2511 - Harnessing the Halo: Network-Level Scheduling to Optimize Start Dates in a Large Radiation Oncology System

Purpose/Objective(s): Clinics implementing proton therapy often experience a “halo” effect, resulting in a 10-25% increase in patient volumes. This increase can exacerbate demands on staff and resources and result in delayed start dates for new patients. We describe here the development and early clinical deployment of a scheduling application designed to minimize time-to-treatment initiation by leveraging distributed resources across a large radiation oncology network while balancing physician workload.

Materials/Methods: A web-based scheduling application was developed to algorithmically determine the earliest feasible verification simulation (VSIM) start dates and times. The system integrates the Mosaiq SQL database to assess real-time machine availability across a network consisting of a main campus and three satellite clinics, and the QGenda REST API to determine physician availability. Treatment machines across the network include MR-Linac, PBS proton therapy, helical photon linacs, c-arm linacs with varying degrees of motion management and SGRT, gamma knife, and HDR brachytherapy. To optimize technology for each disease site, treatment machines across the network were scored based on capability to perform 3D, IMRT/IMPT, SBRT, and SRS treatments using a rule-based capability matrix. The general treatment planning workflow was decomposed into sub tasks for dosimetrists, physicists, and physicians, with times unique for each disease site and plan type assigned to each task. Hisptorical prior authorization approval times for insurance carriers were incorporated. For each new patient, the application calculates the earliest feasible start date based on age, disease site, optimal technology, physician availability, payer, and geographical preference. A soft launch (February 1-20, 2026) enabled therapists to schedule new patients using the application during CT simulation.

Results: Thirty-two patients were scheduled during the three-week pilot. Three (9.4%) were assigned a non-attending provider for VSIM, one was directed to a satellite site to achieve an earlier start date, 19 (59.4%) demonstrated agreement between the application-selected VSIM start time and the final Mosaiq-scheduled appointment, 21 (65.6%) demonstrated concordance for combined VSIM and treatment duration with Mosaiq, and 29 (90.6%) were scheduled on the machine recommended by the application, demonstrating high clinical acceptability of algorithmic recommendations.

Conclusion: We implemented a network-wide scheduling platform to mitigate proton therapy-associated demand growth. Early deployment demonstrated strong adherence to machine recommendations and concordance with operational constraints, supporting the role of predictive scheduling in improving resource utilization, reducing treatment delays and distributing workload across multi-site radiation oncology networks.