Government Case Study: Introduction | Bubonic and Pneumonic Plague

Demonstrating System Dynamics as a Valuable Approach to the Nonlinear Modeling of Human Dynamics: Bubonic and Pneumonic Plague

Problem Statement: Contagious diseases can have significant human, economic, and social impacts on society. How the disease moves through a society influences people’s choices of behavioral responses. In turn those behavioral responses impact the course of the disease. We were asked to apply system dynamics modeling to explore and simulate the dynamic interplay (or feedback relationships) between 1) an outbreak of a virulent disease (plague: which can occur naturally or as an agent of bioterrorism) and 2) the likely behaviors that would both emerge from and contribute to the subsequent course of that outbreak.

Approach: We used our Ladder of Engagement to structure our exploration, including our collaboration with experts in relevant fields. We identified and developed:

• What we KNOW of the dynamics of the spread of plague (epidemiology of transmission, availability and effectiveness of medical interventions, and the role of human behaviors (flight, fight, and freeze) in affecting that transmission)?
• What we UNDERSTAND about the feedbacks between behavioral responses and the course of the disease through the population? How do the evolution of the outbreak itself and the population’s response to that outbreak mutually influence each other?
• How we can use the insights generated by that understanding to design and evaluate intervention strategies with which to INFLUENCE such an outbreak?

A report of the early work on this project is provided here.

KNOWLEDGE

What do we know about the “behavior(s)” of the system?

We focused on two aspects of the disease system:

• We first familiarized ourselves with the basic dynamics of transmission of contagious disease. Here well-established medical data provided guidance on the characteristics of the disease, its stages of progression, and rates of infection. These data provide the foundation for an epidemiological model with which to track the development of the disease with which to explore the effects of population density, size of the initially infected population, and availability and application of antibiotic treatment and prophylaxis.
• We then expanded that analysis by incorporating typical human behavioral responses to risks (flee = leave the affected community; fight = seek or accept medical intervention; freeze = self-quarantine) in the context of plague. Published analyses of a recent (1994) pneumonic plague outbreak in India provided the means to incorporate those behaviors in the basic model. See map of these basic dynamics below.

 

Stock and Flow Map of a Basic Epidemiological Model of Pneumonic Plague (in blue) with the Relevant Behavioral Responses (medical treatment, self quarantine, Susc Flee, and Inf Flee) Depicted in Red

UNDERSTANDING

What “drives” the behavior of the system?

Our first level of modeling produced simulations that projected (and replicated) the human toll of the Indian outbreak. To achieve that agreement with the reality, however, our epidemiological model had to be supplemented by including all three behavioral responses. Particularly valuable was the model-generated insight that a significant level of self-quarantine (“freeze”) was essential to the model’s ability to match the reality. Self-quarantine was a behavior almost totally overlooked in previous published analyses of this outbreak! [ ISDC paper].

That discovery led us to a far more ambitious challenge: identifying and incorporating into the model as endogenous drivers those psychological factors that underlie people’s behavioral decisions. We focused on deeper, more important, levels of understanding in the project’s second phase, utilizing current psychological theories (e.g. the “psychometric paradigm” and “social trust”) to allow the model itself to calculate the pathways and rates by which

• people become aware of the risk or threat posed by the disease (risk communication = Flow "1" (red)),
• people determine that a behavioral response is required (risk perception = Flow "2" (black)), and
• people define which response (of the three available) to utilize (risk response = Flows "3" (green))

Our analysis here did not extend to precisely defining the final logical step of this behavioral response process (returning to "normal" behavior = Flows "4" (orange)).

 

 

Concept Map of Primary Social Behaviors Exhibited During a Pneumonic Plague Outbreak

The resulting model “endogenizes” the behavioral responses so that they are generated by the evolving course of the outbreak and, in turn, affect that dynamically changing outbreak. A relatively small and generic simulation, consciously designed to be transferable to other risk scenarios, models each behavioral “step” in this dynamic scenario. Applying these models to the Indian scenario, we were able to define a “low social trust” submodel that resulted in a decision by a sizable portion of the population to disregard the authorities and to self-quarantine. On a broader perspective, our model suggests how those behaviors would develop and vary in different populations and in response to different diseases.

INFLUENCE

Can this understanding be used to design and evaluate policies to better manage the system?

The resulting simulations provide vehicles for medical and emergency management personnel and policy makers to use in exploring a wide variety of responses to outbreaks in a variety of user defined settings. Each behavioral element of the scenario can be explored in isolation, or applied to other risk scenarios. Users with no previous experience with the modeling software can utilize the model as a “flight simulator” to manipulate the entire simulation to explore extremely wide ranges of “what-if” scenarios.