20  Introduction

This module demonstrates how to use EpiModel to integrate a statistical model of a dynamic contact network (using Temporal Exponential Random Graph Modeling) with a dynamic individual-level model for infectious diseases.

This is the point in the course where the two halves come together. Modules 1 through 3 built up the network side: what a network is, what an ERGM does, and how to estimate one from data. Here we put an epidemic on top of one.

We will start with a brief lecture (slides) below to describe the types of modeling we will be doing in this module, and then move to a hands-on demonstration of the process using EpiModel.

20.1 Slides

20.2 What Is in This Module

The module runs across five chapters, in this order.

SIS Epidemic Across a Dynamic Network introduces the core EpiModel workflow using the simplest case, a single homogeneous population. We estimate a temporal ERGM for the contact network, build a network-based SIS (susceptible-infected-susceptible) epidemic model on top of it, run the simulation, and then extract, summarize, and plot the output.

A Simple Vaccine Intervention is a short detour showing how to add a built-in intervention that reduces the probability of infection.

Lab 1 is your turn: you re-specify the formation model to prohibit concurrency, and work out what that does to the epidemic.

Working with Nodal Attributes in Network Models introduces population heterogeneity through the special group nodal attribute, which lets nodes differ systematically in their behavior and in their epidemic parameters. We motivate it with a heterosexual network in which the groups represent women and men, and build an SIR (susceptible-infected-recovered) model on a network where concurrency differs by group.

Lab 2 is your turn again: you vary the epidemic parameters by group, and then, if you have time, push into the mixing model itself.

Along the way you will plot static snapshots of the network at a few time steps. Those snapshots freeze what is really a moving network; Module 5 picks the thread back up and animates it, so you can watch the epidemic and the contact structure change together.

20.3 Learning Objectives

By the end of the module, you should be able to:

• Explain how a network epidemic model differs from a compartmental one: where the per-act transmission probability (inf.prob) and the act rate (act.rate) go, and why the number of partners is not a parameter at all.
• Describe conceptually how to integrate a TERGM with an individual-level epidemic model using EpiModel.
• Parameterize a basic network-based SIS epidemic model, diagnose the model, and run the epidemic simulation in EpiModel.
• Extract and analyze the numerical data from the EpiModel object.
• Plot and visualize the epidemic from the EpiModel object.
• Apply a built-in network intervention that reduces the probability of infection, using the param.net arguments inter.eff (intervention efficacy) and inter.start (start step).
• Parameterize population heterogeneity in the network degree distribution by dichotomous groups.
• Encode a mixing assumption with nodematch, and work out the logical constraints that assumption places on the group-specific mean degrees.
• Parameterize an SIR epidemic model in EpiModel.
• Summarize and visualize group-specific epidemic outcomes from the SIR model.
• Visualize a transmission tree of the epidemic with a phylogram.