Daily Briefing

What Changed

Health-system outcomes give sepsis AI deployments a concrete evidence benchmark [1].

Clinical Practice | A deployed sepsis AI system was associated with shorter stays and lower mortality. The NEJM AI Grand Rounds podcast reported that a deployed sepsis AI system at Cleveland Clinic was associated with an average two-day reduction in length of stay and a 4% to 5% absolute mortality reduction across about 20 to 21 hospitals and 25 emergency departments [1].

Industry & Products | Epic unveiled Agent Factory for hospitals to build AI agents. MedCity News reported that Epic introduced Agent Factory, a platform that lets hospitals build and monitor their own AI agents for healthcare workflow automation, despite an ongoing FTC antitrust probe [2].

Medical Imaging | Four commercially available AI solutions for chest radiography did not improve diagnostic accuracy. An Academic Radiology prospective crossover study found no accuracy gain for pulmonary infiltrates, pleural effusions, mediastinal masses, pneumothorax, or pulmonary nodules, yet three of five residents experienced median interpretation time reductions of 6 to 17 seconds per case with AI assistance [3].

Diagnostics & Pathology | Bile extracellular vesicle analysis with a four-protein machine learning model plus CA19-9 achieved high diagnostic performance. A JHEP Reports study showed that a four-protein machine learning model (COPA, ATP5H, VTNC, PRDX2) combined with serum CA19-9 achieved an AUC of 0.993 for detecting PSC-CCA, outperforming CA19-9 alone (AUC = 0.801) in a cohort of primary sclerosing cholangitis patients [4].