No One Is Paying for Quality: Radiology AI and Value-Based Care, Jacob Johnson

Abstract: More than 70% of diagnostic pathways pass through radiology, yet almost nothing in the payment system rewards a radiologist for getting the diagnosis right. Radiologists are paid per study, by RVU, in a model where revenue is volume times rate, and the quality measures that do exist such as radiation dose, incidental finding follow-up, fluoroscopy reporting carefully avoid the question of diagnostic accuracy. Meanwhile, per-study reimbursement continues to fall while volumes climb. Primary care, cardiology, and several surgical specialties have value-based care mechanisms. Radiology has almost none.

This is the central obstacle facing radiology AI, and also the opportunity. Quality can be measured credibly, and when it is, it pays for itself: steering patients to high-quality imaging centers, catching overlooked vertebral compression fractures before they become hip fractures, improving prostate MR at the source. Most radiology AI products target detection, segmentation, or triage; using AI to evaluate and demonstrate the quality of radiology itself, at scale, is largely unexplored.

Bio: Jacob Johnson, MD, MS, is SVP of AI & Clinical Product at Covera Health, where he leads a team of data scientists, ML engineers, and clinical product specialists building AI systems that measure radiology quality at scale. He earned a BS in Applied Mathematics, Engineering & Physics and an MS in Biomedical Engineering from UW–Madison, where he was a member of the Breast MR group, and received his MD from the University of Wisconsin School of Medicine and Public Health. During medical school he co-founded CoRead, an imaging AI company that began with CT triage models and grew into quality assurance and missed-finding detection. CoRead was acquired by Covera Health in 2023.