum0(pop),status="Ever vaccinated"),month])
p=rbind(p,ons[status%like%"Second|Third|Fourth",.(dead=sum0(dead),pop=sum0(pop),status="Dose 2+"),month])
p=rbind(p,ons[status%like%"Third|Fourth",.(dead=sum0(dead),pop=sum0(pop),status="Dose 3+"),month])
p[,status:=factor(status,c("Unvaccinated","First dose","Second dose","Third dose","Fourth dose","Ever vaccinated","Dose 2+","Dose 3+"))]
dead=setDT(readxl::read_excel("dailydeaths2014to2023england.xlsx",sheet=4,range="A6:CP3293"))
dead=dead[,.(date=as.Date(paste0(Year,"-",Month,"-",Day)),age=rep(0:90,each=.N),dead=unlist(.SD[,-(1:3)],,F))]
pop=fread("http://sars2.net/f/englandpop.csv")[,date:=as.Date(paste0(year,"-7-1"))]
pop=pop[,with(spline(date,pop,xout=unique(dead$date)),.(date=`class<-`(x,"Date"),pop=y)),age]
p=merge(p,merge(dead,pop)[age%in%60:69,.(basedead=sum(dead),basepop=sum(pop)/365),.(month=substr(date,1,7))])
p=rbind(p,p[,.(dead=sum(dead),pop=sum(pop),basedead=sum(basedead),basepop=sum(basepop),month="Total"),status])
p[,cmr:=dead/pop][,base:=basedead/basepop]
m=p[,tapply((cmr-base)/ifelse(cmr>base,base,cmr),.(status,month),c)]
disp=p[,tapply((cmr/base-1)*100,.(status,month),round)]
m2=p[,xtabs(pop~status+month)]
max1=9;max2=max(m2[,-ncol(m2)]);exp1=.6;exp2=1
m[m==-Inf]=-max1
pal1=hsv(rep(c(21/36,0),4:5),c(1,.8,.6,.3,0,.3,.6,.8,1),c(.3,.65,1,1,1,1,1,.65,.3))
pal2=sapply(seq(1,0,,256),\(i)rgb(i,i,i))
pheatmap::pheatmap(sign(m)*abs(m)^exp1,filename=paste0("i1.png"),display_numbers=disp,
cluster_rows=F,cluster_cols=F,legend=F,cellwidth=19,cellheight=19,fontsize=9,fontsize_number=8,
border_color=NA,na_col="gray90",number_color=ifelse(abs(m)^exp1>max1^exp1*.53,"white","black"),
breaks=seq(-max1^exp1,max1^exp1,,256),colorRampPalette(pal1)(256))
pheatmap::pheatmap(m2^exp2,filename=paste0("i2.png"),display_numbers=ifelse(m2<10,m2,kimi(m2)),
cluster_rows=F,cluster_cols=F,legend=F,cellwidth=19,cellheight=19,fontsize=9,fontsize_number=8,
border_color=NA,na_col="gray90",number_color=ifelse(m2^exp2>max2^exp2*.42,"white","black"),
breaks=seq(0,max2^exp2,,256),pal2)
sub="\u00a0 Source: ONS datasets \"Deaths by vaccination status, England\", \"Daily deaths by date of occurrence, 1st June 2014 to 31st May 2023 by single year of age, England\", and \"Estimates of the population for the UK, England, Wales, Scotland, and Northern Ireland\".
Data for the first three months of 2021 is taken from the July 2022 edition of the ONS dataset which was linked against the 2011 census, and data for other months is taken from the August 2023 edition which was linked against the 2021 census.
The ONS dataset for mortality by vaccination status was released in August 2023, and the dataset for deaths in the general English population was released in July 2023, so both datasets are missing some deaths because of a registration delay in 2023 (but the dataset for deaths in the general population is missing a larger percentage of deaths in May 2023, which explains why ever vaccinated people get positive excess mortality here in May 2023)."
system(paste0("mogrify -trim i[12].png;w=`identify -format %w i1.png`;magick -size $w \\( -pointsize 50 -font Arial-Bold caption:'ONS dataset for mortality by vaccination status, ages 60-69' -splice x10 \\) -pointsize 45 -font Arial \\( caption:'Excess CMR percent relative to CMR the same month among the general population of England' i1.png -splice x24 \\) \\( caption:'Population size in person-years' i2.png -splice x24 \\) \\( -pointsize 42 -font Arial caption:'",gsub("'","'\\\\'",sub),"' -splice x20 \\) -append -trim -bordercolor white -border 32 1.png"))
So Kirsch was wrong that "the vaccinated start dying in droves exactly when Omicron rolls out". The total mortality rate of vaccinated people remained roughly stable even after Omicron emerged, but Kirsch's table only showed that the mortality rate of people with two but not more doses went up in December 2021. But it wasn't because of Omicron but because the third dose was rolled out.