Rmd source

Total excess mortality

The Data

Deaths data from Eurostat (demo_r_mwk_ts):

## Warning in eurotime2date(x, last = FALSE): Unknown time code, W. No date conversion was made.
## 
##             Please fill bug report at https://github.com/rOpenGov/eurostat/issues.

We remove all countries which failed to provide full report in 2015–2019 (actully it is one such country, Ireland):

## summarise (nn=n()) returns number of weeks 
## we assume at least one country provides complete report;
## so max(nn) is the number of weeks from 2015--2019
non.full.reporting <- z %>% filter ( year >= 2015 & year < 2020) %>%
  group_by(geo) %>%
  summarise (nn = n()) %>% 
  mutate(mx = max(nn)) %>%
  filter (nn != mx)

Just to be sure:

as.data.frame(non.full.reporting)
##   geo nn  mx
## 1  IE 12 261

Remove them:

incompleted.countries <- as.vector(non.full.reporting$geo)
z <- z %>% filter (! geo %in% incompleted.countries)

We define (weekly) excess mortality as a difference between number of deaths in 2020–2021 (z1) and 5 year average (2015–2019) for corresponding weeks (z0). The absolute excess mortality (exm) is defined as z1 - z0 and the relative excess mortality (exp) is (z1 - z0)/z0 · 100%:

## mean weekly deaths 2015--2019
z0 <- z %>% filter ( year >= 2015 & year < 2020) %>% 
  group_by(geo, week) %>% 
  summarise (d0 = mean(value, na.rm=T))
## weekly deaths 2020--2021
z1 <- z %>% filter ( year > 2019 ) %>% 
  group_by(geo, year, week) %>% 
  summarise ( d1 = sum(value))

## join z0 z1 and compute differences
zz <- left_join(z0, z1, by=c("week", "geo")) %>% 
  drop_na(d0,d1) %>%
  mutate (exp = (d1 - d0)/d0 * 100, exm = d1 - d0 )
zz <- left_join(zz, nn, by=c('geo'='id'))

Compute last week each country reported:

## if NA then the country stop reporting in 2020
zz.last.week <- zz %>% filter (year == 2021) %>% 
  group_by (geo) %>% 
  summarise (lw = last(week)) %>%
    drop_na(lw)
latestweek <- max (zz.last.week$lw)

The latest data is from 49 week, the oldest data is from 12 week. Countries which did not report in 2021 (UK for example, were removed)

We summarise all weeks (exm) for each country. Countries which reported last earlier than for 43 week are excluded.

There are 29 countries which reported last for week 43 or later.

Excess mortality by country

Total excess mortality 2020–2021

Per 1 mln population

Percent of `normal’ deaths:

The same as above with Poland highlighted

Dynamics of excess mortality 2020–2021

Faceted line plots

Poland and the rest of Europe:

Polar coordinates (I like Nightingale rose chart)

Excess deaths vs COVID deaths

We use Covide19 deaths/cases from OWiD project (https://github.com/owid/covid-19-data/tree/master/public/data) From the dataset we select country iso code and name (location) and population, daye, number of cases and deaths. Then compute year, week number in a year (using date2ISOweek/do not use as.Date conversion). Next compute summaries for weeks and finally compute per 1mln ratios for cases and deaths.

fileC <- "../../COVID/OWiD/owid-covid-data.csv"
mil <- 1000000

c19d <- read.csv(fileC, sep = ',',  header=T, na.string="NA" ) %>%
  filter (continent == 'Europe') %>%
  select (iso_code, 
          location, date, 
          new_cases, new_deaths, 
          pop=population
  ) %>%
  mutate (yyyy = as.numeric(substr(date, 1, 4)),
          ww = as.numeric( substr(date2ISOweek(date), 7,8))
  ) %>%
  group_by(iso_code, yyyy, ww) %>%
  summarise(
    pop = last(pop),
    nd = sum(new_deaths, na.rm=T),
    nc = sum(new_cases, na.rm=T),
  ) %>%
  ## zamień na 1mln
  mutate (
    ## rok-tydzień-dzieńtygodnia
    date = ISOweek2date(sprintf("%i-W%02i-1", yyyy, ww)),
    date_string = sprintf("%i-%02i-1",yyyy, ww),
    nd1m = nd / pop * mil,
    nc1m = nc / pop * mil
  )

Dynamics of excess deaths vs COVID deaths

Difference between excess deaths and COVID deaths

Total difference between excess deaths and COVID deaths

Excess deaths as percent of COVID related deaths

Scatter-plot for excess deaths vs covid deaths (red line is a diagonal)

Excess mortality by age groups

The Data

Deaths data from Eurostat (demo_r_mwk_10_ts; Deaths by week, sex and 10-year age group; https://ec.europa.eu/eurostat/databrowser/view/demo_r_mwk_10/default/table?lang=en):

## Warning in eurotime2date(x, last = FALSE): Unknown time code, W. No date conversion was made.
## 
##             Please fill bug report at https://github.com/rOpenGov/eurostat/issues.
##  [1] "TOTAL"  "UNK"    "Y_GE80" "Y_GE90" "Y_LT10" "Y10-19" "Y20-29" "Y30-39"
##  [9] "Y40-49" "Y50-59" "Y60-69" "Y70-79" "Y80-89"

The age groups: TOTAL, UNK, Y_GE80, Y_GE90, Y_LT10, Y10-19, Y20-29, Y30-39, Y40-49, Y50-59, Y60-69, Y70-79, Y80-89. We skip all sex levels (F, M, T) except T (ie total). We combine Y_LT10–Y50-59 into one group (00-59) and recode the names of the remaining groups. Group Y_GE80 is redundant so we skip it:

There are rlevels(z$age)` age groups now.

Veryfing data consistency

Check how big is UNK (unknown) group:

There are rtotal.unknown` unclassified deaths. We can safely skip this age group.

We compare total deaths and sum of the deaths from other age categories. There should be no difference if the data is consistent:

Some countries have problems:

geo td mfd diff
AL 146985 146929 56
DK 802986 802984 2
EE 363776 363101 675
EL 858096 858046 50
ES 8671762 8671418 344
FR 5405483 5392285 13198
HR 1137121 1136946 175
HU 2891654 2891421 233
LV 665998 665898 100
PT 2363331 2363050 281
RS 2249907 2248710 1197
UK 3693775 3693483 292

French is strench (as usual):

year total 90-and-more 00-59 60-69 70-79 80-89
2013 567712 136887 71014 69627 94020 196163
2014 557102 137442 68552 70308 90667 190133
2015 604310 160424 70529 75123 95559 202674
2016 590836 160662 66894 73957 94733 194590
2017 604105 171141 65825 73933 97866 195340
2018 607786 175503 65481 73432 100728 192642
2019 611297 178731 63662 72160 103186 189473
2020 678679 203626 65065 76458 119094 209300
2021 583656 172219 57116 67969 109192 173185

There are no data for age groups for France in 2020. So age-group analysis for this country is not possible.

Total excess deaths by age groups

Poland and the rest of Europe:

Dynamics of excess mortality 2020–2021