413 Background: Current health information systems lack a robust framework for longitudinal integration of internal experiences and external evidence into existing workflows. The result is labor intensive, error-prone and fragmented systems plagued with process gaps. Learning health systems (LHS) can be a potential solution to address these gaps. We created and implemented a novel LHS framework in radiation oncology (RO). We hypothesize that this framework can be leveraged to identify and implement opportunities for improvement using a patient-oriented endpoint of time from radiation simulation to treatment start (RSTS). Methods: An RO-LHS was developed through a collaboration between an academic, multi-facility radiation oncology department and an industry partner (Medlever). The end-to-end clinical workflow was embedded into the Medlever platform following an interdisciplinary, iterative analysis with representatives from key teams (clerical, finance and clinical). A novel nomenclature was created to label standardized, customizable templates, which choreograph logic-driven tasks that cover all sub-steps of RSTS. The primary endpoint of RSTS was compared pre- and post-implementation, stratified by key variables, such as facility, staff member, and radiation treatment modality. Results: The iterative workflow analysis identified >700 variables and decision points necessary for >200 cancer and benign conditions routinely managed in RO. >450 treatment protocols were templated.180 logic-driven tasks were created and linked to key parameters, resulting in a choregraphed precision workflow for each patient. Using the first iteration of the build, a total of 512 patients were managed during a 50-day period (03/01/2024-04/20/2024) and were compared to 2875 patients treated prior to implementation. Post-implementation, RSTS was shortened by 36.4% (p<0.001). We identified significant differences in RSTS by radiation modalities (protons vs all others), across facilities, and by providers (p<0.01 for all). Analysis of fractional domain of RSTS, 56% of the time was driven by physician-directed tasks, while 44% was comprised of physics and dosimetry-driven work. Physician-directed tasks had the greatest heterogeneity in time to completion. The biggest variation was noted for the contouring completion task. Significant differences were identified by provider (p<0.001) in nearly each step of the process, including steps not directly tied to physician’s work. Conclusions: Through an inter-disciplinary implementation of the first iteration of a LHS in our RO department, we have significantly shortened RSTS. The implementation and analysis has also enabled us to better understand underlying heterogeneities and identify further opportunities for optimization of RSTS. An LHS framework provides a generalizable solution that could be adapted in other RO clinics.