This paper presents a general approach to scaling outputs of an agent-based model through the use of network analysis. By isolating the underlying network of connections in an agent-based model, we are able to utilize network metrics to predict the scaling of simulation outputs. Our results provide the basis for scaling of agent-based models, with a specific application of the method presented that uses the average shortest path length metric in a disease spread model, and a demonstration of its usage across a variety of population network types.