TTA’s May Holiday Triple Feature: VA’s $840M ‘need for speed’ in the EHRM budget, Commure’s $70M raise, Innovaccer buys CaduceusHealth, Doximity vs. OpenEvidence, and two Perspectives on AI

May 23, 2026 | by | No Comments

Friday 23 May 2026

Leading up to two holidays–Memorial Day in the US and the UK late May bank holiday–healthcare news remains light. Our roundup includes Congressional hearings on VA’s need for speed–needing 25% more in the EHR budget, an update on the recent VA fraud indictment, two fundings/M&A, and a long Must Read on the ongoing Doximity-OpenEvidence feud worthy of the Corleones and the Barzinis. Rounding it out are two Perspectives: the first on managing the risk of hallucinating AI chatbots and the second on moving AI tools from pilots to full operations.

Please feel free to comment on the articles and pass along this Alert. Let me know if this is worth it to you!

Holiday weekend roundup: VA asks for ‘cyberspeed’ 25% EHR budget bump, update on EHRM fraud indictment; Commure raises $70M; Innovaccer buys Caduceus, lays off staff; Doximity, OpenEvidence slugfest gets hot

Perspectives: AI Hallucinations in Behavioral Health–Why Access Needs Better Infrastructure, Not Better Chatbots

Perspectives: The Next Phase of Healthcare AI Will Depend on Operational Execution

Last Week’s Headlines

A Must-Read potpourri: the ‘math’ of AI data center builds, healthcare AI failures, telehealth in schools, Hippocratic AI’s problems, the loss of empathy.

US Senate Committee on Aging hearings on senior safety 20 May–available online

Plus…

Character.AI sued by Pennsylvania on its chatbots posing as licensed physicians and psychiatrists

Oracle steps back from the AI debt brink with $16.3B financing for MI data center, the Project Jupiter ‘clean energy’ experiment in NM, and a major Federal DOW contract

Is the health tech business neglecting validated deep learning medical AI models versus less proven LLMs and generative AI?

* * *
Advertise on Telehealth and Telecare Aware
Support not only a publication but also a well-informed international community.

Contact Editor Donna for more information.

Help Spread the News

Please tell your colleagues about this free news service and, if you have relevant information to share with the rest of the world, please let me know!

Donna Cusano, Editor In Chief
donna.cusano@telecareaware.com

Telehealth & Telecare Aware – covering news on latest developments in telecare, telehealth and eHealth, worldwide.

Perspectives: AI Hallucinations in Behavioral Health–Why Access Needs Better Infrastructure, Not Better Chatbots

May 21, 2026 | by | No Comments

TTA has an open invitation to industry leaders to contribute to our Perspectives non-promotional opinion and thought leadership area. Today’s topic is about the use of AI in the mental health area–how it is uniquely exposed to risk from LLMs and generative AI–and where best to use them. The author, Hari Prasad, is co-founder and CEO of  Yosi Health, a full-service technology ecosystem that connects patients with their providers through the entire care journey before, during and after the visit, modernizing the entire healthcare patient experience. 

One in three U.S. adults has now used an AI chatbot for health information in the past year, according to the most recent KFF tracking poll. Among adolescents and young adults, one in six has turned to a large language model specifically for mental health advice. That second number should stop every behavioral health leader in their tracks, because when those tools get it wrong, the consequences are not a bad product recommendation. They are clinical.

We are not witnessing patients casually Googling symptoms anymore. They are having extended, emotionally charged conversations with generative AI tools about medications, crisis management, and care decisions, often weeks or months before they ever speak to a licensed provider. And the technology they are confiding in has a well-documented failure mode that the industry still has not adequately addressed: the hallucination.

The Problem Is Not That AI Sounds Wrong. It Is That It Sounds Right.

An AI hallucination occurs when a model generates a response that is fluent, confident, and completely fabricated. In most consumer contexts, that is an inconvenience. In behavioral health, it is a patient safety event waiting to happen.

Consider the real scenarios already playing out: a generative AI tool confidently suggesting an incorrect dosage for a mood stabilizer. A chatbot offering therapeutic advice that directly contradicts evidence-based protocols for trauma. A model providing reassurance to a patient in acute crisis when the clinically appropriate response is immediate escalation. These are not hypothetical edge cases. They are the predictable output of systems designed to be helpful and conversational above all else.

What makes this uniquely dangerous is the packaging. These models are engineered to sound empathetic and authoritative simultaneously. A patient in a vulnerable mental state has no reliable way to distinguish a clinical fact from a statistically plausible guess. And once that trust is broken, the damage extends beyond the misinformation itself. A patient who receives hallucinated guidance during a moment of crisis may lose willingness to engage with the legitimate care system at all.

Why Behavioral Health Is Uniquely Exposed

Two structural realities make behavioral health the highest-stakes frontier for AI hallucination risk: subjectivity and scarcity.

First, subjectivity. Unlike a fracture that shows on an X-ray, mental health concerns are communicated through nuance, context, and subtext. Generative AI is exceptional at mimicking tone. It is incapable of the clinical judgment required to understand the weight behind a patient’s words. There is no lab value to cross-reference, no imaging to confirm. The entire diagnostic framework depends on the kind of human interpretive skill that cannot be approximated by next-token prediction.

Second, scarcity. As of late 2025, 137 million Americans live in a Mental Health Professional Shortage Area, and only about 27% of need is being met in those regions. Appointment wait times range from three weeks to six months depending on location and specialty. For the roughly 29.5 million adults with a mental health condition who received no treatment last year, the barrier was not awareness. It was access. When the pathway to a licensed professional is blocked by a three-month wait, an AI chatbot becomes the path of least resistance, not by preference, but by default.

This is the uncomfortable truth the industry must confront: patients are not turning to AI because they trust it more than their providers. They are turning to AI because the system has not given them a better option.

The Fix Is Infrastructure, Not Disclaimers

The standard industry response to AI hallucination risk has been to add disclaimers and guardrails to the models themselves. That approach treats the symptom while ignoring the disease. The reason patients are having clinical conversations with chatbots is that the operational infrastructure of most practices still makes it easier to talk to a machine at 2 AM than to get a timely appointment with a human being.

The real solution is to shift AI’s role from clinical logic to operational logistics. Practices need to deploy technology that removes the administrative barriers driving patients toward unregulated alternatives in the first place. That means rethinking three operational layers.

Deterministic intake over conversational intake. When an LLM “chats” its way through a patient intake, it introduces the same hallucination risk we are trying to eliminate. The alternative is structured, deterministic intake systems that gather discrete clinical data without improvising questions or advice. Symptoms, history, social determinants of health, all captured through validated frameworks and delivered into the EHR as clean, fact-based data. The clinician gets a head start. The patient gets accuracy.

Precision navigation over generic triage. AI’s greatest operational strength is processing complex variables at scale. That capability should be pointed at routing, not counseling. If a patient’s digital intake surfaces indicators of acute risk, the system should not offer a supportive quote. It should trigger immediate escalation to a crisis line or emergency clinician. The technology’s job is to get the right patient to the right level of care at the right time, not to play therapist in the interim.

Intelligent follow-up over passive waiting. The period between appointments is where behavioral health patients are most isolated and most vulnerable. This is where AI can add genuine value, not by providing care, but by acting as a monitoring layer. Structured check-ins, flagging of concerning patterns in patient-reported outcomes, and automated alerts to clinical teams when intervention thresholds are crossed. The AI serves as a tripwire, not a therapist.

From Advice to Access: The Shift That Matters

The practices getting behavioral health engagement right are not the ones deploying the most sophisticated AI interfaces. They are the ones using technology to collapse the administrative distance between a patient’s first expression of need and their first clinical encounter. When scheduling, intake, and insurance verification happen before the patient walks in, the clinical encounter starts on solid ground. That is what patient engagement actually looks like in 2026.

The hallucination problem is not an argument against AI in healthcare. It is an argument for precision about where AI belongs. Every time we deploy AI in the clinical layer without adequate safeguards, we introduce risk. Every time we deploy it in the operational layer to accelerate access, we reduce risk. The distinction is not subtle, and the stakes are too high to keep blurring it.

Behavioral health already has a trust deficit driven by stigma, scarcity, and systemic friction. The last thing this field needs is a technology layer that erodes trust further by giving patients confident answers that turn out to be wrong. The opportunity in front of us is to use AI to rebuild that trust by making the system itself faster, smarter, and more responsive. Not by replacing the clinician, but by making sure the patient actually gets to one.

Related article from TTA

Character.AI sued by Pennsylvania on its chatbots posing as licensed physicians and psychiatrists

 

Is the health tech business neglecting validated deep learning medical AI models versus less proven LLMs and generative AI?

May 6, 2026 | by

Eric Topol, MD answers his own startling question, contrasting medical imaging with decision support for both clinicians and patients. His recent Substack ‘Ground Truths’ article (link below, free access) will make you think harder about what is being sold as ‘medical AI’ and what has actually been validated through multiple studies. 

Imaging AI is the Undiscovered–but Mapped Out–Country. Deep Learning (DL)-based AI models developed using medical imaging have substantial validation over more than a decade, and they are accelerating. There have been multiple validated studies using information from retinal scans as predictors of future medical conditions such as Parkinson’s, heart disease, stroke, and Alzheimer’s Disease. The retina is apparently a diagnostic gateway to nearly every organ; many studies have focused on it as scans are fairly routine. Other AI-assisted models have used deep learning to detect multiple health conditions: thymus, cardiovascular conditions, through mammography, colonoscopy, and importantly, detecting pancreatic and other cancers from computed tomography (CT) images done for other reasons. “Opportunistic AI” alone is being used in detection for a long list of health conditions. Dr. Topol’s point is that none of these new diagnostic methods have made it into standard practice, despite being used in other countries like China (PANDA) and with at least four companies developing uses for retinal AI to detect specific diseases.

Medical LLMs and Generative AI, on the other hand, are building what may be Castles In The Air.  Seemingly everyone is developing, funding, and selling a LLM-based chatbot, LLM-aided diagnosis, management, patient triage, and direct patient use. Unfortunately, they’re being sold without real, continuous evidence through rigorous studies over time. What studies there are, are generally simulations, small-scale studies, or individual case studies which need further real-world validation. The clinical trials, the infrastructure, and the monitoring for safety, effectiveness, and cost are simply not there yet, and it’s past time. (Raj Manrai quoted in Science). In addition, generative AI keeps changing making studies harder to track results over time. Dr. Topol’s conclusion: “In summary, there is very little evidence for LLMs benefiting patients or doctors for health outcomes.”

That is not to say, as Dr. Topol does, that AI won’t grow in usefulness in areas such as medical research and chart summaries, discharge instructions, translations, administrative work such as documentation of billing codes, clinical workflow, and insurance authorization. AI has already worked its way into RCM where no respectable company does not have an AI-enabled tool. The American Medical Association (AMA) study he cites indicates both current use and growing acceptance by physicians. (To this Editor, it resembles the telehealth usage graphs of a decade ago, and she expects the same progress.) 

He calls it a paradox between imaging AI and LLMs. This Editor calls it a shame that healthcare technology and investment keep chasing what’s easy, ‘sexy’, and can generate fast revenue/ROI. Not what is more difficult but proven, and that can have a potential huge impact on health outcomes.

Dr. Topol’s closing is fitting:

Let’s fix this paradox of medical AI implementation. It’s a two-fold and major undertaking. Amping up the use of medical AI where it’s proven and performing the clinical trials required to justify wide-scale adoption where pivotal evidence is lacking.

 
The Paradox of Medical AI Implementation

AI failing–at present–to lower costs, grow revenue, improve efficiencies. Yet it’s full speed ahead: Deloitte, PwC surveys

January 27, 2026 | by

When the business process outsourcing (BPO) leaders pour lukewarm water over AI, one hears the air leaking from a bubble. BPOs have been a key part of the hype around AI as a business solution. The McKinseys, Genpacts, Deloittes, and PwCs for years have touted AI and as a result, made large consultancy fees. AI now proliferates for every business problem. Whether it’s generative, (still kicking around) machine learning, NLP, LLMs, agentic, robotic process, and now sovereign AI (domestically developed and powered)–it’s been positioned as the solution for simplifying processes and reducing administrative burden. Of course, a fair chunk of this involves getting rid of those pesky human factors in overseeing whether these new systems and software actually work, or reducing them to the lowest cost possible, to pay for all the AI spend.

Unfortunately for the BPOs, their customers are telling them that AI Is Not Quite All That. In fact, for the money they have spent, it hasn’t performed. Yet. But they remain optimistic, a neat bit of cognitive dissonance or perhaps justification.

The Deloitte global survey of 3,235 business and IT leaders confirms the gloomy news to date–yet it’s full speed ahead. Only 20% have experienced revenue growth as a result of AI. Transformation is coming along slowly; 25% of those surveyed believe that AI is transforming their organizations, which corresponds to 84% not redesigning jobs or work around AI capabilities. In this area, there’s a lot of resistance. While 55% of workers are reportedly open to AI technology, only 13% of workers are highly enthusiastic about AI, 21 percent would prefer to avoid it, and 4% actively distrust it. There’s also a lot of pilot-itis. Only 25% report shifting 40% or more of their AI experiments into live use, though optimistically they project that will increase to 54% in three to six months.

Yet they’re justifying AI. Totally. 66% reported that it improves productivity and efficiency, which contradicts the low revenue growth. 58% of companies are already using it to some extent, with adoption to hit 80% within two years. 74% of companies plan to deploy agentic AI within two years, even though only 23% are using it now and 21% have a model for governance of autonomous agents–a high risk level. 42% believe their strategy is ‘highly prepared’ for AI adoption. Another part of AI adoption has surfaced–sovereign AI, to reduce dependency on foreign sourcing, vendors, and infrastructure. 83% reported that this was at least moderately important to them. The Register 21 Jan, Deloitte’s State of AI in the Enterprise report (PDF, January 2026) 

PwC’s larger survey of 4,454 business leaders in their 29th Annual Global CEO Survey contains gloomier and more detailed feedback for AI advocates. “Most CEOs say their companies aren’t yet seeing a financial return from investments in AI.” Only 30% reported increased revenue and 26% saw lowered costs. More than half–56%–did not see either lower costs and higher revenue. 22% reported an increase in costs due to AI.

Another finding is that isolated AI projects aren’t delivering value. Companies lack a clear strategy in building AI foundations such as clearly defined road maps and sufficient levels of investment​​.

A relatively small proportion of their surveyed CEOs say they’re applying AI to a large or very large extent to areas such as demand generation (22%); support services (20%); the company’s products, services, and experiences (19%); direction setting (15%); or demand fulfilment (13%). In a previous survey, only a tiny minority of workers–14%–are using generative AI daily. PwC’s report goes on to identify many other factors reshaping global business and influencing growth, in context confirming that depending on AI as a quick fix is not paying off.  The Register 20 January, PwC 29th Annual Global CEO Survey (January 2026).

Reality tends to bite. Many of last year’s corporate layoffs were attributed to heavy AI investments that weren’t paying off, but books needed to balance by year’s end and it was taken out of human capital. Layoffs are projected to continue across all industries in 2026. Books balance another way, though. The AI bubble is deflating from Inflated Expectations into the early stages of the Trough of Disillusionment. How long it will take to move to the Slope of Enlightenment is anyone’s guess–two years, five, a decade? The useful tool of the Gartner hype cycle strikes again–as it did with telehealth and health tech. Separately, we’ll be looking at OpenAI’s ChatGPT for Healthcare and Anthropic’s Claude for Healthcare.

News roundup: Walgreens’ $350M opioid settlement, only 30% of healthcare AI pilots reach production, Medicare RPM usage up 10-fold despite benefit limitations

April 24, 2025 | by

Walgreens continues to clean up on Aisle 9 before it goes private. Walgreens settled the Federal allegations around illegally filling invalid prescriptions for opioids and seeking payment from Federal programs for $300 million. There’s an additional $50 million tagged onto it if the company is sold, merged, or transferred prior to fiscal year 2032. Since Walgreens has ‘done the deal’ with Sycamore Partners, the settlement amount will be the full $350 million. According to the Department of Justice’s press release, the settlement was based on Walgreens’ ability to pay. There was no statement on when the $350 million will be due.

This settles the complaint filed on 16 January (amended 18 April) in the US District Court for the Northern District of Illinois by the Department of Justice (DOJ), the Drug Enforcement Administration (DEA), and the Department of Health and Human Services Office of Inspector General (HHS-OIG). In the suit, Walgreens faced civil penalties of up to $80,850 for each unlawful prescription filled in violation of the Controlled Substances Act (CSA), plus treble damages and applicable penalties for each prescription paid by Federal programs in violation of the False Claims Act (FCA) for over 10 years–approximately August 2012 through March 1, 2023 The red flags included prescriptions for the ‘trinity’ of an opioid, a benzodiazepine and a muscle relaxant. If Walgreens had been found guilty, the penalty could have been billions. 

Given the numbers that in January presented a large impediment to a sale, settling rather than fighting makes sense. The projected Sycamore Partners closing is only two quarters away (Q4, TTA 11 Mar). The 35-day ‘go shop’ period has closed with no other offers. The DOJ has moved to dismiss its complaint in Illinois, while Walgreens will also move to dismiss a related declaratory judgment action filed in the District Court for the Eastern District of Texas. In addition to Illinois, the District of Maryland, the Eastern District of New York, the Middle District of Florida, and the Eastern District of Virginia participated in the complaint.

In addition to the settlement, Walgreens’ pharmacy operations are now under Federal scrutiny, based on multiple agreements with DEA and HHS-OIG attached to the settlement, addressing what they and the DOJ saw as compliance violations in dispensing controlled substances. From the release:

  • Walgreens and DEA entered into a memorandum of agreement that requires the company to implement and maintain certain compliance measures for the next seven years. 
  • Walgreens must maintain policies and procedures requiring pharmacists to confirm the validity of controlled substance prescriptions prior to dispensing controlled substances, provide annual training to pharmacy employees regarding their legal obligations relating to controlled substances, verify that pharmacy staffing is sufficient to enable pharmacy employees to comply with those legal obligations, and maintain a system for blocking prescriptions from prescribers whom Walgreens becomes aware are writing illegitimate controlled substance prescriptions.
  • Walgreens has also entered into a five-year Corporate Integrity Agreement with HHS-OIG, which further requires Walgreens to establish and maintain an extensive compliance and training program

Crain’s Chicago Business, Healthcare Finance, Settlement Agreement

Healthcare AI continues to be more show than go. A report by Bessemer Venture Partners surveying payers, pharma, and providers states that 95% of respondents said GenAI will be transformative, with 85% of provider and 83% of payer leaders expecting it to reshape clinical decision-making within three to five years. Yet only 30% of AI pilots — what the report calls “internally and externally developed GenAI proof of concept (POC) projects” make it to production. 

Generative AI applications are being developed by the organizations’ IT teams, building their own tools by partnering with horizontal AI labs (i.e., Anthropic), Big Tech companies, or going to current and new vendors. The impediments they face are cybersecurity, data readiness, integration costs, and limited in-house expertise. Procurement is shifting toward co-development; 64% of execs are open to co-developing with early-stage partners. A contradiction is that only 32% of executives believe GenAI solutions from startups are superior to those from large tech incumbents. yet 48% prefer working with startups over established players.

Yet with these “aspirations and expectations” in the 80-90th percentile, only half of the surveyed organizations surveyed have a clear AI strategy with 57% having an AI governance committee, with payers in the lead. Those in it are putting real money behind it and seeing some meaningful ROI (54%), however. 

Among the three surveyed segments, Bessemer identified 59 jobs-to-be-done. Yet 45% of these jobs are still in the ideation or POC phase, with far fewer actually in production. These jobs clustered as follows: 22 for payers (claims, network, member, pricing), 19 for pharma (preclinical, clinical, marketing, sales), and 18 for providers (care delivery, revenue cycle management). The survey was performed by Bessemer with Amazon Web Services and Bain & Company, across 400 leaders in the three segments. 

Remote patient monitoring (RPM) Medicare usage growing steadily, despite limitations on clinical effectiveness. This new report from the Peterson Center on Healthcare tracks how RPM usage has grown among Medicare (traditional, Medicare Advantage) and Medicaid beneficiaries since 2019, when CMS enabled Medicare codes for reimbursement. For Medicare, despite only 1% of beneficiaries using RPM, it grew exponentially–10-fold for traditional Medicare between 2019-2023 and 14-fold for MA between 2019 and 2022.

  • Top chronic conditions are hypertension (57%) with diabetes far behind at 13%. Also growing but much smaller is remote therapeutic monitoring (RTM), dominated by musculoskeletal (MSK) disorders.
  • Traditional Medicare spent $194.5 million on RPM and $10.4 million on RTM in 2023.
  • Clinical effectiveness tends to be short-term. In hypertension, RPM is most valuable within the first six months, when blood pressure medications are actively monitored. For diabetes, the prime target is likely patients with the highest starting HbA1c levels and those
    who are at critical transition points in their care plan, such as starting insulin. For RTM, the most effective gains occur in 2-4 months.
  • Yet utilization is increasing. The duration of continuous RPM in traditional Medicare in 2023 was 5.2 months, with 22% over nine months. For hypertension, the average is 6.6 months. For MSK RTM, the average was below effectiveness-only 1.7 months.

The Peterson Center’s policy conclusions advocate a reset:

  • Coverage and reimbursement need to be better aligned to actual clinical value
  • Adoption of high-impact remote monitoring services and minimizing or eliminating the use of poorly performing digital applications
  • Improved data collection for remote monitoring services for evidence-based coverage and reimbursement decisions

Evolving Remote Monitoring

Are patients and physicians ready for generative AI? How will it be most acceptable?

August 22, 2024 | by

As the flood of news will ebb over the next two to three weeks or so (and your Editor takes annual leave), some reading for your pondering.

Gimlet EyeBain and Company, the well-known (and well-feared when they come to your company) management consultants (and investors), recently published results from March’s “US Frontline of Consumer Healthcare Survey“ taken from 500 (undefined) respondents. Unfortunately, with the article, there is no opportunity to download the full survey or review the methodology. From their featured toplines, though, we can savvy that there are still many patient and clinical doubts around generative AI. 

Bain’s article generally spins positives from the results. For instance, their lead is that patients are ‘more comfortable with generative artificial intelligence (AI) analyzing their radiology scan and making a diagnosis than answering the phone at their doctor’s office.’  Their second lead is that clinicians generally have ‘a positive view, recognizing generative AI’s potential to alleviate administrative burdens and reduce clinician workload”.

Your Editor takes a more Gimlety view. First, let’s review consumer comfort with generative AI in five functions:

  • No ‘comfortable’ response is above 37%. Adding in ‘neutral’ at 18%, the only area breaking 50% is “taking notes during appointments to send follow-ups” at 55%. Consumers are not comfortable with generative AI “providing medical advice, treatment plans, and prescriptions” at 21% (11% and 10% respectively).
  • ‘Not comfortable’ is lowest for notes for follow up at 45%. The highest is for medical advice at a eye-blinking 79%. In between, the range is 52% to 68%–indicating strong consumer resistance across the board.
  • As to analyzing radiology scans, there is a lot more comfort with that report going to the doctor for review than AI making a diagnosis. For reports to the doctor, the ‘comfortable’ response is 31% but falls to 21% for diagnosis. ‘Not comfortable’ notches from 52% to 62%
  • For answering calls for providers or insurers, there is definite unacceptance at 68% with only 33% ‘neutral’ or ‘comfortable’.

Provider perspectives are split between physicians and (undefined) administrators. (No neutral in these responses)

  • They both believe that administrative burden and workload will be reduced, with admins far more hopeful than physicians at 43% and 35% by area respectively.
  • What neither group likes is the potential to undermine the patient-provider experience: 19% negative for the physicians and the admins not far behind at -17%.

The takeaway: Generative AI is following the telehealth curve in initial low acceptance. The responses and proportions resemble the early days of telehealth and to a lesser degree, remote patient monitoring. Information didn’t fit into workflows, wasn’t seen as critical, and increased administrative burdens. With acceptance languishing for over a decade, it took a black swan event called the Covid Pandemic to overcome–and in the end it reshaped telehealth, as Teladoc and Amwell have learned.

But in present time, if your Editor as a consultant were presenting this to an AI developer, a physician group, or investors, she would advise preparing for a long, hard road. A road which needs validation, real revenue models, demonstrated accuracy, acceptances, and proven value in solving crucial problems and cost reductions. Natural-language processing (NLP) is being touted as a tool for most of this–but it is only part of the picture. 

Views at variance: Healthcare IT News (NLP interest), MedCityNews (LLMs losing to humans on medical knowledge), MedCity News (on proving value)

Mid-week roundup: UK startup Anima gains $12M, Hippocratic AI $53M, Assort Health $3.5M; Abridge partners with NVIDIA; VillageMD sells 11 Rhode Island clinics; $60 for that medical record on the dark web

March 21, 2024 | by

It may be a little chilly out, but it feels like Springtime For Early Round Funding and Big Partnerships.

Anima, a London-based startup fresh out of Y Combinator, now has a $12 million Series A raise. It was led by Molten Ventures, with participation from existing investors Hummingbird Ventures, Amino Collective and Y Combinator. Its platform combines online consultation with productivity tools for integrated care enablement in one dashboard for primary care. Their founders position it as a single source for patient truth across care settings, avoiding missed diagnoses. As of today, Anima is deployed in over 200 NHS clinics in England caring for a combined 2 million patients and a monthly request volume of over 400,000 requests. They also claim to halve the time the time practices spend on coding, processing, and filing documents and resolve 85% of patient inquiries within a day. Shun Pang, co-founder and CEO of Anima, who trained as a doctor at Cambridge University, told TechCrunch. “The entire clinic collaborates in a real-time multiplayer dashboard, like Figma, and can ping cases to each other, and chat with a Slack-like UX.” he said. He also added that Anima’s processing system can “autonomously ingest any document, like handwritten, diagrams, imaging, and output a summary, with structured fields.” Anima has not entered the US market yet. Anima blog/release, Tech.EU

Hippocratic AI raised a jumbo $53 million Series A for what they term the first safety-focused Large Language Model (LLM) for healthcare. AI of course is the hottest funding area in healthcare. With two previous rounds raised in mid-2023, their total funding is $118 million (Crunchbase), creating a valuation estimated at $500 million. Investors were co-led by Premji Invest and General Catalyst with participation from SV Angel and Memorial Hermann Health System as well as existing investors Andreessen Horowitz (a16z) Bio + Health, Cincinnati Children’s, WellSpan Health, and Universal Health Services (UHS). Their product is a novel staffing marketplace where health systems, payors, and others can “hire” auto-pilot generative AI-powered agents to conduct low-risk, non-diagnostic, patient-facing services to help solve the massive healthcare staffing crisis. This is now being released for phase three safety testing with 5,000 licensed nurses, 500 licensed physicians, and the company’s health system partners. Release

San Francisco-based startup Assort Health now has a seed round of $3.5 million to advance its generative AI approach to healthcare call centers. Its goal is to eliminate front desk stress and call center/service holds. Their system in development uses AI and NLP (natural language processing) to understand a caller’s intent, then to integrates with the medical providers’ EHR, including Epic, to resolve patient inquiries without human intervention. Funding was led by Quiet Capital (!) joined by Four Acres, Tau Ventures, and a number of angel investors from tech companies. Release

Another generative AI company with a substantial Series C under its belt, Abridge, is partnering with super-hot NVIDIA.  The partnership also comes with undisclosed funding from NVIDIA’s VC arm, NVentures, to add to last month’s $150 million raise. Abridge is developing conversational AI technology using LLM and speech recognition to ease the burden of taking notes during the doctor’s appointment, with fluency in 14 languages across 55 medical specialties. Abridge’s technology is designed to capture clinician-patient conversations and structure the scribing. NVIDIA’s partnership will give Abridge access to NVIDIA’s computing resources, foundation models, and expertise in efficiently deploying AI systems at scale. Release

Another episode in the continuing Walgreens Restructuring Saga has VillageMD selling 11 practices to Arches Medical Partners. The practices are located in the Providence metro area of Rhode Island and consist of three urgent cares and eight offices with a total of 50 physicians and 75,000 patients. It is unusual because it is the first time that VillageMD sold their practices instead of closing the offices, which they are doing with 85 to 90 offices. Transaction cost was not disclosed but closed on 2 March. Arches is based in Cambridge, Massachusetts. They acquired these practices but also deploy software from its wholly-owned technology subsidiary, New Era Medical Operations (NEMO), to enable IPAs to negotiate and manage global risk contracts. Arches release, Becker’s, Crain’s Chicago Business

Wondering why ransomwareistes, their affiliates, and hackers in general are attracted to healthcare? It’s the value of a medical record. Going rates on the ‘dark web’ are now topping $60, according to CNBC’s source, a cybersecurity researcher Jeremiah Fowler. By comparison, Social Security number are a bargain $15 and a credit card number but $3. It’s also easier to hack than ever due to affiliate relationships termed ransomware-as-a-service or RaaS. The ransomware is supplied, the affiliate hackers do the work, and they share in the rewards–most of the time (see ‘notchy’ being scammed by BlackCat/ALPHV on the Change Healthcare cyberattack TTA 5 Mar). But this doubles or triples the potential for company extortion, with multiple ‘actors’ attacking a company, extorting a ransom, and then keeping healthcare data and selling it through their channels.

The article concludes that healthcare execs need to get very, very serious about protecting their data. Yet this year has marked healthcare downsizing IT departments in order to save money. This is as security software has proliferated–but has to be purchased and managed. Another distressing fact: this Editor only last week attended a major NYC conference on cybersecurity. Healthcare was mentioned only in passing as a market. Worse, till this Editor questioned a speaker from the floor, was the massive Change Healthcare attack even mentioned–and unfortunately she knew more about it than the speaker!

Weekend reading: AI cybersecurity tools no panacea, reality v. illusion in healthcare AI, RPM in transitioning to hospital-at-home, Korean study on older adult health tech usage

February 23, 2024 | by

A potpourri of current articles. Hope you don’t feel like Pepper the Robot after you read them!

AI won’t boost cybersecurity, that’s cutting corners (Cybernews)

AI tools that make cybersecurity more effective and faster in response are increasingly available. They are estimated in a Techopedia article rounding up multiple studies to be a global market of over $133 billion by 2030. IBM claims that organizations with AI cybersecurity took 100 days less to identify and contain data breaches. Yet AI can also leave organizations more vulnerable to cyberattack. Hackers and ransomwareistes have been using AI for years in phishing and vishing (phone-based social engineering) attacks–now using OpenAI. What’s vulnerable? Large language models (LLMs) used in generative AI (AI with the ability to create content) can be corrupted and fed false information [TTA 7 Feb] or create deepfake images–Google Gemini is the latest example (not in article). FTA: “We need human critical thinking to use AI to solve and prevent problems. We’re adopting AI far faster than we have the ability to understand how to adopt it properly.” Another approach is to think like a cybercriminal and use AI to better understand how criminals can break into your systems.

What is real and what is illusion with healthcare AI? (03:16 video, Healthcare IT News)

This is a preview of a HIMSS24 talk on 11 March by Dr. Jonathan Chen, assistant professor at the Stanford Center for Biomedical Informatics Research. Patient care and outcomes are dependent on discerning what is real and what is not, especially in the use of chatbots in patient notes. Generative AI can be very convincing even if it’s not accurate, and that is not what is wanted in patient care. We are at the Gartner Peak of Inflated Expectations when it comes to AI–and we’ve been there before.

RPM strategies for moving from discharge to hospital-at-home care (Healthcare IT News) 

How can the home be better treated as a fundamental care setting? Understanding this is key to transitioning patients from in-hospital acute care to hospital-at-home, which is in reality not being discharged and requires managing a significant number of complex layers. Interview with Cindy Gaines, RN, chief clinical transformation officer at Lumeon, a clinical automation company.

Tailor fit digital health tech to the elderly’s needs: study (Mobihealthnews)

This summarizes a South Korean study that compared the usage of digital devices, such as smartphone apps, health apps, and wearables, among healthy and pre-frail/frail Koreans aged 65+. Smartphone use is nearly universal in South Korea, but wearables are only lightly used. Frailer respondents used social media more than healthy ones and used more healthcare apps on their phones. From the study: “There was a notable difference in the services used by pre-frail and frail respondents compared to healthy respondents. Therefore, when developing digital devices for pre-frail and frail older adults, it is crucial to incorporate customized services that meet their unique needs, particularly those services that they frequently use.”

505 participants completed the survey, with 153 (30.3%) identified as pre-frail or frail and 352 (69.7%) as healthy. Full study in the Journal of Korean Medical Science 27 November 2023

Two studies: telehealth’s ‘generation gap’ and $22B target for healthcare generative AI–by 2032

October 11, 2023 | by

J.D. Power notices that older users aren’t all that comfortable with telehealth. On a 1,000 point scale, pre-boomers (!) and Boomers, score a 671 while Gen Y and Gen Z score 714 for an average of 698. Those surveyed liked telehealth for convenience (28%) and receiving care quickly (17%). 

Issues for the older group are trust, digital channels, and appointment scheduling. The latter two are, in this Editor’s view, interface related, with many telehealth providers neglecting mobile and tablet-friendly platforms, making typefaces large enough, and backgrounds contrasty enough.

CVS leads in satisfaction, surprisingly, in direct to consumer telehealth providers (744), with MDLIVE (Evernorth/Cigna) coming in at 741 and Amwell 739. CVS’ telehealth is provided by Amwell. Where telehealth is provided by a health plan, the numbers were extremely close. UnitedHealthcare scored the best (702), with Kaiser Foundation Health Plan immediately behind at 701 and Humana at 695. UnitedHealthcare uses Included Health’s Doctor on Demand, Teladoc for Kaiser. The June-July survey included over 5,400 telehealth users within the last 12 months. Healthcare Finance, J.D. Power study page (subscription required for report).

Generative AI for healthcare projected to be a $22 billion business by 2032 from $1 billion today. Generative AI is defined as AI that produces text, images, and other media, based on text, audio, and image data supplied to it. The PYMNTS and AI-ID “Generative AI Tracker” points to current uses in complex drug discovery acceleration and medical researcher capabilities. To realize its potential in other healthcare areas, tech companies must team up with payers, providers, and others to train large language models on healthcare-specific data and establish robust benchmarks. The PYMNTS study is available for download here. Healthcare IT News

AI news: GE HealthCare’s 510(k) for Precision DL (+ GE stake sale), Samsung adopts care.ai for in-facility patient monitoring, Mayo Clinic-Google Cloud generative AI, Wolters Kluwer buys Invistics for drug diversion detection

June 8, 2023 | by

GEHC receives FDA clearance for Precision DL (deep learning) image processing software. It improves image quality on GEHC’s PET/CT, Omni Legend, which enables faster scanning time and improved small lesion detection. Deep learning as part of AI is a subset of machine learning (ML), which uses a neural network with three or more layers that simulates the human brain in processing and ‘learning’ from large amounts of data and drawing judgments from it. (See our recent Perspectives for a more nuanced explanation.)  According to GEHC’s presentation brochure on Precision DL, it is trained with thousands of PET images made using multiple reconstruction methods. Mobihealthnews

GEHC was spun off from parent General Electric (GE) in January. GE retained about 19% of its stock at the time with the remaining being distributed to GE shareholders, but on Monday announced that it would sell 25 million shares, or about $2 billion in value, in a debt-for-equity exchange. The debt is held by affiliates of Morgan Stanley which would then receive the stock, which has done well. This would reduce GE’s stake in the spinoff considerably.  Reuters, Yahoo Finance

Samsung partnering with care.ai for facility ‘smart care’. Orlando-based care.ai’s Smart Care Facility Platform monitors for conditions and learns from patient behaviors. It can be used for infection prevention and control, patient and protocol monitoring, workforce optimization, and virtual care. The AI-powered platform will be integrated into Samsung displays for clinician use, including virtual care. The system will be utilized in hospitals, nursing homes, and care facilities. care.ai release, 

Mayo Clinic is also jumping on the AI bandwagon with Google Cloud. Google Cloud’s Enterprise Search in Generative AI App Builder (Gen App Builder) will be used to make it easier for clinicians and researchers to find the information they need and improve the efficiency of clinical workflows to ultimately improve patient outcomes. According to the release, Enterprise Search in Gen App Builder unifies data across dispersed documents, databases, and intranets, making it easier to search, analyze, and identify the most relevant results. Mayo is an early adopter of the system. Google Cloud release  

Wolters Kluwer Health has acquired Atlanta-based Invistics. Invistics’ Flowlytics tracks medication in hospitals and other patient care settings through ML-based systems. The most critical ‘hot button’ use is for detecting drug diversion, which is when a healthcare worker illegally obtains or uses prescription drugs intended for a patient. This is done by reconciling drug transactions from purchase to patient, with their system being used to rapidly and accurately identify patterns of behavior consistent with drug diversion. More routine usage is for automating controlled substance compliance. This will fit in with Wolter Kluwer’s existing products Simplifi+ and Sentri7 in their Clinical Surveillance, Compliance & Data Solutions unit. Information on transaction cost and management transitions were not disclosed. Release

Hat tip to HIStalk’s new AI News feature 7 June for both Mayo-Google Cloud and WK-Invistics.