With the rapid development of large language models and related technologies, various forms of patient modeling have emerged, including virtual patients, synthetic patients, artificial intelligence patients, and digital twin patients. Although these forms all share the core objective of patient simulation, they exhibit significant differences in theoretical ontology, data sources, modeling logic, and application orientation. Based on the abstraction levels of simulation objects and system mapping logic, this study systematically compares five paradigms—standardized patients, virtual patients, synthetic patients, AI patients, and digital twin patients—to clarify their conceptual boundaries and technical positioning. The findings indicate that these paradigms represent distinct modeling pathways: behavioral reproduction, scenario simulation, data generation, cognitive interaction, and individual system mapping, respectively. By constructing a multidimensional comparative framework, this study provides a theoretical foundation for the conceptual standardization of patient modeling and the strategic deployment of intelligent healthcare systems.