Design and Implementation of an Agent-Based Model of Pertussis with Performance Considerations

Abstract

Pertussis, also known as Whooping Cough, is an airborne communicable disease caused by the Bordetella pertussis bacterium. Symptoms include fever, runny nose, and a cough that typically progresses to the point where it interferes with breathing, producing the characteristic whoop from which the common name is derived. Complications, which disproportionately affect infants, include bacterial pneumonia which can lead to death. Pertussis is vaccine-preventable and vaccination programs exist in most countries yet a recent resurgence has been observed in jurisdictions with high vaccine coverage, including Alberta and Canada. Simulation modeling has a long history in the study of epidemiology, including that of pertussis, but most of such work has employed compartmental models. Agent-based models (ABMs) allow differentiation down to the individual level, which cannot be done in aggregate compartmental models, as well as simpler specification of heterogeneity and interaction patterns which can be tedious to implement in aggregate compartmental models. These benefits come at the cost of increased computational burden. This thesis seeks to design and implement an ABM representing the epidemiology of pertussis in Alberta, Canada, and apply that model to evaluate vaccination during pregnancy as a potential intervention strategy to reduce pertussis incidence in infants. In support of this objective, data structures will be explored to improve performance for large ABMs developed using AnyLogic software.