Advancing value-based breast cancer screening in a safety-net health system: A multimodal approach to improving population health and system sustainability.

Abstract

11114 Background: Evaluation of breast cancer screening (BCS) adherence in our large safety-net health system revealed wide variability and late-stage diagnosis rates that exceeded national benchmarks. A multidisciplinary taskforce (primary care, radiology, cancer center, and patient advocates) identified barriers, co-designed, and implemented a multi-pronged intervention to address screening uptake and adherence. Methods: Mixed-methods evaluation across 18 primary care sites revealed wide variability in BCS rates (60-82%); lack of systematic management of screening bottlenecks (ordering, scheduling, no-shows/cancellations); and unmet patient needs (transportation, screening hesitancy). We co-developed and implemented a phased intervention from March–December 2025 to address identified barriers through: (1) radiology navigator outreach to primary care patients at low-performing sites (March), (2) electronic medical record patient portal self-scheduling (July), (3) AI-assisted rescheduling after no-shows/cancellations (December), and (4) assessment and navigation of social needs and screening hesitancy (March). Outcomes measured were outreach, engagement, and screening completion. We performed multivariate regression to examine demographic predictors of screening. Results: Of 57, 041 HEDIS-eligible patients, 15,971 (28%) were overdue at baseline (mean age 58 yo, 77% Hispanic or non-Hispanic Black, 74% Medicaid or Dual). Overall, the intervention resulted in 1,610 additional screens over 9 months, improving systemwide BCS rates from 72 to 76%. Of the additional screens, radiology navigator achieved 452 screens (29%), patient portal self-scheduling achieved 1,097 screens (70%), and AI-assisted outreach of no-show/cancelled achieved 61 screens (4%). Engagement and show rates varied: navigator outreach had the highest scheduling engagement (74%) but lower show rates (46%); patient portal self-scheduling showed moderate engagement (44%) with higher show rates (76%); and AI-assisted outreach had low engagement (4%) with moderate show rates (56%). Demographics did not predict screening outcomes. Screening hesitancy (37%) and transportation (10%) were the most common barriers, addressed through social and peer navigation, respectively. Revenue from additional screenings fully offset navigator costs. Conclusions: A coordinated, resource-efficient screening outreach strategy combining digital tools, human navigation, and AI assistance—co-designed with patients and clinical stakeholders—improved BCS adherence, addressed unmet needs of underserved populations, and proved itself financially sustainable across a safety net health system. Future efforts will focus on further scaling of population health approaches to cancer screening programs.