Comments about COVID statistics (part 2)

The authors seem to have calculated the number of cardiac arrest deaths by multiplying the figure for the "number of cardiac arrests for which ALS resuscitation efforts were attempted" with the rounded percentage of people who didn't survive until they were discharged from hospital alive, which gave them a figure of 1121 deaths in 2020 (from 1350*(100-17)/100)). However the authors could've derived the exact figure by simply subtracting the number of patients who survived to hospital discharge from the number of treated patients (1350-234 = 1116).

The authors also incorrectly entered the survival rate in 2021 as 18% even though it should've been 16%, but I didn't find other errors in their Table 1 when I compared it to the original PDF reports:

The cells for 2023 above are marked with an asterisk because Hulscher et al. calculated them as a linear projection of the data for 2021 and 2022. In the real data for 2023 that was published in September 2024, the number of deaths was slightly lower than Hulscher's projected number of deaths for 2023:

I think the authors should've just omitted 2023 from their paper instead of including projected data for 2023, or in their plots they should've somehow visually differentiated the projected data from the actual data. Many people tweeted a screenshot of one of their figures where it wasn't indicated anywhere that the data for 2023 was projected.

The original PDF reports go back to 2003 but reports for data before 2005 are missing the table which shows the number of treated and survived cardiac patients. Hulscher et al. only included data from 2015 onwards, which might be because there was a big jump down in the number of deaths between 2014 and 2015. I didn't find any explanation for the jump in the report for 2016 which featured the data for 2015. I don't know if it's because there was a high number of deaths in 2014, or if for example before 2015 the King County EMS used to serve more parts of the Seattle metropolitan area outside of King County, or if there was a change to case definition.

I'm not a fan of using z-scores to express the level of excess deaths, because people don't know how to interpret them correctly, and I think it's clearer to just express excess deaths as a percentage of the baseline. The z-scores can often change drastically depending on the range of years used for the baseline fitting period. In the King County cardiac arrest data the years 2015 to 2020 all fall close to the baseline but 2014 is far from the baseline, so here when I used a 2014-2020 baseline instead of a 2015-2020 baseline, my z-score for excess deaths in 2021 fell from about 13.2 to about 4.7:

But anyway, the long-term trend in the number of deaths in the EMS data seems to be curved upwards, so here my excess number of deaths relative to the 2015-2020 linear baseline was even higher in 2010 than in 2021 or 2022:

	Hulscher Table 1	Original PDFs
Year	Treated	Survived	Dead	Treated	Survived	Dead
2005				1124	190 (17%)	934
2006				993	174 (18%)	819
2007				1035	191 (18%)	844
2008				1046	199 (19%)	847
2009				1072	206 (19%)	866
2010				1069	218 (20%)	851
2011				1047	223 (21%)	824
2012				1134	252 (22%)	882
2013				1135	235 (21%)	900
2014				1246	267 (21%)	979
2015	1114	20%	891	1114	221 (20%)	893
2016	1228	24%	933	1228	288 (24%)	940
2017	1215	21%	960	1215	251 (21%)	964
2018	1298	22%	1012	1298	289 (22%)	1009
2019	1308	19%	1059	1308	253 (19%)	1055
2020	1350	17%	1121	1350	234 (17%)	1116
2021	1499	18%	1229	1499	242 (16%)	1257
2022	1598	18%	1310	1588	288 (18%)	1300
2023	1697 *	18% *	1392 *	1669	311 (19%)	1358

library(data.table);library(ggplot2)

d=data.table(x=2005:2023)
d$y=c(934,819,844,847,866,851,824,882,900,979,893,940,964,1009,1055,1116,1257,1300,1358)
d$z="Actual deaths"

p1=d[,.(x,y=d[x%in%2015:2020,predict(lm(y~x),d)],z="2015-2020 linear trend")]
p2=d[,.(x,y=d[x%in%2006:2019,predict(lm(y~poly(x,2)),d)],z="2006-2019 quadratic trend")]
p=rbind(d,p1,p2)[,z:=factor(z,unique(z))]

ybreak=pretty(p$y,9);xstart=min(p$x);xend=max(p$x)
color=c("black",hsv(22/36,1,.8),"#00aa00")

ggplot(p,aes(x,y))+
geom_line(aes(color=z),linewidth=.3)+
geom_point(aes(alpha=z,color=z),stroke=0,size=.9)+
geom_line(data=p[z==z[1]],aes(color=z),linewidth=.3)+
labs(x=NULL,y=NULL,title="Yearly cardiac arrest deaths in King County EMS data")+
scale_x_continuous(limits=c(xstart-.5,xend+.5),breaks=xstart:xend)+
scale_y_continuous(limits=range(ybreak),breaks=ybreak)+
scale_color_manual(values=color)+
scale_alpha_manual(values=c(1,0,0))+
coord_cartesian(clip="off",expand=F)+
theme(axis.text=element_text(size=7,color="black"),
  axis.text.x=element_text(angle=90,vjust=.5,hjust=1),
  axis.ticks=element_line(linewidth=.3,color="black"),
  axis.ticks.length=unit(3,"pt"),
  axis.ticks.length.x=unit(0,"pt"),
  legend.background=element_blank(),
  legend.box.background=element_rect(linewidth=.3),
  legend.justification=c(0,1),
  legend.key=element_blank(),
  legend.key.height=unit(9,"pt"),
  legend.key.width=unit(17,"pt"),
  legend.margin=margin(4,4,4,4),
  legend.position=c(0,1),
  legend.spacing.x=unit(1,"pt"),
  legend.spacing.y=unit(0,"pt"),
  legend.text=element_text(size=7,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.3),
  panel.grid.major=element_blank(),
  plot.margin=margin(5,5,5,5),
  plot.subtitle=element_text(size=7,margin=margin(,,4)),
  plot.title=element_text(size=7.6,face=2,margin=margin(1,,4)))
ggsave("1.png",width=3.5,height=2.4,dpi=400*4)
system("mogrify -resize 25% 1.png")

The long-term trend in all-cause deaths in King County seems to be similarly curved upwards, even though the trend in the 10 years before COVID seemed roughly linear:

library(data.table);library(ggplot2)

t=fread("http://sars2.net/f/wonderkingcountyallcause.csv")
t=merge(t[year%in%2010:2019,.(year=1999:2023,base=predict(lm(dead/pop~year),.(year=1999:2023))),age],t)

a=t[,.(dead=sum(dead)),year]
a$linear=a[year%in%2015:2019,predict(lm(dead~year),a)]
a$quadra=a[year<2020,predict(lm(dead~poly(year,2)),a)]

p=a[,.(x=year,y=unlist(.SD[,-1]),z=rep(c("Actual deaths","2015-2019 linear baseline","1999-2019 quadratic trend"),each=.N))]
p[,z:=factor(z,unique(z))]

ybreak=pretty(p$y,9);xstart=1999;xend=2023
color=c("black",hsv(22/36,1,.8),"#00aa00")

ggplot(p,aes(x,y))+
geom_line(aes(color=z),linewidth=.3)+
geom_point(aes(alpha=z,color=z),stroke=0,size=.9)+
geom_line(data=p[z==z[1]],aes(color=z),linewidth=.3)+
labs(x=NULL,y=NULL,title="Yearly deaths in King County, Washington")+
scale_x_continuous(limits=c(xstart-.5,xend+.5),breaks=xstart:xend)+
scale_y_continuous(limits=range(ybreak),breaks=ybreak)+
scale_color_manual(values=color)+
scale_alpha_manual(values=c(1,0,0))+
coord_cartesian(clip="off",expand=F)+
theme(axis.text=element_text(size=7,color="black"),
  axis.text.x=element_text(angle=90,vjust=.5,hjust=1),
  axis.ticks=element_line(linewidth=.3,color="black"),
  axis.ticks.length=unit(3,"pt"),
  axis.ticks.length.x=unit(0,"pt"),
  legend.background=element_blank(),
  legend.box.background=element_rect(linewidth=.3),
  legend.justification=c(0,1),
  legend.key=element_blank(),
  legend.key.height=unit(9,"pt"),
  legend.key.width=unit(17,"pt"),
  legend.position=c(0,1),
  legend.margin=margin(4,4,4,4),
  legend.spacing.x=unit(1,"pt"),
  legend.spacing.y=unit(0,"pt"),
  legend.text=element_text(size=7,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.3),
  panel.grid.major=element_blank(),
  plot.margin=margin(5,5,5,5),
  plot.subtitle=element_text(size=7,margin=margin(,,4)),
  plot.title=element_text(size=7.6,face=2,margin=margin(1,,4)))
ggsave("1.png",width=3.5,height=2.7,dpi=400*4)
system("mogrify -resize 25% 1.png")

Excess deaths may have been due to COVID

The EMS data included all cardiac arrest deaths regardless of cause, so it's similar to deaths by multiple cause of death at CDC WONDER. So part of the excess deaths in 2021 and 2022 might be due to COVID, because from the next plot which shows monthly cardiac deaths in Washington State, you can see that spikes in MCD cardiac deaths coincided with COVID waves, and the number of excess MCD cardiac deaths got much lower when I subtracted deaths that also had MCD COVID:

library(data.table);library(ggplot2);library(lubridate);library(ggtext)

cul=\(x,y)y[cut(x,c(y,Inf),,T,F)]

t=fread("http://sars2.net/f/wonderwashingtoncardiac.csv")[age>=45]

pop=fread("http://sars2.net/f/uspopdeadmonthly.csv")
t=merge(t,pop[,.(pop=sum(pop)),.(month=date,age=cul(age,seq(45,85,10)))])

t=t[,month:=as.Date(paste0(month,"-1"))]
t[,dead:=dead/days_in_month(month)]

months=sort(unique(t$month))
base=t[!grepl("COVID",cause)&year(month)%in%2011:2019][,dead:=nafill(dead,,0)]
t=merge(base[,.(month=months,base=predict(lm(dead/pop~month),.(month=months))),.(age,cause)],t)
t[,base:=base*pop]

p=t[,.(base=sum(base),dead=sum(dead)),.(date=month,cause)]

and=fread("http://sars2.net/f/wonderwashingtoncardiacandcovid.csv")
and=and[,date:=as.Date(paste0(month,"-1"))][,.(date,andcovid=dead/days_in_month(date))]
p=rbind(p,merge(and,p[cause%like%"Multiple"])[,.(date,base=NA,cause="MCD I00-I99 and not MCD COVID",dead=dead-andcovid)])

p[,cause:=factor(cause,unique(cause)[c(2,1,3)])]

xstart=as.Date("2015-1-1");xend=as.Date("2024-1-1")
p=p[date>=xstart&date<=xend]
xbreak=seq(xstart,xend,"6 month");xlab=c(rbind("",2015:2023),"")
ystart=0;yend=p[,max(base,dead,0,na.rm=T)*1.02];ybreak=pretty(p[,c(base,dead,0)],8)

color=c("black","#ff2222","#ffaaaa")
lab=c("Actual deaths","Baseline");lab=factor(lab,unique(lab))

ggplot(p,aes(x=date+14,y=dead,color=cause))+
geom_vline(xintercept=seq(xstart,xend,"year"),color="gray85",linewidth=.25)+
geom_hline(yintercept=c(ystart,yend),linewidth=.25,lineend="square")+
geom_vline(xintercept=c(xstart,xend),linewidth=.25,lineend="square")+
geom_line(linewidth=.3)+
geom_point(stroke=0,size=.6,show.legend=F)+
geom_line(aes(y=base),linetype="42",linewidth=.3)+
labs(title="Washington State, ages 45+: Monthly deaths with cause I00-I99 (diseases of\nthe circulatory system) divided by number of days in month",x=NULL,y=NULL)+
scale_x_continuous(limits=c(xstart,xend),breaks=xbreak,labels=xlab)+
scale_y_continuous(limits=c(ystart,yend),breaks=ybreak)+
scale_color_manual(values=color)+
scale_linetype_manual(values=c("solid","42"),labels=lab)+
coord_cartesian(clip="off",expand=F)+
theme(axis.text=element_text(size=7,color="black"),
  axis.ticks.x=element_line(color=alpha("black",c(1,0))),
  axis.ticks=element_line(linewidth=.25,color="black"),
  axis.ticks.length=unit(.2,"lines"),
  axis.title=element_text(size=8),
  legend.background=element_blank(),
  legend.box.just="left",
  legend.spacing=unit(5,"pt"),
  legend.justification=c(0,0),
  legend.key=element_blank(),
  legend.key.height=unit(9,"pt"),
  legend.key.width=unit(17,"pt"),
  legend.position=c(0,0),
  legend.spacing.x=unit(2,"pt"),
  legend.spacing.y=unit(0,"pt"),
  legend.box.background=element_rect(fill="white",color="black",linewidth=.3),
  legend.margin=margin(4,4,4,4,"pt"),
  legend.text=element_text(size=7,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  plot.margin=margin(5,5,5,5),
  plot.title=element_text(size=7.6,face=2,margin=margin(,,4)))
ggsave("1.png",width=4.2,height=2.8,dpi=400*4)

sub="\u00a0     Source: wonder.cdc.gov/mcd.html. Age groups below 45 were excluded because they had some months with less than 10 deaths so the number of deaths was suppressed.
      In order to calculate the dashed baseline, first a linear trend was calculated for CMR within each 10-year age group in 2021 to 2019, and then the projected trend was multiplied by the population size of each age group to get the monthly expected deaths for each age group.
      Monthly resident population estimates were downloaded from www2.census.gov/programs-
surveys/popest/datasets/2020-2023/national/asrh/nc-est2022-alldata-r-file0{1..8}.csv, and from the corresponding files in the 2010-2020 directory. In order to get rid of a sudden jump in population size after the switch from the 2010-2020 to 2020-2023 estimates, the 2010-2020 estimates were multiplied by a linear slope so that they matched the 2020-2023 estimates at the point where the estimates were merged in April 2020 (see sars2.net/ethical.html#Files_
for_US_population_estimates_by_single_year_of_age_up_to_ages_100)."
system(paste0("f=1.png;magick 1.png -resize 25% 1.png;mar=32;w=`identify -format %w $f`;magick \\( $f -gravity southwest -chop 0x20 \\) \\( -size $[w-mar*2]x -font Arial -interline-spacing -3 -pointsize 38 caption:'",sub,"' -gravity southwest -splice $[mar]x20 \\) -append 1.png"))

Washington State had a low number of COVID deaths in 2020

The paper by Hulscher et al. showed that there was a much higher number of excess deaths in 2021 than 2020. However compared to other US states, Washington State had an unusually low number of COVID deaths in 2020 relative to 2021 and 2022, and the PCR positivity rate in Washington State was also low in 2020:

Washington State is one of the states with the highest percentage of UCD COVID deaths in 2021 and 2022 relative to UCD COVID deaths in 2020:

This shows how in the year 2020 deaths with UCD COVID accounted for only about 5% of all deaths in Washington State, which was one of the lowest percentages out of any state:

library(data.table);library(ComplexHeatmap);library(circlize)

t=fread("http://sars2.net/f/wonderstatecovidvsall.csv")[year>=2020][,year:=factor(year,unique(year))]
t=rbind(t[,.(dead=sum(dead),year="Total"),.(cause,state)],t)
t=rbind(t,t[,.(dead=sum(dead),state="Total"),.(cause,year)])

me=merge(t[cause=="covid",.(year,state,covid=dead)],t[cause=="all",.(year,state,dead)])

m=me[,xtabs(covid/dead*100~year+state)]

states=fread("https://github.com/cphalpert/census-regions/raw/master/us%20census%20bureau%20regions%20and%20divisions.csv")
states[,order:=match(Division,strsplit("Total,New England,Middle Atlantic,East North Central,West North Central,South Atlantic,East South Central,West South Central,Mountain,Pacific",",")[[1]])]
states[,Division:=sub("([WE]).* (South Central)","\\1 \\2",Division)]
div=rbind(states[order(order),.(State,Division)],list(State="Total",Division=""))

m=m[,div$State];disp=round(m)

colnames(m)[colnames(m)=="District of Columbia"]="DC"

png("0.png",w=ncol(m)*32+2000,h=nrow(m)*32+1000,res=120)
ht_opt$COLUMN_ANNO_PADDING=unit(0,"mm");ht_opt$ROW_ANNO_PADDING=unit(0,"mm")

Heatmap(m,
  column_split=factor(div$Division,unique(div$Division)),
  row_split=rep(letters[1:2],c(1,5)),
  row_title=NULL,
  column_title_gp=gpar(fontsize=16),
  column_gap=unit(8,"pt"),
  row_gap=unit(8,"pt"),
  border="gray60",
  width=unit(ncol(m)*32,"pt"),
  height=unit(nrow(m)*32,"pt"),
  show_column_names=F,
  show_row_names=F,
  cluster_columns=F,
  cluster_rows=F,
  show_heatmap_legend=F,
  rect_gp=gpar(col="gray80",lwd=0),
  top_annotation=columnAnnotation(text=anno_text(gt_render(colnames(m),padding=unit(c(3,3,3,3),"mm")),just="left",rot=90,location=unit(0,"npc"),gp=gpar(fontsize=17,border="gray60",lwd=1))),
  left_annotation=rowAnnotation(text=anno_text(gt_render(rownames(m),padding=unit(c(3,3,3,3),"mm")),just="right",location=unit(1,"npc"),gp=gpar(fontsize=17,border="gray60",lwd=1))),
  right_annotation=rowAnnotation(text=anno_text(gt_render(rownames(m),padding=unit(c(3,3,3,3),"mm")),just="left",location=unit(0,"npc"),gp=gpar(fontsize=17,border="gray60",lwd=1))),
  col=colorRamp2(seq(0,max(m),,256),sapply(seq(1,0,,256),\(i)rgb(i,i,i))),
  cell_fun=\(j,i,x,y,w,h,fill)grid.text(round(disp[i,j]),x,y,gp=gpar(fontsize=16,col=ifelse(abs(m[i,j])>=.45*max(m),"white","black"))))
dev.off()

system("mogrify -trim 0.png;magick -gravity north -font Arial-Bold \\( -size `identify -format %w 0.png`x -pointsize 34 caption:'CDC WONDER: Percentage of deaths with underlying cause COVID out of all deaths' \\) \\( 0.png -splice 0x18 \\) -append -trim -bordercolor white -border 20 1.png")

Decrease in population estimates between 2020 and 2021

The paper by Hulscher et al. also included this plot which showed that the mid-year resident population estimates of King County decreased by about 0.94% between 2020 and 2021:

dlf=\(x,y,...){if(missing(y))y=sub(".*/","",x);for(i in 1:length(x))download.file(x[i],y[i],quiet=T,...)}
dlf("https://www2.census.gov/programs-surveys/popest/tables/2020-2023/counties/totals/co-est2023-pop-53.xlsx")
dlf("https://www2.census.gov/programs-surveys/popest/tables/2010-2020/intercensal/county/co-est2020int-pop-53.xlsx")
dlf("https://www2.census.gov/programs-surveys/popest/tables/2010-2019/counties/totals/co-est2019-annres-53.xlsx")
dlf("https://www2.census.gov/programs-surveys/popest/tables/2000-2010/intercensal/county/co-est00int-01-53.csv")
dlf("https://www2.census.gov/programs-surveys/popest/tables/2000-2009/counties/totals/co-est2009-01-53.csv")
dlf("https://www2.census.gov/programs-surveys/popest/tables/2000-2010/intercensal/county/co-est00int-01-53.csv")

library(data.table);library(ggplot2);library(readxl)

agecut=\(x,y)cut(x,c(y,Inf),paste0(y,c(paste0("-",y[-1]-1),"+")),T,F)

kim=\(x)ifelse(x>=1e3,ifelse(x>=1e6,paste0(x/1e6,"M"),paste0(x/1e3,"k")),x)

pop1=fread("co-est2009-01-53.csv",skip=1)
p=pop1[V1==".King County",.(year=2009:2000,pop=as.integer(gsub(",","",.SD)),z="2000-2009"),.SDcols=2:11]

pop2=fread("co-est00int-01-53.csv",skip=1)
p=rbind(p,pop2[V1==".King County",.(year=2000:2010,pop=as.integer(gsub(",","",.SD)),z="2000-2010 intercensal"),.SDcols=c(3:12,14)])

pop3=read_excel("co-est2019-annres-53.xlsx")
p=rbind(p,list(year=2010:2019,pop=as.integer(dplyr::filter(pop3,pop3[[1]]==".King County, Washington")[4:13]),z="2010-2019"))

pop4=read_excel("co-est2020int-pop-53.xlsx")
p=rbind(p,list(year=2010:2019,pop=as.integer(pop4[pop4[[1]]%like%"King",3:12]),z="2010-2020 intercensal"))

pop5=read_excel("co-est2023-pop-53.xlsx")
p=rbind(p,list(year=2020:2023,pop=as.integer(pop5[pop5[[1]]%like%"King",][3:6]),z="2020-2023"))

p=rbind(p,data.table(year=1999:2023,pop=c(1729058,1737034,1754090,1758685,1763440,1775297,1795268,1822967,1847986,1875020,1912012,1931249,1969722,2007440,2044449,2079967,2117125,2149970,2188649,2233163,2252782,2274315,2252305,2266789,2266789),z="CDC WONDER"))

p=rbind(p,data.table(year=2000:2022,pop=c(1737045.79,1755487.08,1777514.01,1788082.03,1800783.04,1814999.24,1845208.98,1871097.95,1891124.97,1909204.99,1931249,1943608.44,1959305.04,1985687.59,2022469.68,2059056.02,2111487.04,2160624.39,2198063.56,2234581.41,2269674.95,2287050.05,2317700.05),z="DoH PIE"))

p=rbind(p,list(year=2020:2023,pop=c(2269675,2287050,2317700,2347800),z="ofm.wa.gov"))

# p=rbind(p,list(year=2015:2022,pop=c(2.127,2.168,2.205,2.228,2.250,2.274,2.252,2.265)*1e6,z="Hulscher et al."))

p[,z:=factor(z,unique(z))]

xstart=2000;xend=2023
p=p[year%in%xstart:xend]

ybreak=pretty(p$pop);ystart=ybreak[1];yend=max(ybreak);ylab=sprintf("%.1fM",ybreak/1e6)

ggplot(p,aes(x=year,y=pop))+
coord_cartesian(clip="off",expand=F)+
geom_vline(xintercept=c(2009.5,2019.5),color="gray80",linewidth=.23)+
geom_line(aes(color=z,alpha=z),linewidth=.3)+
geom_point(aes(color=z,shape=z,size=z),stroke=.3)+
labs(title="Resident population estimates of King County, Washington",subtitle="Sources: www2.census.gov/programs-surveys/popest/tables, CDC WONDER,
phskc-apde.shinyapps.io/PopPIE (Washington State Department of Health
population interim estimates; based on OFM in 2020-2022), ofm.wa.gov/
washington-data-research/population-demographics/population-estimates/
estimates-april-1-population-age-sex-race-and-hispanic-origin. OFM and PIE
are estimates on April 1st but other sources are mid-year estimates.",x=NULL,y=NULL)+
scale_x_continuous(limits=c(xstart-.5,xend+.5),breaks=seq(xstart-.5,xend+.5,.5),labels=c(rbind("",xstart:xend),""))+
scale_y_continuous(labels=ylab,breaks=ybreak,limits=c(ystart,yend))+
scale_color_manual(values=c(hsv(12/36,c(.8,1),c(.8,.4)),hsv(22/36,c(.8,1),c(1,.5)),hsv(30/36,.8,1),"black",hsv(4/36,.5,.6),"red"))+
scale_shape_manual(values=c(16,16,16,16,16,4,1,3))+
scale_alpha_manual(values=c(1,1,1,1,1,0,0,0))+
scale_size_manual(values=c(.7,.7,.7,.7,.7,1,1,1))+
guides(color=guide_legend(nrow=3,byrow=F))+
theme(axis.text=element_text(size=7,color="black"),
  axis.text.x=element_text(angle=90,vjust=.5,hjust=1),
  axis.ticks=element_line(linewidth=.2,color="black"),
  axis.ticks.length=unit(3,"pt"),
  axis.ticks.length.x=unit(0,"pt"),
  legend.background=element_blank(),
  legend.box.spacing=unit(0,"pt"),
  legend.justification="left",
  legend.key=element_blank(),
  legend.key.height=unit(8,"pt"),
  legend.key.width=unit(17,"pt"),
  legend.margin=margin(,,4),
  legend.position="top",
  legend.spacing.x=unit(2,"pt"),
  legend.spacing.y=unit(1,"pt"),
  legend.text=element_text(size=7,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.23),
  plot.margin=margin(5,5,5,5),
  plot.subtitle=element_text(size=6.8,margin=margin(,,3)),
  plot.title=element_text(size=7.3,face=2,,margin=margin(1,,4)))
ggsave("1.png",width=3.8,height=2.9,dpi=350*4)
system("magick 1.png -resize 25% 1.png")

I don't know if the OFM population estimates are more accurate than the Census Bureau's population estimates or not. At first I thought that a large decrease in population size between mid-2020 and mid-2021 wasn't supported by data for births minus deaths plus net migration. On CDC WONDER I got 23,256 births and 14,779 deaths between July 2020 and June 2021 among residents of King County. [https://wonder.cdc.gov/natality-expanded-current.html, https://wonder.cdc.gov/mcd.html]

There's a discontinued Census Bureau dataset for migration by county which includes data up to the year 2020, where King County had an estimated net migration of -6,437 people in the year 2020: [https://fred.stlouisfed.org/series/NETMIGNACS053033]

Assuming the same net migration between mid-2020 and mid-2021, 23256-14779-6437 would give 2,040 more people in mid-2021 than mid-2020, which is a much smaller increase than in the OFM population estimates where there's 17,375 more people in 2021 than 2020. (The OFM population estimates are for April 1st and not July 1st, but it shouldn't make much difference when you're comparing the difference between two adjacent years.)

However later I found that the Census Bureau has published net migration estimates by county here in the file co-est2021-comp-53.xlsx: https://www2.census.gov/programs-surveys/popest/tables/2020-2021/counties/totals/. In the file King County had a total population change of -20,266 between July 1st 2020 and July 1st 2021, where there were 23,082 births, 16,444 deaths, and net migration of -26,818. So if the net migration estimate is accurate, then it would explain a large decrease in population size between mid-2020 and mid-2021.

1236% increase in excess deaths

The paper said: "Excess cardiopulmonary arrest deaths were estimated to have increased by 1,236% from 2020 to 2023, rising from 11 excess deaths (95% CI: -12, 34) in 2020 to 147 excess deaths (95% CI: 123, 170) in 2023." I think it was misleading that McCullough's headline was not even based on actual data but on a projected number of deaths for 2023.

But anyway there were about 11.8% excess deaths in 2023 (1392/1245-1) and about 1.0% excess deaths in 2020 (1121/1110-1). So I guess an increase of 1236% sounds more exciting than an increase of 10.8 percentage points:

library(data.table);library(ggplot2)

d=data.table(year=2005:2023)
d$dead=c(934,819,844,847,866,851,824,882,900,979,893,940,964,1009,1055,1116,1257,1300,1358)
d$dead2=c(rep(NA,10),891,933,960,1012,1059,1121,1229,1310,1392)
d$base=c(rep(NA,10),884,929,974,1019,1064,1110,1155,1200,1245)

p=d[,.(x=year,y=c(dead,dead2,base),z=rep(c("Original PDFs","McCullough paper","McCullough baseline"),each=.N))]
p[,z:=factor(z,unique(z))]

anno=p[x%in%c(2020,2023)&z!=z[1]]
minus=4
seg=as.matrix(anno[,.(x=x-minus,xend=x,y,yend=y)])
seg=rbind(seg,t(matrix(t(anno[order(x)][,x:=x-minus][,-3]),4))[,c(1,3,2,4)])

lab=cbind(anno[1:2,1:2],base=anno[3:4][[2]])
lab=lab[,.(x,y=(y+base)/2,label=paste0("(",y," - ",base,") / ",base," ≈ ",sprintf("%.1f",(y/base-1)*100),"% excess deaths"))]

note="(1392/1245-1)-(1121/1110-1): increase of about 10.8 percentage points
(1392-1245)/(1121-1110)-1: increase of about 1236% (misleading headline)"

xstart=2005;xend=2023;ybreak=pretty(c(0,p$y),7);ystart=0;yend=max(ybreak)
xbreak=seq(xstart-.5,xend+.5,.5);xlab=ifelse(xbreak%%1==0,xbreak,"")

color=c("black","#ff5555","#ff5555")

cap="Sources: kingcounty.gov/en/dept/dph/health-safety/health-centers-programs-services/emergency-
medical-services/reports-publications, Hulscher et al. 2024, \"Excess Cardiopulmonary Arrest and
Mortality after COVID-19 Vaccination in King County, Washington\"."

sub=str_wrap("The number of deaths in the paper by McCullough foundation is slightly different from the original PDFs because in the paper the number of dead patients was calculated by multiplying the number of treated patients by the rounded percentage of patients who didn't survive, and not by subtracting the number of patients who didn't survive from the number of treated patients. In the paper the number of deaths in 2023 was a linear projection of the number of deaths in 2021 and 2022, but the actual number of deaths in 2023 which was published in September 2024 was slightly lower than the projection. The McCullough paper also erroneously used 18% instead of 16% as the percentage of survived patients in 2021.",80)

ggplot(p,aes(x,y))+
geom_hline(yintercept=c(0,yend),linewidth=.3,lineend="square")+
geom_vline(xintercept=c(xstart-.5,xend+.5),linewidth=.3,lineend="square")+
geom_line(aes(linewidth=z,color=z,linetype=z))+
geom_segment(data=as.data.frame(seg),aes(x=x,xend=xend,y=y,yend=yend),lineend="square",linewidth=.3,color=color[2])+
geom_text(data=lab,aes(x=x-minus-.2,label=label),hjust=1,color=color[2],size=2.4)+
annotate(geom="label",x=2023,y=500,label=note,hjust=1,color=color[2],label.r=unit(0,"pt"),label.padding=unit(4,"pt"),label.size=.3,size=2.4)+
annotate(geom="label",x=2023,y=500,label=note,hjust=1,fill="transparent",label.r=unit(0,"pt"),label.padding=unit(4,"pt"),label.size=0,size=2.4)+
geom_point(aes(color=z,shape=z,alpha=z,size=z),stroke=.4)+
labs(x=NULL,y=NULL,title="Yearly cardiac arrest deaths in King County EMS data",subtitle=sub,caption=cap)+
scale_x_continuous(limits=c(xstart-.5,xend+.5),breaks=xbreak,labels=xlab)+
scale_y_continuous(limits=c(ystart,yend),breaks=ybreak)+
coord_cartesian(clip="off",expand=F)+
scale_color_manual(values=color)+
scale_linetype_manual(values=c("solid","solid","42"))+
scale_linewidth_manual(values=c(.4,.4,.4))+
scale_shape_manual(values=c(16,16,16))+
scale_size_manual(values=c(.9,.9,1.4))+
scale_alpha_manual(values=c(1,1,0))+
theme(axis.text=element_text(size=7,color="black"),
  axis.text.x=element_text(angle=90,vjust=.5,hjust=1),
  axis.ticks.length=unit(0,"pt"),
  legend.background=element_blank(),
  legend.box.spacing=unit(0,"pt"),
  legend.direction="horizontal",
  legend.justification=c(1,.5),
  legend.key=element_blank(),
  legend.key.height=unit(10,"pt"),
  legend.key.width=unit(18,"pt"),
  legend.margin=margin(1,,4),
  legend.position="top",
  legend.spacing.x=unit(1.5,"pt"),
  legend.spacing.y=unit(0,"pt"),
  legend.text=element_text(size=7),
  legend.title=element_blank(),
  panel.background=element_blank(),
  plot.margin=margin(4,4,4,4),
  plot.subtitle=element_text(size=6.5,margin=margin(,,3)),
  plot.caption=element_text(size=5.7,margin=margin(4),hjust=0),
  plot.title=element_text(size=7.5,face="bold",margin=margin(1,,4)))
ggsave("1.png",width=3.8,height=3.8,dpi=400*4)
system("magick 1.png -resize 25% 1.png")

The paper was published in a journal of the predatory publisher Longdom

Were there 98% vaccinated people in 2023?

The paper by Hulscher et al. said: "Approximately 98% of the King County population received at least one dose of a COVID-19 vaccine by 2023." The authors took the percentage of vaccinated people from this website: https://data.tennessean.com/covid-19-vaccine-tracker/washing-ton/king-county/53033/. I got an error when I tried to access the website outside of the US, but I was able to access it with a free US VPN server from Proton VPN: https://protonvpn.com. The number of vaccinated people in the dataset seems to be identical to this CDC dataset: https://data.cdc.gov/Vaccinations/COVID-19-Vaccinations-in-the-United-States-County/8xkx-amqh. However the percentage of vaccinated people out of the total population is different. For example on January 4th 2022 the Tennessean dataset has 1,903,416 vaccinated people which matches the CDC data, but it has 87.99% vaccinated people which doesn't match the CDC data:

I estimated that Tennessean uses a fixed population size of about 2163216 for King County by taking the average value across three different dates for the number of vaccinated people divided by the proportion of vaccinated people:

Here you can see how both the Census Bureau and OFM population estimates are higher than the Tennessean population estimate:

Date	Vaccinated people	Vaccinated percent	Estimated population size (from people divided by percent)
2021-03-22	510930	23.62	2163124
2022-01-04	1903416	87.99	2163219
2022-12-07	2120471	98.02	2163304

library(data.table);library(ggplot2);library(ggtext)

xstart=as.Date("2020-12-1");xend=as.Date("2023-7-1")

t=fread("d/cd/COVID-19_Vaccinations_in_the_United_States_County_20240227.csv.gz")
p=t[FIPS==53033,.(x=as.Date(Date,"%m/%d/%Y"),y=Administered_Dose1_Recip,z="Vaccinated people (CDC)")]
p=rbind(p,t[FIPS==53033,.(x=seq(xstart,xend,1),y=Census2019[1],z="Mid-2019 population (CDC)")])
p=rbind(p,p[,.(x=seq(xstart,xend,1),y=2163216,z="Tennessean population")])

pop=data.table(z="OFM population",x=as.Date(paste0(2020:2023,"-4-1")),y=c(2269675,2287050,2317700,2347800))
pop=rbind(pop,data.table(z="Census Bureau population",x=as.Date(paste0(2020:2023,"-7-1")),y=c(2274282,2252980,2265311,2271380)))
p=rbind(p,pop)

p[,z:=factor(z,unique(z))]
xbreak=seq(as.Date("2021-1-1"),xend,"year");xlab=2021:2023
ystart=0;ybreak=pretty(p$y,10);yend=max(ybreak)

cap="Sources:
data.cdc.gov/Vaccinations/COVID-19-Vaccinations-in-the-United-States-County/8xkx-amqh
data.tennessean.com/covid-19-vaccine-tracker/washing-ton/king-county/53033/
ofm.wa.gov/washington-data-research/population-demographics/population-estimates/
    estimates-april-1-population-age-sex-race-and-hispanic-origin
www2.census.gov/programs-surveys/popest/tables/2020-2023/counties/totals/co-est2023-pop-53.xlsx"

ggplot(p,aes(x,y))+
geom_vline(xintercept=seq(as.Date("2021-1-1"),xend,"3 month"),color="gray85",linewidth=.26)+
geom_hline(yintercept=ybreak,linewidth=.26,color="gray85")+
geom_vline(xintercept=seq(as.Date("2021-1-1"),xend,"year"),color="gray65",linewidth=.3)+
geom_vline(xintercept=c(xstart,xend),linewidth=.3,lineend="square")+
geom_hline(yintercept=c(ystart,yend),linewidth=.3,lineend="square")+
geom_line(aes(color=z,linewidth=z))+
geom_point(aes(color=z,alpha=z,shape=z),stroke=.4,size=1.2)+
labs(title="King County, Washington: Vaccinated people vs population estimates",subtitle=paste0("A CDC dataset for vaccinations by county used mid-2019 resident population estimates for each year, and Tennessean used population estimates from an unknown source. Here they are compared to vintage 2023 population estimates published by the US Census Bureau and 2020-2023 population estimates published by the Washington State Office of Financial Management."),x=NULL,y=NULL,caption=cap)+
scale_x_date(limits=c(xstart,xend),breaks=xbreak,labels=xlab)+
scale_y_continuous(limits=c(ystart,yend),breaks=ybreak,labels=\(x)ifelse(x==0,0,sprintf("%.1fM",x/1e6)))+
scale_color_manual(values=c("black",hsv(c(12,23)/36,1,.8),"red",hsv(3/36,1,.7)))+
scale_shape_manual(values=c(1,1,1,4,3))+
scale_alpha_manual(values=c(0,0,0,1,1))+
scale_linewidth_manual(values=c(.4,.4,.4,0,0))+
coord_cartesian(clip="off",expand=F)+
theme(axis.text=element_text(size=7,color="black"),
  axis.text.x=element_text(hjust=0),
  axis.ticks.length=unit(0,"pt"),
  legend.background=element_rect(fill="white",color="black",linewidth=.3),
  legend.box.spacing=unit(0,"pt"),
  legend.direction="vertical",
  legend.justification=c(1,.5),
  legend.key.height=unit(10,"pt"),
  legend.key.width=unit(18,"pt"),
  legend.key=element_blank(),
  legend.margin=margin(3,5,3,4),
  legend.position=c(1,.5),
  legend.spacing.x=unit(1.5,"pt"),
  legend.spacing.y=unit(0,"pt"),
  legend.text=element_text(size=7),
  legend.title=element_blank(),
  panel.background=element_blank(),
  plot.margin=margin(4,4,4,4),
  plot.subtitle=element_textbox_simple(size=6.5,margin=margin(4,,8)),
  plot.caption=element_text(size=5.8,margin=margin(2,,1),hjust=0),
  plot.title=element_text(size=7.5,face="bold",margin=margin(1,,4)))
ggsave("1.png",width=4,height=3.1,dpi=380*4)
system("magick 1.png -resize 25% 1.png")

The Tennessean dataset only includes data up to December 7th 2022, when it says there were 98.02% vaccinated people, so Hulscher et al. used the last value included in the dataset as their figure for the percentage of vaccinated people in 2023. However the CDC dataset includes data up to May 10th 2023 when it has 2,150,280 vaccinated people in King County, which would be about 99.4% of the population size used by Tennessean (2150280/2163216).

I don't think the percentage of vaccinated people would be as high as 98% anyway, because for example the percentage of vaccinated children under ages 10 is not very high. In the CDC dataset for COVID vaccinations by county state, there's many counties that have more vaccinated people than total people. So I don't know if for example the dataset includes vaccinated people who have died or moved outside the county.

Error in percentage increase of deaths between 2020 and 2023

Hulscher et al. wrote: "Specifically, the number of cardiopulmonary arrest deaths increased from 891 in 2015 to 1,110 in 2020, representing a 24.6% increase. In 2021, deaths jumped to 1,229 and continued to rise to 1,310 in 2022. The projection for 2023 suggests 1,392 cardiopulmonary arrest deaths in King County, WA, indicating a sharp 25.4% increase since the onset of COVID-19 vaccination campaigns"

However in their Table 1 the number of deaths in 2020 was 1121 and not 1110. But 1110 was the baseline value for 2020, so the authors seem to have accidentally calculated the increase between 2020 and 2023 using the baseline value instead of the number of deaths for 2020.

But the number of deaths actually increased from 1121 in 2020 to 1392 in 2023, so it was an increase of about about 24.2% and not 24.6%. But the baseline also increased by about 12% from 1110 in 2020 to 1245 in 2023, so the excess deaths only increased by about 10.8 percentage points: (1392/1245)-(1121/1110).

However the figure for 2023 wasn't even the real number of deaths but a linear projection of the deaths in 2021 and 2022. And the real number of deaths in 2023 was 1358, so the excess deaths between 2020 and 2023 increased by only about 8.1 percentage points relative to the baseline: (1358/1245)-(1121/1110).

But the baseline might have also been too low in 2023 because the long-term trend in deaths seemed to be curved upwards. And the authors also didn't exclude COVID deaths and there were still some COVID deaths in 2023.

Excess mortality in European countries compared to percentage of vaccinated people

However a prepandemic average baseline for raw deaths overestimates excess mortality in Western European countries relative to Eastern European countries. And Eastern European countries have a lower percentage of vaccinated people which explains the positive correlation in 2023.

In the years before COVID Western European countries had on average a steeper increase in the number of deaths per year than Eastern European countries. And in Eastern European countries the population size was reduced more by excess deaths in 2020 and 2021, so in the plot below where I calculated the green line for the expected number of deaths by multiplying the pre-COVID trend in deaths for each age by the population estimates for each age, Eastern European countries have a bigger drop in the baseline by 2022:

library(data.table);library(ggplot2)

q=\(...)as.character(substitute(...()))

# sars2.net/stat.html#Eurostat
t=fread("http://sars2.net/f/eurostatpopdead.csv.gz")[year%in%2010:2022]
t=t[location%in%na.omit(t)[,.N,location][N==max(N),location]]
t=t[!location%in%q(DE_TOT,EFTA)]

base=t[year%in%2010:2019,.(year=2010:2022,base=predict(lm(dead/pop~year),.(year=2010:2022))),.(age,location)]

p=merge(t,base)[,.(dead=sum(dead),base=sum(base*pop)),.(year,name)]
p=merge(p,p[year%in%2016:2019,.(ave=mean(dead)),name])
lab=c("Actual deaths","2016-2019 average","2010-2019 trend in CMR by age × population")
p=p[,.(x=year,y=c(dead,ave,base),z=factor(rep(lab,each=.N),lab),group=name)]

west=q(Iceland,Norway,Finland,Sweden,Denmark,Ireland,Netherlands,Belgium,Luxembourg,Germany,Liechtenstein,Switzerland,Austria,France,Portugal,Spain,Italy,Malta,Greece,Cyprus)
east=q(Estonia,Latvia,Poland,Czechia,Slovakia,Hungary,Slovenia,Croatia,Serbia,Romania,Bulgaria)
tot=c("West average","East average")

p=merge(p,p[z==z[1]&x%in%2016:2019,.(base=mean(y)),group])[,y:=y/base*100]
p=rbind(p,p[,.(y=mean(y),base=mean(base)),.(z,x,group=ifelse(group%in%west,tot[1],tot[2]))])
p[,group:=factor(group,c(west,east,tot))]

xstart=2010;xend=2022;p=p[x%in%xstart:xend]

ggplot(p,aes(x,y))+
facet_wrap(~group,ncol=5,dir="v")+
geom_hline(yintercept=0,color="gray60",linewidth=.4)+
geom_line(aes(color=z,linewidth=z,linetype=z))+
geom_point(aes(alpha=z,color=z),stroke=.5,size=1.5,shape=1)+
geom_text(data=p[rowid(group)==1],aes(label=group),color=ifelse(p[rowid(group)==1,group]%in%c(west,tot[1]),"#4444ff","#ff4444"),y=max(p$y),x=xstart,hjust=0,size=3.2,vjust=1.1)+
labs(title="Eurostat: Yearly deaths as percentage of 2016-2019 average",x=NULL,y=NULL)+
scale_x_continuous(limits=c(xstart-.5,xend+.5),breaks=seq(xstart+1,xend,2))+
scale_y_continuous(limits=extendrange(p$y,,.03),breaks=pretty,labels=\(x)paste0(x,"%"))+
coord_cartesian(clip="off",expand=F)+
scale_color_manual(values=c("black","black","#00aa00"))+
scale_linewidth_manual(values=c(0,.5,.6))+
scale_linetype_manual(values=c("solid","31","solid"))+
scale_alpha_manual(values=c(1,0,0))+
theme(axis.text=element_text(size=11,color="black"),
  axis.text.x=element_text(angle=90,vjust=.5,hjust=),
  axis.ticks=element_line(linewidth=.4,color="black"),
  axis.ticks.length=unit(3,"pt"),
  axis.title=element_text(size=8),
  legend.background=element_blank(),
  legend.box.spacing=unit(0,"pt"),
  legend.justification="right",
  legend.key=element_blank(),
  legend.key.height=unit(13,"pt"),
  legend.key.width=unit(25,"pt"),
  legend.margin=margin(,,6),
  legend.position="top",
  legend.spacing.x=unit(2,"pt"),
  legend.text=element_text(size=11,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.4),
  panel.spacing.x=unit(0,"pt"),
  panel.spacing.y=unit(0,"pt"),
  plot.margin=margin(7,7,7,7),
  plot.title=element_text(size=11.5,face=2,margin=margin(1,,4)),
  strip.background=element_blank(),
  strip.text=element_blank())
ggsave("1.png",width=7,height=9,dpi=300*4)
system("magick 1.png -resize 25% -colors 256 1.png")

In the plot above I used Eurostat's population estimates on January 1st for each year, so for example in Spain the green line drops between 2020 and 2021 because the excess deaths in 2020 didn't affect the population estimates I used until 2021. If I would've used mid-year population estimates or the average population throughout the year, the green line in Spain would've already dropped in 2020.

I treated Finland, Greece, and Cyprus as Western European even though they are geographically part of Eastern Europe. But they had a higher percentage of vaccinated people than any eastern bloc country in my analysis, and before COVID their yearly number of deaths was also increasing at a steeper slope than in most eastern bloc countries.

Here my correlation between the percentage of vaccinated people and excess mortality in 2022 dropped from about 0.34 to about -0.16 when I switched from the 2016-2019 average baseline to a 2013-2019 linear regression for ASMR:

download.file("https://covid.ourworldindata.org/data/owid-covid-data.csv","owid-covid-data.csv")

library(data.table);library(ggplot2)

o=fread("owid-covid-data.csv")

# sars2.net/stat.html#Eurostat
t=fread("http://sars2.net/f/eurostatpopdead.csv.gz")[year%in%2013:2022]
t=merge(t,t[location=="EU27_2020"&year==2020,.(age,std=pop/sum(pop))])
t=t[location%in%na.omit(t)[,.N,location][N==max(N),location]]
t=t[!location%in%c("DE_TOT")]

a=t[,.(dead=sum(dead),asmr=sum(dead/pop*std)*1e5),.(location,name,year)]
a$base=a[year<2020,predict(lm(asmr~year),.(year=2013:2022)),location]$V1
a$base2=a[year%in%2015:2019,predict(lm(dead~year),.(year=2013:2022)),location]$V1
a=merge(a,a[year%in%2016:2019,.(base3=mean(dead)),location])
p=a[year==2022,.(x=(asmr/base-1)*100,y=(dead/base3-1)*100),name]

p=merge(p,o[year(date)==2022,.(vaxpct=mean(people_vaccinated_per_hundred,na.rm=T)),.(name=location)])
p=na.omit(p)

lab=c("2016-2019 average for raw deaths","2013-2019 linear regression for ASMR")
p=p[,.(name,x=vaxpct,y=c(y,x),facet=factor(rep(lab,each=.N),lab))]

q=\(...)as.character(substitute(...()))
east=q(Bulgaria,Croatia,Cyprus,Czechia,Estonia,Greece,Hungary,Latvia,Poland,Romania,Serbia,Slovenia,Slovakia)
lab=q(Eastern,Western)
p$z=factor(ifelse(p$name%in%east,lab[1],lab[2]),lab)

levels(p$facet)=p[,cor(x,y),facet][,sprintf("Baseline is %s (r ≈ %.2f)",facet,V1)]

xbreak=pretty(p$x,7);ybreak=pretty(p$y,7);ylim=extendrange(p$y,,.02)

ggplot(p,aes(x,y))+
facet_wrap(~facet,dir="v",scales="free_x")+
coord_cartesian(clip="off",expand=F)+
geom_vline(xintercept=0,linewidth=.4,color="gray60")+
geom_hline(yintercept=0,linewidth=.4,color="gray60")+
geom_smooth(method="lm",formula=y~x,linewidth=.5,se=F,color="black",linetype="42")+
geom_label(data=p[rowid(facet)==1],aes(label=stringr::str_wrap(facet,40)),x=xbreak[1],y=ylim[2],lineheight=.9,hjust=0,vjust=1,size=3.7,label.r=unit(0,"pt"),label.padding=unit(6,"pt"),label.size=.4)+
geom_point(aes(color=z),size=.7)+
ggrepel::geom_text_repel(aes(label=name,color=z),size=3,max.overlaps=Inf,segment.size=.3,min.segment.length=.2,box.padding=.07,show.legend=F)+
labs(title="Correlation between percentage of vaccinated people in 2022 and\nexcess mortality percent in 2022",x="Average percentage of vaccinated people throughout 2022 at OWID",y="Excess percentage of deaths in 2022 at Eurostat")+
scale_x_continuous(breaks=xbreak,limits=range(xbreak),labels=\(x)paste0(x,"%"))+
scale_y_continuous(breaks=ybreak,limits=ylim,labels=\(x)paste0(x,"%"))+
scale_color_manual(values=c("#ff4444","#5555ff"))+
theme(axis.text=element_text(size=11,color="black"),
  axis.text.x=element_text(margin=margin(3)),
  axis.ticks=element_line(linewidth=.4,color="black"),
  axis.ticks.length=unit(4,"pt"),
  axis.title=element_text(size=11),
  axis.title.x=element_text(margin=margin(3)),
  axis.title.y=element_text(margin=margin(,2)),
  legend.background=element_blank(),
  legend.box.spacing=unit(0,"pt"),
  legend.key=element_blank(),
  legend.key.height=unit(13,"pt"),
  legend.key.width=unit(16,"pt"),
  legend.position="top",
  legend.justification="right",
  legend.spacing.x=unit(2,"pt"),
  legend.margin=margin(,,6),
  legend.text=element_text(size=11,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.4),
  panel.grid.major=element_blank(),
  panel.spacing=unit(4,"pt"),
  plot.margin=margin(6,22,5,5),
  plot.subtitle=element_text(size=11),
  plot.title=element_text(size=11.5,face=2,margin=margin(1,,5)),
  strip.background=element_blank(),
  strip.text=element_blank())
ggsave("1.png",width=6,height=6,dpi=300*4)

sub="Source: covid.ourworldindata.org/data/owid-covid-data.csv and Eurostat datasets demo_magec and demo_pjan. ASMR was calculated by single year of age up to ages 100+ so that the 2020 EU population estimates were used as the standard population. Countries that had deaths or population estimates missing for any combination of age or year in 2013-2022 were omitted. Romania was omitted because it was missing vaccination data at OWID."
system(paste0("mogrify -trim 1.png;w=`identify -format %w 1.png`;magick 1.png \\( -size $[w]x -font Arial -interline-spacing -3 -pointsize $[44*4] caption:'",gsub("'","'\\\\'",sub),"' -splice x100 \\) -append -resize 25% -bordercolor white -border 30 -colors 256 1.png"))

Here Western European countries are also more likely to be above the diagonal, which means that the 2016-2019 average baseline likely overestimates their excess mortality in 2022 compared to ASMR:

library(data.table);library(ggplot2)

eu=fread("http://sars2.net/f/eurostatpopdead.csv.gz")[year%in%2013:2022] # sars2.net/stat.html#Eurostat
eu=merge(eu,eu[location=="EU27_2020"&year==2020,.(age,std=pop/sum(pop))])
eu=eu[!location%in%c("DE_TOT","EFTA")]
eu=eu[location%in%na.omit(eu)[,.N,location][N==max(N),location]]

a=eu[,.(dead=sum(dead),asmr=sum(dead/pop*std*1e5)),.(name,year)]
a$base=a[year<2020,predict(lm(asmr~year),.(year=2013:2022)),name]$V1
a=merge(a,a[year%in%2016:2019,.(base2=mean(dead)),name])

p=a[year==2022,.(x=(asmr/base-1)*100,y=(dead/base2-1)*100,name)]

east=c("Bulgaria","Croatia","Czechia","Estonia","Hungary","Latvia","Poland","Romania","Serbia","Slovenia","Slovakia")
lab=c("Western Europe","Eastern Europe")
p[,z:=factor(ifelse(name%in%east,lab[2],lab[1]),lab)]

breaks=p[,pretty(c(x,y),9)];lims=p[,extendrange(range(c(x,y)),,.04)]

ggplot(p,aes(x,y))+
coord_fixed(clip="off",expand=F)+
geom_vline(xintercept=0,linewidth=.4,color="gray60")+
geom_hline(yintercept=0,linewidth=.4,color="gray60")+
geom_abline(linetype="dashed",color="gray60",size=.4)+
geom_point(aes(color=z),size=.6)+
ggrepel::geom_text_repel(aes(label=name,color=z),size=3,max.overlaps=Inf,segment.size=.1,min.segment.length=.2,box.padding=.07,show.legend=F)+
labs(title="Eurostat: Excess ASMR vs excess raw deaths in 2022",x="Excess ASMR in 2022 relative to 2013-2019 linear trend",y="Excess raw deaths in 2022 relative to 2016-2019 average")+
scale_x_continuous(limits=lims,breaks=breaks,labels=\(x)paste0(x,"%"))+
scale_y_continuous(limits=lims,breaks=breaks,labels=\(x)paste0(x,"%"))+
scale_color_manual(values=c("#5555ff","#ff4444"))+
theme(axis.text=element_text(size=11,color="black"),
  axis.text.x=element_text(margin=margin(3)),
  axis.ticks=element_line(linewidth=.4,color="black"),
  axis.ticks.length=unit(-4,"pt"),
  axis.title=element_text(size=11),
  axis.title.x=element_text(margin=margin(3)),
  axis.title.y=element_text(margin=margin(,2)),
  legend.background=element_blank(),
  legend.box.spacing=unit(0,"pt"),
  legend.key=element_blank(),
  legend.key.height=unit(13,"pt"),
  legend.key.width=unit(26,"pt"),
  legend.position="top",
  legend.justification="left",
  legend.spacing.x=unit(2,"pt"),
  legend.margin=margin(,,6),
  legend.text=element_text(size=11,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.4),
  panel.grid.major=element_blank(),
  plot.subtitle=element_text(size=11),
  plot.title=element_text(size=11.5,face=2,margin=margin(1,,5)))
ggsave("1.png",width=6,height=6,dpi=300*4)

sub="ASMR was calculated by single year of age up to ages 100+ so that the 2020 EU population estimates were used as the standard population. Deaths and population estimates by single year of age were compiled with this script: sars2.net/stat.html#Compile_a_CSV_file_for_deaths_and_population_estimates_by_single_year_of_age_in_various_countries. Countries that had deaths or population estimates missing for any combination of age or year in 2013-2022 were omitted. Western Europe includes geographically Eastern European countries that were not part of the eastern bloc."
system(paste0("mogrify -trim 1.png;w=`identify -format %w 1.png`;magick 1.png \\( -size $[w]x -font Arial -interline-spacing -3 -pointsize $[44*4] caption:'",gsub("'","'\\\\'",sub),"' -splice x100 \\) -append -resize 25% -bordercolor white -border 30 -colors 256 1.png"))

In the previous plots I used the Eurostat dataset demo_magec which has yearly deaths by single year of age: https://ec.europa.eu/eurostat/databrowser/product/page/demo_magec. However it was still missing deaths for 2023 as of December 2024, so in the next plot I looked at the Eurotat dataset for weekly deaths by five-year age groups instead which also has data for 2023 and 2024: https://ec.europa.eu/eurostat/databrowser/product/page/demo_r_mwk_05. In the top panel of the plot below if you look at the black line that includes both Western Europe and Eastern Europe, the average correlation in 2023 is close to zero:

download.file("https://covid.ourworldindata.org/data/owid-covid-data.csv","owid-covid-data.csv")
download.file("https://ec.europa.eu/eurostat/api/dissemination/sdmx/2.1/data/demo_r_mwk_05?format=TSV","demo_r_mwk_05.tsv")

library(data.table);library(ggplot2)

isoweek=\(year,week,weekday=7){d=as.Date(paste0(year,"-1-4"));d-(as.integer(format(d,"%w"))+6)%%7-1+7*(week-1)+weekday}
q=\(...)as.character(substitute(...()))
ma=\(x,b=1,f=b){x[]=rowMeans(embed(c(rep(NA,b),x,rep(NA,f)),f+b+1),na.rm=T);x}

o=fread("owid-covid-data.csv")

euro=fread("demo_r_mwk_05.tsv")
meta=fread(text=euro[[1]],header=F)
pick=meta[,V2%like%"^Y"&V3=="T"];meta=meta[pick];euro=euro[pick]
meta[,age:=as.integer(sub("\\D+(\\d+).*","\\1",sub("Y_LT5",0,V2)))]
eu=meta[,.(loc=V5,age,date=rep(names(euro)[-1],each=.N),dead=as.integer(sub("\\D+","",unlist(euro[,-1],,F))))]
eu[,year:=as.integer(substr(date,1,4))][,week:=as.integer(substr(date,7,8))]
eu[,date:=isoweek(year,week,4)]
eu=eu[loc%in%eu[year%in%2013:2023,.N,loc][N==max(N),loc]]
eu=eu[year>=2013]

pop=fread("http://sars2.net/f/eurostatpopdead.csv.gz")[year%in%2013:2023] # sars2.net/stat.html#Eurostat
a=pop[,.(pop=sum(pop)),.(location,name,date=as.Date(paste0(year,"-1-1")),age=pmin(age,90)%/%5*5)]
a=a[location%in%na.omit(a)[,.N,location][N==max(N),location]]
a=a[,spline(date,pop,xout=unique(eu$date),method="natural"),.(location,age,name)]

a=merge(eu,a[,.(date=`class<-`(x,"Date"),pop=y,loc=location,name,age)])
a=merge(pop[location=="EU27_2020"&year==2020,.(pop=sum(pop)),.(age=pmin(age,90)%/%5*5)][,.(age,std=pop/sum(pop))],a)
a=a[loc%in%na.omit(a)[year%in%2013:2023,.N,loc][N==max(N),loc]]

a=a[,.(dead=sum(dead),asmr=sum(dead/pop*std/7*365e5)),,.(name,date,year,week)]
slope=a[year%in%2013:2019&week%in%15:35,mean(asmr),.(year,name)]
slope=slope[,{x=.(year=2013:2024,slope=predict(lm(V1~year),.(year=2013:2024)));x$slope=x$slope/x$slope[x$year==2016];x},name]
weekly=a[year%in%2013:2019,.(weekly=mean(asmr)),.(week,name)]
a=merge(slope,merge(weekly,a))
a=merge(a[year%in%2016:2019,.(deadbase=mean(dead)),.(name,week)],a)
p=a[,.(pct=(asmr/(slope*weekly)-1)*100,pct2=(dead/deadbase-1)*100),,.(name,date)]

east=q(Estonia,Latvia,Poland,Czechia,Slovakia,Hungary,Montenegro,Slovenia,Croatia,Serbia,Romania,Bulgaria)

p=merge(p,o[,.(date,vax=nafill(nafill(zoo::na.approx(people_vaccinated_per_hundred,na.rm=F),"locf"),,0)),.(name=location)])
p[,z:=ifelse(name%in%east,"Eastern Europe","Western Europe")]
p=rbind(p,copy(p)[,z:="Total"])
p[,z:=factor(z,unique(z))]

lab=c("Baseline is seasonality-adjusted linear trend for ASMR in 2013-2019","Baseline is average raw deaths on corresponding week number in 2016-2019")
p=na.omit(p)[,.(y=c(cor(pct,vax),cor(pct2,vax)),group=lab),.(x=date,z)]
p[,group:=factor(group,lab)]

p[,y:=ma(y,2),.(z,group)]

xstart=as.Date("2020-12-1");xend=as.Date("2025-1-1")
xbreak=seq(as.Date("2021-1-1"),xend,"6 month");xlab=ifelse(month(xbreak)==7,year(xbreak),"")

sub="\u00a0    Weekly excess deaths by 5-year age groups are from the Eurostat dataset demo_r_mwk_05. Weekly population estimates were spline interpolated from population estimates on January 1st from the Eurostat dataset demo_pjan. ASMR was calculated by 5-year age groups up to ages 90+ so that the 2020 EU population was used as the standard population.
     The slope of the baseline for ASMR was determined by doing a linear regression for mean ASMR on weeks 15-35 of each year, where winter weeks were excluded due to higher variability in mortality during winters. In order to adjust the baseline for seasonality, each week number has its own intercept.
     The variable people_vaccinated_per_hundred at OWID was used for the percentage of vaccinated people. Missing values were interpolated linearly and the percentage was treated as zero before the earliest reported value. The percentage of vaccinated people on each Thursday was used as the percentage for the week.
     Andorra, Albania, Germany, Lithuania, and UK were omitted because data for deaths or population estimates was missing.
     Western Europe includes Finland, Greece, and Cyprus."

ggplot(p,aes(x,y))+
facet_wrap(~group,ncol=1,dir="v",scales="free_x")+
geom_hline(yintercept=0,color="gray60",linewidth=.4)+
geom_vline(xintercept=seq(as.Date("2021-1-1"),xend,"year"),color="gray60",linewidth=.4)+
geom_line(aes(color=z),linewidth=.6)+
labs(title=paste0("Correlation between excess deaths at Eurostat and vaccinated percent at\nOWID for ",a[,length(unique(name))]," European countries (±2-week moving average)"),x=NULL,y=NULL)+
scale_x_continuous(limits=c(xstart,xend),breaks=xbreak,labels=xlab)+
scale_y_continuous(limits=c(-1,1),breaks=seq(-1,1,.5))+
scale_color_manual(values=c("#5555ff","#ff5555","black"))+
coord_cartesian(clip="off",expand=F)+
theme(axis.text=element_text(size=11,color="black"),
  axis.text.x=element_text(margin=margin(3)),
  axis.ticks=element_line(linewidth=.4,color="black"),
  axis.ticks.x=element_line(color=alpha("black",c(1,0))),
  axis.ticks.length.x=unit(0,"pt"),
  axis.ticks.length.y=unit(4,"pt"),
  axis.title=element_text(size=8),
  legend.background=element_blank(),
  legend.box.spacing=unit(0,"pt"),
  legend.justification="right",
  legend.key=element_blank(),
  legend.key.height=unit(13,"pt"),
  legend.key.width=unit(25,"pt"),
  legend.margin=margin(),
  legend.position="top",
  legend.spacing.x=unit(2,"pt"),
  legend.text=element_text(size=11,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.4),
  panel.spacing=unit(0,"pt"),
  plot.margin=margin(7,7,7,7),
  plot.title=element_text(size=11.5,face=2,margin=margin(1,,5)),
  strip.background=element_blank(),
  strip.text=element_text(size=11,face=2))
ggsave("1.png",width=6.2,height=4.6,dpi=300*4)
system(paste0("mogrify -trim 1.png;w=`identify -format %w 1.png`;magick 1.png \\( -size $[w]x -font Arial -interline-spacing -3 -pointsize $[46*4] caption:'",gsub("'","'\\\\'",sub),"' -splice x100 \\) -append -resize 25% -bordercolor white -border 30 -colors 256 1.png"))

In the following code I made a model for each country, where at the start of the model each age had the same population size as Eurostat's population estimate for January 1st 2019. Each year I killed the same proportion of people each age that died in 2017-2019, I incremented the age of surviving people by one, and I added the same number of people aged zero as in Eurostat's population estimate for 2019. Then I used the number of deaths produced by the model in 2022 as the expected deaths for each country:

The output shows that when I took the number of deaths produced by the model in 2022 and divided it with the average number of deaths in 2016-2019, the ratio was much higher on average in Western European countries than Eastern European countries, which again shows how Eurostat's 2016-2019 average baseline exaggerates excess deaths in Western European countries relative to Eastern European countries:

Steve Kirsch: Claim that Singapore had the highest excess mortality in 2023

However at Mortality Watch Singapore doesn't have as high excess CMR percent in 2023 as Qatar, Macao, and Turkey:

Macao has about 260% excess deaths in January 2023 on OWID. The first big COVID wave in China was in December 2022 to January 2023 after the zero COVID policy was dropped.

Singapore had the first big COVID wave in October to November of 2021, so Singapore was still essentially in its second year of COVID for most of 2023. But if the excess deaths were caused by the vaccines then why was there were low excess deaths in the first half of 2021 when most people got vaccinated?

Clare Craig: Correlation for excess mortality between first and second half of the year across US states

However Craig fitted her baseline against deaths in 2010-2019, so 2023 was further away from her fitting period than 2021. So by 2023 her baseline was probably further off from the actual long-term trend on average than in 2021.

In the following code I fitted a linear regression against deaths in June of 2010-2019 in each US state and DC. But in the early 00s my baseline was further off than in the late 00s, so my correlation between excess deaths in the first and second half of the year was much higher in the early 00s than the late 00s:

Steve Kirsch: COVID deaths in Israel and Palestine

Kirsch posted the image below and wrote: "If the Covid vaccine worked so well, then why wasn't there a divergence in these curves? Palestine was severely under vaccinated compared to Israel, so why does Israel have more Covid deaths per capita than Palestine? They basically tracked each other!!" [https://x.com/stkirsch/status/1887186858326790420]

However in early 2021 there was a big spike in all-cause deaths in Palestine but not Israel, but at the time Israel already had a high percentage of vaccinated people but in Palestine almost no-one was vaccinated:

The plot above is not even adjusted for age but the population of Israel is much older than the population of Palestine. Kirsch's plot for COVID deaths per capita was not adjusted for age either.

Humanspective: Yearly deaths registered in Western Australia

p$poly=p[year<2020,predict(lm(dead~poly(year,2)),p)]
p=data.table(year=1981:2024,dead=c(8180,8287,8550,8706,9042,9503,9082,9698,9718,9614,9753,10163,10491,10466,10570,11251,10905,10741,11074,10858,10751,11711,11520,11437,11504,11821,12581,13011,12855,12927,13001,13605,13628,14040,14705,15071,14754,14873,15281,15258,16106,17516,17734,18303))

p$trend=p[year%in%2010:2019,predict(lm(dead~year),p)]
p$trend2=p[year%in%2016:2019,predict(lm(dead~year),p)]

png("1.png",1800,1200,res=300)
par(mar=c(4.9,2.7,1.8,.7),mgp=c(0,.6,0),adj=0,lend="square",tck=-.03)

tit="Yearly deaths registered in Western Australia"
leg=c("Deaths","2010-2019 linear trend","2016-2019 linear trend")
color=c("black","blue","#9999ff")

ybreak=pretty(c(0,p$dead,p$trend));ylim=range(ybreak)
xstart=min(p$year);xend=max(p$year)
xbreak=seq(1982,2024,2)

plot(p$year,p$trend,type="n",main=tit,xlab=NA,ylab=NA,xaxs="i",yaxs="i",xaxt="n",yaxt="n",ylim=ylim,xlim=c(xstart-.5,xend+.5),cex.main=1)
axis(1,at=xbreak,las=2)
axis(2,at=ybreak,las=1,labels=ifelse(ybreak==0,0,paste0(ybreak/1e3,"k")))
lines(p$year,p$dead,lwd=1.5)
points(p$year,p$dead,pch=20,cex=.6)
lines(p$year,p$trend,type="l",lty=2,lwd=1.5,col=color[2])
lines(p$year,p$trend2,type="l",lty=2,lwd=1.5,col=color[3])
legend("topleft",legend=leg,lty=c(1,2,2),lwd=1.5,col=color)
mtext(text="Source: https://www.wa.gov.au/organisation/department-of-justice/the-registry-
of-births-deaths-and-marriages/statistics-births-deaths-and-marriages-registered",side=1,line=3.5,adj=1,col="gray50",cex=.95)
dev.off()

Jikkyleaks: Is Short-Term Mortality Fluctuations data synthetic?

Jikkyleaks claimed that the weekly data for deaths in the Short Term Mortality Fluctuations dataset is synthetic, because he said the US chart looks too smooth compared to other countries (never mind that the United States has a bigger sample size of deaths than other countries in his plot): [https://x.com/Jikkyleaks/status/1888460292923023454]

However STMF uses the same set of 5 age groups for each country (0-14, 15-64, 65-74, 75-84, and 85+). So in countries where data using the STMF age groups is not available, the deaths are estimated by modifying the source age groups to match the set of 5 age groups used by STMF: [https://mortality.org/File/GetDocument/Public/STMF/DOC/STMFmetadata.pdf]

In STMF's data for the United States, the "Split" column is set to 1 on weeks before week 2 of 2018, because the deaths in the age groups 0-14 and 15-64 were estimated by splitting the age group 0-24 into two parts: [https://mortality.org/File/GetDocument/Public/STMF/DOC/STMFNote.pdf]

From week 2 of 2018 onwards STMF takes data for United States from CDC WONDER, which has weekly deaths available by single year of age, so the number of deaths in ages 0-14 and 15-64 is an integer:

Gibo et al.: Paper about cancer ASMR in Japan

Yearly ASMR

The paper featured the plot below where cancer ASMR went up between 2020 and 2021. However I found a spreadsheet of yearly cancer ASMR values on a website of the Japanese National Cancer Center: https://ganjoho.jp/reg_stat/statistics/data/dl/index.html (second link, fifth sheet, second row). The spreadsheet includes the ICD codes C00-C97 and it uses the 2015 standard population by 5-year age groups up to ages 95+, so it should match the paper by Gibo et al. in both ways. But the red circles here show that the cancer ASMR in the spreadsheet went down between 2020 and 2021 and not up like in the paper by Gibo et al.:

Comparison of monthly ASMR to other variables

This plot from the Japanese paper shows that several months that have local maxima in excess cancer deaths coincide with a month that has a local maximum in the number of vaccine doses administered:

In the plot above there's three vaccination waves in 2022, but the yellow line shows that all three of them coincided with a COVID wave. However the plot above only shows COVID cases and not COVID deaths, and CFR was much higher in 2021 than 2022 so the yellow line for cases is very low in 2021, so you can't easily see that a COVID wave also coincided with the period with elevated cancer deaths in August to September 2021:

I wasn't sure if the Japanese paper looked at deaths by underlying cause of death or multiple cause of death. At first I thought the paper might have shown MCD deaths which would have explained why there were spikes in cancer deaths during COVID waves. In the United States there are spikes in MCD cancer deaths during COVID waves, but the spikes disappear if you exclude deaths with MCD COVID, or if you look at deaths with UCD cancer instead of MCD cancer.

However Table 1 of the Japanese paper shows that there were 1,575,936 deaths from all causes in 2023, out of which 382,492 deaths had the cause malignant neoplasms (C00-C97). It's almost the same as the number of deaths in file 2 under volume 2 here: https://www.e-stat.go.jp/stat-search/files?page=1&layout=datalist&toukei=00450011&tstat=000001028897&cycle=7&year=20230&month=0&tclass1=000001053058&tclass2=000001053061&tclass3=000001053065&result_back=1&cycle_facet=tclass1&tclass4val=0. The file had 1,576,016 deaths from all natural causes in 2023, but the deaths under all individual 3-letter ICD codes added up to 1,537,930, and the deaths under ICD codes starting with C added up to 382,504. So since the number of deaths under individual codes added together was nearly the same as the total number of deaths, you can tell that the file shows deaths by underlying cause of death and not by multiple cause of death, which means that Table 1 in the Japanese paper also shows UCD deaths.

However it's still possible that there were elevated UCD cancer deaths during COVID waves if COVID contributed to the cancer deaths even though it was not listed as the underlying cause. But cancers normally take time to develop, so unless there were a lot of people who died of a fast-acting turbo cancer soon after vaccination, I don't think vaccines would explain why the peaks in cancer deaths occurred around the same months as peaks in new vaccine doses administered.

Monthly excess ASMR

library(data.table);library(ggplot2);library(lubridate)

t=fread("http://sars2.net/f/japancancerbyage.csv")
t[,date:=as.Date(paste0(date,"-1"))]
t=t[,.(dead=sum(dead),pop=sum(pop)),.(date,age=pmin(age,95))]

std=data.table(age=0:19*5)
std$std=c(978+4048,5369,5711,6053,6396,6738,7081,7423,7766,8108,8451,8793,9135,9246,7892,6306,4720,3134,1548,423)*1e3
std[,std:=std/sum(std)]

a=merge(t,std)[,.(asmr=sum(dead/pop/days_in_month(date)*std*365e5)),date]
a$base=a[year(date)%in%2010:2019,predict(lm(asmr~date),a)]
a[,month:=month(date)]
a=merge(a,a[year(date)%in%2010:2019,.(monthly=mean(asmr-base)),month])

lab=c("ASMR","2010-2019 linear baseline","Seasonal baseline","Excess ASMR")
p=a[,.(x=date,y=c(asmr,base,base+monthly,(asmr/(base+monthly)-1)*100),z=factor(rep(lab,each=.N),lab),facet=factor(rep(facet,c(.N*3,.N)),facet))]

p=rbind(p,p[z==z[1]&x==max(x)][,x:=seq(x,,"month",2)[2]])

xstart=as.Date("2010-1-1");xend=as.Date("2025-1-1")
xbreak=seq(xstart,xend,"6 month");xlab=ifelse(month(xbreak)==7,year(xbreak),"")

ggplot(p,aes(x=x,y))+
facet_wrap(~facet,dir="v",scales="free")+
geom_vline(xintercept=seq(xstart,xend,"year"),color="gray83",linewidth=.4)+
geom_segment(data=data.table(facet=facet[2]),x=xstart,xend=xend,y=0,yend=0,linewidth=.4,lineend="square")+
geom_step(data=p[z==levels(z)[1]],aes(color=z))+
geom_line(data=p[z%in%levels(z)[2:3]],aes(x=x+14,color=z))+
geom_rect(data=p[z==levels(z)[4]],aes(xmin=x,xmax=x%m+%months(1),ymin=pmin(y,0),ymax=pmax(y,0)),color="gray20",fill="gray50",linewidth=.15)+
geom_text(data=p[,.(y=max(y)),facet],x=pmean(xstart,xend),aes(label=facet),size=3.8,fontface=2,vjust=1.4)+
labs(x=NULL,y=NULL,title="Age-standardized cancer mortality rate per 100,000 person-years in Japan")+
scale_x_continuous(limits=c(xstart,xend),breaks=xbreak,labels=xlab,expand=expansion(0,0))+
scale_y_continuous(breaks=\(x)pretty(x,7),labels=\(x)ifelse(x<10,paste0(x,"%"),x),expand=expansion(.03,0))+
scale_color_manual(values=c("black","#8888ff","blue"))+
coord_cartesian(clip="off")+
theme(axis.text=element_text(size=11,color="black"),
  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(5,"pt"),
  legend.background=element_blank(),
  legend.box.background=element_blank(),
  legend.box.spacing=unit(0,"pt"),
  legend.direction="horizontal",
  legend.justification="left",
  legend.key=element_blank(),
  legend.key.height=unit(11,"pt"),
  legend.key.width=unit(26,"pt"),
  legend.margin=margin(,,6,-2),
  legend.position="top",
  legend.spacing.x=unit(3,"pt"),
  legend.spacing.y=unit(0,"pt"),
  legend.text=element_text(size=11,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.4),
  panel.spacing=unit(2,"pt"),
  plot.margin=margin(6,6,1,5),
  plot.title=element_text(size=11.5,face=2,margin=margin(2,,4)),
  strip.background=element_blank(),
  strip.text=element_blank())
ggsave("1.png",width=6.3,height=4.8,dpi=300*4)
sub="\u00a0    Cancer deaths are deaths with underlying cause of death malignant neoplasms (C00-C97). Cancer deaths by 5-year age groups were compiled from table 1-6 for each month here: www.e-stat.go.jp/stat-search/files?page=1&layout=datalist&toukei=00450011&tstat=000001028897
&cycle=1&tclass1=000001053058&tclass2=000001053060&cycle_facet=tclass1&tclass3val=0. Population estimates on October 1st of each year were spline interpolated to monthly population estimates: www.e-stat.go.jp/en/stat-search/files?page=1&toukei=00200524&tstat=000000090001 (annual reports for 2021-2023 combined with 2000-2020 time series reports). The 2015 Japanese standard population was used by 5-year age groups up to ages 95+: www.mhlw.go.jp/content/
10700000/000557741.pdf.
     The seasonality-adjusted baseline was calculated by adding the average difference between the linear trend and actual ASMR in January 2010-2019 to the baseline for each January, and similarly for other months."
system(paste0("f=1.png;mar=100;w=`/usr/local/bin/identify -format %w $f`;/usr/local/bin/magick \\( $f \\) \\( -size $[w-mar*2]x -font Arial -interline-spacing -3 -pointsize $[42*4] caption:'",gsub("'","'\\\\'",sub),"' -gravity southwest -splice $[mar]x80 \\) -append -resize 25% -dither none -colors 256 1.png"))

The average of my monthly ASMR values was about 273.9 in 2021 and 275.1 in 2020, so I again got lower ASMR in 2021 than 2020 like in the spreadsheet at ganjoho.jp.

In the next plot I overlaid the same monthly excess ASMR values that are shown in the previous plot on top of Gibo et al.'s plot. I don't know if Gibo et al. manually modified their excess ASMR results to get a better correspondence between peaks in excess ASMR and months when vaccines were administered, because the correspondence was not nearly as impressive in my plot:

From the plot above you can also see that Gibo et al. got much lower excess ASMR each February than me, so I don't know if they forgot to adjust their excess ASMR calculation for the difference in the length of months. But maybe that's not the case, because Gibo et al. also got much lower excess ASMR each January than me.

However Gibo et al. also used a different method to calculate the baseline than me, and they only linked a generic page for Japanese vital statistics as the source of their mortality data, so I don't know which specific tables under their link they used to calculate ASMR, or if they used precalculated mortality rates to calculate ASMR or if they combined one dataset for deaths with another dataset for population estimates like me. They should publish the code they used to calculate the baseline and mention which exact tables of data they used.

Excess deaths by age and month

If the spikes in cancer deaths which coincided with vaccination waves were caused by vaccines, you would expect the spikes to occur earlier in older age groups that got vaccinated earlier and later in younger age groups that got vaccinated later. However such a pattern is not apparent in this plot:

t=fread("http://sars2.net/f/japancancerbyage.csv")
t[,date:=as.Date(paste0(date,"-1"))]
a=t[,.(dead=sum(dead),pop=sum(pop)*days_in_month(date)),.(date,age)]
a[,cmr:=dead/pop]

a=merge(a,a[year(date)%in%2010:2019,.(date=unique(t$date),base=predict(lm(cmr~date),.(date=unique(t$date)))),age])
a[,month:=month(date)]
a=merge(a[year(date)%in%2010:2019,.(monthly=mean(cmr-base)),.(month,age)],a)

a=a[year(date)>=2020,.(dead=sum(dead),base=sum((base+monthly)*pop)),.(age=agecut(age,c(0,40,6:9*10)),date)]
a[,date:=factor(substr(date,1,7))]

a=rbind(a,a[,.(age="Total",dead=sum(dead),base=sum(base)),date])
a=rbind(a,a[,.(date="Total",dead=sum(dead),base=sum(base)),age])

m1=a[,tapply(dead,.(age,date),c)]
m2=a[,tapply(base,.(age,date),c)]

m=(m1-m2)/ifelse(m1>m2,m2,m1)*100
disp=round((m1/m2-1)*100)

exp=.8
maxcolor=max0(m)

pheatmap::pheatmap(abs(m)^exp*sign(m),filename="1.png",display_numbers=disp,gaps_row=nrow(m)-1,
  gaps_col=seq(12,ncol(m),12),
  cluster_rows=F,cluster_cols=F,legend=F,cellwidth=14,cellheight=14,fontsize=9,fontsize_number=8,
  border_color=NA,na_col="gray90",
  number_color=ifelse(!is.na(m)&abs(m)^exp>maxcolor^exp*.4,"white","black"),
  breaks=seq(-maxcolor^exp,maxcolor^exp,,256),
  colorRampPalette(hsv(rep(c(7/12,0),5:4),c(.9,.75,.6,.3,0,.3,.6,.75,.9),c(.4,.65,1,1,1,1,1,.65,.4)))(256))

system("w=`identify -format %w 1.png`;pad=20;magick -pointsize 43 -font Arial-Bold \\( -size $[w-pad] caption:'Monthly percentage of excess cancer deaths in Japan (underlying cause of death malignant neoplasms, C00-C97)' -splice $[pad]x24 \\) -pointsize 40 -font Arial \\( -size $[w-pad*2] caption:'First a seasonality-adjusted linear trend in CMR was calculated by each 5-year age group in 2010-2019, and the projected trend was multiplied by the population size of the age group each month to get monthly expected deaths. The age groups shown here were aggregated from 5-year age groups.' -splice $[pad]x8 \\) 1.png -append -dither none -colors 256 1.png")

In the plot above some age groups have consistently positive excess mortality and others have consistently negative mortality. But it might be because there is some degree of error introduced to the slope of the baseline due to random variation in mortality during the prediction period, or because the long-term trend in mortality was not actually linear, or because I calculated the excess mortality by 5-year age groups and not by single year of age.

Comparison to excess deaths from all natural causes

My previous plot showed that the peak monthly excess cancer ASMR was about 3.5%. But it's actually fairly small in comparison to the excess all-cause ASMR shown below, which has remained at a sustained level of over 5% each month since February 2022, and which reached above 20% during two months of 2024:

library(data.table);library(ggplot2);library(lubridate)

t=fread("http://sars2.net/f/japanacmbyage.csv")
t[,date:=as.Date(paste0(date,"-1"))]
t=t[,.(dead=sum(dead),pop=sum(pop)),.(date,age=pmin(age,95))]

std=data.table(age=0:19*5)
std$std=c(978+4048,5369,5711,6053,6396,6738,7081,7423,7766,8108,8451,8793,9135,9246,7892,6306,4720,3134,1548,423)*1e3
std[,std:=std/sum(std)]

a=merge(t,std)[,.(asmr=sum(dead/pop/days_in_month(date)*std*365e5)),date]
a$base=a[year(date)%in%2010:2019,predict(lm(asmr~date),a)]
a[,month:=month(date)]
a=merge(a,a[year(date)%in%2010:2019,.(monthly=mean(asmr-base)),month])

lab=c("ASMR","2010-2019 linear baseline","Seasonal baseline","Excess ASMR")
p=a[,.(x=date,y=c(asmr,base,base+monthly,(asmr/(base+monthly)-1)*100),z=factor(rep(lab,each=.N),lab),facet=factor(rep(facet,c(.N*3,.N)),facet))]

p=rbind(p,p[z==z[1]&x==max(x)][,x:=seq(x,,"month",2)[2]])

xstart=as.Date("2010-1-1");xend=as.Date("2025-1-1")
xbreak=seq(xstart,xend,"6 month");xlab=ifelse(month(xbreak)==7,year(xbreak),"")

ggplot(p,aes(x=x,y))+
facet_wrap(~facet,dir="v",scales="free")+
geom_vline(xintercept=seq(xstart,xend,"year"),color="gray83",linewidth=.4)+
geom_segment(data=data.table(facet=facet[2]),x=xstart,xend=xend,y=0,yend=0,linewidth=.4,lineend="square")+
geom_step(data=p[z==levels(z)[1]],aes(color=z))+
geom_line(data=p[z%in%levels(z)[2:3]],aes(x=x+14,color=z))+
geom_rect(data=p[z==levels(z)[4]],aes(xmin=x,xmax=x%m+%months(1),ymin=pmin(y,0),ymax=pmax(y,0)),color="gray20",fill="gray50",linewidth=.15)+
geom_text(data=p[,.(y=max(y)),facet],x=pmean(xstart,xend),aes(label=facet),size=3.8,fontface=2,vjust=1.4)+
labs(x=NULL,y=NULL,title="All-cause ASMR 100,000 person-years in Japan")+
scale_x_continuous(limits=c(xstart,xend),breaks=xbreak,labels=xlab,expand=expansion(0,0))+
scale_y_continuous(breaks=\(x)pretty(x,7),labels=\(x)ifelse(x<100,paste0(x,"%"),x),expand=expansion(.03,0))+
scale_color_manual(values=c("black","#8888ff","blue"))+
coord_cartesian(clip="off")+
theme(axis.text=element_text(size=11,color="black"),
  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(5,"pt"),
  legend.background=element_blank(),
  legend.box.background=element_blank(),
  legend.box.spacing=unit(0,"pt"),
  legend.direction="horizontal",
  legend.justification="left",
  legend.key=element_blank(),
  legend.key.height=unit(11,"pt"),
  legend.key.width=unit(26,"pt"),
  legend.margin=margin(,,6,-2),
  legend.position="top",
  legend.spacing.x=unit(3,"pt"),
  legend.spacing.y=unit(0,"pt"),
  legend.text=element_text(size=11,vjust=.5),
  legend.title=element_blank(),
  panel.background=element_blank(),
  panel.border=element_rect(fill=NA,linewidth=.4),
  panel.spacing=unit(2,"pt"),
  plot.margin=margin(6,6,1,5),
  plot.title=element_text(size=11.5,face=2,margin=margin(2,,4)),
  plot.subtitle=element_text(size=10.5,margin=margin(,,4)),
  strip.background=element_blank(),
  strip.text=element_blank())
ggsave("1.png",width=6.3,height=4.8,dpi=300*4)
sub="\u00a0    The monthly number of deaths from all causes was compiled from table 1-6 here: www.e-stat.go.jp/stat-search/files?page=1&layout=datalist&toukei=00450011&tstat=
000001028897&cycle=1&tclass1=000001053058&tclass2=000001053060&cycle_facet=tclass1&
tclass3val=0. Population estimates on October 1st of each year were spline interpolated to monthly population estimates: www.e-stat.go.jp/en/stat-search/files?page=1&toukei=00200524&tstat=
000000090001 (annual reports for 2021-2023 combined with 2000-2020 time series reports). The 2015 Japanese standard population was used by 5-year age groups up to ages 95+: www.mhlw.go.jp/content/10700000/000557741.pdf.
     The seasonality-adjusted baseline was calculated by adding the average difference between the linear trend and actual ASMR in January 2010-2019 to the baseline for each January, and similarly for other months."
system(paste0("f=1.png;mar=100;w=`/usr/local/bin/identify -format %w $f`;/usr/local/bin/magick \\( $f \\) \\( -size $[w-mar*2]x -font Arial -interline-spacing -3 -pointsize $[42*4] caption:'",gsub("'","'\\\\'",sub),"' -gravity southwest -splice $[mar]x80 \\) -append -resize 25% -dither none -colors 256 1.png"))

So cancer deaths in Japan are not even elevated that much compared to deaths from other causes.

The last author Masanori Fukushima may have also published a fake survey

The last author of the Japanese cancer paper was Masanori Fukushima, who is a Japanese alt media celebrity who is frequently featured in videos of press conferences that appear similar to Füllmich's press conference larping operations. While I was researching Twitter bots that promote content from the controlled alternative media, I saw many of them posting videos of Fukushima. Videos by the first author Miki Gibo have also been posted by bots that promote Miles Guo. [https://x.com/search?q=miki%20gibo&f=live]

Masanori Fukushima was also the last author of a paper from December 2024 about a Japanese vaccination survey. The paper about the survey was published by Cureus which also published the previous version of the Japanese cancer paper. However the paper about the survey might have been AI-generated: [rootclaim2.html#Japanese_survey_for_self_reported_incidence_of_COVID]

Comments about COVID statistics (part 2) - sars2.net

Contents

McCullough Foundation: Paper about cardiac arrests in King County EMS data

Comparison to data from original PDFs