gshartnett / network_vaccination

Contains code used to simulate simple epidemic models on network datasets.

Network Vaccination

This repository contains the code and data used in two recent publications that used mobile device data to model the spread of COVID-19 throughout a partially vaccinated population:

• Protecting the Most Vulnerable by Vaccinating the Most Active (RAND Perspective report)
• Modeling the Impact of Social Distancing and Targeted Vaccination on the Spread of COVID-19 through a Real City-Scale Contact Network.

Project description

Motivated by the (then) imminent availability of COVID-19 vaccines, this repository was used to compare the efficacies of different network-based vaccination strategies. It is a well-known result in network science that targeted strategies, where more connected nodes are vaccinated first, can sometimes far outperform the strategy where nodes are vaccinated uniformly at random. ¹ It also seems plausible that vaccinating nodes with low connectivities is an even worse strategy.

We investigated these hypotheses using a real-world person-to-person contact network derived from mobile device data obtained from the company Uber Media. Rather than attempting to faithfully simulate the spread of COVID-19 on this network, we instead simulated two rather simple contagion models (SIR and SEIR). Consequently, we are simulating a very idealized contagion model on a realistic network - in future work we would like to improve this by considering more realistic contagion models.

How to use this repository

Installation

• Download the repo: git clone https://github.com/RANDCorporation/network_vaccination
• Navigate to repo: cd network_vaccination
• Install required packages: pip install -r requirements.txt

It is recommended to use a conda environment, in which case the command sequence should be modified to

• conda create -n network_env
• conda activate network_env
• git clone https://github.com/RANDCorporation/network_vaccination
• cd network_vaccination
• pip install -r requirements.txt

Data

The mobile device data was used to create 3 different anonymized contact networks:

• G_pre capturing interactions before social distancing measures were enacted
• G_post capturing interactions during social distancing
• G_superposition representing a combined network which is described in the Perspective report in more detail

The weighted adjacency lists for these anonymized contact networks are located in the data directory.

Code

The repository is primarily organized into multiple Jupyter notebooks, each accomplishing a different task.

• Contact Network Analysis.ipynb: This notebook analyzes the person-to-person contact networks used in the simulations.
• Code used in Protecting the Most Vulnerable by Vaccinating the Most Active:
  • SIR Model.ipynb: This notebook carries out the SIR simulation over G_superposition.
  • SIR Data Analysis.ipynb: This notebook analyzes the results of the simulation and makes some plots.
• Code used in Using Mobile Device Data to Measure The Impact of Social Distancing Measures and Optimal Vaccination Strategies for COVID-19:
  • SEIR Model.ipynb: This notebook carries out the SEIR simulation over the G_pre and G_post contact networks.
  • SEIR Data Analysis.ipynb: This notebook analyzes the results of the simulation and makes some plots.

The utils.py folder also contains some useful functions used by the various notebooks.

Contact

Code developed and maintained by Gavin Hartnett ([email protected]).

1 Pastor-Satorras, Romualdo, and Alessandro Vespignani. "Immunization of complex networks." Physical review E 65.3 (2002): 036104. ArXiv link. ↩