The management and control of medical migration for major diseases is one of the major concerns of Chinese medical and health system reform. Existing research in the field of health management lacks a global perspective of modeling medical migration for major diseases. Based on the patient flow data of medical migration, this study proposes to use random networks to model and analyze medical migration. The stationary probability of nodes in a random network reflects a global attribute of the network, and the properties shown by the random network model have strong practical significance. This study also discusses the results of random network models under different non-migrant preference levels and different node scales, and analyzes the sensitivity of stationary probability to interaction intensity. The random network model and the corresponding optimization goal proposed in this paper lay a foundation for theoretical and methodological studies on the management and control of medical migration for major diseases.In the first part, this paper presents the patient flow data of medical migration from a real-world database. The data analysis results indicate that while the local attributes of the network can capture the presence of large nodes within the network, they fail to consider the heterogeneity in contributions of different edges to the nodes. Therefore, random networks are used to model and analyze medical migration. Specifically, each city is considered a node of the random network and a state of the Markov chain, and the nodes are linked with each other with certain transition probabilities. The movement of patients between cities can be viewed as the transfer of states between network nodes. The stationary probability of the Markov chain captures the long-run proportion of visits to each node, and thus reflects the concentration of patients in each city.In the second part, based on the patient flow data of medical migration, this paper establishes and analyzes specific random network models. Specifically, the paper considers the impact of different local treatment preferences on the random network model, as well as the random network model under different node scales. It proposes an optimization objective for the random network model and establishes a general framework for solving its optimization problem. The sensitivity of stationary probabilities to interaction intensities is also analyzed. The numerical results show that the stationary probability of nodes in random network model can reflect and predict the rise of “sub-center” for medical migration earlier. Management decision makers should use different network configurations depending on the degree of local treatment preference and scale of concern. The original distribution of patients can be used as a criterion for evaluating and optimizing random network models of medical migration. Referring to the sensitivity analysis results, management decision makers can identify and verify key patient flow data, thereby improving the robustness of the random network model to data errors.In summary, this article takes a global perspective and utilizes mobile data on cross-regional medical visits to model and analyze the phenomenon. The random network is used to model and analyze cross-regional medical visits, where the steady probabilities of the nodes can reflect and predict the emergence of “sub-centers” in cross-regional medical visits at an earlier stage. Depending on the degree of local medical visit preferences and node scale, decision-makers should choose different random network models. By referring to the sensitivity analysis results, decision-makers can identify critical patient flow data and verify them to enhance the robustness of the random network model against data errors. Compared to existing research, this paper makes three main contributions: 1) Introducing a random network to model medical migration and providing the specific form of transition probabilities using a data-driven approach. 2) Proposing an optimization problem framework for medical migration in the random network, exploring the impact of management decisions on optimizing objectives. 3) Analyzing the properties exhibited by the medical migration random network model and providing practical explanations and managerial insights of significant relevance.On the basis of the random network model for medical migration, in-depth researches can be continued from the following two directions in the future. The first is to quantitatively analyze the core driving factors affecting the interaction strength in random networks. How to filter and fuse feature factors from high-dimensional data and improve the efficiency of learning quantitative driving mechanism is the difficulty of this problem. The second is to optimize the decision variables (i.e., the core driving factors) in the random network model, targeting the original distribution of patients. The learning efficiency of the interaction strength quantitative driving mechanism and the optimization efficiency of the random network model are closely related and coupled with each other. How to reasonably allocate computing resources to balance learning efficiency and optimization efficiency is the difficulty of this problem. Researches in these two directions will provide scientific basis and theoretical methods for the formulation of medical and health policies, such as the establishment of regional medical centers.