Agent-Based Modeling and System Dynamics in Healthcare
Healthcare simulation models have attracted significant offered important insights in to health policy selection. More complete accounting for the cost and health implications of upstream interventions is hindered by the need to consider impact on, and interactions between, multiple comorbidities. Within this paper, we explore several distinct approaches for representing comorbidities, some of them at the aggregate level, and some of them at the individual level. All of these representations have the virtue of being declarative, in that they allow the user to focus on what is to be characterized, rather than how it is to be implemented. Our exploration suggests that while several aggregate representations of comorbidities are possible, they suffer from a variety of shortcomings, ranging from low fidelity to combinatorial blowup. While individual-level representations impose a heavy performance load, greater difficulties in calibration and less rapid analysis, such representations do offer greater transparency, modifiability, scalability, and modularity, and ease of representing transmission and influence networks. With much to recommend each approach, further research is needed to shed additional light on the tradeoffs and identify situations where one representation is preferable to another.
While there have been considerable success stories in predictive healthcare modeling and cost implications of health interventions for specific conditions, the tight coupling between different conditions makes it desirable to examine simultaneous policy impact on broad sets of conditions. At the same time, the representation of multiple comorbidities – like the representation of other heterogeneities – can add considerable complexity to the modeling process. This paper seeks to explore some options for representing such comorbidities, as well as relevant risk factors.