Optimization of medical resource allocation plays an important role to ensure the efficient operation of cloud healthcare systems, which is a new internet healthcare system. However, this optimization problem can be presented as an uncertain optimization problem of allocating core doctor service time due to the multi-organizational collaboration, uncertainty of referral, and randomness of diagnosis and treatment time. Based on the robust optimization theory, this paper presents a medical resource allocation model with the objective function of minimizing the maximal medical serve cost in a cloud healthcare system. In this model, two different uncertain factors, which are the patient diagnosis and treatment time and the patient referral probability, can be considered simultaneously according to the control parameters given by decision-makers. From the experimental analysis, it can be concluded that the presented model can reduce the impact of these uncertain factors of patient demands upon the total medical cost, and ensure the robustness of cloud healthcare system operation.