guibuzi / covid_multi_strain

Code for 'Impact of vaccinations, boosters and lockdowns on COVID-19 waves in French Polynesia'

Impact of vaccinations, boosters and lockdowns on COVID-19 incidence

This repository contains code and data for the analyses in the preprint 'Impact of vaccinations, boosters and lockdowns on COVID-19 waves in French Polynesia'. The aim of this work is to estimate the counterfactual impact of vaccinations, boosters and lockdowns on incidence of COVID-19 cases, hospitalisations and deaths in French Polynesia. We do this by fitting an age-structured multi-strain model of SARS-CoV-2 transmission to data on daily numbers of COVID-19 confirmed cases, hospitalisations and hospital deaths in French Polynesia from July 2020 to May 2022, and data from two sero-surveys conducted in February 2021 and November-December 2021 [1]. The model is written in the odin syntax and compiled to work with the dust discrete-time C++ simulation engine using odin.dust [2]. The code also uses the particle filter tools of the mcstate R package, but implements a bespoke adaptive MCMC algorithm for fitting the model.

Requirements

The eigen1, odin, dust, odin.dust and mcstate R packages are required to run the code. Since these packages are in constant development it is probably best to use the same versions of these packages as were used for this analysis:

eigen1 0.1.1
odin 1.3.2
dust 0.11.24
odin.dust 0.2.16 
mcstate 0.9.0

These can be installed in R with:

remotes::install_github(c(
  "mrc-ide/[email protected]",
  "mrc-ide/[email protected]",
  "mrc-ide/[email protected]",
  "mrc-ide/[email protected]",
  "mrc-ide/[email protected]"))

Installation

Clone/download this project onto your machine.

Data

All data required to run the code are contained in the data folder and are also available on Zenodo: doi:10.5281/zenodo.8320333

Running the code

The odin model code is contained in inst/odin/covid_multi_strain.R. If required, the model can be modified by editing this code, and then recompiled by calling

odin_dust("inst/odin/covid_multi_strain.R")

as in fit.R.

The whole analysis (fitting + counterfactual simulations) can be run by running the top-level script run.R:

source("run.R")

This runs the fitting function run_fitting() and the function for running the counterfactual simulations without vaccinations, boosters, and lockdowns, and with different lockdown timings run_simulations(). It produces output called MCMCoutput<run>.RDS and cntfctl_output<sim_run>.RDS in the output folder, where <run> and <sim_run> are the MCMC run and simulation sim_run numbers set in run.R, along with trace and posterior plots of the parameters and model fit produced by plot_fit.R and counterfactual outcome plots.

The breakdown of population immunity over time (Fig. 4 in the preprint) can be plotted with:

source("R/plot_immunity.R")

setting the run number in plot_immunity.R to match that in run.R.

Note that the fitting code takes ~2.5hrs to run 4 MCMC chains of 50,000 iterations in parallel (in R 4.1.0 on an 8-core MacBook Pro with an M1 chip and 16GB RAM).

Built With

• R version 4.1.0 (2021-05-18)

Authors

• Lloyd Chapman: [email protected]

License

This project is licensed under the GNU General Public License v3.0 - see the LICENSE.txt file for details

Citation

Chapman LAC, Aubry M, Maset N, Russell TW, Knock ES, Lees JA, Mallet H-P, Cao-Lormeau V-M, Kucharski AJ. Impact of vaccinations, boosters and lockdowns on COVID-19 waves in French Polynesia. medRxiv. 2023.

References

1. Aubry M, Maset N, Chapman L, Simon A, Olivier S, Bos R, Chung K, Teiti I, Kucharski A, Mallet H-P, Cao-Lormeau V-M. Seroprevalence of sars-cov-2 antibodies in french polynesia and perspective for vaccine strategies. Preprints, 2022. doi:10.20944/preprints202212.0386.v1

2. FitzJohn RG, Knock ES, Whittles LK, Perez-Guzman PN, Bhatia S, Guntoro F, Watson OJ, Whittaker C, Ferguson NM, Cori A, Baguelin M, and Lees JA. Reproducible parallel inference and simulation of stochastic state space models using odin, dust, and mcstate. Wellcome Open Research, 5:288, 12 2021. doi:10.12688/wellcomeopenres.16466.2