Healthcare AI Daily Briefing
New to OpenRounds? Get the free daily email or see how we work.
Research & Evidence
Survey of 524 health care professionals in the United States and United Kingdom finds AI use common but varies by role and country.
Journal article (JMIR formative research)
A cross-sectional survey of 524 health care professionals in the United States and United Kingdom showed that AI use is common but varies by professional role and country. Respondents emphasized that evidence, organizational guidance, training, privacy safeguards, and workflow integration are key factors for responsible adoption. [1]
Diagnostics & Pathology
Llama 3.3 70B pipeline extracts oncology registry variables with high F1 scores.
Preprint, not peer reviewed
At a Singapore cancer centre, a privacy‑preserving pipeline based on Meta Llama 3.3 70B, deployed on‑premises with a Hide‑In‑Plain‑Sight de‑identification step, extracted more than 50 registry variables from 6,700 oncology notes without task‑specific fine‑tuning. On 200 clinician‑adjudicated notes, F1 scores ranged from 97.2 % for diagnosis to 88.1 % for treatments. A structure‑aware evaluation distinguished omissions from fabrications, and early testing on 50 external notes showed strong performance for core diagnostic variables but reduced accuracy for list‑valued fields. [2]
Diagnostics & Pathology
Interview: Amazon Pharmacy integration hurdles highlight need for easier care navigation.
Podcast from Raising Health
The interview explains that Amazon Pharmacy had to retrofit an ecosystem never intended for protected health information or HIPAA compliance. It cites an example where General Medicine arranged a biopsy within three days, illustrating the gap between abundant health advice and the difficulty of acting on it. [3]
Industry & Products
Interview: Japanese medical center teams with OpenAI to let clinicians query records via prompts.
Podcast from KFF's The Business of Health with Chip Kahn
In a KFF interview, a Japanese medical center describes a partnership with OpenAI that runs OpenAI for Healthcare on its clinical data, letting clinicians prompt the record with questions such as what happened with a patient recently, what might be missing in a differential diagnosis, or why a patient is not on aspirin. [4]
Policy & Governance
Interview: Author contrasts flawed physician credentialing with continuous AI simulation testing.
Podcast from Health & Veritas
The Health & Veritas interview reports the author's argument that current physician credentialing is narrow and episodic, whereas AI systems could be continuously evaluated using realistic simulations that expose them to difficult cases and allow revision on failure. [5]
Sources
- Artificial Intelligence Use Among Health Care Professionals in the United States and United Kingdom: Cross-National Survey. · JMIR formative research Original source
- Privacy-Preserving Large Language Model Deployment for Oncology Registry Abstraction: Structure-Aware Evaluation in a Real-World Clinical Setting · medRxiv Original source
- Building the Store for Healthcare | General Medicine · Raising Health Original source
- Building the AI Hospital · KFF's The Business of Health with Chip Kahn Original source
- Long COVID, Regulating AI Medical Advice, and the Hidden Costs of Prior Authorization · Health & Veritas Original source
Daily healthcare AI brief
Get the free healthcare AI briefing
One concise briefing built from reporting, research, policy, blogs, videos, and full podcast transcripts.
Free. Get a sign-in link by email to finish subscribing. An account is created if needed. Unsubscribe anytime. Already subscribed? Sign in to save and personalize.