This is a preview of the October 15 edition of Access Health—Tap here to get this newsletter delivered straight to your inbox on Thursdays.
Good afternoon. Ahead of Newsweek’s AI Health Summit a few weeks ago, we hosted an executive roundtable to discuss how AI is changing the way hospitals operate. One question from Dr. Ashley Beecy has stuck with me.
Beecy—a cardiologist and clinical informaticist who currently serves as chief AI officer at Sutter Health, based in Sacramento, California—was describing the increasingly expensive vendor pitches crossing her desk.
“Every time I hear a vendor presentation, I'm thinking, ‘That's half a million dollars. That's nine hundred thousand dollars,’” Beecy said. “I'm like, I could build that. But do I want to build that?”
She raised a dilemma I suspect more hospital executives will face as AI development becomes faster and more accessible. Technology that once required a substantial investment in outside vendors might now be within reach of an organization's own developers.
But the actual building is only one part of the build-buy equation.
“Who are you calling at 2 a.m. if it breaks?” Beecy asked. “Do you shift your whole IT model to then have a support infrastructure so you can do more development?”
Her comments sparked a broader discussion about the maintenance requirements for AI tools after deployment. Dr. Jennifer Litton, chief clinical research officer at MD Anderson Cancer Center, summed up the challenge: "Easy to build, but hard to sustain." Others questioned whether health systems were introducing new complexity by layering AI onto already fragmented workflows.
“That’s where we went wrong with the electronic health records at the beginning as well,” added Dr. Nabil Chehade, chief clinical transformation, innovation and strategy officer at The MetroHealth System, based in Cleveland, Ohio.
I heard a similar argument this week from Michael Meucci, president and CEO of health care data company Arcadia, which recently launched Ascent, an AI-enabled platform designed to help health systems and health plans translate data into improvements in clinical and financial performance.
"Even though it has become easier to build software than it has ever been, it doesn't mean that you should," Meucci told me.
He encouraged health system leaders to consider partnerships with established technology companies and invest in the organizational changes required to make new tools work.
"We spent more money on change management than we did on EHR tech in 2009 to 2013," he said, and now, “we’re in the same wave."
Read on for more from my conversation with Meucci in this week's Pulse Check.
And as always, I’d love to hear from you: How is your health system weighing the costs of building AI tools against the responsibility of maintaining them? Email me at a.kayser@newsweek.com and let me know.
In Other News
Major health care headlines from the week
- Chicago-based Rush University System for Health and Endeavor Health announced plans to merge, creating an integrated academic health system with 12 hospitals, more than 400 care sites and over 46,000 employees and clinicians. The combined organization would serve more than 2 million patients annually across northern Illinois and Northwest Indiana.
- Per the proposed agreement, Endeavor Health will fund an additional $1 billion investment in community health initiatives, academic programs, research and care innovation. A significant portion will support partnerships addressing social determinants of health and improving access to care in underserved communities.
- The combined organization will operate under the Rush name, with Endeavor Health President and CEO J.P. Gallagher taking the helm. Rush CEO Dr. Omar Lateef will become CEO emeritus and strategic adviser following the merger, which is expected to close in early 2027, pending regulatory approval.
- The Centers for Medicare & Medicaid Services released its first report on how states are allocating funding through the $50 billion Rural Health Transformation Program, which runs from fiscal year 2026 through 2030. The report details funds obligated through July 31, 2026, toward first-year initiatives, with spending expected by September 2027.
- States are directing funding toward a wide range of priorities, including workforce recruitment, hospital transformation, chronic disease prevention and technology upgrades. For example, Iowa has allocated millions toward advanced diagnostic equipment and robotic surgery, while Georgia is investing in value-based care readiness, workforce development and rural health technology.
- The pace of funding commitments varies considerably across states. As of July 31, Georgia and Hawaii had obligated their full first-year awards, while Florida had committed just $1.8 million of its nearly $210 million allocation and Idaho had obligated about $430,000 of its nearly $186 million award. CMS cautioned that the figures do not represent a complete picture of state progress and plans to update the data quarterly.
- View the full report here.
- Function Health and Google Health announced a partnership Monday to integrate lab testing results with wearable health data and AI-powered wellness guidance. Beginning later this fall, U.S. users will be able to share their Function lab results directly with the Google Health app, bringing clinical biomarkers and daily activity tracking into one platform.
- The integration will combine Function's more than 160 lab tests with data from Google wearables, including sleep patterns, resting heart rate and heart rate variability. The companies say this will help users better understand how daily habits, such as exercise and nutrition, relate to changes in cardiovascular, metabolic, hormonal and other health markers over time.
- Google's AI-powered Health Coach will use lab results and personal health records to deliver personalized wellness recommendations based on users' biology, routines and goals. The companies say the technology could help users identify health changes earlier and have more informed conversations with clinicians, although Google cautions that its AI guidance is not intended for diagnosis, treatment or disease monitoring.
- CMS announced Wednesday that it identified 189,000 people receiving subsidized Affordable Care Act coverage in 2026 who were also enrolled in Medicare, Medicaid or the Children's Health Insurance Program, or were listed as deceased. The agency said it has moved to terminate subsidies for those enrollees, projecting approximately $1.5 billion in taxpayer savings over the next year.
- The announcement follows a similar effort in 2025, when CMS ended ACA subsidies for approximately 551,000 people who were also enrolled in Medicaid. CMS Administrator Dr. Mehmet Oz said the agency has saved nearly $6 billion through its broader efforts to eliminate duplicate enrollment, although the release did not detail how that figure was calculated.
- Individuals flagged for duplicate coverage are given 30 days to dispute the determination before their subsidies are terminated. They can also appeal afterward or demonstrate that their Medicaid or CHIP coverage is ending to regain eligibility for ACA subsidies.
- In a separate move Wednesday, Health and Human Services Secretary Robert F. Kennedy Jr. formally delegated new enforcement authority to the CMS administrator, allowing the agency to pursue certain discretionary exclusions from federal health care programs. The change is intended to consolidate investigative and enforcement responsibilities within CMS, but preserves the HHS inspector general's existing authority and does not grant CMS the power to conduct hearings or issue regulations.
Pulse Check
Executive perspectives on key industry issues
...
Last week, health care data company Arcadia launched Ascent, a new AI-enabled platform designed to help health systems and health plans move beyond analyzing performance to taking action. Its five solution suites address everything from identifying gaps in patient care and coordinating outreach to managing financial risk, evaluating provider networks and delivering predictive insights.
The launch builds on Arcadia's 2024 acquisition of CareJourney, which expanded its access to national claims data and provider performance benchmarks, and a 2025 deal that made private equity firm Nordic Capital its majority investor. The investment was intended to accelerate Arcadia's expansion and development of tools that help health care organizations translate data into improvements in clinical and financial performance. Ascent represents the latest step in that strategy.
On Tuesday, I spoke with Arcadia President and CEO Michael Meucci about the company's evolving strategy and what it might mean for the broader industry.
Meucci believes advances in AI could finally help health care deliver on decades of technological promises. But with operating margins under pressure, he warned that leaders will need to balance immediate financial challenges with longer-term investments in data infrastructure and organizational change.
Find a portion of our interview below.
Editor's Note: Responses have been lightly edited for length and clarity.
As we reflect on 2026 and look ahead to 2027, what are the main goals you're hoping to accomplish at Arcadia in the coming year?
When we started the business, the vision was really simple. We wanted every health care interaction to be influenced by excellent insights. In 2008 and 2009, that meant connecting a bunch of systems and trying to use a mesh network of data to create better insights.
But there was always a challenge in that you could only scale insights as quickly as you could scale resources. At the same time, the U.S. health care ecosystem was faced with a number of challenges, whether that was the large-scale transformation through the implementation of EHRs, clinical labor shortages, the pandemic or the shocks to costs and the way people navigate care that followed.
While we've made really good progress on that vision, there's still a lot of care delivered where care teams don't have the insights they need to provide the best possible care. That's not because they don't want to. It's because technology in health care has often lagged behind the rest of the economy.
What I'm looking forward to, and why this ties into the Ascent launch, is that we live in an era where technology has gotten to a point where it can scale us almost infinitely.
With that comes a lot of practical applications in health care. One is a real, bona fide activation of the consumer with good insights. No longer is it, "Doctor Google told my patient that I have X, Y or Z." It's, "I uploaded my medical records into Claude, and I, as a patient, can have a really informed conversation with the provider."
Oftentimes, I might have better access to my own medical records than the provider does, especially if I'm getting care across sites.
As we think about moving forward, it's how we use great data to make sure the insights being surfaced are excellent insights. That enables care teams and consumers to collaborate in shared decision-making, using big datasets to help identify the most effective, cost-appropriate and consumer-satisfying treatments for disease.
I'm an optimist. I think we're going to see continued advancements in curative therapies over the next decade. If you look at what's happened with cell and gene therapy and diseases like sickle cell, the advancements around type 1 diabetes or the conversations about vaccines and curative treatments for cancer, these are all remarkable advancements in medicine. But they require massive amounts of data and insight.
You need to connect that layer of advancement and R&D down to the hundreds of thousands of care providers and care sites in the country.
As we look at where Arcadia is going with Ascent, it's about being that connectivity layer, that data platform, so that as you deploy AI, it's on top of clean, trusted data, and you're not amplifying missingness or gaps in data.
I want to go back to Arcadia's acquisition of CareJourney in 2024, because it seems like we're seeing that strategy come to fruition with Ascent. What did CareJourney bring to the table, and how is that influencing the platform today?
First and foremost, we worked really closely with the CareJourney team for years. Aneesh [Chopra] and Dan [Ross], who started the business, are good personal friends and great colleagues. They started their business with a very similar vision that data and insights could propel better care outcomes.
We approached the problem differently at Arcadia. We approached it by connecting our customers, which are large health systems and health plans, and connecting their individual systems to create a view of what care looks like inside their four walls. And when I say inside the four walls, it actually extends well beyond the hospital into community practices and affiliates.
At CareJourney, they said, instead of building interfaces across the ecosystem, let's take preaggregated, really large datasets and start to tell stories about what networks look like and what provider and specialist performance looks like, based on large publicly available and proprietary datasets.
When we worked together, I would often go to CareJourney to get benchmarks or insights about a given network. CareJourney would come to us when those networks said, "That's great that you know about us from the outside. We need to know about ourselves from the inside."
We realized there was a much stronger platform if you brought together Arcadia's first-party insights, what we can tell customers about themselves based on their own data, and CareJourney's third-party insights, what you can tell customers based on data the market has on them.
As we brought the two companies together, our view was to create a 360-degree view, not just of the patient journey, but of the system. How can they perform better?
That's particularly relevant today as you think about the transformation that's started in rural health care communities. A lot of these rural hospitals are actually great sites of care. But people drive right by their regional hospital and go to an urban center, which isn't necessarily the best thing for the patient from a convenience perspective. It's terrible for rural communities when you have this drive-by care concept.
Our dataset can demonstrate that a rural community hospital can be just as effective, and sometimes higher quality, than an urban academic medical center.
For areas where they're not as high quality or effective, the opportunity to connect rural and urban providers and create much better closed systems of data and patient transitions is a reality, not a dream.
How long will it take until patients and consumers can see these performance insights and access them for themselves?
Today, Arcadia has made a commitment to the CMS Health Tech Ecosystem. Part of that commitment includes providing access points for consumers to access their own medical records. That's a hallmark of Amy Gleason's [CMS advisor and acting administrator for the Department of Government Efficiency (DOGE)] Ecosystem: empowering the consumer.
The other part is liberating data between sites of care, and that's a big part of what we already do today.
I would say consumers see the outcomes of this data today. If you pick up U.S. News and look at the rankings of the best ambulatory surgical centers, that's our data powering those rankings. If you're working with certain health plans and looking at provider finders with quality or efficiency scores, that's oftentimes powered by our dataset or a dataset like ours.
At a tactical level, Arcadia is connected to 50,000 sites of care, and many of those providers use Arcadia to identify patients who are rising risk or who haven't engaged with the health care system.
There are a lot of people who need care but don't engage with the health care system. When you don't engage, oftentimes the health care system doesn't know you exist.
I'll give you an example. If you have a patient with a high family history of breast cancer, maybe their mother or sister had it, but they're otherwise relatively healthy and don't engage with primary care, they might not have seen a doctor in two or three years. Maybe they have a different job and different insurance now.
That new health plan doesn't necessarily know the patient has a family history of breast cancer and needs advanced screening, but we might. Networks like ours do.
We can notify the patient that they're missing critical preventive care services. We can also notify their health plan or provider network, if they're Arcadia customers, that there's a patient they haven't seen in years who needs care.
The best part of what we're building is that it's not just a report dropped on someone's desk.
With the advancement of agents, we sent a million care reminders to patients over the last year. These aren't just upcoming appointment reminders. They're about services patients need that we can help them get, whether that's vaccine reminders or emergency care alerts.
We did a lot of alerting during the last wildfire season on the West Coast, helping patients with high social needs and respiratory conditions get to air-conditioned cooling centers.
These are the types of capabilities deployed today across tens of thousands of users.
Aside from sending notifications about needed services, are there other examples of decisions or actions Arcadia's AI agents can take autonomously that previously required human intervention?
In the patient-provider interaction today, we're being really careful. We're on a glide path to autonomy, but right now most of our activities are supervised.
It's a risk tolerance we've aligned on with our clinical advisory committee, which comprises our chief medical officer and many of our customers.
Everyone wants to get to full autonomy, but with so much rapid development in the space, we've aligned on the importance of understanding which cohorts are being contacted, about what and how often. Having that extra set of eyes is important.
That doesn't mean a human is looking at every single interaction. It means there's oversight at the macro level.
Where we have unlocked autonomy is in solving what I think is one of our industry's biggest problems, and one we don't talk about enough.
The health care industry generates more data than any other industry in the world. I think 36 or 37 percent of the world's data comes from health care. I've been in health care data for almost two decades, and I would not say we've solved the underlying data issues we have as a system.
Data is messy. It's not always fast. It's verbose.
If we don't solve the data foundation problem, we'll never realize the huge benefit of AI in health care. AI is a huge amplifier of really good things, but it also amplifies garbage.
We've talked in health care data for two decades or longer about garbage in, garbage out. In traditional analytics, you have a human looking at this stuff, and you can generally catch the garbage. With AI, because you can scale a person 10 or 100 times, you often can't catch it until it's really bad.
We believe strongly that this is the moment to fix the health care data foundation and get in front of the garbage-in, garbage-out problem.
A lot of where we've gotten fully autonomous has been on that problem.
We maintain thousands of interfaces to clinical systems, claims systems and ancillary systems every day to stitch together that data fabric for our customers.
This year, we launched a series of autonomous agents that monitor every one of those interfaces every minute of every day. They're self-curing and self-healing.
As interfaces go down, they'll try to get them back up. If a customer upgrades their core system and the data model changes, the agents can inspect that data model, look across the thousands of other interfaces we run and identify patterns.
They can essentially say, "We've seen this pattern before. Here's a recommended change. I'm going to go code that change."
In a lot of ways, we have an army of data engineers that are AI agents we've built ourselves, and those are fully autonomous.
That's changed our ability to partner with customers around data quality in ways that, two years ago, I would have dreamed of. Two years ago, it would have required me to increase the size of my data engineering team tenfold.
We've executed a pretty material internal AI autonomy transformation, which has allowed us to deliver a higher-quality data foundation that's more trustworthy and more recent.
How do you expect AI agents to affect health care costs, workforce pressures and some of the other challenges we haven't seen dramatic improvements in since AI arrived?
You're speaking my language.
Last week was our user group meeting in Philadelphia, and I opened my session quoting A Tale of Two Cities: "It was the best of times, it was the worst of times." The spring of hope and the winter of despair, all at the same time.
The initial deployments of AI across the health care ecosystem are pretty inflationary.
We've seen this movie before. It's exactly what happened when EHRs were implemented. There was this big pitch: Implement an EHR, and it's going to make your clinicians' lives better and help bring down the cost of care because we'll reduce duplicate services.
I don't think anybody made that up. Our incentives, however, in the U.S. health care system are that you get paid for what you bill, and billing is adjusted by acuity and complexity of encounters.
You increase acuity and complexity with more documentation. EHRs made it easier to document more. Therefore, costs went up because people were coding appropriately or upcoding, depending on who you talked to.
We're seeing the same thing with the deployment of AI.
The biggest use cases for the first deployments of AI in health care have been revenue cycle management and payment integrity, and both of those things were inflationary.
The Blue Cross Blue Shield Association put out a paper last month saying it believes AI has driven up claims costs by about a billion dollars over the last 12 months. Looking at my data, I think that's true. You have generally the same amount of health care being delivered, but the cost has gone up.
That said, I'm an optimist.
We have 100,000 nurses in this country who process prior authorizations. We have teams that generate bills. It costs hundreds of dollars to generate a surgical bill in this country. It costs about six bucks in Canada.
Some of that's the complexity of our system, but all of that is administrative burden that could be affected by AI.
What I've challenged my customers and my team to rally around is that value-based care, where we spend a lot of our time, is helping hold cost trends relatively in check, although they're still growing rapidly.
We spend a trillion and a half dollars in the health care system on administrative expense. Transaction costs as a percentage of the transaction in health care are 20 to 25 percent. They're two to three percent in financial services.
I know it's different. Everyone yells at me when I compare health care and financial services, but fundamentally, merchants don't accept a transaction fee of 5, 6, 7 or 10 percent in financial services, and we accept 25 percent in health care.
If we really want to bend the cost curve and reduce the overall burden of the health care economy, it's going to be through AI transformation.
Whether you focus solely on cutting administrative costs in half or start looking at where you can shift sites of service, activate more virtual care or scale programs like the CMS ACCESS Model, I think there's about $1,200 per beneficiary per year in impactable costs that we'll see unlocked over the next decade through the implementation of AI.
One of my calls is that we've spent all this time over the last decade working on prior authorization, making it faster and more consumer-friendly. You don't really need it anymore.
I mean that tongue in cheek, because we can bring real-time insights into the point of care, when a provider is doing shared decision-making with a patient, to say whether something is covered by the benefit plan or whether the data needed for approval is being captured.
A lot of the back-and-forth in prior authorization is about data and information. If we can get in front of it, we can deliver a much less burdensome health care system and redeploy those nurses.
We have about 100,000 nurses working on prior authorizations, and the clinician shortage we're facing in the U.S. is about 100,000 clinicians.
We have the workforce. They're just not working on the right things.
That's where I think we're going to start to see the advancement of agentic AI solve this workforce gap that's creating access issues and preventing us from investing in prevention.
What advice do you have for leaders hoping to avoid the mistakes made during EHR implementation and actually realize AI's value?
A lot of advice.
Health care is terrible at implementing technology.
McKinsey did a great paper that looked at [labor] productivity over the last 25 years. Across industries, the U.S. workforce has become [more than 55] percent more productive. Over the same time horizon, health care has become [approximately] one percent less productive.
Part of that is regulation, whether you're at Cleveland Clinic, a federally qualified health center or CVS. Very few of us can influence that, although we can influence it through lobbying.
I think you've got an administration right now that's really focused on reducing administrative burden.
The second thing is, if you go back in time before meaningful use, a lot of health systems built their own technology.
I'm a technology company. I have excellent engineers. You know what I don't build? I don't build Salesforce. I don't build NetSuite. I buy the best-of-breed solutions from the market.
Health care needs to realize that even though it's become easier to build software than ever before, that doesn't mean you should.
McKinsey presented some really great slides about how payers, providers and life sciences organizations are approaching AI implementation, and a lot of them said they're partnering.
I would really encourage people to partner with technology leaders who know how to drive innovation, create the right operating committees where you can have fast, closed feedback loops and not sleep on the investment in change management.
The chief information officer of Montefiore talked about their implementation of ambient AI. I think he said they had 80 change management consultants working across their system to roll this out.
Someone in the audience said, "Oh, we just turned it on."
One implementation worked a lot better than the other, and it was the one where they had change management professionals.
We spent more money on change management than we did on EHR technology from 2009 to 2013. We're in the same wave [now].
The other piece of feedback is that we have the most complicated data foundation of any industry. We've never simplified it. We've never fixed it. We still have faxes flying around all day, every day.
You can't put AI on top of a fax machine.
You need a data foundation. Your EHR is not the platform to play AI on top of. You need a data strategy, and your data strategy needs to be more than having access to your EHR database.
You need someone who owns it. You need governance around it. You need interoperability into and out of it. You need a set of SDKs and APIs to connect into it. You need a platform for this.
As we think about implementation, it comes down to change management, getting your data foundation in place and having the right partners who can work with you.
Just because you can build something doesn't mean you're best equipped to do it.
Do you think we're headed toward a health technology market dominated by a small number of comprehensive platforms rather than specialized vendors?
I think you're going to have a middleware layer dominated by platform players that are organizing, collecting, cleaning and normalizing data.
That's our strategy: Let's collect all that data and make it really easy for people to plug into it, so that a bunch of point-solution vendors can do that.
I'll give you an example. We're not an oncology navigation business. A lot of our customers need oncology navigation. Those navigators need great data and connectivity to other sites of care.
There's no reason we should build 18 different data layers. They should be able to plug into ours, get information easily and write back to it.
Our strategy is to become a platform and onboard a bunch of different applications.
That's a big part of the CMS Health Tech Ecosystem, creating these networks of connected data that's trusted and high quality, where people can plug in their point solutions.
I think we'll look like a lot of other industries, and frankly, that's what we look like today.
You've got big platforms like Workday, Epic and Cerner. You've got a lot of applications that plug into them.
I think you're potentially in the middle of a paradigm shift where the core platform, the core system of a health system, isn't the EHR anymore. It's a data platform.
It's not dissimilar in the health plan space. The core system may not be QNXT or HealthEdge anymore. It might be the data layer that connects all the different systems.
C-Suite Shuffles
Where health care leaders are coming and going
- Wayman Cummings has a new title at Ochsner Health: chief security and infrastructure officer.
- Cummings has served as the New Orleans-based system’s vice president and CISO since November 2024. Five months ago, the organization asked him to tack on the role of CTO, and now they’re making the combination permanent.
- “The future of security and technology is not separation, but integration,” Cummings said on LinkedIn. “Looking forward to merging both organizations into one.”
- Michel Lentz was named vice president, provider partnerships, execution and performance at UnitedHealthcare.
- Lentz has been with UnitedHealth Group in various roles since 2003, save for a nine-month leave to direct government quality programs at Highmark Health, beginning in 2016.
- Since 2019, she has served as an executive director for UnitedHealthcare, according to her LinkedIn profile.
- Sarasota Memorial Health Care System promoted Chief of Staff Ashley Abbondandolo Cole to the CEO position.
- The Florida health system includes two hospitals—one being its flagship 897-bed campus—along with a cancer center and a network of outpatient centers.
This is a preview of the October 15 edition of Access Health—Tap here to get this newsletter delivered straight to your inbox on Thursdays.