In partnership with
🤖AIBytes
Two clinical studies that deserve a closer look.
Can AI Coaching Match the Human Experience?
In this paper, researchers compared how patients engaged with and experienced an automated AI diabetes prevention program versus one led by human coaches.
Methods
This prespecified secondary analysis included 368 adults with prediabetes and overweight or obesity:
183 received the AI-led program.
185 were referred to a CDC-recognized program delivered by human coaches through live video sessions.
The AI program used reinforcement learning to personalize lifestyle recommendations. Outcomes included time to start, engagement, acceptability, satisfaction, and program preference. A total of 272 participants completed the final survey.
Results
Participants started sooner with AI:
11 days with AI.
26 days with human coaching.
P<.001.
Engagement was more evenly distributed in the AI group.
However, human coaching received higher ratings for overall acceptability, satisfaction, information, and perceived health impact.
Nearly half of participants assigned to AI, 46.5%, said they would have preferred a human coach, compared with 31.5% in the human-led group who would have preferred AI.
Key Takeaways
AI reduced delays and supported sustained engagement.
Human coaching was still viewed as more satisfying and supportive.
Greater engagement with the AI program did not translate into higher satisfaction or acceptability.
Hybrid or preference-based programs may offer a better balance.
🔗 Lalani B, Eaton C, Zade D, et al. Patient engagement, acceptability, and preference of artificial intelligence versus human coaching for diabetes prevention. npj Digit Med. Published online 2026. doi:10.1038/s41746-026-03063-w
Can AI Improve Blood Smear Analysis?
In this paper, researchers looked at whether AI-assisted digital morphology could help laboratory professionals review peripheral blood smears more accurately and in less time than conventional manual microscopy.
Methods
This prospective, multicenter, randomized paired study included 1,570 blood smears from three hospitals in China.
Each sample was reviewed with both AI-assisted digital morphology and conventional microscopy. Expert review was used as the reference for cell morphology, while flow cytometry was used for platelet estimates.
Results
AI-assisted review improved the classification of most nucleated blood cells and performed as well as or better than manual microscopy across all 13 red blood cell morphology categories.
Acute promyelocytic leukemia screening sensitivity was 100% with AI-assisted review versus 96% with manual microscopy.
Schistocyte sensitivity improved from 82.7% to 89.9%.
Median review time decreased from 852 to 328 seconds, a 61.5% reduction.
Key Takeaways
AI-assisted review made blood smear analysis faster and more consistent.
It improved the detection of several clinically important abnormalities.
The system supported lab professionals rather than replacing them.
More validation is still needed across different laboratories, equipment, staining methods, and workflows.
🔗 Lu SS, Guo YH, Liu DM, et al. AI-assisted digital peripheral blood morphology multicenter randomized paired method clinical validation study. npj Digit Med. Published online 2026. doi:10.1038/s41746-026-03089-0
🧬AIMedily Snaps
Fast updates worth knowing
5 physician leaders on what comes next in healthcare (Link).
Former Mayo Clinic research director sues system over alleged retaliation for raising AI practice concerns (Link).
Stanford Medicine researchers awarded $20 million for AI-guided research facilities (Link).
76% of Gen Z and 63% of millennials turn to AI for first line of primary healthcare (Link).
Microsoft: Teaching AI to speak the language of pathology (Link).
A paper on the Impact of AI on U.S. Health care costs and spending growth (Link).
🧪Research Signals
New papers worth your time.
Nature: Large language model-assisted referral triage automation in a tertiary hospital (Paper).
The Lancet: Identifying patients with retroperitoneal sarcoma who benefit from radiotherapy (Paper).
ArXiv: High-Stakes decisions with LLM: Insights from emergency triage (Paper).
JAMA: Consumer perspectives on trust in and benefits of AI in health care (Paper).
BMJ: The use of ambient voice technology for clinician–patient consultations in healthcare practice (Paper).
NEJM AI: Benchmarking the brain’s blood vessels — why we need specialized models in medical AI (Paper).
🦾TechTools
Useful tools to explore
MAST
A public benchmark from ARISE that compares medical AI models across clinical reasoning, safety, radiology, and imaging tasks, offering a more realistic view than general benchmark scores.
Counsel Health
A virtual primary care platform that combines medical AI guidance with the option to add a licensed physician to the conversation when needed.
📈 Productivity AI Tool of the Week
ZeroTwo
An AI workspace that organizes research, documents, and notes, then helps you turn them into clear, finished outputs.
That’s all for today.
Thank you for being here.
Medical AI is easier to follow when we learn together. Feel free to share this issue with a colleague who may find it useful.
Itzel Fer, MD PM&R
Forwarded this email? Subscribe free.
Every headline satisfies an opinion. Except ours.
Remember when the news was about what happened, not how to feel about it? 1440's Daily Digest is bringing that back. Every morning, they sift through 100+ sources to deliver a concise, unbiased briefing — no pundits, no paywalls, no politics. Just the facts, all in five minutes. For free.