migp11/epicommute

Simulate an epidemic metapopulation model with mobility-reducing containment strategies

EpiCommute

Simulate an epidemic on a metapopulation network with commuter-type mobility, and potential mobility-reducing containment strategies.

The model is used and defined in the following publication:

"COVID-19 lockdown induces structural changes in mobility networks -- Implication for mitigating disease dynamics", Frank Schlosser, Benjamin F. Maier, David Hinrichs, Adrian Zachariae, Dirk Brockmann, (https://arxiv.org/abs/2007.01583)

Install

python setup.py install

Usage example

>>> import numpy as np
>>> from EpiCommute import SIRModel
>>> # Create dummy data
>>> M = 10 # Number of subpopulations
>>> mobility = np.random.rand(M, M) # mobility matrix
>>> subpopulation_sizes = np.random.randint(20,100,M) # subpop.-sizes
>>> # Run simulation
>>> model = SIRModel(mobility, subpopulation_sizes)
>>> results = model.run_simulation(VERBOSE=True)
Starting Simulation ...
Simulation completed
Time: 0min 3.35s

More examples are given in the notebooks at /examples.

Model description

The code simulates an SIR epidemic on a subpopulation network, where subpopulations are connected by commuter-type mobility.

A detailed descriptions of the model is given in the mauscript linked above.

Mobility

Movement of individuals between subpopulation is implemented using commuter-type dynamics. This means that each individual lives at a home location i, and works at a work location j.

How the individuals are distributed among the compartments is determined by an origin-destination mobility matrix mobility of size M x M, which contains the number of individuals commuting between pairs of locations.

The population in the system is then distributed into M x M compartments, where compartment ij includes those individuals living at i and working at j.

Infection dynamics

Epidemic spread is simulated using the SIR model, consisting of susceptibles S, infecteds I and recovereds R.

The infection step is subdivided in two phases of equal length:

• In the home phase, each individual has a chance to get infected at their home location i.
• In the work phase, infections can take place at the work locations.

Containment/lockdown effects

The model can consider changes in absolute mobility flux (for example due to lockdown effects). For this, it is a assumed that a matrix mobility is provided with the current (possibly reduced) number of commuters, and a matrix mobility_baseline with the number of commuters during normal times.

Changes in mobility flux are taken into account in two different scenarios:

• In the isolation scenario, it is assumed that a reduction in mobility means that individuals are effectively removed from the system.
• In the distancing scenario, a reduction in mobility instead leads to a reduction in the effective transmission rate in the system.

A more detailed description of the scenarios and the model can be found in the publication.