Changing travel patterns in China during the early stages of the COVID-19 pandemic

Authors: Hamish Gibbs*¹, Yang Liu*¹, Carl AB Pearson¹, Christopher I Jarvis¹, Chris Grundy¹, Billy Quilty¹, Charlie Diamond¹, LSHTM CMMID COVID-19 working group¹, Rosalind M Eggo¹
* authors contributed equally
Affiliations:
¹ Department of Infectious Disease Epidemiology, London School of Hygiene & Tropical Medicine, Keppel Street, London. WC1E 7HT. UK

Open respository of code supporting the analyses in the paper: Changing travel patterns in China during the early stages of the COVID-19 pandemic. Available on medRxiv.

Analyses are conducted using a symmetrical matrix of movement flux data collected from the Qianxi Baidu 2020 movement map. The full dataset is included in this repository at data/china_prf_connectivity_0101_0301.rds.

This research also relies on Covid-19 case count data and prefecture level hospital resource data from China, C. D. C. Public Health Science Data Center, as well as prefecture level population data from China Statistical Yearbook 2018. Covid-19 case count data, prefecture level hospital resource data, and prefecture level population data require applications for use.

Software versions

R code is compatible with R ~3.6. Python code is compatible with Python ~3.7.

Dependencies

R packages:
ggplot2 3.3.0
dplyr 0.8.3
readr 1.3.1
ggrepel 0.8.2
cowplot 1.0.0
data.table 1.12.8
cluster 2.1.0
factoextra 1.0.7
mapview 2.6.11
magrittr 1.5
ggpubr 0.3.0
viridis 0.5.1
sf 0.9

Python packages:
glob 10.7
pandas 1.0.3
igraph 0.8.2
leidenalg 0.8.0
re 3.8.3
numpy 1.19.0

Installation

Clone this repository and use Python or R to run the source files. See the file reference below for details on required datasets and expected outputs.

Overview

-calc_Access2Care.R calculates access to healthcare and healthcare pressure on the prefecture level. -calc_Connectivity_ByPopQ.R contains stratified analyses of inter-prefecture travel patterns.
-extract_ts.R extract origin or destination specific travel time-series.
-reduce2_ntile.R reduces dimension of connectivity matrix from 366*366 dimensions to n*n based on population.
-rank_variability.R compare the travel surge in Wuhan to the rest of mainland China before Lunar New Year.
-trajectory_clustering_analysis.R cluster normalized outflow trajectories from Chinese prefectures. -create_cluster_piv.R create a long version of clustering analysis results.
-create_cluster_sum.R create a summary of clustering analysis results.
-leiden.py compute daily prefecture communities.
-all_module_Q_breakdown.R compute Q of all identified communities.
-community_sizes.R compute size of all identified communities.

Instructions for use:

After downloading this repository, load the datasets required to conduct each analysis. Please see the file reference below.

Demo:

Because of the application requirement for some datasets used in this analysis, it is not possible to provide a full demonstration of the output of each file. Please see the expected output reference below.

File reference:

calc_Access2Care.R
Datasets required:
Movement matrix, case count data, hospital location data.
Expected output:
Plots of changes in traveler volume and timeseries box plots of changes in healthcare pressure.

calc_Connectivity_ByPopQ.R
Datasets required:
Movement matrix, population data.
Expected output:
Plot of in- and out-bound traveler volume by population quartile.

extract_ts.R
Datasets required:
Movement matrix.
Expected output:
Plot of in- or out-bound travel volume for a single origin or destination.

rank_variability.R
Datasets required:
Movement matrix, population data.
Expected output:
Plot of peak outflow deviation from 2019 outflow values for all prefectures.

trajectory_clustering_analysis.R
Datasets required:
Movement matrix.
Expected output:
Line plot and silhouette plot of clustered outflow trajectories for a single origin prefecture for a given number of clusters.

create_cluster_piv.R
Datasets required:
Movement matrix.
Expected output:
"Long" output from trajectory_clustering_analysis.R.

create_cluster_sum.R
Datasets required:
Movement matrix, case count data, population data.
Expected output:
Summary of prefecture cluster labels, population, and date of first case arrival.

leiden.py
Datasets required:
Movement matrix.
Expected output:
Dataset of community labels over time computed using the Leiden algorithm.

all_module_Q_breakdown.R
Datasets required:
Movement matrix.
Expected output:
Plots of community modularity over time.

community_sizes.R
Datasets required:
Movement matrix.
Expected output:
Dataset of prefecture community sizes over time.