Prospective deployment of AI-based risk stratification to enable expedited mammography workflow in a safety-net setting

Delays between screening and diagnostic mammography contribute to disparities in breast cancer outcomes and reduce screening adherence. We prospectively evaluated the operational feasibility and impact of integrating AI risk stratification into screening workflows at an urban safety net facility. In this HIPAA-compliant, IRB-approved, controlled study, Mirai 1-year risk scores were generated in real time from screening mammograms. Patients in the top 10% of risk were flagged as high-risk and, on enrollment days, were offered immediate screening interpretation and same-day diagnostic evaluation when indicated. Outcomes included operational feasibility, care delivery timelines for screening results, diagnostic evaluation, and biopsy, and cancer detection rate, compared with high-risk controls on non-enrollment days. Among 4145 screening mammograms, 525 were flagged as high-risk and 100 women consented to expedited care, of whom 94 percent received immediate reads and 26 received same-day diagnostic evaluation. Time to screening results, diagnostic evaluation, and biopsy were significantly shorter in expedited patients versus control patients with screen-detected cancers, with reductions of 99.1%, 99.1%, and 87.2%, respectively. The expedited cohort demonstrated a cancer detection rate of 60/1000 compared with 2.3/1000 among non-high-risk participants. Integrating AI risk models into clinical workflows substantially improved care delivery timelines and provides a translational framework to improve care and reduce disparities.