Healthcare AI Daily Briefing

Research & Evidence

Machine‑learning score PROGRESS stratifies mCRPC risk using three routine labs.

Journal article (JCO clinical cancer informatics)

A machine‑learning derived score (PROGRESS) using PSA, alkaline phosphatase and AST was built from 2,035 mCRPC patients and externally validated in 1,239 patients across therapeutic settings; it showed strong discrimination in training (ROC‑AUC 0.89, C‑index 0.69) and consistent prognostic significance in all validation cohorts (all P<.0001). [1]

Policy & Governance

Interview: slow tech rollout and reactive cooling‑tower oversight drive legionella deaths, former health chief says.

Podcast from Health & Veritas

In a Health & Veritas interview, the former NYC health commissioner said legionella deaths persist because government rolls out technology too slowly and relies on building owners to police cooling towers, an outdated reactive model; more than twelve died this summer, and he urged modernizing with wastewater analytics and syndromic surveillance for proactive, hyperlocal investigations. [2]

Industry & Products

OpenEvidence and Doximity settle legal battle over alleged corporate espionage.

OpenEvidence and Doximity settled dueling lawsuits that included claims Doximity executives used NPI numbers to pose as physicians and extract proprietary code from OpenEvidence’s platform. OpenEvidence dropped trade‑secret claims and refiled under a federal anti‑hacking statute; Doximity countersued for defamation and harassment. Both free U.S. physician tools remain rivals. [3]

Policy & Governance

Provider groups tell FDA developers should bear responsibility for generative AI medical devices.

The AHA, AMA and American Telemedicine Association urged the FDA to keep manufacturers responsible for safety, performance and ongoing monitoring of generative AI devices, warning that the monitoring load would overwhelm small and rural practices and that limited model transparency hampers providers' ability to spot flaws. [4]