Tabletop ABM for Davis County, Utah

This repository presents a tabletop agent-based model (ABM) for H5N1 spread in Davis County, Utah. The model utilizes age mixing data to simulate H5N1 transmission dynamics under various intervention scenarios, including isolation, quarantine, and post-exposure prophylaxis (PEP).

This is not a real situation and is intended for educational purposes only.

Description of the model

The implemented model is a Susceptible Exposed Infectious Hospitalized Recovered (SEIRH) model with mixing and quarantine features. The model features what we call entities, which are subgroups of the population defined by age groups and schools. School-age agents are assigned to schools based on the population data, and the rest of the population is assigned to age groups. The mixing patterns are given by an age-based contact matrix based on the Polymod study.

flowchart LR
    S[Susceptible] --> E[Exposed]
    E --> I[Infected]
    I --> H[Hospitalized]
    H --> R[Recovered]
    I --> R

Loading

The model features contact tracing, isolation of detected cases, and quarantine of contacts. Detection happens with uncertainty (80% success rate), and agents isolate and move to the quarantine states with certainty (for now). The model also allows for the implementation of post-exposure prophylaxis (PEP) for contacts of detected cases. The PEP is implemented as a baseline tool that reduces the probability of becoming infected and transmitting H5N1.

The quarantine process is as follows:

flowchart LR
    Start((Start)) --> infected{"Already<br>quarantined<br>or isolated?"}
    infected -->|Yes|End((End))
    infected -->|No|Infected{"Infected<br>(infectious)?"}
    Infected -->|Yes|WillToIsolate{"Willing to isolate"}
    Infected -->|No|vax
    WillToIsolate -->|Yes| Isolate((Isolate))
    WillToIsolate -->|No| End
    vax{"Will use PEP"}
    vax -->|Yes|End
    vax -->|No|WillQuarantine
    WillQuarantine{"Willing to<br>Quarantine?"} -->|No|End
    WillQuarantine -->|Yes|Quarantine((Quarantine))

Loading

Limitations of the model

The model has various assumptions that may not hold in real-world scenarios:

1. Like most, this model does not include a behavioral change from the agents’ perspective. This is something we cannot predict accurately. For example, parents deciding to send or not their kids to school, or how they might respond to public health messaging.

2. The disease affects equally all age groups, which is not the case for many diseases. This may be important since we are incorporating a component of the population that is school-age, and they may have different susceptibility and infectiousness profiles compared to adults.

3. We are not using real social network data, but rather a contact matrix that gives us the average number of contacts between age groups. This means that we are not capturing the heterogeneity in contact patterns that may exist in the real world. In previous research it has been demonstrated that social networks (clustering) may play an important role in the spread of infectious diseases, and this is not captured in our model.

Running the model

To execute this model, it is recommended to run it in a high-performance computing environment due to its computational intensity, especially when simulating multiple scenarios. The model is fast, but the Davis county population is large (over 350,000), and running multiple simulations can be time-consuming on a standard personal computer.

The following diagram illustrates the compartments and transitions in the SEIRH model:

flowchart TB

    %% Disease progression states
    subgraph Main[Disease Progression]
        direction TB
        S[Susceptible]
        E[Exposed]
        In[Infected]
        H[Hospitalized]
        R[Recovered]
    end

    S --> E
    E --> In
    In --> H
    H --> R
    In --> R

    %% Quarantine states
    Dh[Detected<br>Hospitalized]
    Qs[Quarantined<br>Susceptible]
    Qe[Quarantined<br>Exposed]
    I[Isolated]
    Ir[Isolated<br>Recovered]

    %% Infected to
    In <==> I
    In --> Ir
    In --> Dh

    %% Isolated to
    I --> R
    I --> Ir
    I --> H
    I --> Dh

    %% Susceptible quarantined
    S <==> Qs

    %% Exposed
    E <==> Qe

    Qe --> I
    Qe --> In

    Dh --> R

    %% Isolated recovered
    Ir --> R

Loading

Summary of results

R0	isolation	quarantine	pep	Probability of Outbreak Size >= 10
1.4	isolation no	quarantine no	PEP no	0.18
1.4	isolation no	quarantine no	PEP yes	0.18
1.4	isolation yes	quarantine no	PEP no	0.04
1.4	isolation yes	quarantine no	PEP yes	0.04
1.4	isolation yes	quarantine yes	PEP no	0.04
1.4	isolation yes	quarantine yes	PEP yes	0.04
1.9	isolation no	quarantine no	PEP no	0.32
1.9	isolation no	quarantine no	PEP yes	0.32
1.9	isolation yes	quarantine no	PEP no	0.14
1.9	isolation yes	quarantine no	PEP yes	0.14
1.9	isolation yes	quarantine yes	PEP no	0.12
1.9	isolation yes	quarantine yes	PEP yes	0.12
2.4	isolation no	quarantine no	PEP no	0.42
2.4	isolation no	quarantine no	PEP yes	0.42
2.4	isolation yes	quarantine no	PEP no	0.21
2.4	isolation yes	quarantine no	PEP yes	0.21
2.4	isolation yes	quarantine yes	PEP no	0.13
2.4	isolation yes	quarantine yes	PEP yes	0.13

Summary of the probability of an outbreak size greater than or equal to 10 for each scenario in Davis County.

Links to the scenarios

The following table links to the reports generated for each of the scenarios run for Davis County:

R0	Isolation	Quarantine	PEP	Link
1.4	no	no	no	View Report
1.4	no	no	yes	View Report
1.4	yes	no	no	View Report
1.4	yes	no	yes	View Report
1.4	yes	yes	no	View Report
1.4	yes	yes	yes	View Report
1.9	no	no	no	View Report
1.9	no	no	yes	View Report
1.9	yes	no	no	View Report
1.9	yes	no	yes	View Report
1.9	yes	yes	no	View Report
1.9	yes	yes	yes	View Report
2.4	no	no	no	View Report
2.4	no	no	yes	View Report
2.4	yes	no	no	View Report
2.4	yes	no	yes	View Report
2.4	yes	yes	no	View Report
2.4	yes	yes	yes	View Report

Links to the scenario reports for Davis County

Software

The simulations used the R package epiworldR version rpackageVersion(“epiworldR”)``, which can be found at https://github.com/UofUEpiBio/epiworldR, and R version R version 4.4.0 (2024-04-24).