518 Translating research into clinical practice: Pre-implementation assessment of an AI-enabled same-day diagnostic workflow for breast screening

Abstract

Objectives/Goals: Asynchronous imaging workflows delay follow-up for abnormal screening mammograms. Real-time artificial intelligence (AI) flags high-risk scans for synchronous, immediate review. We present pre-implementation qualitative and quantitative assessments to inform AI-enabled workflow redesign and improve outcomes. Methods/Study Population: We conducted a workflow study in the largest of 14 breast imaging centers in our health system. A multidisciplinary team of experts in breast imaging, implementation science, and AI/informatics conducted four clinical shadowing visits held on different weekdays (August 2023) and a time-motion study (February 2024) to identify opportunities for optimization in the current breast cancer screening imaging workflows. To determine the optimal time for potential AI workflow integration, we used statistical process control (SPC) charts to visualize patient volumes based on 2022 to 2023 scheduling data. Results/Anticipated Results: The selected site had an average daily patient volume of 67, including 29 screening mammography patients, 3 of whom with abnormal findings prompting a diagnostic workup. During shadowing, observers recorded time stamps for each breast imaging exam step, from check-in to check-out, to calculate exam duration and estimate waiting room capacity for diagnostic and screening mammography or ultrasound. We analyzed workflow to identify unnecessary tasks and radiologists’ disruptions that, if changed, could create the space for real-time reading without more difficult changes to workflow, for example, the number of radiologists and their deployment. SPC charts indicated that selected clinic days would be most promising for testing the new real-time workflows without major disruption to the system. Discussion/Significance of Impact: A detailed understanding of the workflow ensures future implementations achieve intended outcomes. This analysis helped identify modest opportunities in radiologist and staff downtime to support the new workflow, potentially improving satisfaction, reducing anxiety, and shortening time to diagnosis.

Ähnliche Arbeiten