In a complex healthcare delivery environment there are unforeseen risks and uncertainties that may unfold when new policies are applied. In silico predictive simulation approaches allow exploration of potential responses of a system to new policy and rule implementations. The validity of such computational models comes into question unless they operate with realistic representations which require significant modeling detail over a large-scale, and with high accuracy. This necessitates a large amount of computing capacity and data management. To address these needs we propose a high-performance computing (HPC) agent-based framework for healthcare system predictive simulations. The framework is designed to emulate a healthcare system modeled at high fidelity and with high resolution data, evaluate its performance in response to different user defined policies, and find polices that maximize outcome measures and system efficiency. The paper details our data preparation procedures, and describes how the framework is implemented and run on a supercomputer to model a healthcare system at an appropriately large scale.