Invited commentary: The virtual epidemiologist—promise and peril

Abstract

The simulation of complex systems has received increasing attention as a useful approach in epidemiology. As discussed by Marshall and Galea in this issue of the Journal (Am J Epidemiol. 2015;181(2):92-99), systems approaches are appealing because they allow explicit recognition of feedback, interference, adaptation over time, and nonlinearities. However, they differ fundamentally from the traditional approaches to causal inference used in epidemiology in that they involve creation of a virtual world. Systems modeling can help us understand the plausible implications of the knowledge that we have and how pieces can act together in ways that we might not have predicted. It can help us integrate quantitative and qualitative information and explore basic dynamics. It can generate new questions that can be investigated through new observations or experiments. The process of building a systems model forces us to think about dynamic relationships and the ways in which they may play a role in the process we are studying. However, the validity of any causal conclusions derived from systems models hinges on the extent to which the models represent the fundamental dynamics relevant to the process in the real world. For this reason, systems modeling will never replace causal inference based on empirical observation. Causal inference based on empirical observation and simulation modeling serve interrelated but different purposes.

Keywords: causal inference; complex systems; methods; simulation.