Graham Walker, MD

Most people tune out when doctors rage about prior auth, opaque pricing, or middlemen sucking value out of the system. Except Mark Cuban.

He reads the regulations, runs the numbers, and builds alternatives.

Where we all see complexity, he sees opportunity.
Middlemen? Rebates? Copays? Deductibles? Don’t need ’em.

Sell drugs at cost + 15%, and you:
1️⃣ Save your own companies money
2️⃣ Save Americans billions
3️⃣ Keep local pharmacies alive
4️⃣ And force bloated PBMs to compete (or die trying).

One recent study found that Mark Cuban Cost Plus Drug Company, PBC saved patients and taxpayers over $3 billion (and that was just from analyzing 77 generics).

It was a real privilege to talk to him last week about what’s broken in healthcare, who’s profiting off the mess, and what it’ll take to rebuild trust in medicine.

He’s real. He’s sharp. And he might be the only person alive who actually enjoys healthcare complexity, because it's another chance for him to knock it all down. 🥊

As always, link to the full 𝙃𝙤𝙬 𝙄 𝘿𝙤𝙘𝙩𝙤𝙧 𝙋𝙤𝙙𝙘𝙖𝙨𝙩 episode in the comments.

🌟 Did UCSD just turn one of healthcare’s most hated, costly, and time-consuming tasks into an easy, automated process? 🌟

Finally catching up on my GenAI reading, and this NEJM AI article is fan-freakin’-tastic. UCSD took the 𝗶𝗻𝗳𝗮𝗺𝗼𝘂𝘀𝗹𝘆-𝗽𝗮𝗶𝗻𝗳𝘂𝗹 sepsis quality bundle (SEP-1) and tasked an LLM with handling the 𝟲𝟯-𝘀𝘁𝗲𝗽 𝗽𝗿𝗼𝗰𝗲𝘀𝘀 required to review each 𝘱𝘰𝘵𝘦𝘯𝘵𝘪𝘢𝘭𝘭𝘺-𝘴𝘦𝘱𝘵𝘪𝘤 patient. The result? A 90% match to human accuracy across 100 patients, with discrepancies that were easily sorted out — and in 4 out of 10 cases, the LLM was right, and the human was wrong! 🤯
 
For me, the takeaway isn’t just the accuracy. This study is a brilliant example of where GenAI should be laser-focused. (And, as an ER doctor, I’ve distilled it into a mnemonic: 𝙲𝙻𝙴𝙰𝚁-𝚁.)

🔹 𝗖𝗼𝘀𝘁-𝗘𝗳𝗳𝗲𝗰𝘁𝗶𝘃𝗲: Slashes costs from endless hours of human labor needed for manual quality reporting.
🔹 𝗟𝗼𝘄-𝗥𝗶𝘀𝗸: Verifiable, safe results that can be reviewed and overseen by humans.
🔹 𝗘𝗳𝗳𝗶𝗰𝗶𝗲𝗻𝘁: Reduces documentation bloat for MDs and RNs — a huge win.
🔹 𝗔𝘂𝘁𝗼𝗺𝗮𝘁𝗲𝗱 & 𝗔𝗱𝗺𝗶𝗻𝗶𝘀𝘁𝗿𝗮𝘁𝗶𝘃𝗲: Tackles data-heavy admin tasks autonomously.
🔹 𝗥𝗲𝗱𝘂𝗰𝗲𝘀 𝗕𝘂𝗿𝗱𝗲𝗻: HCWs 𝘩𝘢𝘵𝘦 this low-value stuff. This could help.
🔹 𝗥𝗲𝘁𝗿𝗼𝘀𝗽𝗲𝗰𝘁𝗶𝘃𝗲: Works on past patient data, so it doesn’t interfere with real-time patient care.

Not only does this approach “Get Rid of Stupid Stuff“ (GROSS), but I’ll take it a step further: imagine an AI-powered automated feedback loop to keep the treatment team informed. Imagine getting a message like: 𝘙𝘦𝘮𝘦𝘮𝘣𝘦𝘳 𝘔𝘳𝘴. 𝘊𝘩𝘶? 𝘚𝘩𝘦 𝘥𝘪𝘥𝘯’𝘵 𝘩𝘢𝘷𝘦 𝘴𝘦𝘱𝘴𝘪𝘴 — 𝘢𝘯𝘥 𝘨𝘰𝘰𝘥 𝘵𝘩𝘪𝘯𝘨 𝘺𝘰𝘶 𝘥𝘪𝘥𝘯’𝘵 𝘨𝘪𝘷𝘦 𝘩𝘦𝘳 3𝘓 𝘰𝘧 𝘧𝘭𝘶𝘪𝘥𝘴, 𝘴𝘩𝘦 𝘥𝘦𝘷𝘦𝘭𝘰𝘱𝘦𝘥 𝘱𝘶𝘭𝘮𝘰𝘯𝘢𝘳𝘺 𝘦𝘥𝘦𝘮𝘢 𝘵𝘩𝘦 𝘯𝘦𝘹𝘵 𝘥𝘢𝘺.

🎩 Hats off to UC San Diego Health for pioneering a practical, impactful use of AI in healthcare! This is exactly the kind of solution we need more of — and that could be deployed 𝘁𝗼𝗱𝗮𝘆. Bravo.

#HealthcareAI #QualityMetrics #CLEAR_R #InnovationInHealthcare #LLM #HospitalQuality
Christopher Longhurst Shamim Nemati Karandeep Singh Aaron Boussina Rishivardhan K Gabriel Wardi Hayden H. Pour Atul Malhotra Mike Hogarth, MD, FACMI, FACP Amy Sitapati Chad VanDenBerg, FACHE Nicholas (Nick) Hilbert, MSN, RN Kimberly Quintero