Using agent-based simulations of infectious disease spread to enhance public health decision support tools

Background: Simulation methods have been used extensively to model the spread and control of infectious diseases in human populations. However, they have rarely been used to help inform policy decisions related to controlling

outbreaks. Public health decision support tools are still in their infancy and much remains to be learned about the advantages and disadvantages of components that enhance situational awareness and forecasting of nonlinear infectious disease dynamics.

Objective: To learn how computer simulations of disease spread can help determine which presentations, visualizations, references, and other components of public health decision support tools best assist epidemiologists and those who must make public health policy with limited time and resources.

Methods: Recognizing the potential for such simulations to inform the design of effective public health decision support tools, we created an agent-based simulation of pertussis transmission in a computer synthesized population and incorporated this simulation in an informatics architecture that provides users both static and dynamic public health decision support. The simulation tracks resources and supports decision-making in the context of resource constraints.

Results: Each simulation run generates time-stamped records at the individual agent level (contacts, symptomatic states, infection states), the public health department response level (line lists, reported cases, contact tracing activities, vaccination, antibiotic prophylaxis), the full simulation level (cases not reported, alternative-world records), and user level (user demographics, public health expertise, comprehensive activity record, situational awareness surveys).

Discussion: Embedding an agent-based simulation into public health decision support devices introduces a stochastic realism that forces decision makers to consider alternative phenomenological etiologies and to reflect on the effectiveness of applying textbook intervention strategies in non-textbook situations---what we call, “informed uncertainty.” This simulated world forces decision support users to consider the real-world convergence of nonlinear disease transmission dynamics with the individual-level behavior that forms the machinery of communicable disease spread.