Skip to content

Tracking century of excess mortality in Switzerland πŸ‡¨πŸ‡­, Sweden πŸ‡ΈπŸ‡ͺ & Spain πŸ‡ͺπŸ‡Έ

License

Notifications You must be signed in to change notification settings

RPanczak/ISPM_excess-mortality-historical

Folders and files

NameName
Last commit message
Last commit date

Latest commit

Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 
Β 

Repository files navigation

Tracking excess mortality in Switzerland πŸ‡¨πŸ‡­, Sweden πŸ‡ΈπŸ‡ͺ & Spain πŸ‡ͺπŸ‡Έ.

Citations and links

Paper

Data & code to reproduce paper:

Staub K*, Panczak R*, Matthes KL, Floris J, Berlin C, Junker C, Weitkunat R, Mamelund SE, Zwahlen M*, Riou J* (2021) Historically high excess mortality during the COVID-19 pandemic in Switzerland, Sweden and Spain. Annals of Internal Medicine. [Epub ahead of print 1 February 2022]. doi:10.7326/M21-3824
* Equal contribution.

Code version 2.1.

Archived on DOI

Preprint

Earlier version of preprint:

Staub K*, Panczak R*, Matthes KL, Floris J, Berlin C, Junker C, Weitkunat R, Mamelund SE, Egger M, Zwahlen M*, Riou J* (2021) Pandemic excess mortality in Spain, Sweden, and Switzerland during the COVID-19 pandemic in 2020 was at its highest since 1918. medRxiv 2021.08.12.21261825. doi: https://doi.org/10.1101/2021.08.12.21261825
* Equal contribution.

Code version 1.0.

Archived on DOI

Online documentation

Details of the data sources, data preparation and main as well as supplementary analyses are available on the project pages.

Visual summary of repo content here.

Content of this repository

Structure

.
+-- analyses
+-- data
+-- data-raw
+-- docs
+-- paper
|   \-- supplementary
+-- R
+-- stan
\-- ubelix

analyses folder

This folder contains set of literate programming R scripts written either in R or inR flavoured markdown. The files are numbered sequentially in order in which they should be executed. The key parts of this folder include:

  • 01_data-prepare.Rmd this script download, prepares data from The Human Mortality Database and also ancillary, country-specific datasets containing updates of monthly death counts, yearly age-specific death counts and population figures from statistical agencies. The output of these analyses is presented here.
  • 02_BfS-data-weekly.Rmd this script prepares and compares weekly data on death counts for Switzerland from two sources: Short-term Mortality Fluctuations and Swiss Federal Statistical Office
  • 03_joint-model-cmdstan.R this script runs the main analyses of the paper; it sources functions from R folder which in turn source and compile Stan programs from stan directory. One main model is fitted and then three sensitivity analyses performed for each country and analysis year combination. The script can be run from command line with a parameter in range 1-3 to obtain country specific results - examples of such calls can be found in ubelix folder scripts; results will be stored in folder outputs_YYYY-MM-DD in data directory. Note that depending on the architecture these analyses might take many hours or days to finish and we strongly recommend running them in appropriate computing environment - server, cloud or HPC. -04_joint-model-boot.R` this script performs additional sensitivity analysis; due to lower computational demands of this analyses they can be performed reasonably fast on a decently equipped laptop.
  • 05_outputs.Rmd this script combines results obtained by running scripts 03 and 04 and prepares outputs (figures and tables) for the paper and supplement. The output of these analyses is presented here
  • 06_comparison_week_month.Rmd this script performs additional sensitivity analysis comparing estimates from weekly models to monthly models.

data-raw & data folders

Raw, unprocessed data are stored and accessed from the data-raw folder; prepared data are stored in data. Both of these folders contain subdirectories with names of the data source, for instance mortality_org folder stores data from the HMD whereas BfS folder stores data from Swiss Federal Statistical Office [Bundesamt fΓΌr Statistik (BfS)]. outputs_YYYY-MM-DD folders in data directory hold results of the analyses runs and can be reused in order to save the computational time. Three files in data directory with names starting with results_ store outputs of the main models that were used for reporting in main figures and tables of the manuscript.

docs folder

This folder stores html files generated from the scripts in the analyses folder, with the names equivalent to the scripts that are used to generate them. These files are used on the project pages.

paper folder

This folder stores figures and tables presented in the paper and in the supplementary materials.

R folder

This folder stores various scripts used for data manipulation or analyses by the scripts from the analyses folder. Most important scripts include:

  • fn_age_serfling_nb_cmdstan.R used for the main analyses (age adjusted, negative binomial model implemented in Stan)
  • fn_global_serfling_nb_cmdstan.R used for the sensitivity analyses (unadjusted, negative binomial model implemented in Stan)
  • fn_global_serfling.R and fn_boot_pi.R used for sensitivity analyses (unadjusted Poisson model)

stan folder

Scripts from R script above make calls to Stan programs stored in this directory. Names of this programs are matching the scripts calling them (age_serfling_nb.stan & global_serfling_nb.stan). Both of the scripts have versions with 21 in the name of the file which are special cases of each script designed to handle incomplete 2021 data. Note that the programs will be compiled on their first run and will create additional files with the same name and extensions depending on the platform (for instance .exe on Windows OS; shell scripts on *NIX systems).

ubelix folder

This folder stores example OpenPBS scripts that can be used to submit the analyses scripts to HPC cluster.

secrets folder

This folder should store username & password for accessing Human Mortality Database in plain text files named username.txt & password.txt.

Computing environment

Details on the version of R software and packages used are presented at the bottom of the output documents stored in docs folder. Script 03_joint-model-cmdstan.R was run on HPC cluster using R version 4.1.0 with packages versions cmdstanr 0.4.0 & cmdstan 2.27.0.

Visual summary of main results

Estimates of excess

Yearly differences

Yearly differences

Monthly differences

Monthly differences

Four pandemics

Four pandemics

Four pandemics - percent

Age effects

Four pandemics

CC BY-NC-SA 4.0