10/27/2026
Read It Back Ep. 26: Who really saves when medicine uses AI?
Who really saves when medicine uses AI? Three posts from Eric Topol, JAMA, and NEJM, read word for word, then explained.
Posts & sources:
• Eric Topol (): https://x.com/EricTopol/status/2114716553921118545
Source: https://www.nature.com/articles/s44360-026-00205-y
• JAMA (): https://x.com/JAMA_current/status/2114750941513572672
Source: https://jamanetwork.com/journals/jama/fullarticle/10.1001/jama.2026.21230
• NEJM (): https://x.com/NEJM/status/2114824423752548852
Source: https://www.nejm.org/doi/10.1056/NEJMp2605590
Follow Medical Boulevard for a new read-back every day.
Discuss it with us on X:
For education, not medical advice. Every post is quoted as published.
This automated news aggregator is powered by cortxt technologies, in cooperation with Medical Boulevard. Build an automated news aggregator for you: https://cortxt.com/?utm_source=readitback&utm_medium=video&utm_campaign=news-aggregator
FULL TRANSCRIPT
HOST: [curious] Who really saves when medicine uses AI? This is Read It Back, from Medical Boulevard, with posts from Eric Topol, JAMA, and the New England Journal of Medicine. The money story depends on what happens to the PATIENT. Let's read it back.
HOST: [thoughtful] Start with a deceptively simple choice… which product name does a chatbot suggest? Eric Topol, the cardiologist at Scripps Research, flagged a new Nature Health study that puts a PRICE on that question. On October twenty-sixth, twenty twenty-six, Topol posted: Drugs recommended by L L Ms differed from general medical practice and often were for higher cost alternatives. There are links in this post. You'll find them in the caption.
READ-BACK: A drug's active ingredient and its product name are different things. Here, researchers supplied the disease and the ingredient, then asked five language models which products to name. Across synthetic cases, the answers kept clustering around the SAME products… and their projected pharmacy acquisition costs often exceeded those suggested by actual prescribing patterns. [serious] The fine print: these were simulated recommendations, not patient purchases. Pharmacy substitution could change what someone actually receives. Manufacturers and payers have a stake in product choice — but this study doesn't establish PAID influence. [emphatic] So, the read-back: agreement between chatbots isn't a value judgment. A repeated answer still needs checking for appropriate options and eventual cost.
HOST: [curious] Now follow that trail to the payment decision. AI can help review whether a service meets coverage rules. JAMA, the medical journal, shared new research on Medicare's Wiser model. On October twenty-sixth, twenty twenty-six, JAMA posted: Following implementation of the Wasteful and Inappropriate Service Reduction (Wiser) model by the Centers for Medicare and Medicaid Services, traditional Medicare use of Wiser-targeted services declined in participating states versus other states during the first six months. The reduction was seen for prior authorization targets including epidural steroid injections and other selected services, while delayed-implementation services did not change, informing ongoing evaluation of the model’s effects on utilization. There are links in this post. You'll find them in the caption.
READ-BACK: Prior authorization means checking coverage BEFORE a service is delivered. Wiser combines technology with clinical review