The UK’s health secretary says that AI-driven health checks – called MOTs for measurements, observations, and tests — could transform care for frail patients who are over 65 by using machine learning and genomics to speed diagnosis, guide treatment, and predict illness. Japan offers a similar early detection program called Ningen Dock, a cash-only program that uses imaging, endoscopy, and lab work to generate personalized risk assessments. That service is also offered to foreign residents in a medical tourism package that is covered by some US insurers, such as Aetna (above).
Google DeepMind CEO Sir Demis Hassabis – who won a Novel Prize in chemistry last year after starting his career as a designer of widely popular video games — predicts in a “60 Minutes” interview that AI will reduce drug development time from years to weeks, making all diseases curable within 10 years.
Agentic coding platform Cursor draws online scorn and customer cancellations after its AI support agent incorrectly blames a user’s inability to run multiple sessions as company policy rather than a software bug. Cursor says that it will start labeling AI-generated responses after users questioned whether it was trying to pass off its “Sam” assistant as human. It also fixed the bug that the user had reported.
Business
Just 30% of healthcare AI pilot projects reach production, a new study finds, most often stalled by security concerns. Providers are much more interested in trying and buying AI solutions than they were with EMRs.
Middle East specialty provider Burjeel Holdings will use Hippocratic AI’s agents for patient-facing, non-diagnostic clinical tasks that will be delivered in multiple languages.
Research
A new study finds that AI models beat PhD-level virologists in solving complex wet lab problems, scoring 44% versus the experts’ 22%. While the findings offer hope for advancing infectious disease control, they also raise alarms about AI’s potential use to develop bioweapons.
Stanford Health Care researchers find that endocrinologists view AI-generated draft responses to patient portal messages as helpful, but see tools that use patient data, such as for triage, as risky. AI was rated most useful for administrative tasks like writing authorization letters and patient education, with the greatest potential use being the management of patient scheduling.
Other
A Spain-based Microsoft engineer who was frustrated by repeated misdiagnoses of his son’s rare condition develops DxGPT, an AI tool that analyzes user-reported symptoms to suggest possible diagnoses. Access is free.
Anthropic adds a research feature to Claude that searches both internal and web-based content, using an agentic approach to run iterative queries, resolve open questions, and deliver well-sourced answers with citations.
A proposed federal bill would create a consistent Medicare reimbursement path for FDA-approved, AI-enabled medical devices by placing them in a new technology ambulatory payment classification under the Hospital Outpatient Prospective Payment System for at least five years, allowing time to collect data before determining if a permanent code should be created.
Research
A University of Florida researcher develops an open source AI tool that analyzes patient movement videos to help doctors detect subtle motor changes.
Other
North Carolina Central University is using Wolters Kluwer’s VRClinicals for Nursing, a virtual reality hospital simulation, to train its nursing students.
A medical writer with a PhD in math says that she doesn’t want AI scribes to write visit notes for her pulmonologist.
My pulmonologist’s notes are much more than a summary of our privileged clinical encounters. Each of his notes is an important and carefully crafted document for my care planning and for coordination with other providers. Equally important, the notes are a communication to me, his patient. As I read his notes, I can feel his acumen and experience as a practitioner of medicine — his interest and understanding, his concern and compassion, his discernment and responsiveness. I don’t think an algorithm can re-create those specifically human experiences.
It’s interesting that the 2023 story about ChatGPT outdiagnosing 17 doctors is suddenly trending again despite no new developments. Maybe the AI has moved beyond diagnosis to ghostwriting clickbait.
Counterforce Health offers no-cost AI tools that write patient appeal letters for insurance denials.
Google Cloud and Seattle Children’s Hospital launch Pathway Assistant, an AI-powered tool that gives clinicians instant access to evidence-based care pathways for 70 pediatric diagnoses.
Spain-based Neurologyca says that its Kopernica AI facial analysis tool can identify 100 emotions, even in crowds, which allows it to detect neurological conditions such as stroke and brain hemorrhages.
New CMS Administrator Mehmet Oz, MD, MBA says in his first town hall meeting that AI-powered avatars may outperform frontline doctors and may be preferable to humans by patients.
Research
An NIH-funded AI screening tool matched provider performance in identifying hospitalized patients who were at risk for opioid use disorder and recommending specialist referrals, while reducing 30-day readmission odds by 47%
A study by the Icahn School of Medicine at Mount Sinai finds that AI models may recommend different treatments for identical medical conditions based on patients’ socioeconomic and demographic backgrounds, highlighting the need for safeguards to ensure equitable AI-driven medical care.
A study reports that implementing an AI-powered video management system enabled comprehensive documentation and analysis of over 10,000 minimally invasive surgeries, supporting the standardized assessment of key surgical steps and the self-assessment of surgeons.
Other
A new AI system for pediatrics that draws on data from 300 clinicians and years of hospital records will be deployed across China to deliver personalized care and interact with patients’ families.
April 2, 2025Healthcare AI NewsComments Off on Healthcare AI News 4/2/25
News
Amazon releases Nova Act, an AI model that can perform actions within a web browser.
An Israel-based medical advocacy group urges the health ministry to develop AI usage guidelines and enforcement, citing cases in which clinicians misused AI tools and harmed patients. It says that faulty AI-generated medication dosing and diagnostic suggestions have caused life-threatening errors.
Apple will reportedly add an AI health coach to an updated Health app as early as 2026.
Business
Layer Health, which offers an AI-powered EHR data abstraction tool, raises $21 million in a Series A funding round.
Research
A new KLAS report finds that many health system executives are wondering if their organizations are falling behind their peers in using AI, but most are just getting started with strategy, governance, and choice of specific products. All executives from large health systems report that their organizations are using AI. EHR vendors are driving AI adoption through integration with their core products. AI use in imaging, operations, security, and revenue cycle is more focused but growing.
A small Dartmouth study finds that people with significant behavioral disorders improved after using a therapy-focused chatbot app. Depression symptoms dropped by 51%, while those with depression, anxiety, and eating disorders saw smaller but still meaningful improvements. The researchers say that chatbot-delivered therapy rivaled the results of traditional outpatient care and could help ease the mental health provider shortage.
Other
A woman in Australia files a complaint with the health regulator after her doctor used ChatGPT to interpret her lab results during a visit, which she called “lazy and unprofessional.”
OpenAI adds an advanced image generator into ChatGPT-4o. The new tool can handle up to 20 objects, learn from user-uploaded images, and create photo-realistic images. I had it reimagine a HIMSS conference session as if it were happening on the deck of a Caribbean cruise ship. This is my brilliant idea for guaranteeing booth traffic by trapping attendees at sea with nowhere to go except the exhibit hall. Think “Monsters of Rock Cruise” but with more AI panels and fewer grandparent-aged musicians stuffed like a sausage into long-mothballed leather pants.
China-based DeepSeek releases V3.1, which is free for commercial use and can run locally on a high-end Mac Studio.
Amazon is testing a health-focused chatbot that answers wellness questions and recommends products, with responses that it offers that have been reviewed by a US-based clinician marked with a “clinically verified” badge.
Business
Cleveland Clinic partners with UAE-based AI firm G42 to develop and apply healthcare AI solutions.
Healthcare AI clinician copilot developer Navina raises $55 million in a Series C funding round.
Research
A Stanford Medicine study finds that an AI tool that was trained on 80,000 EHR-based nutrition orders for premature infants could reduce medical errors, save time and cost, and improve care in low-resource settings. The AI identified 15 standardized TPN formulas that meet the needs of most patients and was able to accurately recommend the best option for each case. Neonatologists consistently preferred the AI-generated orders.
A fascinating Harvard Business School study of Procter and Gamble commercial and R&D employees looks a how AI might change worker performance, expertise sharing, and social engagement:
Individual employees who used AI produced work of similar quality to two-person teams without AI, suggesting that AI can function as a “cybernetic teammate.”
Teams that used AI submitted more balanced solutions that crossed functional boundaries, unlike non-AI teams whose ideas reflected only their own area of expertise.
Participants who used AI reported increased excitement and energy, indicating that AI can replicate the social and motivational aspects of teamwork.
AI enabled less-experienced employees to achieve output quality that was comparable to that of seasoned innovators.
AI-assisted individuals and teams spent up to 16% less time on tasks while producing longer and more comprehensive solutions.
AI-supported teams were three times more likely to generate top-quality solutions.
Other
A Duke University Medical Center study finds that ChatGPT accurately routed radiology procedure requests when it had been trained on team members, contact info, and schedules, but struggled with out-of-scope requests like those during evenings and weekends.
China’s military is using DeepSeek AI in its hospitals, where it provides treatment suggestions.
A New York Times article explores how doctors are using AI to repurpose existing drugs to treat rare diseases, 90% of which lack approved treatments and draw little interest from drug companies due to limited profit potential.
A woman asks ChatGPT about her anxiety and grief but rejected its suggestion of blood cancer, certain that her symptoms were caused by her father’s recent death. As her condition worsened, she saw a doctor who found that she had Hodgkin’s lymphoma.
March 19, 2025Healthcare AI NewsComments Off on Healthcare AI News 3/19/25
News
Nvidia releases GR00T N1, an open source foundation model for generalist humanoid robots.
Google unveils new health-focused features at its The Check Up healthcare event:
What People Suggest, a new search feature that summarizes online discussions from individuals with the same condition.
FHIR support for Health Connect, which is now API-enabled to read and write medical data in FHIR format.
Loss of pulse detection, an FDA-cleared Pixel Watch 3 feature that automatically calls emergency services if the wearer’s heart stops.
Korea-based LG AI Research announces Exaone Deep, an LLM with advanced reasoning capability that the company says beats DeepSeek R1 on math benchmarks while being 95% smaller.
Researchers find that Metat’s open source AI model Llama outperformed OpenAI’s GPT-4 in diagnosing complex medical cases. Open source LLMs offer healthcare advantages such as auditability, bias detection, and domain-specific tuning for medical literature and patient records. They integrate more easily into healthcare systems, reduce costs, prevent vendor lock-in, and can be deployed locally or in secure environments to protect patient data.
Illinois lawmakers advance bills to restrict online behavioral health providers from using AI for therapy sessions and to limit insurer use of AI to deny coverage.
India’s Apollo Hospitals will increase its investment in AI with the goal of freeing up 2-3 hours daily for doctors and nurses. AI is being used to suggest diagnoses, recommend tests and treatments, transcribe physician notes, and generate daily schedules from nurse documentation.
Business
R1 and Palantir launch R37, an AI lab that will focus on developing automation solutions for healthcare reimbursement.
Urgent care operator CityMD will use Notable’s AI platform and agents for patient scheduling, messaging, registration and intake, and payments. CityMD is owned by VillageMD, whose majority owner Walgreens Boots Alliance is expected to seek a buyer for the business after its own acquisition by a private equity firm.
Other
A JAMA Viewpoint article argues that since lawmakers are unlikely to expand the FDA’s oversight of AI standards, private governance mechanisms — such as licensing agreements between AI developers and health systems — will be necessary. The authors also warn that litigation over AI-caused patient harm will rise, but that won’t drive safety improvements because physician users will bear the brunt of liability.
OpenAI introduces a developer platform for building AI agents that includes tools to perform web and file searches and to perform web-based tasks similar to its Operator browser.
NHS England is deploying an AI tool that can predict a patient’s risk of falling with 97% accuracy. The software, which was developed by Cera, is also being used to predict deterioration in home care patients.
A study finds that patients slightly preferred AI-generated responses to their portal questions over human-written ones, but reported lower satisfaction when told that the response came from AI. The authors conclude that patients should be told that AI was used since it didn’t reduce satisfaction significantly. They also polled patients on their preferred wording of the disclosure, with the winner being, “This message was written by Dr T. with the support of automated tools.”
Business
Memorial Sloan Kettering Cancer Center completes a pilot of Abridge’s AI ambient documentation and plans a broad rollout over the next two years.
AI drug discovery company Insilico Medicine deploys a “bipedal humanoid” to train AI systems on the tasks performed by laboratory scientists. They are also using the robot, called “Supervisor,” to assist with lab tours, telepresence, and lab supervision.
Research
Researchers find that LLMs show promise in reducing pediatric medication dosage errors. A medication ChatGPT and Claude were more accurate and faster than pediatric and neonatal nurses, while Llama performed poorly due to an apparent weakness in its calculation logic. The authors recommend evaluating specific LLMs rather than treating all of them as equally capable.
LLMs exhibit “anxiety” when processing emotional mental health topics like interpersonal violence and accidents. Researchers found that mindfulness-related prompts could help regulate the model’s responses, similar to how human therapists manage their emotional reactions while maintaining empathy.
Stanford researchers use AI to identify a naturally occurring prohormone that is as effective as Ozempic in weight loss without the side effects of nausea, constipation, and loss of muscle mass.
Other
Patients are using LLMs to analyze their hospital bills for charges that exceed state and national averages. New startup OpenHand is offering similar analysis, after which the company negotiates with providers to lower the bill.
TikTok users report that AI-generated deepfake doctors are spreading medical advice on topics like surgery, diet, and cosmetic procedures. Some use the Captions app to create and edit AI videos that can be easily replicated with different messages, which is how the users noticed the fakes.
Epic and Agentic AI
A reader asked for my take on Epic’s plans for agentic AI. I have no inside knowledge, so this is pure spitballing.
Some background. Agentic AI acts independently to achieve goals without human oversight while responding dynamically to its environment. Think self-driving cars. It renders robotic process automation (RPA) obsolete, as RPA relied on rigid rules and predefined inputs. It’s hard to believe it’s been just four years since Olive was health tech’s hottest startup.
Non-agentic AI, by contrast, requires human direction. Chatbots are an example. They answer questions and retrieve information but don’t take external actions like scheduling appointments. In between are limited function, app-specific copilots that assist users without initiating decisions.
The business case for agentic AI is workflow automation, reduced labor costs, real-time monitoring (cybersecurity, throughput, resource allocation), and rapid feature deployment. Instead of modifying core systems via traditional coding, testing, and releases, AI can introduce new functionality faster and allow customization at the client level. It also streamlines integrations with external systems. All of this is theoretical, of course, and is heavily dependent on the vendor and user organization.
Epic has already embedded non-agentic AI across its platform, with use cases like drafting patient replies, simplifying documents, automating prior authorizations, and enabling voice control. These are quickly becoming table stakes with AI’s ubiquity.
Agentic AI is the logical next step, and Epic seems to be out front, although Oracle Health’s plans aren’t quite clear yet either. Early implementations will likely focus on low-risk back-office tasks, then expand into clinical support, population outreach, and automated reminders. Unlike third-party AI vendors that rely on brittle workarounds like screen scraping, Epic can integrate AI natively and provide scalability and stability.
AI’s role in clinical decision support is gaining acceptance, as long as a human remains in the loop as FDA requires to avoid inviting regulation as a medical device. Future AI applications could preassemble patient histories, flag care gaps, match patients to clinical trials, and pull relevant literature. AI could also be used to personalize the patient’s treatment and communication.
Few vendors have the resources to develop and support AI agents that have unknown ROI. Reputational risks from AI errors and regulatory scrutiny will be a deterrent for some companies. Another possible barrier is the willingness of a developer-focused software company to allow an AI agent to take over software flow but still support normal user interaction.
Epic benefits from its homogeneous customer base and a track record of incremental software development. It doesn’t need to chase AI-jazzed investors, so it can roll out tools when it’s ready in an Minimum Viable Product-type approach.
Epic also has advantages such as its Cosmos data repository, the ability to integrate deeply with its existing products, and the market power to influence what partners and competitors do.
I would expect Epic to deploy both agentic and non-agentic AI initially to reduce clinician burden and surface relevant insights within workflows. It will probably have another group working on reducing the health system labor that is needed to basically push (electronic) paper that someone outside the health system requires. It will eventually use AI to adapt its underlying software to user preferences. It will probably tread lightly at first with clinical functions, making sure to allow opt-outs and human overrides when the AI’s confidence is low.
On the big-picture operational side, Epic will position itself as offering an intelligent, proactive platform for hospital management, which people have been talking about for years. That will be a significant development assuming that early adopters show measurable improvement in moving from “tools” to “systems.”
Success depends on Epic’s ability to build new expertise in AI and determine the level of cloud dependency its customers will accept. It’s likely already working with an early adopter cohort, though we won’t hear much outside of UGM presentations. By August, we should have a clearer picture of its direction. Anything in the meantime is speculation, which I wouldn’t have offered if the reader hadn’t asked. Your thoughts are welcome.
March 5, 2025Healthcare AI NewsComments Off on Healthcare AI News 3/5/25
News
Microsoft announces Dragon Copilot, a voice assistant that is designed for ambient documentation, task automation, information retrieval from trusted medical sources, and AI-powered document generation.
Talkdesk announces AI-powered healthcare contact center agents that can schedule appointments, verify benefits and prior authorizations, and manage prescription refills, all in any language.
Salesforce announces pre-built healthcare AI agents that include provider search and scheduling, care coordination, benefits verification, and customer service.
Google Cloud enhances its Vertex AI Search for healthcare to allow searching tables, charts, and diagrams. Meditech is using Vertex AI Search in Expanse and will implement the new capability.
Mednition develops an AI-powered sepsis detection model with 95% sensitivity and 96% specificity. The company’s AI platform has received FDA Breakthrough Device designation.
Research
Researchers suggest that clinical AI systems should replicate the collaborative approach of multidisciplinary medical teams to enhance trust. They note that clinicians focus on high-level concepts rather than data details when communicating, such as a dermatologist who talks about mole characteristics rather than image pixels.
Other
Health authorities in China’s Hunan province ban telehealth doctors from using AI to generate prescriptions. Online providers must connect to the province’s electronic prescribing system and verify their identities to document that they are delivering their services personally.
February 26, 2025Healthcare AI NewsComments Off on Healthcare AI News 2/26/25
News
OpenAI expands access to its recently released Deep Research tool, which generates in-depth reports, to all paying ChatGPT users. It previously required a Pro-level subscription at $200 per month. The output above comes from my request to compare atrial fibrillation symptom relief and quality-of-life outcomes between cardiac ablation and rate-control drugs. The tool provided a running narrative of insights from 22 reputable sources before compiling a comprehensive report, which took several minutes to generate.
Meanwhile, XAI makes Grok 3 – which has similar DeepSearch agentic search functionality – available for free to all users.
Arizona lawmakers unanimously pass a bill that prohibits insurers from using AI alone to reject claims, deny prior authorization, or make other decisions that require medical judgment. The legislation mandates that insurers assign a clinician to review AI-generated decisions.
A survey of 10,000 people across 20 countries on AI replacing human jobs finds that their top concerns are doctors and judges.
Amazon adds AI capabilities to Alexa, enabling it to make reservations and appointments, play music, order food or restaurant delivery, and book service providers. A new mobile app allows seamless conversations across Echo devices, the web, and the app. Users can also share documents like schedules, study materials, and emails for reminders, summaries, or actions. Alexa+ costs $19.99 per month but is free for Prime members. Early rollout begins in the coming weeks, with priority given to recent Echo Show models.
Business
Charta Health, whose founders sold their previous AI company to OpenAI, raises $8.1 million in a seed funding round. Its product uses AI to automate patient chart reviews to find missed billing codes.
OpenEvidence, which offers a clinical decision support chatbot for providers, announces $75 million in funding at a $1 billion valuation. It recently signed a content agreement with The New England Journal of Medicine.
A physician reviews how Hims uses its MedMatch AI system to drive business:
The AI analyzes basic patient data to suggest optimal medications.
Physicians can prescribe faster with personalized treatment recommendations and improved outcomes.
Patient trends help identify custom or compounded medications that Hims can sell at higher margins than generics, creating a competitive advantage. More than half of Hims patients use personalized medications, distinguishing the company from generic drug sellers.
The system can justify selling compounded versions of drugs like GLP-1s for weight loss by recommending doses that vary from commercially available products by more than 10%, which avoids FDA oversight.
Research
A small study of hospital clinicians finds that use of the DAX Copilot ambient scribe that is integrated tool with the Epic Haiku mobile EHR app was associated with greater efficiency, lower mental burden, and a greater sense of engagement with patients.
Researchers develop a framework for radiologist reading of chest X-rays that follows the gaze of the radiologist, then focuses on the image areas that drew the most attention.
Other
An AMA physician survey finds that 61% worry that payer use of AI will increase denials of their prior authorization requests. One-third of respondents say that their PA requests are often or always denied, while 82% say that the process at least sometimes forces patients to receive something other than the physician’s preferred treatment.
Another AMA survey finds that physician use of AI jumped from 38% last year to 66% now. Common use include creating billing and visit notes documentation, creating discharge instructions and progress notes, language translation, and diagnostic support.
February 19, 2025Healthcare AI NewsComments Off on Healthcare AI News 2/19/25
News
Google launches Co-Scientist, a Gemini 2.0-powered research LLM that researchers say is like having an expert collaborator. Users provide a plain language research goal, after which the tool provides a summary of published literature and cited suggestions for new hypotheses and possible experiments.
Healthcare accreditor URAC announces plans to release a healthcare AI accreditation program later this year.
CVS Health redesigns its app with AI-driven features to provide a personalized health concierge experience. The app allows users to manage prescriptions for their entire family across CVS Pharmacy, CVS Caremark mail order, and CVS Specialty pharmacies. It also allows them to open locked display cabinets and pick up prescriptions using a barcode. Future enhancements include conversational AI for checking refill status and tailored recommendations for chronic condition management.
Business
Innovaccer launches telephone-based AI voice agents for patient scheduling, protocol intake, referral, authorization, care gap closure, HCC coding, and patient access.
HP acquires wearable device maker Humane AI for $116 million, gaining its employees and software while discontinuing the device. The widely hyped product, criticized as a solution in search of a problem, will be retired. The year-old, $699 Humane AI Pin — which also required a $24 monthly subscription – will become dysfunctional next week when its cloud service shuts down. The startup had raised $230 million and sought a $1 billion buyout.
Ambient documentation vendor Abridge raises $250 million an a Series D funding round. The company was previously valued at $2.5 billion.
Crunchbase lists five healthcare-related companies whose new fundraising rounds suggest a Unicorn Club valuation of at least $1 billion:
Neko Health (body scanner), $1.8 billion.
Hippocratic AI (AI agents), $1.6 billion.
Aragen Life Sciences (drug discovery), $1.4 billion.
Truveta (healthcare data from its provider owners), $1 billion.
Cera (digital-first home health), $1 billion.
Research
A UK-based company develops an AI-powered “super test” for prostate cancer screening that offers greater accuracy and sensitivity than traditional tests like PSA. The multi-omics test uses AI to analyze blood and urine samples for the presence of specific genes and proteins that have been clinically associated with the disease.
University of Michigan and other organizations are using a federal grant of up to $25 million to develop an AI-equipped van that can help medical generalists deliver hospital-level services in rural areas.
February 12, 2025Healthcare AI NewsComments Off on Healthcare AI News 2/12/25
News
Nine hospitals in the UK are using AI to prioritize high-risk patients who are stuck in the NHS’s 7.5 million-long waiting list. The C2-Ai system has flagged 1,000 patients who are at risk of deterioration during their wait to be seen, which reduced surgery-related complications and shortened inpatient stays by four days.
Mass General Brigham and IBM partner through the IBM Sustainability Accelerator to develop an AI tool that will predict extreme heat events, identify at-risk patients, and send warnings to those who are at risk. Previous research indicates that heat event days will cause 235,000 ED visits, 56,000 hospital admissions, and $1 billion in costs, much of that in urban areas.
OpenAI CEO Sam Altman predicts in a blog post that AI will soon reach human-level problem-solving ability (AGI), a transformation that will be comparable to the invention of the transistor. Key points:
AI intelligence scales predictably with investment since it correlates with the log of resources used.
AI costs drop 10x every 12 months, far outpacing Moore’s law.
Exponential AI investment will persist due to its super-exponential socioeconomic value.
AI agents will handle most tasks of mid-level professionals but won’t generate big ideas, will require human oversight, and will still fail in some areas.
AGI’s greatest impact will likely be in science.
Former FDA Commissioner Scott Gottlieb, MD urges the FDA not to classify AI as a medical device if it simply synthesizes and presents information, warning that excessive regulation could hinder AI integration into EHRs and limit its ability to generate clinical insights. He emphasizes that AI is most effective when it is embedded in physician workflows with access to EHR data.
Business
Australia-based Harrison.ai, which offers AI diagnostic solutions for radiology and pathology imaging, raises a $112 million Series C funding round. The company will use the proceeds to expand its US operations.
Cedars-Sinai nurses are piloting an AI tool that was developed by Aiva Health, which was developed through the hospital’s accelerator program. Nurses dictate into the nurse assistant app, which then populates Epic fields upon their approval.
Research
An American Medical Association survey finds that two-thirds of physicians recognize the benefit of AI. Use of AI in practice doubled in one year to 66%, while 57% see administrative burden as its biggest opportunity.
Other
A student-run media site profiles how Brown University Health physicians are using AI, including simplifying the language of surgical consent forms and using OpenAI’s Voice Engine to provide patients who have lost their voice with a text-to-speech tool that sounds like them.
Harvard Medical School psychiatrist and bioethics professor Rebecca Brendel, MD, JD ponders whether AI should be used in end-of-life decision-making, raising these issues:
Clinicians will respect the wishes of a patient as long as they are competent to make decisions and don’t want something that is medically contraindicated.
AI might help after the initial diagnosis if decisions will be made between treatment and palliation.
It might also help in situations where patients are incapacitated, have few relationships, and have avoided treatment.
It’s not enough to decide from the odds of survival alone since people who have experienced a sudden tragedy wouldn’t necessarily have the same wishes as someone who has battled a chronic illness over time.
Decisions of high consequence should always be made by humans.
Information asymmetry has always caused healthcare professionals to be held in high esteem, but AI might outperform humans in some aspects, which would shift professional emphasis to how they use information and interact with patients.
February 5, 2025Healthcare AI NewsComments Off on Healthcare AI News 2/5/25
News
China-based TikTok parent ByteDance releases OmniHuman-1, an AI tool that transforms a single photo into a lifelike video with natural singing and speech. The technology is raising concerns about more advanced and easily generated deepfakes.
Biomedical researcher Derya Unutmaz, MD uses ChatGPT’s new Deep Research tool to review two cancer cases. He reports, “Both reports were simply impeccable, like something only a specialist MD could write. There’s a reason I said this is a game-changer.” He has previously said that AI will be a standard, required tool in medical practice within two years, and electing not to use it might be considered malpractice. Deep Research is available to ChatGPT Pro subscribers and will be rolled out eventually to Plus subscribers. Users of X say that a key will be if OpenAI can figure out how to provide access to paywalled medical journals.
Britain’s NHS launches a 700,000-patient trial of using AI to screen for breast cancer. The Royal College of Radiologists has urged the government to embrace healthcare AI to help offset a clinician shortage.
Business
Beacon Health System will implement AI-driven medical necessity reviews using the generative AI tool of Xsolis. The health system began using company’s AI to manage people with chronic illness in May 2019.
ScribeAI launches an ambient documentation solution. Co-founder and board chair Kyle Robertson remains CEO of Zealthy. He was fired as CEO of behavioral telehealth startup Cerebral in May 2022 due to federal scrutiny of the company’s controlled substances prescribing practices, after which the company paid $3.6 million in November 2024 to settle charges of encouraging the unauthorized use of controlled substances.
OpenAI founder and CEO Sam Altman says that the company is considering developing a healthcare-focused version of ChatGPT that would comply with HIPAA.
Research
A JAMA Network discussion of using a secure instance of ChatGPT to draft clinician responses to patient messages makes these points:
Patients preferred the AI-generated messages.
A key factor was that the AI messages were more detailed than the quick responses that clinicians wrote in their available time.
The authors note that appointment lead times are so long that patients will be using AI tools to ask medical questions, even in specialty areas that are not underserved.
The authors say that healthcare, like other industries, will need to decide between improving clinician working conditions based on freed-up time versus just increasing their workload.
University of Cincinnati associate professor Dong-Gil Ko, PhD proposes a telehealth billing model that rewards experienced, efficient physicians instead of using time-based metrics that pay all providers equally. Ohio’s billing system compensates clinicians based on time spent. He cautions that a downside of the proposed model is that patients may avoid seeking virtual care due to uncertain costs.
Researchers find that ChatGPT effectively de-identifies clinical notes and generates synthetic data, maintaining a balance between data utility and privacy in clinical research.
A study finds that AI-analyzed ECG readings can estimate biological age, potentially linking heart health to cognitive function.
Other
A Health Affairs opinion piece urges the FDA to implement post-market surveillance for healthcare AI applications to detect unintended bias that may only emerge after widespread deployment.
A small Deloitte survey of health system executives finds that 40% report moderate to significant returns on AI investments. Over 80% expect AI to have a substantial impact on their organizations in 2025 but see government regulation as necessary.
A state representative in Oregon submits a bill that would ban the use of “nurse” and similar titles to refer to AI agents.
A woman turns to ChatGPT when doctors are stumped by the cause of fevers and rashes in her five-month-old son. ChatGPT correctly diagnosed him with Kawasaki disease.
The local TV station highlights BayCare St. Anthony’s Hospital’s AI pilot project using Aiva Health’s documentation tool. The system enables conversational EHR access, remote room control, facility-wide communication, and alerts.
China-based AI startup DeepSeek introduces a low-cost, open-source AI model that rivals leading US technologies, sending US tech stocks sharply down. DeepSeek, which is free to use and can be run locally on modest hardware, has prompted discussions about US technology export controls, the contrast in cost in AI development and training. DeepSeek was launched in July 2023 by a billionaire hedge fund operator who previously worked with using AI in investing. He stockpiled some lower-power Nvidia chips which were eventually banned from US export to China.
Meanwhile, Chinese technology company Alibaba releases a new version of its own LLM that it says outperforms DeepSeek-V3 and ChatGPT-4o.
OpenAI launches a research preview of Operator, an AI agent that can analyze a webpage and interact with it to perform tasks that involve typing, clicking, and scrolling.
A Health Affairs commentary piece – published as part of the National Academy of Medicine’s Vital Directions for Health and Health Care: Priorities for 2025 initiative — calls for the incoming administration to focus on four health AI-related priorities:
Ensure safe, effective, and trustworthy AI use.
Develop an AI-competent workforce.
Invest in AI research to support the science, practice, and delivery of health and healthcare.
Promote policies and procedures to clarify AI liability and responsibilities.
New Vatican ethical guidelines call for using AI in healthcare to enhance, rather than replace, the relationship between provider and patient.
Stat reports that the White House has indefinitely cancelled meetings of the Health Information Technology Advisory Committee and its workgroups, which operate under HHS’s ASTP/ONC. The committee, which was established under the 21st Century Cures Act, makes recommendations to ASTP on a variety of technology issues, including AI and interoperability.
Mayo Clinic President and CEO Gianrico Farrugia, MD tells the World Economic Forum that his organization has access to 320 algorithms, but their biggest challenge is that today’s technology that can’t support the tools. He says that he would not want to receive specialty care without the clinician using AI, which will require a new architecture to deliver at scale. The panel discussion was titled “Healthcare Innovation at Davos 2025: Cracking the Code of Digital Health.” The panel’s key points involved interoperability, platform-based models, health worker AI training, data liquidity and sharing, and building trust.
Also at the World Economic Forum, the CEO of AI company Anthropic predicts that AI could double human lifespan within five to 10 years. A skeptical response might be:
Attaining that goal would require the US lifespan to increase to 155 years almost immediately.
Even AI-enhanced drug research is unlikely to make a difference of that magnitude given the need to conduct clinical trials, earn FDA approval, and figure out the economics that would be required to make a solution universally available.
Raising the average lifespan would require applying the principles to nearly every American, which has never happened.
His statement is more aspirational than based on biology and medical research.
He also says that AI systems will outperform humans in nearly everything by 2026 or 2027.
His company sells AI.
Spain-based AI medical imaging company Quibim raises $50 million in a Series A funding round. The company hopes to use AI to extract information from medical imaging to identify phenotypes that can predict outcomes, an advanced imaging field known as radiomics.
AI startup Retro Biosciences is reportedly raising a $1 billion funding round. OpenAI CEO Sam Altman provided the company’s initial funding of $180 million and will participate in the latest round. The company’s goal is to add 10 years to the healthy human lifespan by using AI to target and reengineer the cellular drivers of aging.
US Rep.David Schweikert (R-AZ) introduces a bill that would allow FDA-approved AI tools to qualify as a medical practitioner, including the ability to prescribe drugs.
Business
Paragon Health IT will divest its consulting business, change the parent company’s name to Strings, and refocus on AI-driven healthcare workload management system.
AI-powered cancer drug discovery startup Manas AI launches with a $25 million seed funding round. The co-founders are LinkedIn co-founder and venture capitalist Reid Hoffman and oncologist and Pulitzer-winning author Siddhartha Mukherjee, DPhil, MD,
Equality AI CEO Maia Hightower, MD, MPH, MBA announces that the company, which focused on trustworthy AI in healthcare, has closed after four years following the White House’s rollback of a previous executive order that called for trustworthy AI development.
Research
Researchers in Scotland are developing AI software that can scan someone’s retina to detect early signs of dementia, which could be used during routine eye exams. The team explains, “The retina holds a whole wealth of information and is a biological barometer of our brain health … Something very simple like a photograph of a retina can now be harnessed to potentially predict brain change later on in life.” They hope to roll the technology out to opticians in 2026.
Other
A Twitter user finds a DeepSeek-generated dataset of 143,000 made-up encounter conversations on the Hugging Face AI website. Each item contains the disease, language, scenario, patient-provider conversation, common errors in diagnosis, differential diagnoses, related diseases, and a summary, all in multiple languages. It was developed for AI training and can be licensed from MIT. Leading the project is endocrinologist Johnson Thomas, MD.
January 22, 2025Healthcare AI NewsComments Off on Healthcare AI News 1/22/25
News
Shares of AI infrastructure companies rise sharply on the White House’s announcement of Stargate, a private sector project in which major technology companies — including OpenAI, SoftBank, and Oracle — will invest up to $500 billion to build AI data centers. Oracle CTO Larry Ellison, who was present during the announcement, said that the “most charismatic” application of AI would involve EHRs, while OpenAI CEO Sam Altman predicted that “we will see diseases get cured at an unprecedented rate.” The project’s key technology partners were announced as Arm, Microsoft, Nvidia, Oracle, and OpenAI. An OpenAI blog post says that SoftBank will run the financial side of the project and OpenAI will lead operations.
Oracle’s Larry Ellison also predicts at the Stargate announcement that AI will be used to create personalized MRNA cancer vaccines. He says that AI can diagnose cancer from a blood test, sequence the tumor’s genes, and then create a vaccine, all within two days. Shares of MRNA vaccine maker Moderna jumped the news coverage.
The White House revokes a previous executive order that required AI developers who create tools for certain users, including public health, to share product safety results with the federal government.
The UK’s NHS will establish a $250 million fund to help trusts invest an diagnostic and predictive AI tools, with an initial focus on medical image analysis and pathology.
Mayo Clinic and Cerebras Systems will collaborate to develop LLMs that can review the medical records of patients to determine how they will respond to treatments based on their genetic makeup. The models will be trained on the genomic data of 100,000 Mayo patients. Cerebras sells AI hardware, software, and AI model creation services.
Business
Precision medicine technology company Tempus AI launches a personal health concierge app that collects a patient’s medical records and applies AI to generate insights. The company says that its Olivia app can connect directly to the EHRs of 1,000 health systems, sync with health devices, and accept manual uploads. Outputs include a profile summary, summaries of physician notes, medical image sharing. and clinical trials matching.
Google DeepMind’s CEO says that the first drugs that were developed with the help of AI will start clinical trials this year.
Other
Hofstra’s medical school is using AI to make the first cut of the applications it receives, which it says is not only efficient, but also objective and consistent. Several other medical schools are using AI similarly to choose a few dozen students from thousands of applications. George Washington University’s medical school says it spends 6,000 hours of faculty time each year performing manual screening of applications.
The CEO of drug maker Sanofi says the company is using AI to decide if a given drug under development should advance to the next phase. He explains, “We’re not used to having somebody without a career at stake in the room at a senior level.”
Unionized nurses from the Albany, NY VA hospital and other hospitals nationwide rally to voice concerns about hospitals prioritizing AI investments over staff recruitment and retention. The union, which organized the January 16 national protest, has called for stricter AI regulation and greater nurse involvement. A member survey revealed that nurses believe AI often compromises patient safety by undermining their clinical judgment. One nurse highlighted concerns about discharging a post-operative patient with only a tablet for AI-guided updates, saying, “You need eyes on a patient.” The VA responded by emphasizing improved performance metrics and assured that AI would be used safely and responsibly to support, not replace, care teams. National Nurses United, an AFL-CIO affiliate, has issued guidelines for AI use in patient care. Over 100,000 of its members are entering contract negotiations.
January 15, 2025Healthcare AI NewsComments Off on Healthcare AI News 1/15/25
News
OpenAI launches Tasks, a beta feature for ChatGPT that allows users to instruct the LLM to execute tasks or reminders at specified times in a Siri-like way. Example: “Every Friday, provide a list of weekend activities based on my location and the forecasted weather.” The rollout of Tasks is the first step of OpenAI’s AI agent strategy, where ChatGPT can go beyond answering questions to actually performing actions selectively.
A survey by Gwynedd Mercy University of American adults finds that 59% expect AI to improve outcomes, 57% believe it will reduce costs, and 77% believe it can reduce healthcare disparities in the next five years.
The Peterson Health Technology Institute launches an AI task force that will assess AI’s current and potential impact on cost and efficiency, the uptake of AI scribe systems, and the need for metrics and evidence.
UK Prime Minister Keir Starmer says that AI analysis of NHS patient records can speed development of treatments and tests, saying that he doesn’t think a “defensive stance” can deliver possible breakthroughs. He cited diagnostic imaging AI and models that identify patients who are likely to miss appointments.
Business
Microsoft’s venture fund invests in RAAPID, which offers an AI-powered risk adjustment coding platform. Founder and CEO Chetan Parikh, MS previously held executive roles at EzDI and Mediscribes.
Mayo Clinic launches Mayo Clinic Digital Pathology, with assistance from Nvidia and Aignostics. The system was trained on 20 million digital slides and 10 million patient records from Mayo. Mayo apparently intends to commercialize it given its announcement at the J.P. Morgan Health Conference.
The American Hospital Association publishes “Building an Implementing an Artificial Intelligence Action Plan for Health Care.” It lists the top use cases that can generate a quick ROI for health systems: predicting claims denial, optimizing OR time, managing supply chain data, and streamlining discharge planning.
Amazon Web Services and investment firm General Catalyst will collaborate to develop and deploy AI solutions. General Catalyst’s portfolio includes Summa Health acquirer Health Assurance Transformation Company, Commure, and Aidoc.
Hippocratic AI launches an AI agent app store for clinician creators, who have developed 300 agents covering 25 specialties in the past few weeks.
UK-based digital-first home healthcare provider Cera raises $150 million in financing, valuing the company at over $1 billion. The company reduces hospitalization through AI-powered fall prediction, vital signs monitoring, and next best action.
Research
Researchers from Penn’s medical school use AI to analyze the EHR data of long COVID patients and identify those who may need care from specific departments and teams.
Other
The Military Health System recaps the work that it completed in 2024 to prepare for AI: inventorying use cases, enhancing collaboration with other government agencies, creating a digital health strategy, and completing an exercise to identify AI’s vulnerabilities, risk, and biases in a military medicine setting.
California’s attorney general reminds healthcare users and developers that AI must be tested, validated, and audited to ensure that its use is safe, ethical, lawful, free of bias, and transparent with regard to system training and provider use in making decisions.
Two Tenet hospitals delete a job fair announcement when people notice that the depicted employees are sporting extra fingers and unrealistic hair due to what appears to be an “AI fail.”
FDA issues draft guidance for supporting development and marketing of AI-enabled devices throughout the Total Product Life Cycle. It also publishes draft guidance for the use of AI to support development of drugs and biologicals.
Women’s health smart ring maker Movano announces EvieAI, a chatbot that was trained on peer-review medical journals. The company said in its CES announcement that the chatbot’s accuracy is 99%.
Law professors suggest that the Supreme Court’s recent overruling of the Chevron Doctrine — which allows federal agencies, rather than courts, to interpret and implement statutes when authorized by Congress — could impact the FDA’s ability to regulate AI, as its approach often relies on non-binding guidance documents and position papers.
Business
Hartford HealthCare implements Aidoc’s AI platform, which includes 17 FDA-cleared algorithms.
UK-based digital pathology vendor Deciphex raises $32 million in a Series C funding round.
Research
Researchers in China develop a health insurance fraud detection model for insurers and auditors.
A study finds that while AI is effective at making a diagnosis when fed exam-style questions, if fares worse when analyzing real-world conversations. The authors make these recommendations for AI developers:
Train and test systems on conversational, open-ended questions as are found in unstructured doctor-patient conversations.
Assess the model’s ability to ask the right questions.
Design models to work across multiple conversations.
Design models that can capture both textual data and images.
Incorporate non-verbal cues such as facial expressions, voice tone, and body language.
Other
A health news site lists ways that North Carolina providers are using AI:
Analyzing lung nodule scans to predict cancer risk (Atrium Health).
Electronic follow-up with patients who have received a hip or knee replacement (OrthoCarolina).
ED scanning of images to detect serious conditions (Novant Health).
AI-drafted responses to patient portal messages (Atrium Health and WakeMed).
Cognitive impairment detection (Wake Forest University School of Medicine).
Flagging patients who are due a follow-up visit or imaging (Wake Forest Baptist).
Early detection of sepsis (Duke Health, UNC Health).
Suicide risk assessment (Novant Health).
Optimizing OR use by predicting the length of surgical procedures (Duke Health).
A popular TV journalist in Israel who lost his voice due to Lou Gehrig’s disease returns to Channel 12 by using AI that was trained on his voice to narrate his stories.
Health authorities in Sudan are hoping that AI can perform some of the work of doctors who have been killed in the country’s civil war.
A bioethicist warns that healthcare AI systems that allow customization could restrict exposure to important information in catering to the user’s preferences and biases. She presents distinctions between systems that are customized for information discovery and those intended for information delivery.
December 18, 2024Healthcare AI NewsComments Off on Healthcare AI News 12/18/24
News
The House Task Force on Artificial Intelligence publishes its findings and recommendations, which include:
Challenges include data availability and quality, incomplete or inaccurate responses, non-individualized recommendations, lack of decision transparency, data privacy, interoperability with existing systems, liability for AI-contributed errors, biased decision making, and deployment for financial gain rather than patient care.
AI can help improve the percentage of drugs that are eventually approved by FDA by streamlining study design, finding study patients, and answering patient questions.
AI can assist radiologists by performing first-pass image screening, cleaning up images, and applying consistent interpretation criteria across all patients and locations.
Clinical decision support can be tailored to an individual patient’s symptoms and medical history and can compare their data to the medical literature or a “patients like this one” model.
AI can use population health data to predict response to initiatives among population groups.
Physician burnout can be reduced by deploying AI-assisted clinical documentation.
AI can reduce the administrative burden of the prior authorization process.
CMS will need to evaluate Medicare payment for algorithms and other AI tools.
AI has potential in medical management, but could create unnecessary denials.
Innovation will require standardization of EHR-specific data formats to be applicable to broader populations and for AI training.
HIPAA may require updating to meet the challenges of provider AI deployment.
Principal Deputy ASTP/National Coordinator Steven Posnack, MS, MHS says that healthcare AI should not be strictly regulated since it is constantly changing and because risk tolerance varies by setting, such as back office support versus patient care versus research.
A technology reporter shares five weeks of her Apple Health information with ChatGPT and finds that with no prompting, it analyzes trends and makes personalized suggestions for improvement, weaving seemingly unrelated statistics into a fitness journey story.
OpenAI makes the ChatGPT search function available to free users (it was previously available only to paying users) and adds voice-integrated search.
ASTP publishes a downloadable inventory of HHS AI use cases, documenting 66% more than a year ago.
Business
Company insiders say that most of the work of EvenUp — which calculates the value of a personal injury case by analyzing customer medical records and case files — is performed by humans rather than the AI that the company touts to investors. Former employees of the startup, which is valued at $1 billion, say that the AI system misses injuries, hallucinates medical conditions, and is not reliable in analyzing doctor visits, leading managers to tell employees not to use it.
Other
Google Cloud’s healthcare solutions director Aashima Gupta, MS predicts the short-term use of AI to be focused on routine tasks such as appointment scheduling and processing forms, clinical documentation, communication, claims processing, marketing outreach, and use of AI agents for member and provider communications. She expects longer-term AI uses to include enhanced access, enhanced imaging screening and early detection, use of AI as a health concierge, multilingual support, medication reminders, and care navigation advice.
The UK’s National Health Service is using AI to identify patients who are at risk for high usage of emergency services, who are then offered coaching and in-home support.
HIStalk Text Ads Big audience, low price. Seven lines on the most talked about site in the industry. Easy - your ad starts in hours and is seen by thousands of visitors each day. more ...
Regarding Medstar - I have to wonder how many decisions were made to go to Cerner over Epic based on…