5 Florida eye practices settle $6M Medicare fraud case; whole milk returns to school lunches; new AI tool screens for cognitive decline – Morning Medical Update

5 Florida ophthalmology practices settle $6M Medicare fraud case

Five Florida ophthalmology practices have agreed to pay nearly $6 million to resolve federal allegations they submitted false claims to Medicare and Medicaid for medically unnecessary transcranial Doppler ultrasounds, according to the U.S. Department of Justice (DOJ). Prosecutors alleged the practices participated in a kickback arrangement with a third-party testing company, billing hundreds of dollars per test based on diagnoses that were not supported by patients’ medical records or test results. The settlement covers claims from 2018 to 2022 and includes alleged violations of the False Claims Act, Anti-Kickback Statute and Stark Law, with more than $1.1 million awarded to a whistleblower and a portion allocated to the State of Florida’s Medicaid program.

Whole and 2% milk return to school lunches

Schools participating in the National School Lunch Program will once again be allowed to serve whole and 2% milk after President Donald Trump signed legislation reversing Obama-era limits that restricted school milk options to fat-free or low-fat varieties. The change follows the release of updated federal dietary guidelines from the U.S. Department of Health and Human Services (HHS) that emphasize whole-fat dairy.

While supporters argue broader milk choices could reduce students’ reliance on sugary drinks, critics note concerns about saturated fat intake. Implementation will depend on supply, cost and local demand, according to school nutrition officials. Whole milk and 2% milk should be available in schools in “just a few weeks,” said Agriculture Secretary Brooke Rollins. CNN Health has more.

AI agents screen clinical notes for early cognitive decline

Researchers at Mass General Brigham say they’ve built a fully autonomous artificial intelligence (AI) system that can screen for cognitive impairment using routine clinical documentation, without human prompting after deployment. In tests, the system reached 98% specificity and showed that when it disagreed with human reviewers, an independent expert sided with the AI 58% of the time. The agentic workflow uses five specialized AI agents that critique and refine one another’s outputs, runs on an open-weight model designed for local hospital infrastructure and does not send patient data to external servers.

Alongside the npj Digital Medicine publication, the team released an open-source tool called “Pythia” to support broader testing and adaptation of autonomous screening approaches.