We propose the use of artificial societies to support health care policymakers in understanding and forecasting the impact and adverse effects of policies. Artificial societies extend the agent-based modeling paradigm using social science research to allow integrating the human component. We simulate individuals as socially capable software agents with their individual parameters in their situated environment including social networks. We describe the application of our method to better understand policy effects on the opioid crisis in Washington, DC, as an example. We document how to initialize the agent population with a mix of empiric and synthetic data, calibrate the model, and make forecasts of possible developments. The simulation forecasts a rise in opioid-related deaths as they were observed during the pandemic. This article demonstrates how to take human aspects into account when evaluating health care policies.