covid-sep-ch

Code and supplementary material to the article

Riou J., Panczak R., Althaus C.L., Junker C., Perisa D., Schneider K., Criscuolo N.G., Low N., Egger M. 2021 "Socioeconomic position and the cascade from SARS-CoV-2 testing to COVID-19 mortality: a population-based analysis of Swiss surveillance data", Lancet Public Health 6(9):e683-e691. doi:10.1016/S2468-2667(21)00160-2 .

Observations:

• This code is not meant to be run directly (as the data cannot provided by default), but rather to better explain the approach and facilitate its adaptation to other contexts and datasets.

• We use R version version 4.0.0 code and rely mainly on the following packages: tidyverse 1.3.0, rstanarm 2.21.1 and sf 1.0.0.

• Extraction, management and formatting of the individual level data available at the Swiss Federal Office of Public health (FOPH) is done in analyses/ folder using the R files with names starting with FOPH-. The objectives of these scripts are to

1. apply exclusion criteria;
2. create variables;
3. match individual geocodes with the closest Swiss neighbourhood index of socioeconomic position (SEP);
4. match individual geocodes with the closest care facility (SOMED);
5. aggregate the individual data by period, SEP decile, age group, sex, and canton.

The original individual-level data is not available to share for confidentiality reasons. The aggregated data is available on motivated request to the authors (julien.riou@ispm.unibe.ch).

• The code for the main statistical analysis is in analyses/run_models.R file. This script is best run from the console (e.g. using a high-performance computer cluster) with command Rscript run_models.R i, where i is the number of the dataset defined in data_files. The outputs are several sets of posterior samples for each relevant combination of model types, numerator and denominator. This script can be adapted to any data set, as long as it is aggregated and formatted in the same way:

Variable	Type	Comments
canton	character	Canton or other geographical division
period	double	First/second wave or other temporal division
sex	double	Binary indicator
age_group_f	factor	Age in 10-year bands
age_group_f2	factor	Same as age_group_f but all age groups below 50 are pooled
ssep_d	int	SEP index in 10 groups divided in deciles
n_test	double	Counts of total tests
n_pos	double	Counts of positive tests
n_hospit	double	Counts of hospitalisations
n_icu	double	Counts of ICU admissions
n_death	double	Counts of deaths
n_pop	double	Counts of population (to be used as a denominator only)

• The scripts paper_outputs.r and all other scripts starting with po_ in folder analyses/ take the raw data and posterior samples to produce formatted results, tables and figures for the main paper and the supplementary.