NIH wants one safety-monitoring standard for ALL the clinical research it funds. 🔬
NIH's draft update to its 1998 Data and Safety Monitoring Policy (Federal Register 11/17) would require an NIH-approved, risk-based monitoring plan for all NIH-supported clinical research, not only clinical trials, and would remove the blanket DSMB requirement for Phase 3 and multi-site trials. NIH notes that almost two-thirds of its clinical research portfolio is non-trial studies. Public comments are open until January 6, 2027.
For care delivery, trust in research starts with participant safety and data integrity. QC Healthcare is watching this closely as we grow pediatric and adult care through My Pediatric Doctor and My Adult Doctor.
Source: osp.od.nih.gov/modernize-the…
cc @NIH @NIH_OSP @NIHDirector_Jay @HHSGov @SecKennedy @US_FDA @TheNASEM @AmerMedicalAssn
QC Healthcare: myadultdoctor.com/about-qc-h…
My Pediatric Doctor: mypediatricdoctor.com ([email protected])
My Adult Doctor: myadultdoctor.com ([email protected])
#ClinicalResearch #HealthPolicy
9
Americans are eating less sodium than a decade ago, but average intake is still above the 2,300 mg/day reference level.
New from @CDCgov's National Center for Health Statistics (NCHS Health E-Stat 124, Nov 18), using NHANES data on mean sodium intake on a given day:
• Ages 1+: 3,385.6 mg (2013-14) to 3,104.0 mg (Aug 2021-Aug 2023)
• Ages 1-19: 2,959.8 mg to 2,745.0 mg
• Adults 20+: 3,528.9 mg to 3,217.3 mg (men 3,705.8 mg, women 2,757.0 mg)
The 2,300 mg Chronic Disease Risk Reduction intake was set by @TheNASEM in 2019. NCHS notes one-day dietary recall data and a change in data collection mode for the 2021-2023 cycle. Authors: Catharine Couch, Nicholas Ansai and Cynthia Ogden.
Context: @US_FDA has voluntary sodium targets for industry, and says a report evaluating progress on its Phase I targets is due later this year.
NCHS report: cdc.gov/nchs/data/hestat/hes…
FDA sodium reduction: fda.gov/food/food-labeling-n…
cc @HHSGov @American_Heart @AmerAcadPeds
At QC Healthcare, we follow the public health data that shapes care for kids and adults. Learn more: myadultdoctor.com/about-qc-h…
My Pediatric Doctor: mypediatricdoctor.com ([email protected])
My Adult Doctor: myadultdoctor.com ([email protected])
#PublicHealth #Nutrition #CDC
5
Understanding the underlying pathways of a condition like endometriosis is how better answers eventually reach patients. The part I keep coming back to is how quickly research like this can shape what happens in a primary care visit.
xitter.gq/NIH/status/21237670106…
NEWS: Large NIH-supported study identifies key underlying pathways in endometriosis.
Read the full story ➡️ bit.ly/4B5g8W7
4
Prevention is where primary care does its best work, and coverage has lagged behind that reality. If Medicare tests food-as-medicine and lifestyle programs, the real question is whether they connect to the clinicians people already trust, and whether access reaches those who need it most.
xitter.gq/statnews/status/212385…
Medicare does not cover many of the lifestyle interventions that could keep people from getting sick in the first place. trib.al/fwh3edt
2
Better imaging that helps clinicians see arterial plaque earlier is the kind of advance that makes prevention more practical. The next step is getting findings like this into everyday primary care, where most people's heart health is actually managed.
xitter.gq/NIH/status/21237924998…
In a first-in-human study, NIH-supported researchers tested a new imaging device designed to help improve the detection and treatment of atherosclerosis, the buildup of plaque in the walls of arteries that causes blood vessels to narrow.
Read to learn more: bit.ly/48vzxDe

ALT A cross-section illustration of an artery with plaque buildup. Text overlay reads: "New imaging device could improve atherosclerosis detection and treatment. NIH-supported researchers completed a first-in-human study of a new catheter-based device that provides real-time information about plaque buildup inside the arteries." Logos for NIH and a related organization are at the bottom.
"Less spiky and more reliable" is the bar that matters for AI in care too. A tool that is brilliant on some tasks and shaky on others is hard to trust with patients. Consistency is what turns a demo into something a clinic can rely on every day.
xitter.gq/GoogleDeepMind/status/…
Our SVP @koraykv believes there won't be a single test or benchmark that tells us we've reached AGI. Instead, progress looks like models becoming less spiky and more reliable across our daily lives.
Listen in as he and @fryrsquared discuss the road ahead ↓
Timecodes:
00:00 Introduction
01:26 Evolution of Google DeepMind
04:41 Products as a path to AGI
13:01 Architectures and models
24:40 Recursive self improvement
30:50 Safety and agent teams
40:35 Benchmarks
Benchmarks like this matter because drug discovery is where AI reasoning gets tested against real scientific judgment. The harder question for healthcare is how quickly that kind of progress reaches the clinic, and how we keep human expertise at the center when it does.
xitter.gq/OpenAI/status/21238563…
MedChemBench is a new benchmark for AI reasoning in medicinal chemistry.
It shows the progress our frontier models have made in their ability to help scientists use AI to identify promising drug candidates.
openai.com/index/introducing…
3
Per Menlo Ventures, U.S. healthcare labor costs (~$1.4T) dwarf legal and software engineering, so it makes sense that AI aimed at clinical and administrative work is drawing so much attention. The real test is whether it gives clinicians time back for patients instead of adding more screens. Curious which workflows people think will move first.
xitter.gq/HealthcareAIGuy/status…
HEALTHCARE IS AI’S BIGGEST VERTICAL.
And could be bigger than legal and coding combined if you look at annual US labor costs..
> Lawyers: $140B
> Software engineers: $250B
> Healthcare workers: $1.4T
1
7
AI's promise in medicine is real, but Bill Gates is right that we have to make deliberate choices about where people stay at the center. For us, that means technology that gives clinicians more time with patients and gives families easier access, with a human accountable for the care. @BillGates
xitter.gq/BillGates/status/21236…
Replying to @BillGates
I believe AI has the potential to cure diseases, improve education, and give people more time to spend on the things that matter most. But we need to make deliberate choices about how it fits into our lives, including where we want people to remain at the center.
5
One in eight cancers tied to infections is a reminder of how much upstream care matters. Vaccination, screening and treating chronic infections early are primary care work, and virtual-first care can make that follow-through easier for families. @EricTopol
xitter.gq/EricTopol/status/21235…
—1 of 8 cancers are associated with infections (Figure)
—Autoimmune diseases and viruses, such as EBV and multiple sclerosis, enteroviruses and Type 1 diabetes.
—Cirrhosis with Hepatitis B and C
How many other diseases linked to viruses and bacteria yet to be identified?
scientificamerican.com/artic…
10
Health tech earns trust by staying power. Patients and clinicians notice who is still showing up for them after the headlines move on, and that consistency is what makes care better. @chrissyfarr
xitter.gq/chrissyfarr/status/212…
I’m a little bummed by the big tech companies all proclaiming healthcare to be the great contribution, on TV, in S-1s, and then so rarely sticking with it. There’s bigger biz in making incumbents the customer, versus trying to fix the problems.
With a few exceptions!
5
Framing AI around better services and healthier, more affordable lives is the right bar. In care, the measure is whether it gives people easier access and more time with a clinician, not just a smarter search box.
xitter.gq/nvidia/status/21235091…
AI is more than a better way to search.
Over a drink with NVIDIA’s Chris Alexiuk, @langdock Co-CEO @DadaJudith shared why AI’s biggest impact may be felt in better services, better products, and healthier, more affordable lives.
Watch their full conversation from GTC Berlin: nvda.ws/4fJhKfE
37
A blood marker for CTE would change a diagnosis that today can only be made after death. Earlier, less invasive detection could change how we counsel patients and families, provided the findings keep holding up under replication. @EricTopol
xitter.gq/EricTopol/status/21234…
We've not been able to diagnose chronic traumatic encephalopathy (CTE) in people alive, only via post-mortem brain path findings. That may likely change with discovery of a protein (eMTBR-tau243) in the blood with independent replication
nytimes.com/2026/11/19/scien…
7
Voice-first documentation keeps getting closer to the clinician's hands. The test for any of these tools is whether it gives time back to patient conversations rather than adding another device to manage. @Berci, what would make a tool like this worth carrying in a clinic?
xitter.gq/Berci/status/212370922…
Is this how physicians will interact with their AI agents in the future?
I just came across this Kickstarter project about Interface, a ring with one button: "press it, talk, and your words become a note, a message or a request to your AI agents, without taking out your phone. It works with NatureOS on iPhone and Android, and with your Mac or Windows computer."
A month before the campaign's end, they already have over a thousand backers, surpassing the original 20.000 USD goal by 460.000 USD pledged.
I think it's a good idea, and it would allow physicians to take notes and use prompts with their AI agents, but not everyone feels comfortable wearing a ring. That could be a downside.
Would you wear one if it could do what this one claims?
2
Patients rarely have a say in how the studies they join are designed, and trust in research and in care both depend on that voice. Treating participants as partners from the start seems like the right default. Which part of trial design would you open to patient input first?
xitter.gq/statnews/status/212374…
Ian Burkhart, founder of a coalition of brain implant trial participants, says device developers should give patients more agency in research. trib.al/Om5J276
1
6
The first major HIPAA Privacy Rule update since the 2013 Omnibus Rule is here (per @HIPAAJournal).
HHS Office for Civil Rights (@HHSOCR) issued the final rule Nov 16; it published in the Federal Register Nov 19. Per HHS, it:
• Cuts the time to respond to patient record requests to 15 calendar days, down from 30, with one 30-day extension
• Sets fee rules, including when copies must be free, and requires advance notice of fees
• Gives patients the right to get electronic PHI through a personal health app when the entity can produce it
• Lowers identity-verification burdens
• Allows PHI sharing with social services and community organizations for care coordination
• Drops the required written acknowledgment of the Notice of Privacy Practices
Effective 60 days after publication; compliance is due 240 days after publication (per @AHAhospitals).
HHS: hhs.gov/hipaa/for-profession…
AHA summary: aha.org/news/headline/2026-1…
OCR Director Paula Stannard leads HIPAA enforcement. cc @HHSGov @SecKennedy @CMSgov @ONC_HealthIT @AHIMAResources @HIMSS @AmerMedicalAssn
At QC Healthcare, we believe access to your own health information matters. Learn more: myadultdoctor.com/about-qc-h…
My Pediatric Doctor: mypediatricdoctor.com ([email protected])
My Adult Doctor: myadultdoctor.com ([email protected])
#HIPAA #HealthIT #PatientAccess
6
Useful frame: the model is only part of the system. The rules, the handoffs to a person and the tests you run before changing them are what make an agent dependable. In healthcare, knowing when an agent hands work to a human is the part that matters most.
xitter.gq/mardehaym/status/21233…
The better the harness, the better the business.
By harness I mean everything you build around the model.
It tells the agent where to get an answer, what it’s allowed to do and when to hand the work to a person. It also includes the tests you run before you change any of those rules.
That matters for two reasons.
#1 What an answer costs.
Say your agent gets 1,000 questions a day. If you can write a rule that answers 700 of them, only 300 need to go to the model. If each question costs about the same, that cuts the model bill by roughly 70%.
The hard part is knowing which 700. For that, you need somebody who does the work and can tell you when the usual rule is not applicable (exceptions).
On a logistics platform we rebuilt, AI cost about $200 per developer per month. What took time was figuring out how the old system calculated customers’ bills and which rules we needed to keep.
#2 What the company gets to keep.
Take the patient-billing voice agent we built for a US healthtech startup. It can give a patient the correct balance and still handle the call badly.
> Did it verify who was on the phone first?
> If the patient disputed the charge, did it hand the call to staff?
> If it scheduled a follow-up, did it check non-working hours when the message actually went out?
Those decisions are part of the harness. The system ran 13,650 tests on every pull request. In production, it handled 61,246 calls, and about 97% of patient conversations finished without a human stepping in.
The company-specific IP is clearest in the exceptions.
When an experienced billing lead explains why two calls that sound alike need different handling, you can turn that distinction into a rule and save both cases as tests.
Now the next engineer can change the workflow, or try a different model, and check whether it still makes the call the billing lead would make.
Your competitor can rent the same model, but they still have to learn which exceptions matter and what to do about each one.
If you run a software company adding AI this year, I’d put the budget there first. Sit with the people who handle the exceptions. Write down how they decide. Give the model the work that still needs judgment.
16
Encouraging to see later-born generations entering old age with better memory and slower cognitive decline. It makes the case for investing early in health, and for making sure everyday primary care reaches people wherever they live.
xitter.gq/Nature/status/21234441…
People born later in the twentieth century have a better memory when they enter old age and experience slower cognitive decline than do those born earlier
go.nature.com/4zu2RFe
5
Couldn't agree more. Clinicians who get to be operators and entrepreneurs tend to build care around what patients actually need. Virtual-first tools can help independent practices reach more patients without giving up that ownership.
xitter.gq/chrissyfarr/status/212…
5
Turning scientific breakthroughs into better health outcomes is the hard part. The gap is rarely the science, it's delivery: getting new tools into everyday primary care and into patients' homes.
xitter.gq/MSFTResearch/status/21…
AI is opening new possibilities in medical research. But what will it take to turn scientific breakthroughs into better health outcomes?
In the latest episode of The Shape of Things to Come, Microsoft Research Technical Fellow and CVP @dcburger brings together:
@avapamini, Principal Researcher Manager at Microsoft Research,
@erichorvitz, Chief Scientific Officer at Microsoft, and
Dr. @EricTopol, Executive Vice President, Scripps Research
to explore how AI is reshaping our understanding of biology and medical care and what it could mean for the future of medicine.
From accelerating scientific discovery to the challenges of real-world experimentation and clinical implementation, the group examines both the promise of AI and the work still ahead.
Could these advances help shift medicine toward earlier intervention and prevention? Watch or listen to the podcast:
msft.it/6019x1y5X
3



















