DISSERTATION A Comparative Analysis of System Dynamics and Agent-Based Modelling for Health Care Reimbursement Systems

System dynamics and agent-based modelling are two methods for modelling dynamical systems. While both emphasize that complex dynamics can evolve from simple rules and relationships, they have also important differences. Agent-based models are considered to be more flexible, but normally have a higher computational demand than system dynamics models. The central goal of this study is therefore to compare both approaches and investigate their similarities and differences. Moreover, it tries to answer the question of under which conditions one of the two methods is preferable or how they could complement each other. The motivation for these research questions comes from the author’s modelling work on the study of reimbursement systems for physicians in extramural health care. A further goal is therefore to investigate how agent-based modelling and system dynamics could be applied to this field of research. The thesis puts both methods on a common system-theoretic basis as it introduces the concept of a stochastic dynamical system, which is able to cover approaches that include randomness, such as agent-based modelling, but also contains deterministic dynamical systems (in particular system dynamics models) as a special case. It is then shown that system dynamics defines a dynamical system via its equivalence to differential equation systems. For agent-based modelling, a formal definition based on the Stochastic Discrete Event System Specification (STDEVS) is given. The thesis then presents a comparative analysis of the two methods regarding several aspects such as their suitability for modelling heterogeneity, the consequences of aggregation, and the representation of feedback. It is shown that the concept of a rate is used to describe change equivalently both in system dynamics and agent-based modelling. The thesis argues that the two methods can complement each other beneficially