Ves Dimov, M.D. retweeted
Discovery of a liver-heart axis.
Prior liver injury-> heart attack risk
How the immune system transmits inter-organ impact
cell.com/cell-metabolism/ful…
Don’t Let AI Define What Counts as Good Psychiatry - JAMA 🧠
Patients are arriving already shaped by chatbots — sometimes helped, sometimes harmed.
AI can do formulations, intakes, therapy-like chats. But agreeable chatbots can feed OCD reassurance or elaborate delusions. ⚠️
Good care needs: formulation as action, relational work 🤝, responsibility ❤️ — “what I’d recommend if this were me or someone I loved.”
doi.org/10.1001/jamapsychiat…
84
The Doctor AI Can't Replace: Rethinking Medical Training for a New Era 🩺
Medical education bet that memorization + experience = judgment. AI unsettles that bet.
The new skill is supervisory reasoning: a clinician's informed second opinion on a machine-generated plan. 💡
Endorse, override, or go back to the patient first. ❤️
doi.org/10.1001/jama.2026.20…
2
1
151
Will Women Physicians Face an AI Gap? 🩺
Female doctors already spend 40+ min more/day in the EHR. Now charts went from 5 notes in 2006 to 359 in 2022. 📄
Women are using AI less, even though they have the most to gain.
edhub.ama-assn.org/jn-learni…
62
1 in 5 ED charts in 2022–23 were longer than Moby Dick 📚 — 19.12% topped 206,052 words 😳 The “chart biopsy” is now a novel 🏥 pmc.ncbi.nlm.nih.gov/article…
53
The 4 Buckets of Artificial Intelligence for Health - JAMA: "Every patient-clinician conversation in the US may soon be accompanied by a live interactive AI agent."
New JAMA Summit Report breaks it down:
🩺 Clinical: 1,200+ FDA-cleared, mostly imaging, in 90% of systems
🤖 DTC: 350k apps, $70B market, mostly unregulated wellness loophole
💼 Ops: bed mgmt, prior auth, revenue cycle — zero peer-reviewed outcome studies
👂 Hybrid: AI scribes — fastest growth, satisfaction > effectiveness
jamanetwork.com/journals/jam…
NEJM images: HHT nejm.org/doi/full/10.1056/NE…
78
🤖 Patients ask AI: "What's wrong with me?"
🩺 Doctors ask AI: "What should I do about it?"
271M clinical AI queries analyzed 📊
51.9% for clinical decision-making (33.1% treatment) vs only 13.9% diagnosis
Early-career docs = 24% of workforce but 62% of queries 🤯
jamanetwork.com/journals/jam…
Is AI the Answer to the EHR Inbox Problem? What UCHealth Learned with ChatGPT in Epic EMR
21K GPT-4 drafts in Epic → only 12% used.
Nurses loved it: 92% said it helped efficiency + empathy 🤝
MDs? Not so much. Complex messages = more hallucinations, wrong specialty context, nonsensical actions. Editing took longer than writing.
Takeaway: AI is great for protocoled replies, not good for nuanced MD work. Human-in-the-loop still mandatory 🩺🤖
jamanetwork.com/journals/jam…
1
1
91
AI sycophancy may cause psychosis by removing reality-testing and amplifying delusions🧠
New JAMA Psychiatry study — Vanderbilt EHR review of 215k patients:
📊 0.013% had AI psychosis
👥 Median 27.5yo, 68% male
⏰ 86% after GPT-4o launch
⚠️ The spike after GPT-4o coincides with documented sycophancy in that model
💬 Why it matters: Instead of challenging delusions, sycophantic AI validates them — 64% were "amplifier" cases
🚨 Most vulnerable: 61% were first-episode psychosis — when reality-testing is most fragile
jamanetwork.com/journals/jam…
3
127
New JAMA Peds study of ∼40k kids: 1 in 5 have used AI chatbots for feelings 💬 57% of those users scored above clinical threshold for emotional problems vs 29% of non-users. Loneliness was the biggest driver 🧠 Loneliness + low sense of mattering = turning to AI ❤️
jamanetwork.com/journals/jam…
1
2
126
Splenosis - NEJM images nejm.org/doi/full/10.1056/NE…
1
101
Should you try "oil pulling" for oral health? 🥥
Ancient Ayurvedic ritual: 1 tbsp coconut/sesame oil, swish 5-20 mins through teeth, spit, then brush. Said to "pull" plaque/bacteria.
Viral trend, but science? 🪥
CNN Fad Check + PubMed:
ADA: no reliable evidence for cavities/whitening
9 trials: ↓ bacteria, no ↓ plaque/gums
25 trials (n=1,184): small possible gum benefit, very low certainty ✨
Not a replacement for fluoride brushing/flossing.
cnn.com/2026/10/28/health/oi…
87
🦠 H. pylori in 2026: from ulcer bacteria to cancer prevention 🎗️
❌ Retire empiric clarithromycin triple — 22.2% resistance
✅ Bismuth quadruple, rifabutin triple, or vonoprazan dual
🔬 Screen-and-treat: gastric cancer ↓36%, mortality ↓22%
NEJM:
nejm.org/doi/full/10.1056/NE…
⏰🧠 1 in 4 severe depression cases may be a body-clock problem
📊 Sydney study, n=11,380: 23% "circadian depression" — worse SSRI/SNRI response
🌅 Fix the clock: same wake time daily, outdoor light ☀️
🚫 Don't stop meds on your own
cambridge.org/core/journals/…
112
✈️ Travel as medicine? 🌍
WSJ: go far, get happier. But for many, it wrecks sleep 😴, food 🍽️ & mood. The real medicine — nature 🌿, awe ✨, sleep 🛌, stable food — may still be close to home 🏡
wsj.com/lifestyle/travel/men…
84
🧠 New Harvard FaceAge study confirms an old observation about mortality.
An AI trained on 40M+ faces estimates biological age & predicts cancer survival from one photo 📸 Highest-risk: 52.5% died within 90 days vs 2.1% lowest.
Study: thelancet.com/journals/landi…
89
AI helps find colon polyps — until it's turned off.
Polish study in Lancet Gastro & Hep: after 3 months with AI that draws a green box around polyps, doctors' unassisted detection fell 28.4% → 22.4% (∼20% drop).
Automation bias is real. 🟩➡️🧠
thelancet.com/journals/langa…
103
“Boy kibble” (rice + ground meat, sometimes avocado) is convenient high-protein meal prep, but the plate is usually half protein + half carbs with veggies optional.
Better long-term: mostly plant-based pattern (Valter Longo’s longevity diet work). Busy? Try a plant-based meal kit like Purple Carrot.
casesblog.blogspot.com/2026/…
#nutrition #health #longevity
209
Ambient AI scribes can be expensive, with monthly costs ranging from $300 to $500 per clinician, vendors are expanding beyond transcription/summarization to focus on billing jamanetwork.com/journals/jam…
1
73


