There are positive and negative examples of AI in medicine, I try to share stories from both camps. It is encouraging to see the good use cases, but we have to keep a critical eye out for AI causing harm or abuse.

A Senate report revealed that top Medicare Advantage insurers used AI to reject post-acute care claims at much higher rates from 2019-2022. Humana’s denials were 16 times higher than its overall rate. 🩺 🛟 🤖

UnitedHealth, CVS and Humana increasingly deploy AI and deny post-acute care claims, Senate report finds

The country’s three largest Medicare Advantage (MA) insurers obstruct seniors’ ability to receive post-acute care, a scathing report from the U.S. | A report from Sen. Richard Blumenthal, D-Connecticu...

🧪 🩺🖥️ #MLSkyjamanetwork.com/journals/jam...

Testing and Evaluation of Health Care Applications of Large Language Models

This systematic review characterizes the current performance of large language models in evaluating clinical health care settings, including uniformity, thoroughness, and robustness and proposes a fra...

🧪 Only 5% of healthcare evaluations of LLMs use real patient data, per a JAMA study. Inconsistencies in bias, fairness, and task range highlight the need for standardized LLM evaluations.🩺🖥️ #MLSky

Study finds health care evaluations of large language models lacking in real patient data and bias assessment

A systematic review reveals only 5% of healthcare evaluations for large language models use real patient data, highlighting gaps in bias and task assessment.

🧪 30% of patients find healthcare unaffordable, with 28% delaying or skipping care due to costs, per PwC’s 2024 survey. 20% support using AI as a physician assistant, but data privacy concerns remain. Only 60% of seniors feel ready for AI in healthcare. 🩺🖥️ 🤖

Affordability, GenAI top-of-mind for healthcare consumers | TechTarget

Learn how healthcare organizations can meet healthcare consumer demands for better healthcare affordability, convenience and a smooth health IT experience.

If you're curious why everybody's username is a domain, it's because every user is essentially a website

* You can modify who can interact with your post (see or reply), and you can also detach a post from quote posts if you like. * Items not available (yet): drafting threaded posts, scheduling posts, editing posts, polls, and longer posts. Hope you find this info useful! 4/4 🧵

Bluesky adds ‘anti-toxicity’ options to limit dogpiling and hostile quote posts

Users can also now hide replies and block list creators.

* DM are default as only people you follow can message you, but you can set it to anyone, or no one. * You can register a domain name as your handle: bsky.social/about/blog/3...bsky.app/profile/prof... 3/4 🧵

Just shipped: Bluesky Direct Messages! - Bluesky

You can now send direct messages (DMs) to people on Bluesky! Say hi to a friend, colleague, or a crush.

* Blocks, likes, posts, and lists are public, but your subscription to a list are private, mutes are also private. * Up to 60 second of video is supported, but there is a 25 video limit per day for now. * Add alt text to images, you can require it for any post w/ media in your settings 2/4🧵

Share video on Bluesky! - Bluesky

Bluesky now has video!

Lots of new people arrived to BlueSky recently, welcome aboard! I figured I throw out some tips on what makes BlueSky different from other sites. * Kicking off, BlueSky makes certain user actions, your likes and blocks, publicly visible to all. Check out clearsky.app to see what I mean. 1/4 🧵

a woman is sitting on a couch in front of a sign that says welcome everything is fine

Alt: a woman is sitting on a couch in front of a sign that says welcome everything is fine