Roundup: Testing Healthcare AI in Space, Increasing Statin Adoption, Epic's move into Scribing, etc

Wellness is entering healthcare, and healthcare is moving into wellness. 

That seems to be the theme of so many of the conversations I’ve been having lately. Peloton hinted in its earnings this week that one of its key plans to grow the business involves expanding its relationship with customers beyond the treadmill and the bike. CEO Peter Stern spoke to the broader “members’ wellness journey,” pointing to “mental wellbeing, sleep, recovery and over time, nutrition and hydration.” Once its users are hooked on getting exercise through Peloton, either through its hardware or software, Stern views wellness as the next logical step. Meanwhile, the Life Time gym chain has added longevity clinics, and Equinox is partnering with Function Health for biomarker collection. 

At the same time, many of the longevity companies are building partnerships with health systems and employers, and there’s massive momentum on both a state and federal level to bolster reimbursement for food as medicine programs to prevent or halt the progress of a chronic disease.

All of this makes perfect sense as a recognition that a health care system shouldn’t only serve to treat sickness, but should also support prevention. Currently, we operate in a sick care system and it’s extremely challenging to build a successful business in the health-tech space in prevention. 

There’s also good business reasons that wellness and medicine are converging. Consumer health and wellness businesses all face the same fundamental challenge: How to acquire users, sell them products or services, and then keep them around from there. That calculus related to a customers’ lifetime value matters deeply. Healthcare offers a potential path to improved unit economics, because consumers tend to convert if some portion of their care is subsidized via insurance. And there might be ways to acquire them outside of super expensive paid marketing campaigns, such as provider referrals, co-marketing with health plans, or programs at the workplace. Likewise, healthcare desperately needs an infusion of consumer DNA, particularly as the systems we have today are failing to reach younger generations. Understanding the consumer better offers a path to vastly increasing utilization of services, and also to lower costs through prevention.

So both sides can benefit from the relationship, but with a few caveats. Treating all things digital or consumer as a “bright, shiny object” without putting real resourcing behind it has rarely been effective for healthcare organisations. Likewise, wellness companies should ensure that they have the right medical expertise are on the team to ensure that they’re not using their platforms to promote procedures and interventions that lack evidence (or worse, do harm).

A quick note from our sponsor of this week’s Roundup, the team at Parsley Health

On a very different “note,” some other huge news that caught my eye this week: Epic’s rumored advancement into the ambient scribing space. This is a playbook we’ve seen before from Epic, the behemoth electronic health record company that is a near-monopoly in the inpatient space. Where it leaves the ambient scribe companies in the category, many of whom have raised hundreds of millions of dollars from investors, is a very big question. Once the Epic scribing tool gets announced, my strong hunch is that there will be a lot of public discourse over which company has the superior product, because Epic’s software won’t be as feature rich when it first hits the market. But that won’t work for more than a year or two, as health systems opt to stay within Epic for reasons of convenience. So in my opinion, the inpatient-focused ambient scribes should move quickly to leverage what they have today: The goodwill from providers frustrated by documentation requirements. From there, it’ll be a race to move into adjacent areas. My view is that it is a very good time to be a startup in the revenue cycle management space, or even better to be one like this that sits in between documentation and billing. I suspect there will be M&A in all directions. The RCM (revenue cycle management) players will like the idea of moving closer to the provider - the “front end” - by scooping up an ambient scribe. And the ambient scribing companies will buy or build their way more deeply into RCM so the service starts to feel more end to end. 

What other companies might do well as this all plays out? I wouldn’t be surprised to see companies that tackle documentation and clinical workflow in other parts of the system, like the outpatient space, also become very attractive acquisition prospects. 

How confident am I that Epic will release its own ambient scribe? Very confident.

So stay tuned. And grab some popcorn because Epic UGM kicks off next week. 

With Annalisa Merelli 

The news: Large non-profit health system Ascension launched a Clinical Innovation Institute (CII), a unit within the company with the stated mission of “improving patient care and clinician experience through technology and innovation.” 

The goal: The primary focus of the new team is to introduce innovation that reduces the administrative burden, freeing up time for providers to focus on patients. Somewhat surprisingly, the announcement does not even mention AI.

What Ascension said: “True clinical innovation goes beyond incremental technology upgrades – it requires a fundamental reimagining of how we deliver holistic, patient-centered care,” said the Chief Clinical Transformation Officer in a statement. 

Our POV: Health systems have had a mixed track record of incorporating technology, particularly anything patient-facing. What I always question when I see these announcements is how much power this new institute will have. Will there be a P&L? Budget? And will startups be able to build relationships with this institute as a fast track to a sale? Some of the initiatives may be internally facing, but I’m curious to see whether there will be programs that are focused on startups. 

The news: Maven Clinic, the virtual reproductive and maternal health clinic, announced two new upcoming offerings, Starting in the fall, some Maven members will be eligible to use a cycle tracker that pairs the insight of data collected through wearables with the insight of a network of specialized physicians. 

Yes, men: The clinic will also begin offering at-home semen analysis kits, and connect men with specialists that can help them figure out next steps. 

The selling point: There are a lot of apps offering semen analysis, and even more tracking periods. Where Maven Clinic hopes to stand out is by leveraging its provider network, and ability to connect patients with specialists. 

Expert POV: Maven has been considering getting into the men’s health space for a while! Its Chief Medical Officer Neel Shah has taken a big interest in the topic, and I’ve seen him speak about it on several conference stages. This seems both like a step into men’s health, a space that has existing players like Posterity Health and Legacy Health, but also brings Maven into the consumer market more so than I’ve seen from the company in the past. 

The study: A group of endoscopists tested their ability to identify precancerous polyps in colonoscopies before and after they started using the help of an AI detection tool. Their detection rates dropped, Katie Palmer reports in STAT.

What happened?: There may be many other factors, but one theory is: Doctors are humans, too. So when they get used to having the help of AI, their brains get a little lazier. When a technology has proven time and again to be effective and accurate (and AI has, with colonoscopies), the human stops checking quite as thoroughly. It’s not dissimilar to what we’ve seen happen with airline pilots that forget certain manoeuvres over time, because the tech has essentially taken over. 

The concern: The idea that AI can make doctors work better is essentially a gamble, says Adam Rodman, a clinical reasoning researcher. “We’re going full speed ahead without fully understanding the cognitive effects on humans,” he said. 

Does it matter?: It will, for the next half a century or so. After that, doctors will likely no longer need the skills that AI can replace. But for now, it would be very worrying if doctors started to defer so much to AI that they stopped checking it for accuracy, even when the technology is known to fabricate and hallucinate. 

The news: The two companies are working on a program to fill the gaps in chronic disease treatment. The first phase is designed to increase the use of statins among eligible patients.

How it works: Humana identifies patients who should take statins, and uses DrFirst to connect with providers who can prescribe the treatment. 

POV: This seems like a no brainer as it makes it extremely easy for providers to prescribe in a single click. And the idea is to reach them in their workflow, so it doesn’t take a bunch of additional busywork to get the job done. 

The news: The fitness company has beaten expectations and reported surprise profits, with a net income of $21 million (the same quarter last year, it lost $30.5 million). 

The action: The company is further reducing its workforce by 6%, a move part of a $100 million saving strategy. The last round of layoffs was last year, when the company cut 15% of jobs. 

The take: One of the smartest things we did during the pandemic was buy Peloton stock, and then sell it once people started to return to offices. We made enough money off of that trade to buy a Peloton (which we both still use despite the return-to-gym trend). The company still faces poor sales of its hardware, so the future is all about looking for ways to engage with people through digital content, like the app. Peloton also has a big opportunity to get into healthcare if it can start expanding the scope of the courses, and that’s a huge opportunity for digital health. 

Space sick: NASA is working on a new era of space exploration, and plans to bring humans back to the moon. But what happens if an astronaut gets ill in outer space? The answer, according to a new Google and NASA partnership, could be the Crew Medical Officer Digital Assistant (CMO-DA), a clinical decision support system that provides medical help when there is no physician as part of a mission.

AI-enabled? Of course. In fact, the program isn’t only about providing support to sickly astronauts, it’s about testing how far — quite literally — AI for clinical needs can go.

Why should we care here on Earth?: As a proof of concept, something that works in outer space will also work in remote areas of the US where healthcare resources are limited. “It’s about pushing the boundaries of what's possible with AI to provide essential care in the most remote and demanding environments,” Google said. 

This is not necessarily a big piece of news, but many facets of health care — from hospitals to biotech companies — are going through restructuring and layoffs (more than 800 people were laid off this past week alone). The health publication Fierce Healthcare is keeping track, so here is a page you may want to bookmark if you want to keep an eye on employment trends in health. 

Heartflow’s IPO surpassed $630 million: The company, which developed a software making 3D models of the heart from scans, started trading publicly on Friday, August 8. By close of business Monday, shares were selling for $30 each, 58% above the initial IPO price of $19. 

Doximity bought Pathway Medical: The deal — $26 million in cash and up to $37 million in equity — is aimed to integrate further AI capability into Doximity, which already offers an AI scribe, by offering a clinical reference tool.

Chai Discovery closed a $70 million Series A: The AI-enabled drug discovery company’s latest funding round was led by Menlo Ventures, and included several new investors. 

Apreo Health raised $130 million for its emphysema device: The oversubscribed Series B supports the development of Breathe, an implant that slowly releases the air trapped in the lungs of emphysema patients. 

Healthcare Outcomes Performance Company (HOPCo) acquired Caro Health: The digital health company sees the partnership with the Dutch conversational AI leader as a way to grow its international presence and integrate further AI capabilities with its existing clinical offering. The terms of the deal were not disclosed. 

SetPoint Medical got $140 million: The funding, raised from investors including Abbott and Northwell Health, is to go toward commercializing SetPoint System, a product to treat rheumatoid arthritis in people who didn’t respond to other treatments. 

EL: Extending the expansion of access to telemedicine for a variety of services (after introducing during COVID-19) is a big deal for digital health companies, which eliminates the biggest sticking point for consumers getting access to care, which is having to show up somewhere. And that’s the complex co-pay and deductibles for a few different reasons. Consumers aren’t comfortable inserting their credit card information into unknown sources. It’s not like it’s Amazon ‘one click’ for paying a co-pay. And they don’t like thinking about paying for care in general if they have insurance. If you eliminate those two frictions, it accelerates the ability for someone to see a doctor. Outside of the disease management companies, patients who are struggling with an issue can now talk to a doctor who will spend the time with them. It won’t take three months to get into a specialist, and that can avoid unnecessary procedures and imaging. So it has a real potential to impact the total cost of care. I think the biggest downside is the way the law was written, it’s up to the plan or the employer. As we’ve seen multiple times, anytime there’s ambiguity it creates more friction to operationalize a regulatory ruling. 

EL: We think just getting in-network is critical before selling directly to employers. Do you want to go it alone, or do you want operational support to get more negotiating power related to rates. And they can focus on the care model versus the operational hoops necessary to work with health plans. I think the strategy of selling into employers directly is declining. The ACA allowed companies to create a market for preventative health services, and that played a role but it led to fragmentation and silos of care delivery. This stuff needs to be aggregated somehow, whether it’s via new innovators like a Thatch through ICHRA or a Transcarent for navigation, someone has to unify the care model. 

EL: Anytime you introduce more consumer choice, it’s exciting. I’d love to see a future where employers don’t need to dictate what an employee has to choose for their benefits, and employees can choose their own adventure. When you combine it with LLM and AI, we’re on the advent of delivering really personalized benefits. That has the potential to transform how care is delivered and paid for. But we’re in the early innings. I’m watching companies like Thatch and XO Health, and I also think Quantum’s acquisition of Embold in the provider quality arena is really interesting. It’s all about getting to the same result. How do you deliver on that personalized experience for members? 

EL: We are seeing a perfect storm between rising costs, uncertainty and more – and that’s really finally opening the door to the alternative health plan space. I think it’s also bolstering some of the navigation players. The question for me, particularly given the job market is softening and there’s more people looking for work, is whether employers will be forced to compete on benefits? I don’t know if there will be a single payer system in America. But I think the market will dictate whether it’s an alternative health plan model, like an individual or defined contribution (perhaps you get $1000 a month and you get to figure out what to do with that money). Or is it super sophisticated benefits designed via large employers to attract better talent? That’s a big question that fascinates me, and I’m continuing to follow it closely. 

In a new series, “One hot take,” we’re inviting operators to share a real world experience, a challenge, a frustration or an observation with Second Opinion’s reader base. Thanks to Shama for sharing the inaugural hot take on the topic of balancing automation and efficiency, which investors are looking for, while also keeping in mind the interests of female patients. 

By Shama Rathi, MD, co-founder of LunaJoy 

I’ve spent the past five years building out critical clinical infrastructure for maternal mental health. Like most digital health founders, my co-founder and I have been eager to explore how agentic AI could help accelerate our mission. The promise is seductive: AI agents that can make decisions, take action, and learn over time have massive upsides for efficiency, scale, and impact.

But when we started developing AI agents to support our clinicians and help coordinate care, we were quick to pause development. After a few weeks, we realized that these agents were scaling the very dysfunctions within our healthcare system that we’re aiming to correct. 

One agent we tested was designed to streamline documentation. But the logic baked into it – a relic of the EHR – flagged postpartum women who missed appointments as “noncompliant” and automatically deprioritized them for follow-up. On paper, it looked like a signal of disengagement. But in reality, these were women who were dealing with childcare breakdowns, trauma, and recovery. The agent didn’t know that. It just moved them further down the queue. 

Across the board, the data clinical agents are being trained on don’t reflect patients’ clinical reality. Take CPT codes, for example: they’re supposed to offer a standard framework for how care is delivered and reimbursed. But many of the conditions we see in women’s health, from postpartum anxiety to perimenopausal dysfunction, don’t even have proper codes. When they are documented at all, they’re squeezed into generic categories like “anxiety” and “depression,” erasing key distinctions that drive effective treatment.

Polycystic ovary syndrome, autoimmune diseases, menopause – these are all conditions that remain vastly under-researched, poorly coded, and massively under-treated.

This is the challenge healthcare innovators are facing right now: we’re being encouraged to move fast and automate everything, but in doing so, we’re automating everything that is wrong with our healthcare system, too.

The incentives driving this aren’t malicious. They’re just misaligned. Venture capital rewards speed and growth. Founders are racing to ship and scale. The funding ecosystem rewards velocity over the caution and nuance healthcare requires, and the result is a flood of AI agents being built on biased or incomplete data and broken infrastructure.

If we’re not careful, we’ll end up automating failure. Patients will pay the price – especially women.

We’ve seen this before. The last wave of healthcare innovation – from broad telehealth platforms to wearable tech – promised to “democratize” care. But much of it was built for the “default male” patient (that is to say, healthcare has historically standardized around the mid-sized male patient as the “average”), and women’s health was once again relegated to the sidelines. It happened because investors overlooked the challenging work of building better care infrastructure for women in pursuit of extravagant demos and the promise of high growth.

Maternal mental health should be a proving ground of this next wave of AI, not a casualty of it. In the U.S., maternal mental health conditions are the most common complication of pregnancy and childbirth, yet our system barely addresses them. As a psychiatrist, I’ve seen firsthand how critical it is to detect and treat these conditions early. Agentic AI has the potential to help: it could surface insights from EHRs, wearable devices, and even socioeconomic data to flag a worsening case before it becomes fatal. It could triage and coordinate care with greater nuance than a time-strapped physician. It could identify patterns we’ve historically missed.

But that will only happen if we build AI that places women at its center from the start.

To do that, we need to think differently about the data we’re using to build agentic systems. Women’s health data is not “niche.” It’s fundamental, and deeply contextual. A woman missing appointments during the postpartum period might be attributable to transportation insecurity or domestic violence. These social determinants of health must be part of the agent’s reasoning – not just footnotes in the medical record.

The onus to correct course before we send our healthcare system flying off the rails is on providers, plans, and life sciences companies, but it’s also very much on investors. Venture capital has a choice right now: fund the flashiest agentic AI platforms promising system-wide transformation, or invest in tools and teams that are working to correct the structural biases baked into the healthcare system first.

As innovators, we cannot afford to get this wrong. The stakes are too high. What gets built now will shape the infrastructure of care for decades.

That kind of oversight isn’t just costly. Moving forward, it’s going to be dangerous. Agentic AI presents us with a chance to get it right – an opportunity to innovate in service of the actual users of the healthcare system. To do otherwise is to codify the same system that has already failed so many.

AI agents are already outlining the contours of tomorrow’s healthcare system. Now is the time to ensure that the system is maximally effective for the people who rely on it most: women.