COVID Campus

This code simulates the spread of COVID-19, as well as the effect of various mitigation measures, on an empirical contact network of college students collected through the Copenhagen Network Study. This work was published in the International Journal of Infectious Diseases. See https://doi.org/10.1016/j.ijid.2021.10.008 for more information.

Cite the Paper: Hambridge HL, Kahn R, & Onnela JP (2021). Examining SARS-CoV-2 Interventions in Residential Colleges Using an Empirical Network. International Journal of Infectious Diseases, 113, 325-330.

Cite the Code:

File Overview

Core Functions

• testing_freq.py: main class and methods for running simulations examining testing frequency
• utils.py: helper functions for constructing weighted adjacency matrices, assigning non-pharmaceutical interventions (NPIs), and incorporating NPIs into dyadic transmission probabilities
• plotfx.py: some basic plotting functions to show metrics over time and by testing frequency
• Simulation_Examples.ipynb: example code showing how to run simulations

Data Processing & Visualization

• CNS_DataProcessing.ipynb: loads, cleans, and analyzes Copenhagen Network Study data; removes weak ties, empty scans, and non-participating devices
• Network_Plots.ipynb: creates network plots for Copenhagen Network Study data with connected components arranged in a grid for each day

Generate Results

• beta003_noNPI_vax_hicom.py: runs simulations for R0 ~ 1.5 with high transmission from the community; no mask wearing or social distancing
• beta006_noNPI_vax_hicom.py: runs simulations for R0 ~ 3.0 with high transmission from the community; no mask wearing or social distancing
• beta009_noNPI_vax_hicom.py: runs simulations for R0 ~ 4.5 with high transmission from the community; no mask wearing or social distancing
• beta003_noNPI_vax_locom.py: runs simulations for R0 ~ 1.5 with low transmission from the community; no mask wearing or social distancing
• beta006_noNPI_vax_locom.py: runs simulations for R0 ~ 3.0 with low transmission from the community; no mask wearing or social distancing
• beta009_noNPI_vax_locom.py: runs simulations for R0 ~ 4.5 with low transmission from the community; no mask wearing or social distancing
• beta006_multiNPI_0vax_hicom.py: runs simulations for R0 ~ 3.0 with high transmission from the community; no baseline immunity; proportion of the population wearing masks and socially distancing
• beta006_multiNPI_20vax_hicom.py: runs simulations for R0 ~ 3.0 with high transmission from the community; 20% baseline immunity; proportion of the population wearing masks and socially distancing
• beta006_multiNPI_40vax_hicom.py: runs simulations for R0 ~ 3.0 with high transmission from the community; 40% baseline immunity; proportion of the population wearing masks and socially distancing
• beta006_multiNPI_0vax_hicom_correlated.py: runs simulations for R0 ~ 3.0 with high transmission from the community; no baseline immunity; proportion of the population wearing masks and socially distancing; mask wearing clustered on the proximity network

Analyze & Visualize Results

• Clean_noNPI_Cluster_Results.ipynb: loads, combines, and restructures output from beta003_noNPI_vax_hicom.py, beta006_noNPI_vax_hicom.py, beta009_noNPI_vax_hicom.py, beta003_noNPI_vax_locom.py, beta006_noNPI_vax_locom.py, and beta009_noNPI_vax_locom.py scripts; results used in noNPI_Plots.ipynb
• noNPI_Plots.ipynb: creates plots showing infections, tests, and isolations over time for R0 ~ 1.5, R0 ~ 3.0, and R0 ~ 4.5 for both high and low community transmission settings; generates plots by percent with baseline immunity
• NPI_Cluster_Results_hicom.ipynb: creates heatmaps comparing average cumulative incidence for various testing frequencies, proportion mask wearing, proportion social distancing under high community transmission scenario
• NPI_Cluster_Results_locom.ipynb: creates heatmaps comparing average cumulative incidence for various testing frequencies, proportion mask wearing, proportion social distancing under low community transmission scenario
• NPI_Cluster_Results_Correlated.ipynb: creates heatmaps comparing random assignment of mask wearing with clustered assignment of mask wearing; social distancing was randomly assigned in both scenarios
• Relative_Effectiveness.ipynb: analyzes relative effectiveness of various mitigation measures using linear regression model

Usage

Below is a basic example of how to use the files to run testing frequency simulations. For a full working example, see Simulation_Examples.ipynb.

1. Import the necessary packages/files.

import pandas as pd
import numpy as np
from testing_freq import *
from utils import *

2. Read in the data. Here we use a cleaned version of the Copenhagen Networks Study data. timestamp is in seconds starting at 0, user_a and user_b are integer user IDs, and rssi is an integer received signal strength indicator. The cleaned data has been scrubbed of empty scans and non-participating users. For more information on how the cleaned file was generated, see CNS_DataProcessing.ipynb.

bt = pd.read_csv('bt_data_clean.csv', header = 0, names = ['timestamp','user_a', 'user_b', 'rssi'])

3. Construct weighted adjacency matrices for each timestep from the data. time_step indicates the number of seconds each time step should represent. data_loops indicates how many extra times the data should be looped through to extend the length of the simulation. dist_thres indicates the RSSI cutoff for what is considered a close contact. Returns a 3D numpy array with one weighted adjacency matrix per time step. Cell entries indicate the number of interactions between a pair of nodes. df is the original dataframe with a time_step ID column added.

adj_mats, df = construct_adj_mat(bt, time_step = 86400, data_loops = 3, dist_thres = -75)

4. Set up simulation parameters. See the utils.py file for additional options.

disease_params = dict()
disease_params['ext_inf'] = 0.002
disease_params['asymp'] = 0.3
disease_params['sigma_a'] = 1/3
disease_params['sigma_s'] = 1/3
disease_params['gamma_a'] = 1/7
disease_params['gamma_s'] = 1/12
disease_params['init_status'] = gen_init_status(n_nodes = adj_mats.shape[1], asymp = disease_params['asymp'], n_init_inf = 1, seed = 1)
disease_params['beta'] = gen_trans_prob_NPI(n_nodes = adj_mats.shape[1], base_beta = 0.006157, p_sd = 0.2, p_fm = 0.2, seed = 1)
disease_params['n_time_day'] = 1

test_params = dict()
test_params['test_freq'] = 7
test_params['spec'] = 0.99
test_params['symp_test_delay'] = gen_symp_test_delay(n_nodes = adj_mats.shape[1], univ_delay = 3)
test_params['nc_schd'] = 0.01 
test_params['nc_symp'] = 0.25
test_params['time_dep'] = True
test_params['time_dep_type'] = 'W'
test_params['false_symp'] = 0.005

quar_params = dict()
quar_params['quar_delay'] = gen_quar_delay(n_nodes = adj_mats.shape[1], univ_delay = 1)
quar_params['quar_comp'] = gen_quar_comp(n_nodes = adj_mats.shape[1], seed = 1)
quar_params['quar_len'] = 10

5. Instantiate the simulation class and run a simulation with testing.

testin = TestFreq(adj_mats, disease_params, test_params, quar_params)
(ia_nodes_byt, is_nodes_byt, test_pos_schd_byt, test_pos_symp_byt, q_schd_byt, q_symp_byt) = testin.sim_spread_test(seed = 1)