The continuous development of online medical has confronted hospitals with the problem of how to allocate physician resources in the joint online and offline medical service system, whose main challenges lie in the time-varying demand and the different service models of online medical. To address the physician scheduling problem in the online medical service system, the online medical service system is modeled as a processor-sharing queue, and time-varying Markov chain and uniformization methods are used to analyze and evaluate the patient sojourn time, queue length, and physician overtime time. Based on the uniformization method, a heuristic algorithm for variable-neighborhood search is proposed. Finally, numerical experimental analysis is carried out based on the actual data of the cooperative hospital, which verifies the accuracy of the time-varying Markov chain based modeling, and proves that the proposed algorithm can obtain better scheduling results relative to the actual scheduling of the hospital, so that it can more reasonably arrange the working time of the physicians, reduce the sojourn time of the patients, and control the number of patients in the system, and has excellent robustness. The study has practical significance for improving the operation and management of online medical service system in China.