The Chinese healthcare referral system is a hierarchical system that includes medical institutions at different levels to ensure high-quality medical resources available and accessible to patients in need. In this research, we formulate a mixed-integer nonlinear programming model to minimize the total cost of patients, including their transportation cost, treatment cost and purchase of home-care services in a novel two-way referral system incorporating telemedicine and home-care services. To deal with the uncertain requests in teleconsultation, an adaptive robust optimization method is applied to reformulate the chance constraint into a linear form. Moreover, the uncertainty set under the measurement of Kullback–Leibler divergence is introduced to use the information and tractably approximate the chance constraint. We have used real-world data to conduct numerical experiments to determine the sensitivity of the results concerning the changes in the parameters such as the government subsidy, the maximum service capacity and the cure rate of diseases in the telemedicine center. We also prove that our optimal strategy of patient allocation in the novel referral system with the current strategy achieves environmental and social sustainability by reducing travel to healthcare facilities(less GHG emissions) and less crowdedness in the comprehensive hospitals compared to the current strategy in the traditional system. Finally, managerial insights are provided to achieve better performance in the healthcare referral system.