<p>In the global effort toward decarbonization, the healthcare sector plays a critical role in reducing emissions. Hospital administrators can mitigate their carbon footprint by optimizing operational processes and improving efficiency. In this paper, we propose a new environmentally friendly elective patient scheduling problem aimed at reducing hospital carbon emissions while maintaining service quality. By reducing the unnecessary number of open operating rooms (ORs) and minimizing the preoperative hospital stays, ORs and bed resources are utilized more efficiently, leading to a reduced carbon footprint for hospitals. To ensure that the scheduling plan aligns with the actual surgery and hospitalization needs, we develop a predict-then-optimize framework to address the impact of patient heterogeneity and uncertainties in patients’ surgery durations and postoperative length of stay (LOS). Specifically, the machine learning methods are used to incorporate the patient feature information into the ambiguity set of the distributionally robust optimization (DRO) model. We then reformulate the DRO model and apply an inexact column-and-constraint generation (i-C&amp;CG) algorithm to solve it efficiently. Extensive numerical experiments based on real-world data are conducted to compare the performance of the proposed DRO model against actual hospital schedules and benchmark methods, providing several managerial insights. The results demonstrate that our DRO model generates high-quality schedules that reduce both hospital costs and carbon emissions.</p>