Agent-based Modeling

Quantifying model uncertainty in agent-based simulations for forecasting the spread of infectious diseases and understanding human behavior using social media

Get Expertise  

"Modeling and simulating infectious diseases is crucial to understanding transmission and developing mitigation strategies that can save lives."

Overview

We use two agent-based simulations—the Dynamic Activity Simulation (DASim) to study activity patterns relevant to disease transmission and the Object-oriented Platform for People in Infectious Epidemics (OPPIE) to study disease spread and the impact of pharmaceutical and non-pharmaceutical interventions. We have analyzed several mitigation strategies, including vaccines and antivirals, as well as social closures, facemasks, hygiene, isolation, quarantine, fear, and other changes in human behavior.

Dynamic Activity Simulation (DASim)

DASim is a parallel, discrete–event, agent-based model that simulates human activity patterns. Daily activity schedules of individual agents are generated based on demographics, utilities, priorities, and time constraints. The schedule planner can account for exogenous events, such as disease or natural disasters, and suggest activities accordingly.

The main components of DASim are:

  • A set of activities for various demographics (e.g., children, adults, and seniors)
  • An objective function consisting of time constraints, utilities, and priorities
  • A planning and re-planning algorithm based on exogenous events (e.g., early or late arrival, disease, natural disasters)
  • An optimization algorithm to plan activities

Snapshot of activity patterns for the Twin Cities population 2.6 million agents.

DASim outputs "demand hours" on an hourly basis. Demand hours represent the total number of people participating in an activity at a given time. This example shows output for six activities over the course of one week for the Minneapolis-Saint Paul region in Minnesota. Notice how some activities (e.g., home) occur with obvious regularity while others (e.g., social recreation) occur more sporadically.

Object-oriented Platform for People in Infectious Epidemics (OPPIE)

OPPIE (previously known as the Epidemic Simulation System – EpiSimS) is a stochastic, agent-based simulation agent that models disease spread in regions, allowing for the assessment of disease prevention, intervention, and response strategies. OPPIE explicitly represents the daily movements and interactions of synthetic individuals in a geographic region, including their interactions with others. It is an experimental test bed for analyzing consequences, feasibility, and effectiveness of response options to disease outbreaks.

OPPIE has four main components:

  • A synthetic population based on U.S. census data
  • Activity patterns for the population based on the National Household Transportation Survey or DASim
  • Geo-referenced locations obtained from Dun & BradStreet
  • A flexible disease progression model

Why was the name changed from EpiSimS to OPPIE?

The Epidemic Simulation System (EpiSimS) began at Los Alamos National Laboratory (LANL) in 2000. EpiSimS originated as an extension of TRANSIMS, a massive agent-based model that simulates traffic, also developed at LANL. In January 2005, EpiSimS split into two independent efforts, one at Virginia Bioinformatics Institute, and the other remaining at LANL. Since then, the LANL EpiSimS team has continued expanding its capabilities and in 2012, the team decided to rebrand it as OPPIE to highlight the significance of these improvements. OPPIE is named after J. Robert Oppenheimer, LANL’s first director and leader of its Manhattan Project effort.

Illustration showing infuenza attack rate in Southern California ranging from mild to severe.

OPPIE-generated clinical attack rate by census tract, ranging from mild (green) to severe (red) for pandemic influenza in southern California. The hotspots are highly correlated (R2 = 0.9) with average household size.