Abstract for: On the characterization of censored survival curves in system dynamic models

System dynamics (SD) models typically use one or more sequential exponential processes to characterize survival or other expiry processes. These processes imply gamma distributions of the time until death, including the special case of exponential distributions for single-stage processes. For many applications, this provides an adequate representation of the survival process. However, in the context of characterizing risks associated with rare events, the shape of the distribution, and in particular the tail, may carry important implications. We explore a number of different ways to characterize survival in an SD framework to fit a set of censored survival data for a medical condition that may lead to low probability-high consequence events of re-introduction of poliovirus after global polio eradication. We find that most survival characterizations reasonably match the limited data, but the choice of characterization can lead to significantly different behavior in the tail of the survival curve, which could result in widely different implications for risk management.