By KIM BELLARD

Science won. At least, that’s one of the takeaways from last week’s midterm elections. If you were one of the 124 million Americans who voted, you probably already know that there was a Blue Wave, with resounding Democrat victories across the country, even in historically red areas.

Georgia not only reelected its Democrat Senator – no surprise — but also elected a Democrat Governor, who will be the first black woman governor in the country, no less. Iowa elected a Democrat Senator and Governor. South Carolina is sending a Democrat to the Senate, and Mississippi – Mississippi!!! – may as well. Even Kansas elected a Democrat Senator, the first such in almost 100 years. Ohio and Texas not only flipped their Senators and Governors but Democrats won all the statewide elections.

Truly strange times.

I’m allowing myself to hope that we’ll finally have a Congress that will fulfill its constitutional duties of acting as a counterbalance to the power of the Presidency, but one thing I’m particularly excited about is the prospect these results will be less about being a Blue wave than what The Washington Post called a “STEM wave.” Among the newcomers to Congress are a record number of physicians and scientists. They’re going to have their hands full.

Much has been written (including by me) about the Trump Administration’s attacks on science. Just last week The New York Times reported on How Trump Is Dismantling America’s Scientific Research Machine. Amid cuts in research grants, resignations and firings of top scientists, departures of leading scientists to other countries, and reductions in Ph.D. students, the Defense Department warned that its “innovation engine is at risk of stalling.” Even people who aren’t defense hawks should realize that spells big trouble for American science in general, and for all of us.

A year ago I wrote about how organizations like 314 Action and Stand Up for Science were leading the resistance to the attacks on science. 314 Action, for example, spent some $23 million on pro-science candidates, doubling its 2024 donations. Shaughnessy Naughton, president of 314 Action, told Alexander Nazaryan of The New York Times that the election results, especially the incoming scientist lawmakers, represent “a real repudiation of what has been two terrible years for science and public health.”

Deborah D. Stine, who served as the executive director of the President’s Council of Advisers on Science and Technology during the Obama administration, told Mr. Nazaryan: “The importance of having scientists, engineers and health professionals in national, state and local political positions cannot be overstated.” Mr. Nazaryan highlights a mathematician, a microbiologist, a chemical engineer, and engineer, and several physicians who are part of the wave. Kudos to the NYT headline writers: These Newly Elected Lawmakers Know Something Trump Doesn’t: Science.

“It’s a new day in Washington,” Gretchen Goldman, president and CEO of the Union of Concerned Scientists, told Adam Kovac of Scientific American. “There’s huge opportunities for science, for democracy, for getting accountability for these attacks on science that we’ve seen. We are keen to get to work pushing back.” Colette Delawalla, founder and CEO of Stand Up for Science, added: “Thinking through how all of this could pan out, this outcome is better than any eventuality that we worked through.”

In another NYT article on the topic, Rep. Zoe Lofgren, who is expected to lead the House Science committee in the new Congress, warned: “In some cases, the damage can’t be undone, But we can prevent it from continuing.” She told Jeffrey Mervis of Science: “My top priority is to help the U.S. science enterprise recover from the blows it’s received from the Trump administration and also get some wins for the American people when it comes to legislation that never was approved.”

Newly elected Representative Jasmine Clark, a microbiologist/scienced educator/political activist, added to Mr. Mervis: “People show what they care about by how they spend their money. And as Democrats we need to be able to show people, show the science community, and show the public health community that we prioritize health and research.”

The NYT article highlighted four ways the STEM wave could impact science policy:

Reclaiming the Grant-Making Machine

“The power of the purse belongs to CONGRESS!” Senator Patty Murray of Washington, the top Democrat on the Senate Appropriations Committee, wrote on X. “It’s NOT Trump’s money. Taxpayer dollars belong to the American PEOPLE.”

New Investigative Targets

“A lot of how the new majority will try to drive change is through investigations and hearings,” said Mitch Benzine, senior counsel in the congressional investigations practice at the law firm Eversheds Sutherland.

A.I. Guardrails

Restrictions on data centers, regulations for A.I. development, and new protections for copyright holders (against LLM data scraping) are all in play.

Safe Space

Already bipartisan support for programs such as NASA and N.I.H, will get even stronger.

I’m excited to see what impact these newcomers can have on science and health policy, but we have to remember that they’re still in the minority among all the lawyers and millionaires in Congress, and that Congress has never shown itself to be very nimble – and certainly less nimble than the Trump Administration. Even more concerning, there are headwinds against science that are not solely attributable to President Trump.

In Nature, Miryam Naddaf illustrates how academic freedom has been declining in recent years in several countries – most notably, the United States. Citing the latest annual report from Scholars at Risk (SAR), there has been a steep decline in academic freedom in the United States, and not just since the start of the second Trump Administration. “The pressures on higher education are not limited to one election [or] one administration,” says Robert Quinn, founding executive director of SAR.

The report warns:

According to the Academic Freedom Index (AFI), since 2020, academic freedom has declined dramatically in the US. In 2019, the AFI rated the US as “fully free” with respect to academic freedom. By 2025, it was rated “moderately restricted.”…The US’s decline in university autonomy was particularly rapid, deteriorating by 50 percent over a 10-year period from 2015 to 2025.

Certainly the first Trump Administration started some of the attacks, and the second has dramatically broadened them, but state legislatures are complicit in many of the restrictions on academic freedom, and many of those legislatures weren’t swamped by the Blue wave.

So let’s be optimistic about the Blue wave. Let’s be hopeful in particular about it being a STEM wave. But let’s not forget that the war on science is not over, that our science infrastructure has taken terrible blows, and that there are many battles yet to come.

Kim is a former emarketing exec at a major Blues plan, editor of the late & lamented Tincture.io, and now regular THCB contributor

The transcript (made by Youtube and Claude so blame AI for any mistakes!)

Matthew Holt, THCB

Matthew Holt, with another THCB Spotlight. I have with me Amy Bucher. Amy is the Chief Behavioral Officer, scientist, and intellectual powerhouse, I think, behind Lirio. Lirio is a company in the behavior change space, so you’re absolutely in the right place. Amy has a long history of hanging out in the world of healthcare design and healthcare media, with stints at what was Health Media before it was J&J, and after it was J&J, and then Mad*Pow, a company I knew well, because Mad*Pow was run by your fellow Amy, Amy Cueva, who I think went on to become chief design officer at UnitedHealthcare at one point, and is now doing something else. I know she just harassed me recently.

But Amy, welcome, and tell me a little bit about what the basic idea is that Lirio is helping with. We hear about behavior change, and we don’t know if it’s possible. Tell me what Lirio is trying to do. What’s the objective?

Amy Bucher, Chief Behavioral Officer, Lirio

Yeah, and thanks for having me. Lirio is really premised on the idea that personalizing how we talk to people, and giving them what they need to overcome their barriers to action, is what will engage patients and actually get them to do the things that are recommended for their health. We know that in healthcare, adherence and patients taking action has been a consistent problem. I’ve been working in this space for over 20 years, and while we’ve solved some challenges, there are still so many that remain.

But now we have the technology, especially with artificial intelligence, where we can look at a huge amount of contextual data and really figure out the right way to talk to each person, at a specific moment in time, about different behaviors. We’ve had approaches in healthcare like segmentation, which are so much better than generic, one-size-fits-all messaging, but which still miss the mark. As much as you might be a certain type of person when it comes to — we were talking earlier about mammograms — mammogram programs, at least for me, that doesn’t mean I’m that same type of person when it comes to my vaccines. Or if I’m actually diagnosed with a serious health condition, that segment won’t necessarily carry through. We haven’t had a mechanism to accommodate that kind of segmentation drift: how do we understand how people are shifting in the moment and changing over time? So what Lirio really does is take in, and in fact create, a large amount of data and figure out the optimal way to communicate with this individual at this point in time.

Matthew Holt, THCB

Broadly speaking, we’ve got quite good outside of healthcare at moving people around, in areas like e-commerce and retail, combining advertising with information. We seem to be getting very good at it — perhaps too good at it in terms of politics, voting, campaigns, misinformation, and God knows what else. But what is the state of the science behind this? Where did you originally come from, in terms of the state of the science of behavior change? How much more effective are we now than we were when it was basic? Go back 100 years, really basic, general generic advertising, then more segmentation and targeting, obviously turbocharged by the internet, and now AI and tons of data. We had the big data era, and now we have AI. How much better have we got? And then we’ll talk about healthcare.

Amy Bucher, Chief Behavioral Officer, Lirio

It’s an interesting question, because it’s sort of how much better could we be, and how much better are we actually? There’s what we could do with the science, and then there’s what we actually do when we get into the messy real world and have to negotiate with legacy technologies and business rules and so on. I do believe, and have seen lots of evidence, that behavioral science makes a difference in how we can activate people around their health. I wouldn’t have devoted my career to this otherwise.

Matthew Holt, THCB

Before we get into health, let me ask you how much better we’ve got outside of health, in these other areas like retail.

Amy Bucher, Chief Behavioral Officer, Lirio

Oh, a lot. And I say this as sometimes a victim of the behavioral science, because being educated in it doesn’t help you. Lord knows how many hours I’ve lost to TikTok, or even Netflix. One of the things those technologies specifically do — or even Facebook, a product which I quit years ago, to the benefit of my mental health, because it’s so good at this stuff — is pick up on situational signals: not just what am I interacting with, but when, and for how long. They use those sources of data too, which are really powerful in getting people to — in their case — spend more time with their product. What we’re trying to do is get them to spend more time actually with their health care. But it’s all that context that I think makes a huge difference. It’s not just what I would call traditional behavioral science, of what are you doing.

Matthew Holt, THCB

If I like putting things on scales of one to ten: generic advertising was a one or a two, and market segmentation, sort of pre-internet, was a three or four. Outside of healthcare, are we now an eight? A ten?

Amy Bucher, Chief Behavioral Officer, Lirio

My instinct is to say an eight.

Matthew Holt, THCB

Okay. So it’s pretty good. The fact that ads follow me around the internet because I expressed an interest in plumbing or something — it’s gotten pretty good at inference. Now let’s go to healthcare, because in the case of other behavioral mechanisms, it’s trying to get us to buy something, vote for somebody, send money somewhere. There’s a transaction involved. Healthcare is a real pain in the rear, because you’re trying to do transactions, but you’re also trying to get people to change how they’re thinking and how they’re behaving, and these changes tend to have the unpleasant artifact of cost or activity being required up front, and the benefit usually coming a long time down the road. We know that’s hard for people in general, and it seems to be getting harder — judging by my analysis of society, which is observational only.

So where’s the state of play in behavior change in healthcare? You’ve been doing this for a long time. How are we doing with people who have chronic diseases, where we’re trying to change diet, we’re trying exercise? We’ve got some drugs that seem to be working on the diet side. But broadly, going back to my scale, where was healthcare, and where is healthcare?

Amy Bucher, Chief Behavioral Officer, Lirio

Oh boy. Were we negative before?

[laughter]

I mean, healthcare certainly wasn’t good. I’m going back about 20 years when I think about this. I would say maybe a two or a three, and I think maybe now we’re at a four or a five. We have some good reasons for that, and some maybe not so good. The good reasons are that we care more about data privacy in healthcare than we do in consumer interactions. The things that I do with my doctor are not available as data as readily as the things that I do on my internet browser, which is sort of for sale to the highest bidder — or honestly, anyone who’s willing to pay anything, probably. So we protect healthcare data more than other sorts of data, and that limits the way we can use it to personalize and engage people. I think that’s probably a good thing overall, but we also sometimes take it too far.

One of the things I see a lot in my research is that patients get frustrated with healthcare because we know we’re sharing our most private, personal data with our doctors, with our pharmacies, with our insurance companies, and then we have an interaction with one of these organizations and they’re asking us all the same questions again. It’s like they don’t talk to each other, and they don’t even seem to remember the things we say to them. For all that people find some of the consumer advertising creepy, we actually expect our healthcare organizations to know things about us, and then feel surprised when they don’t. So I think we have a lot of latitude to use the data that we do have a little bit more assertively. Yes, data privacy matters a lot, and we need to be really thoughtful about how we protect that data, but even the organizations that already have it are not using it the way they could be. So that’s one part of the problem: can we get the data?

Matthew Holt, THCB

Speaking as somebody who’s been involved in a lot of surgery and a lot of healthcare visits lately: my name and my birth date, I know very well. I’ll say them six times in one visit. That’s just for identification, so they know they’ve got the right person. They have to be careful, and obviously there’s lots of checking going on. But yes, there’s a lot that clearly could be done better by those organizations, because of data limits.

But there’s the other side of the equation, which is that you’re asking people to do hard things, right? Change their behavior, eat better, take their medications, get their vaccine, whatever it happens to be. Life gets in the way, or, hey, that chocolate bar is really good. There are these other forces, the people who make high fructose corn syrup and whatever else, who are counteracting what would be good for me. And we know the numbers. I was judging an event yesterday where there was a company doing remote heart monitoring that had gotten hypertension control up to, I think, 85%. That’s great, because the California statewide average for people with hypertension, I think, is 50% at goal, out of people who have it — let alone the people who have it and don’t even know they have it. So is five okay, and maybe we get to a six or a seven, but we’re never going to get to an eight or nine? How hard is this problem?

Amy Bucher, Chief Behavioral Officer, Lirio

Again, it’s a hard problem that is actually kind of easy in some ways. We have some of the tools we need to be better at this. Some of the work we’re doing at Lirio, I’ve been blown away by. You just mentioned hypertension control: we’re working with Cone Health in North Carolina on hypertension control. They have a series of nine clinics for high-risk hypertension patients, and they’ve been doing this whole big effort, for which they’ve gotten a national award from The Joint Commission, around how they get, in particular, their Black patients to hypertension control. The goal is to have 70% of them reach control. They’ve done well in the past with their overall patient population, but if you break it down by race, it tends to be the white patients who are doing well. So they’ve really focused on the health equity piece of this, received their national award, and Lirio is part of their work.

What we’re basically doing is trying to nudge the people who don’t engage after they’ve gotten the patient portal messages. They’re overdue for their doctor’s appointments. Part of what Cone is trying to do is have the doctor talk to you at the appointment about making the next appointment, but if you never go to the appointment, obviously the doctor isn’t having that conversation with you. We’ve been able to get upwards of 60% of these patients to show up for at least one doctor’s visit. So we’re using some data that we have about them, and we’re using behavioral science to speak to them and hopefully address the actual reasons why they haven’t been showing up to the clinic. Otherwise, these might have been patients who were left behind. That’s not 100%, that’s not getting us to ten out of ten, but I think having these sorts of things to address the pockets of the population that haven’t responded in the past can get us a lot further along the way. I don’t want to say it’s easy, but we know how to do it.

Matthew Holt, THCB

Go a layer deeper into how you do it. What kind of data is Lirio collecting? What’s the secret sauce behind the algorithms, the wizardry? What are you collecting, what’s going on internally? And then I’d like to see what it looks like on the other end. But first, explain what is new about what Lirio is doing in behavioral science.

Amy Bucher, Chief Behavioral Officer, Lirio

Sure. When we start, we can work with whatever type of data our client organization has available to give us. At a minimum, we need to know who we’re communicating with, we need to have consent to communicate with them on the channels, and then we need to know who’s eligible for what, which is typically some degree of health background. With the hypertension example, we know who has a diagnosis of hypertension, when they last visited their doctor, and when they were supposed to come in, so that we can say: yes, this is a patient who’s overdue for that visit. Some organizations will give us a lot of data on the population we’re nudging, which we can definitely use to make more informed first outreaches, and some will give us the absolute minimum. It doesn’t really matter, though, ultimately, because our secret sauce is that we’re creating new data every time we nudge somebody.

My team builds a logic model for every set of nudges. Basically: what are the barriers that prevent somebody from taking this action, and what are the behavioral science techniques we’re going to use to overcome those barriers? We write all the content that uses those techniques. Then when we send a nudge and somebody responds or doesn’t respond to it, we’re learning whether this is a technique that’s effective for this person, for this behavior. That includes the context of, hey, we text messaged on Wednesday at 2 in the afternoon about a mammogram, and Amy opened that text, clicked the link in the text message or didn’t, scheduled the mammogram or didn’t, and then ultimately, did she show up for that mammogram? So we’re tying together the behavioral science outreach to the individual’s behavioral response, and that’s what’s really allowing us to develop this picture of what a person responds to in terms of outreach.

Matthew Holt, THCB

How much does it matter that you know a lot? You get a lot from what you’re doing. Do you get pretty quickly to where you’re pretty smart about what to do next from just your own data, or does it help if you know a decent amount more about people and their health status, and is that information easily available?

Amy Bucher, Chief Behavioral Officer, Lirio

It depends on the behavioral space, and it’s also because we bring trained AI models to our implementations. If it’s a space where we’ve done a lot of work — vaccinations is a place where we have nudged millions of people over four years — we can nudge a new population with minimal data and be pretty effective very quickly. If it’s a new behavioral space for us — I’m trying to think of something we don’t do — oh boy — actually, cancer treatment. We do a lot of cancer prevention and screening, but we don’t do anything yet in the space of cancer treatment. So let’s say we were doing that. It would probably take us longer to be effective.

We do typically see a sort of hockey stick in our behavioral conversion data, where to start, we do fine. We’ve actually done tests where it’s just the behavioral science, no AI, no personalization — let’s just use our behavioral science messaging against a standard of care — and we outperform that. As you would know, behavioral science messaging does well compared to just flat health messaging. But the AI takes a little bit of time to learn, and I guess that curve gets pushed out a little bit if we’re in a new space with less data.

Matthew Holt, THCB

When Lirio gets engaged by a health plan, hospital, pharmacy — a big healthcare organization trying to reach consumers — are you generally regarded as messaging outreach that is online, text, phone, whatever? Or is there a change that has to happen when the doctor and the patient, or the person in the call center, are reaching the patient as well? Is this a bolt-on, or are you a fundamental change in how ABC Health System is trying to figure out how to deal with people?

Amy Bucher, Chief Behavioral Officer, Lirio

Our aspiration is to be the latter. We really see ourselves as an intelligence capability, and part of what we’re doing with the behavioral science is not just putting content into the product. When we implement, we go through a process we call behavioral discovery and design. Actually, you mentioned that I used to be part of Mad*Pow; it’s very similar to the work we would do at Mad*Pow. We want to map the patient action paths, we want to understand what we’re asking people to do, and offer feedback to help make that a smoother process for people. And then there are parts of our product that we can also use to support people, like a call center rep or a provider, but it depends on how the client wants to use us.

We have some clients who say, “We want you to just take over our mammography outreach, because we’re really struggling there.” In that case, it does feel a little bit more like a bolt-on, perhaps. But we work with a national retail pharmacy chain, and we’re doing all of their healthcare communications. In that case, it’s really an integral part of how they interact with their pharmacy customers, and what we’re seeing is that, because of that, the learning system is superpowered. Just from when we first went online with prescription refills, we were able to raise the percentage of people refilling their meds from their existing customer base by several raw percentage points. Out of the box, that’s pretty incredible. And we’ve had experience nudging their pharmacy patients about vaccines for years now. So that’s really where the power of what we do gets unlocked, at that sort of full integration.

Matthew Holt, THCB

So why don’t we take a look at what this looks like in actual reality. I think you have a couple of examples of things that you’ve been doing.

Amy Bucher, Chief Behavioral Officer, Lirio

Absolutely. One of the things I’ll emphasize is that we work really closely with our clients to make sure we’re embedded in their existing patient experiences. We white label the product. Patients have not heard of Lirio. We want to leverage that trusted brand and create a consistent experience across their touchpoints. So what I’m going to show you could look different depending on who we work with. This is the generic demo.

Matthew Holt, THCB

I get it. Other people will look at this too, so I want to make sure they get it. Who is this health system we’re discussing here?

Amy Bucher, Chief Behavioral Officer, Lirio

It’s our ABC Health, or whatever it is.

[laughter]

Matthew Holt, THCB

Okay, great.

Amy Bucher, Chief Behavioral Officer, Lirio

So this is precision nudging for mammography, which I mentioned. It’s actually our OG precision nudging intervention, which we first launched in 2019. It’s geared toward getting women who are overdue for their screening mammograms to come in and do that. What you see here on the left is what a patient would get, and this is an email. On the right is a visualization of the behavioral intelligence behind the scenes. It shows what barrier is being addressed and which behavior change technique we’ve chosen to use to address it. The “BCTTv1” thing refers to the COM-B model and the Behaviour Change Wheel, which is the framework behind the scenes that we use to structure all of this.

Matthew Holt, THCB

Is that well known in behavior change, or is this a behavior change technique that you’re adopting?

Amy Bucher, Chief Behavioral Officer, Lirio

It’s well utilized. It’s all made publicly available, and one of the reasons we like it is that it’s just an organizing framework. It still leaves you a lot of space under the top level to do your art, but it creates this replicable, evidence-based approach.

Matthew Holt, THCB

Yeah, it’s good.

Amy Bucher, Chief Behavioral Officer, Lirio

So here’s the email. It’s personalized: you can see here, it’s been over a year since your last screening. One thing you don’t see in this particular image is that we can demographically tailor the images, which I know from all the way back in my health media days helps to engage people — seeing folks similar to yourself in the images. The call to action is really to book the mammogram. Like I mentioned, that’s going to be the key piece of data that we train on: whether the person actually did that or not. We could also do this in SMS if that was indicated as better for this person, or if that was the only thing they had consented to. Or we could do something like an app push notification. What you see here is that across all three channels, the behavioral science methods remain consistent. We build our content in a way where we can flex to the form factor instead of changing the message entirely.

Matthew Holt, THCB

So the message is slightly different, but all related, and basically the same technique.

Amy Bucher, Chief Behavioral Officer, Lirio

Right, exactly. There are things you need to change about it. For example, we also do MyChart messaging, and even though that looks like an email, it doesn’t have a subject line, so what they do instead is make the subject line the first part of the message. We structure the messages differently for different form factors. But I’m going to say “not scheduled” here, because I want you to see what happens: she’ll get another nudge. Typically when we nudge somebody, the frequency is something we can use the AI to determine, or sometimes our clients will say, we want to message people once a week until they schedule. But if somebody doesn’t respond, the next message they get will be different. It will be selected by the AI, knowing that whatever we sent last time wasn’t the thing, or at least not the thing at that time. So here you see a different barrier being addressed, a different behavior change technique, and again it could be in any of these form factors. It could also be in different languages. We work across cultures and across languages; I just spent the last week in Europe doing some work over there.

This time I’m going to have her schedule. She’d book the mammogram, and it would deep link her into a scheduling page, or potentially inside MyChart. Sometimes our clients use phone numbers here. We’ll have her confirm this appointment, and she receives a commitment message. I’m just going to happy-path this and have her say she’ll be there, and then she attends. Basically her precision nudging for mammography journey has ended for now, and you can see a recap of what she — I say she, it’s my name — received and completed.

One of the things we actually measure: people often become eligible for precision nudging when they’re overdue for something. Not always, but a lot of our clients have basically said, “We want you to work with the people who are not doing what they’re supposed to do.” If someone gets nudged by us and goes and has the mammogram, for example, and then they don’t become eligible next year because they’ve proactively scheduled and don’t go overdue, that’s a big win, even though they don’t show up in our outreach anymore.

Matthew Holt, THCB

I think that’s great, and that seems fairly obvious, right? I get a lot of text messages. Most of the text messages I get are saying, “Please donate to some political campaign,” because of the time of year we’re in at the moment, coming up to the midterms. They’re obviously trying lots of different things, trying, “please just give a dollar.” It’s a standardized behavioral change technique that’s been around for a long time, but there’s a lot going on behind the scenes here, right? Obviously you’re changing the message based on this. For something that’s relatively simple, like booking an appointment for something you should have once a year, what’s the level of complexity going on, and how much did AI, and what we’ve learned in the last few years, add to this?

Amy Bucher, Chief Behavioral Officer, Lirio

The level of complexity, I think of that, first of all, in terms of the logic model that my team is building. In general, for most of these behaviors, we’re identifying 15 to 25 barriers that are broadly applicable. We’re looking at the literature, we’re doing some of our own research, but basically, if you look at a population of people, what’s the majority of reasons why they might not be doing this? Then we build out message assets, at least one message asset per each one of those barriers. So you’re looking at libraries that have potentially, I don’t know, five, six, 700 combinations just from those base pieces.

Matthew Holt, THCB

Even for what is a relatively, quote-unquote, basic transactional thing, like, can I get you to sign up for a moment? There’s a lot going on behind the scenes.

Amy Bucher, Chief Behavioral Officer, Lirio

Well, maybe not transactional so much, because we do sometimes nudge things like enrolling in the patient portal. Those don’t tend to have as many or as serious barriers as something like prostate cancer screening.

Matthew Holt, THCB

But it’s a one-time thing. And I know you have some other examples which are longer journeys.

Amy Bucher, Chief Behavioral Officer, Lirio

Well, and again, because it’s AI, it’s a learning technology, and the more data that’s available, the easier it is to learn. So we are best suited to behaviors that recur more often. That’s kind of why our capabilities are best with that whole intelligence-layer implementation instead of a bolt-on. Even if each behavior only happens once a year — colonoscopies once every ten years, if you’re lucky — if we’re nudging you about a couple of different things, we can learn more about you anyway.

Matthew Holt, THCB

Okay, I see. So perhaps you can show me one that’s a little bit different, where you’re talking more about a patient journey.

Amy Bucher, Chief Behavioral Officer, Lirio

Yeah, for sure. I’m going to show something we’ve built for the life sciences space, so it’s really focused on medication adherence. You’ll see that it appears there’s much more going on, but that’s because the intelligence view on the right is built out in much greater detail, showing some of the system inputs.

Matthew Holt, THCB

Who am I, looking at this? I’m probably somebody in a call center who’s helping do adherence management or something like that.

Amy Bucher, Chief Behavioral Officer, Lirio

Well, there’s a piece of it where yes, you would be that person. This is for a fictional specialty pharmaceutical product, and as you know, these are incredibly expensive medications with high consequence to non-adherence, so these organizations do have telephone support representatives for their patients. Part of what we built here is something that they can look at when they’re on the phone with the patient.

I went through Maria’s demo a little bit just so that I could populate this. What it’s doing is recapping what Maria has actually done in the system. She’s received her medication, she’s done one refill, but she hasn’t completed her labs, which is a risk for non-adherence. And it’s creating a non-adherence risk score for her. It’s using data that we were perhaps able to pull in, like her physician enrolled her in this program and provided information about her insurance and her health history. Then there’s third-party data available that we could use around things like social determinants of health. We derive data all the time from things like her zip code: we can tell income level, and how far she is from the nearest pharmacy, and things like that. All of that goes into the non-adherence risk score. The engagement score is based on her interactions with the digital product.

The phone rep could be looking at this and saying, “Hey, Maria’s at really high non-adherence risk and she’s only so-so engaging with digital. This is somebody who it’s worth outreaching to by phone, or trying to get involved in something that’s a little more high-touch.” So it tells the life sciences company where they might want to invest the patient support money they have. And on the left, it has some feedback that the patient support rep could be using on the phone with the person. A self-efficacy phrase that resonates with Maria might be, “It takes effort some days.” She’s told us in her interactions with the system that this isn’t easy, which we know, of course, but now we know that this language is going to resonate back with her. We want to make the intelligence our system is gathering human-readable for the people who are supporting the patients in the real world.

Matthew Holt, THCB

So these data and these scorings — there’s a combination of AI-based inference from data you’ve got and data surrounding the person, and then some direct connections back and forth, some notes taken, kind of like a medical record, from the conversation. And what’s the messaging look like? Can we go back to the message she received?

Amy Bucher, Chief Behavioral Officer, Lirio

Yeah, absolutely. Again, this is for life sciences, so you would be seeing this inside an authenticated experience, and we want to really limit the self-report here, because adverse event reporting is a concern in this area. Here it’s letting her know she has to have a blood test in about two weeks, and it’s asking her to let us know: did you schedule this or not yet? She can say “not yet.” It also has celebratory sorts of things: your first month on Clarenza is almost done, that’s a real milestone. And we give her the opportunity to ask to connect with the pharmacy if she wants, so she can schedule a callback. I’m not going to click that, because it works.

[laughter]

Down here, this is just for operating-the-demo purposes, to say whether she received the refill or not. In the real world, we’d be receiving this data from a database that we connect to. As much as possible, we want to be getting that sort of objective, real-world data. There are cases where we do ask people to self-report, because how motivated are you feeling is not something you can really get from anything except self-report. Here it’s asking an adherence question: how many days out of the last seven were you able to take it as prescribed? I’m going to have her say “most days.” It gives her a little bit of feedback, but now this is something the phone rep can use as well, to say, “Hey, I’ve noticed you said you’re not doing this every single day. Can we work on that? Can we talk about that?”

Matthew Holt, THCB

So in general, you’re essentially combining vast amounts of data from different places: some inference, some direct, some self-reported, some from, did a refill get issued, that kind of stuff. And obviously there have been a lot of complications about getting this data right, but there’s much more of this data around now than there ever was before. Then you’re essentially giving informational support to a phone rep or somebody who’s working directly with the patient, and obviously motivational support to the patient.

What’s the most important part of this? What moves the needle on patient behavior the most? We hear that some people like human contact, some people prefer getting texts or messages. If you’ve got something that’s complex — we hear these terrible numbers that 50% of people on chemotherapy don’t finish their courses, even though it doesn’t work if they don’t, and it costs the pharma company a fortune in lost revenue — what is working here? What’s moving the needle?

Amy Bucher, Chief Behavioral Officer, Lirio

I think part of it is actually addressing the barriers that people experience in a direct way. You don’t see that a lot in healthcare communications. When you do see healthcare communications that try to address barriers, they don’t select what might be most applicable to an individual. I have some provider organizations that I’ve received care at, and I get these emails from them that are literally two, three pages long. I have a high need for information and I’m willing to process things, and even I’m like, ugh, I can’t do it. Your average person is not going to read through all of that. But we’re sending very short messages that directly address something we believe is relevant to that person, and I do think that makes a difference, that it doesn’t feel generic.

We do a lot of user testing, and we consistently hear that people feel like, “This cares about me. This seems to think about me as a person.” So it’s giving them something a little different from what they’re used to getting from healthcare organizations. I think that cuts through the noise and helps it stand out. And I do think that’s probably one of the most important things, because if you don’t get people’s attention, it doesn’t really matter what else you put into the message. It’s not going to do anything.

Matthew Holt, THCB

For a program like that, where you’re going to an ABC pharma company and helping them with a high-cost drug that obviously has to be taken to work, what are the kind of numbers you’re seeing? Where were they, and where do they end up after the program?

Amy Bucher, Chief Behavioral Officer, Lirio

Well, we just submitted a paper about one program that we’re doing with a health system. They were sending out an email to people around their birthdays that was, welcome to your new year, here’s the preventive healthcare you’re going to need in the next year. We took that on starting January 1st and added our personalization and behavioral science to what they were doing, and only nine months in, we are at 2.5 times the number of preventive care appointments and screenings that people have had compared to last year. So it’s our nine months over their twelve months, and it’s 2.5x.

Matthew Holt, THCB

Has this gone from 30% to 70%, or has it gone from three to six?

[laughter]

Amy Bucher, Chief Behavioral Officer, Lirio

More like the former. Part of it, too, is that these are not necessarily people who weren’t getting the preventive care, because the way we’re measuring it is that we have to have sent them the message, and then they have to have scheduled within a time window. I think we looked at 30 days after the message. So we’re not looking at the full year, like, did they ever come in and get their thing. But it wasn’t a super sad-sack case. With vaccinations, we’re not going from zero to two.

Matthew Holt, THCB

No, this is more like 30 to 60 type of deal. So what you’re saying is that this behavioral science-based approach, plus AI, plus what you guys are doing, is not only moving the needle but moving it on a large swath of people, which is the point, right?

Amy Bucher, Chief Behavioral Officer, Lirio

Yeah. And the one I’m really excited about — I keep coming back to vaccinations and our pharmacy partner, because we’ve been working with them now for four years — when you look at the year-over-year, that’s where you really see the effect. We’re nudging all their vaccines, and every year up until this year, we’ve done a treatment versus control, where we use their old standard messaging on a subset of the patients and measure. In the fall season — flu, COVID, vaccination season — we doubled their control messaging. Then starting January 1st, we released everybody back into Lirio messaging, and we were able to convert 30% of those folks who hadn’t gotten vaccinated yet in an average of 1.1 messages. The year following, I think it was 40% of the people who got vaccinated did so within the first two messages we sent. Our personalization is getting really, really good, because we’re just so knowledgeable about the population.

Matthew Holt, THCB

I’m not sure if Walgreens has managed to convince me to get the RSV vaccine yet.

[laughter]

But it works on the COVID and the flu shot. So I think you have one more demo to show me, which is around a wider preventive journey.

Amy Bucher, Chief Behavioral Officer, Lirio

Yeah, I can quickly show you this as well. This showcases some of our conversational AI capabilities, which you haven’t seen in the others as much. Conversational AI is increasingly the way people interact with these tools, and it’s something we have in our toolset. When I say conversational AI, that’s both chatbot and voice. We’re doing all the automated voice communications as well for our pharmacy partner — I can’t say who they are, but if you ever call them, hopefully — I’m sure there’s a pharmacy chain in America somewhere that people have heard of; this is the one.

Matthew Holt, THCB

Yes. It could be one that’s down the street from you. If you see one with really good communications, that’s the one.

Amy Bucher, Chief Behavioral Officer, Lirio

If you like the communications, it’s the one that we designed. Great. This one’s designed for the health insurance space. Health insurance companies don’t always know a lot about their members. They want to keep people healthy, and they compete every year, since people usually re-enroll in their insurance plans, so health insurance companies want to differentiate themselves in terms of offering good service and being the one that people choose. This is an onboarding experience we’ve built to help with that.

I’m going to start here with Sarah, who is a brand-new member. Again, on the right, you’re going to see the visualization of the intelligence behind the scenes, and the left is the patient experience. Sarah is getting a text message from her brand-new health plan, letting her know she has something called the Health Empowerment Index available to her. We’ve built this to be bite-sized interactions with the system that give us some information that allows for future personalization and for recommending preventive care activities, which, of course, people like. It’s a value that your health plan can add for you. The health plan likes it because if people can stay healthier, they’re lower-cost members over time. If Sarah were to click on this, it’s an email letting her know about this new experience. If she gets started, it will populate her in here. It’s a very short, conversational experience, I think three questions. It’s using some behavioral science and asking her to self-report some things that we would not otherwise know about her, which will then let her health plan communicate with her in a more personalized and meaningful way.

It starts with stages of change: what feels most true about your health right now? She’s going to say that she’s working on some changes right now. Over here, her behavioral profile is starting to populate based on what she says. And then in terms of the insights, I like this bit: she’s already motivated, so we don’t need to educate her about why change is important or try to put a carrot in front of her. We just need to help her do what she’s already decided to do. So those are the implications for further messaging. She gets a little congrats for that, so we can point her toward the right resources, and she gets a few suggestions: members like you ask. One thing I mentioned before is that we receive data from our clients, and part of what we can use that for is to see who people are like, sort of like Amazon does.

Here she’s going to say, “I want to lose weight.” She’s going to type in something different, and it will ask her, hey, that’s great, but there are two different pillars of lifestyle medicine that connect to that. Do you want to work on eating better or moving more? I’ll have her start with eating better. Behind the scenes, there’s a database we’ve built of the different benefits the health plan offers, and it’s going to curate suggestions for her. It will only offer two choices. Behind the scenes, I think there are seven or eight per pillar of lifestyle medicine, and they vary in terms of intensity, the commitment level required, whether they’re virtual or in person, all that good stuff. Here she’s offered either MyFitnessPal or a telehealth visit. She can select one of those, and it’s going to urge her to sign up for the app, which, let’s pretend she did that. Then it will drop her into this Health Empowerment Index experience.

The intention is that the health plan can use this to reach out over time. Most health plans don’t really have a good, meaningful reason to reach out to their members and interact with them beyond an explanation of benefits or, you know, we’re denying your care. So this is something that’s hopefully a little more upbeat than that, and lets people reach out and have a productive conversation. If for some reason Sarah wanted to self-report her way through the other pillars of lifestyle medicine, she could do that here. We’re not going to hold somebody back if they want more recommendations, and she can also see all of the different activities available to her. So again, it’s a way to collect some of that behavioral information that might not otherwise be available, connect people with preventive care resources, and hopefully get them healthier and differentiate the health plan from their competition.

Matthew Holt, THCB

In this case, are you actually taking over almost all of the health plan communications? How much of the website and the app are you in, in these cases?

Amy Bucher, Chief Behavioral Officer, Lirio

We don’t typically run the website or the app ourselves. This usually comes from working with the client to advise on changes they’re making to those properties. We can also use our intelligence layer inside of those types of assets, but typically they would own the presentation layer of that, and of course they’re also the ones offering the engagement endpoints. We’re not doing, like, the MyFitnessPal app; they need to have that available. I sometimes say Lirio’s like connective tissue. Those pieces need to be there, but if they’re there, we can do all the amazing routing in between and get people to the right destination.

Matthew Holt, THCB

There are obviously a number of companies doing patient engagement websites in general for those companies, and obviously a ton of apps they’re buying. I was chatting with a friend the other day about what Cigna was currently offering me, which I’ve now remembered: I think seven, eight, or nine different mental health companies, and six or seven different substance use disorder companies. That’s presumably very personalized to me, I would assume, because that’s what they think I need.

[laughter]

But there’s a lot going on there, and it’s tended to be a very messy experience for people, because they’ve Frankensteined these websites and applications over the years. Helping just get people to the right place and pointing people around is probably a very valuable service, for you to be the connective tissue doing that.

Amy Bucher, Chief Behavioral Officer, Lirio

Yeah. Just in talking with some health plans, they’ll talk in terms of offers: “We have these offers for our members,” and they’ll say we have 150 offers for any given member, and it’s almost like internal politics which three or four get highlighted on each person’s dashboard. There’s a better way. I understand that every part of the business wants their initiative highlighted, and we all think all these things are important, but at the same time, it’s about the patient or member, and what they need, and what they’re going to do. We’ll be more effective as a system if we’re getting people to that point.

Matthew Holt, THCB

What you’re saying is that companies like Lirio, and the science you represent, are over time helping figure out the mess of applications, data, and stuff that has been launched at health plans, providers, and everyone else in the business system: how do you sort through the mess and make it more impactful, and get us higher up the scale of helping people? Are there particular types of people for whom it’s easier? We talked about the persona of the woman out for drinks with her besties in the video — this is the early part of the night, before it gets to the substance use disorder later — but you also mentioned Black populations in the South dealing with hypertension. We’ve heard a lot about selection bias in the early days of a lot of these tools. We had a lot about that at Health 2.0: that’s great, you’ve convinced the person who was already the triathlete, eating well, to go on your program, but the person who was pre-diabetic, sitting around watching video games, eating chips, you couldn’t reach. Is there a population that’s harder to reach, easier to reach? And are there things we can do for those harder-to-reach populations, who are the ones at risk of getting the chronic disease or who already have it?

Amy Bucher, Chief Behavioral Officer, Lirio

Well, by accident, we sort of specialize in the harder to reach, because we’re a newer company, and a lot of our early clients were like, we’ll give you a try, but we’re going to give you this population.

Matthew Holt, THCB

This horrible thing to do.

Amy Bucher, Chief Behavioral Officer, Lirio

Yes. One of our very early case studies was a population of about 6,000 people with diabetes who had not been seen by the health system in three years. These were value-based contracts, so big deal. We got 60% of them, over 18 months, to come in for at least one provider visit. We do breakdowns of the populations that we nudge around things like what type of insurance they have, and some of our populations are 80% Medicare and Medicaid: older, lower income, higher disability. We look at similar things in terms of race and education level, and pretty consistently, the populations that we’re messaging in these sorts of scenarios are the more disadvantaged ones.

We do better when we can message an entire population that includes your triathletes and your A students, who are going to do it anyway. But we do really well with those populations who struggle, and again, I think a lot of it is that we’re approaching them differently. We’re really trying to put empathy in the middle of the way we communicate. We’re trying to be personal. We’re really trying to say, hey, we know this is hard for you. We’re not assuming that you should just do this thing; you have a real reason why you haven’t taken action, and we want to help you with that reason.

Matthew Holt, THCB

It strikes me that what you’re trying to do is something we’ve failed miserably at in healthcare. My favorite one-liner in healthcare ever was from an expert workshop I ran in the 1990s, when I was at the Institute for the Future. A guy called John Mattison, one of the smartest guys I ever met, who was the CMIO of Kaiser Southern California for years, said: the problem with healthcare is that we know what to do. We don’t know how to do it.

When I think about the examples you mentioned — hypertension, hard-to-reach Medicaid populations and so on — I love the example of a guy called Bob Matthews, who talks about the medical group he’s associated with, which is PriMed Physicians in Dayton, Ohio. Whenever I talk about this, I say to my audience: this is a medical group in a city you’ve never heard of, in a state you’ve never been to. I’m always talking to the Silicon Valley or New York crowd or whatever. They’ve got hypertension control rates in their Black population — majority-Black population clinics in Dayton, Ohio — above 90, control at over 95%. They’ve done it. They’ve figured out how to do it. Can we replicate that in the rest of the country? Apparently not.

So what you’re doing is sort of taking that and promoting it at scale, saying we’ve got all these organizations that have got access to people, whether it’s the pharmacy or the health plan or whatever, who’ve really not used it for good. There are a couple of health plans I could introduce you to that still need to figure out how they’re communicating with their population.

Amy Bucher, Chief Behavioral Officer, Lirio

Listen, I am currently covered by one of them.

[laughter]

There’s a lot of, can we get the average up a little bit. But there is great stuff going on out there. We have these amazing programs and amazing organizations that have figured it out, and we somehow haven’t used the ability to scale that.

Matthew Holt, THCB

I think that’s clearly where you are. So let’s end with hearing a bit more about Lirio, the company. How big are you? You raised some money a while back. How long have you been around? Give me a sense of the state of play: how many people are there, how many clients are you working with, where is Lirio?

Amy Bucher, Chief Behavioral Officer, Lirio

We call ourselves about a seven-year-old organization. If you look on Crunchbase or something, you’ll see Lirio extending back further, but we dedicated ourselves to healthcare, and most of the current leadership team started five to seven years ago. That’s the modern Lirio. We’re about 80 people full-time, and we also work with contractors and extended teams, especially as we do more international business. For example, I will always have at least one behavioral science contractor who’s on the ground in the client location, who works directly with my team to make sure we’re acculturating. But 80 full-time folks, based in Tennessee, with a hybrid workforce. I live in Boston, a lot of folks live remotely, and we travel and see each other either in the office or on client sites.

Matthew Holt, THCB

Give me a sense of how many clients and what the structure is. You mentioned pharma companies, pharmacies, hospitals, and health plans. Is it spread across the board?

Amy Bucher, Chief Behavioral Officer, Lirio

Health systems were our first clients, so I think we still have the most clients in that category, and they’re our longest-tenured clients as well, since we’re still working with some of our very first clients. But it’s increasingly becoming the life sciences companies and health plans, and then also international. We’re working in Japan, we’re working in Costa Rica, and I just got back from the UK, where we’re doing some work. Those organizations maybe don’t fit as neatly into the US categories, but they’re becoming a bigger part of our business.

Matthew Holt, THCB

And then finally, we’ve mentioned AI a lot, but we haven’t mentioned the big AI companies, who seem to be very much in the behavior change space in some ways. People are using ChatGPT as a therapist, people are getting Claude to design their life for them. It’s certainly happening in my household: my daughter’s running both her school projects and, when she wants to fix something in the garden, she runs it through these things. What is the role of behavior change from healthcare organizations using someone like Lirio to message, when you’ve got these brand-new but heavily used consumer tools, which are all kinds of other influencing and informational supply, and now the Google search has become the Gemini answer? As a behavioral scientist, what do you think is going on there, and how are these two things potentially being integrated?

Amy Bucher, Chief Behavioral Officer, Lirio

Well, we can work on directly integrating them. When we do conversational AI, for example, we don’t have an LLM at Lirio. We sometimes call our intelligence a large behavior model, but we can integrate with the sorts of things our clients are using. But I think the question you’re asking is really: when you have a patient who’s using these tools, what is the role of the healthcare organization? I think there still is one. We are consistently finding in our research with patients that — maybe not so much the organization, especially in the US; we’re seeing things like trust in the CDC really fall. We used to be able to say, “The CDC recommends,” and we’ve moved completely away from that in the last two years. But people do still trust their doctors, they still trust their pharmacists, and I think they’re probably trying to cross-reference what they’re learning from a Claude or a ChatGPT with what they’re hearing from these individuals, or the organizations that employ them.

So there’s something about us making the personal outreach and saying, “Hey, that thing you’re talking to Claude about is relevant to you when it comes to your pharmacy, and we’re here to provide that service.” It’s that final mile of what you’re talking to the AI about: we’re here to do that.

Matthew Holt, THCB

Do you think, as scientists, that the direct connection to the big LLMs is helpful in general, or is it more misleading?

Amy Bucher, Chief Behavioral Officer, Lirio

When I say we’re doing that, I’m talking about, for example, if we’re working with a client that has a chatbot that’s powered by ChatGPT, we can be behind the scenes in that instance, basically providing structure to the responses that happen. I hope it’s helpful. We’re doing it because we think it’s helpful. In that case, that chatbot, although it’s using the general LLM, is designed for a specific purpose that is aligned with what we do, so it’s trying to get people to take some sort of action. This is something we’re doing in Costa Rica, where their health insurance system is extremely different from ours. We’re trying to get people connected with the right wellness benefits for them in the system they’re in, and the chatbot is designed to understand what they care about in terms of their health, what their health status is, and based on that, where they’re going to get the most value for their money and the best care. The guardrails keep it to that sort of conversation. The person isn’t going to be able to ask about, “I feel like I’m having a heart attack.” And I think that helps, too.

Matthew Holt, THCB

And then finally, going back to where I started: we’ve been really bad at doing behavior change in healthcare, we have all these not-great outcomes, and we haven’t moved the needle on much of this stuff. How optimistic are you? Take off your Lirio hat for a second and put on your behavioral scientist hat. How optimistic are you that things are going to get better in terms of the big issues, chronic disease, lifestyle choices, diet, exercise, all that stuff? You look at America and go, it can’t be fixed in many different aspects. In that one, do you think this is fixable using the new tools and the clinics we have?

Amy Bucher, Chief Behavioral Officer, Lirio

I think the right ingredients are there. I think the system we’re in right now doesn’t have the right incentives in place for things to get better. I think we need to see some major structural change. I don’t know how into politics I want to get, but I am really optimistic about things like GLP-1s. We haven’t even seen the long-term health effects of those, but from the early results, it seems like they might contribute to longevity — healthier, not just living longer, but living longer at quality. That’s incredible. But right now, the economics of GLP-1s isn’t there for most people, and there’s a lot that would have to change in the system for that to change. So I’m optimistic that if we can kind of blow things up and rebuild, we could do better. But can we blow things up and rebuild? I don’t know.

Matthew Holt, THCB

You’re joining me on the barricade. Knock it all down and have concierge care for all. That’s my spiel. That’s a separate podcast I’m about to launch, which is not so much about companies, more about techniques in primary care. I think you know a lot, and there’s some interesting stuff being uncovered in some of the interviews I’ve been having: for instance, people now with AI managing patient populations, the doctor with the AI raising issues, the doctor coming in and having a three-way conversation between the patient, the AI, and the doctor. There are some interesting ways we can think about doing this. But yes, in general, we need to blow up how we do healthcare delivery, and by the way, blow up how we do food and a bunch of other stuff, which probably won’t happen, because things are difficult to change. But there we go.

[laughter]

Amy Bucher, Chief Behavioral Officer, Lirio

You never know. Sometimes things do change. That’s why we all do what we do, too. We’re trying to change stuff, and hopefully our little work on the margins will have a bigger effect.

Matthew Holt, THCB

Which is why I’ve been asking you about this great stuff. All right, I’ve been talking with Amy Bucher. She is the Chief Behavioral Officer — did I get your title right?

Amy Bucher, Chief Behavioral Officer, Lirio

That’s correct. Chief Behavioral Officer of Lirio.

Matthew Holt, THCB

A company which is doing a lot of interesting work, as you’ve seen, in the behavior change space for healthcare organizations. Amy, thank you very much for your time.

Amy Bucher, Chief Behavioral Officer, Lirio

Thanks, Matthew. Nice talking to you.

By MANUEL PINILLA

Every few months a new study reports that an AI model has outperformed doctors on a set of diagnostic cases. The latest one I’ve read came out in Science in October: Microsoft’s MAI-DxO, tested on 304 complex cases from the New England Journal of Medicine and run as step-by-step simulations, reached the correct diagnosis in about 83% of them. Primary care doctors and general internists, working on their own with reference materials, got 36%. Microsoft says the next step is a study with real patients.

I build clinical AI, so these results matter to my work, but they answer a different question from mine. In the setting we work on, emergency departments and primary care, a doctor will be in the room, and that doctor signs the diagnosis and answers for it. While that is the case, a score of AI against doctors tells me little about whether a tool helps. What I need to know is whether the doctor, with the AI, gets the diagnosis right more often, and what happens when the AI is wrong.

There is a second gap. A published case arrives already written up: the history is complete, the relevant findings are there and the question is clear. A real patient tells the story out of order, leaves out what they don’t think matters and sometimes can’t put into words what they feel. Getting from that to a good differential is a large part of a doctor’s job, and a benchmark mostly skips it. How a tool behaves in a specific setting, with the information that setting really provides, matters as much as what the model knows.

Where the AI shows up in the consultation

I’m an engineer, not a physician, and when I think about these tools my first concern is the interaction more than the model. To use a chatbot during a consultation, a doctor has to stop, summarize the case, write a question, read the answer and judge how far to trust it before turning back to the patient. In a ten-minute primary care visit or a crowded emergency department there is rarely time for that, so in practice the tool tends to be consulted afterwards, to check a diagnosis that is already more or less settled. By then the patient has left, and with them the chance to ask a different question or examine something that was missed.

That is the problem we set out to work on at Medsys AI, a startup in Madrid. Our assistant listens to the consultation and, as it goes on, suggests diagnostic hypotheses, questions worth asking and things to check in the physical examination. The suggestions are updated as new information comes up, and the doctor doesn’t have to type anything. What to do with them is the doctor’s call. The picture I had in mind was an attending standing behind a resident who is unsure, available to doctors at any stage of their career.

A couple of design decisions follow from that. The assistant shows several hypotheses with the reasoning behind each, drawn from more than a thousand clinical guidelines and checked by more than one model before they reach the screen. And the system is built so that we can change the underlying model and rerun our evaluations when a better one appears.

Google and Beth Israel Deaconess published a study in The Lancet in which about a hundred patients with an upcoming primary care visit for a new problem chatted with Google’s AMIE beforehand, with a physician supervising every conversation live. The doctors received a transcript and a summary before seeing the patient, and none of the conversations had to be stopped for safety. The physicians said it moved the visit from collecting information towards checking it and deciding with the patient. It is a different design from ours, before the consultation instead of during it, and I think it is a good example of how to do this: the AI does part of the work and the doctor keeps the decision.

What we measured

We evaluated our assistant with doctors using it. Thirteen board-certified family physicians in Spain conducted 260 voice consultations with virtual patients, played by a voice model held to the details of each case. There were 40 cases, chosen for their diagnostic difficulty, such as a pulmonary embolism or an autoimmune disease presenting with vague symptoms. Each case was randomly assigned to be done with or without the assistant, and three adjudicators who didn’t know which arm a consultation came from scored the diagnoses. The results were published in JMIR Formative Research.

The correct diagnosis was among the doctor’s top three in 62.3% of unassisted consultations and in 74.6% of assisted ones, a difference of 12.3 percentage points. Counted as misses, that is 32.6% fewer, and 44.8% fewer in the hardest cases. Physicians rated the tool 4.38 out of 5 for usefulness.

The study has clear limits. The patients were simulated, thirteen doctors is a small sample, and in the unassisted arm physicians had no external resources (no UpToDate, no search engine), which is not how they normally work. These are preclinical results. We are now running a clinical trial with real patients in the emergency department of a major hospital in Madrid, authorized by a research ethics committee and by the Spanish Agency of Medicines and Medical Devices.

Two (out of the, probably, many) questions we haven’t answered

The first is time. Assisted consultations took about 35 seconds longer, roughly 10.7%. In an emergency department with a full waiting room, 35 seconds per patient adds up over a shift. We don’t know yet where the acceptable limit is, or how much of that time goes into questions that change the diagnosis and how much into reading suggestions that don’t.

The second is overreliance. We also looked at what happened in the consultations where the AI’s diagnostic suggestion was wrong. The system generated suggestions in both arms, but only the assisted physicians could see them. In those consultations with wrong suggestions, assisted physicians submitted the same wrong diagnosis 55.8% of the time, compared with 38.8% for unassisted physicians, who never saw it. The difference was not statistically significant, so the study does not show that the assistant pulled doctors toward wrong answers. It doesn’t rule it out either, and with a gap of that size in a small study, I read it as something to follow closely. Showing several hypotheses with their reasoning doesn’t by itself stop a doctor from anchoring on the first one. The clinical trial measures this again, with real patients.

Why diagnosis

Most of the AI investment in healthcare over the last two years has gone into ambient scribes. They give doctors time back, and an imperfect note carries relatively little risk, so that makes sense. We see our work as complementary: a scribe writes up the conversation, and we try to help while it is still happening.

Diagnostic error is where a large share of the harm is. AHRQ’s 2022 review estimated that about 1 in 18 emergency department patients in the US is misdiagnosed, around 7.4 million people a year. Emergency medicine societies disputed how that figure was extrapolated, and the real number may well be lower. Even a fraction of it is a large problem, and documentation tools do little about it.

The regulatory path

In Europe, software that contributes to a diagnosis is a medical device, and not in the lowest risk class. To get CE marking we need a technical file covering benefits and risks, assessed by a notified body. For us that realistically means some time in 2027. It is also the process that has us running a trial with real patients and reporting risk measures like the one above.

What I’d like to see measured

For the kind of system we are building, standalone model performance is not enough to know whether it helps. I’d like to see more studies, ours included, that measure whether doctors using the AI are more accurate, whether they are safer when the AI is wrong, and how much time it costs them, first in simulation and then with real patients.

Even that is a step on the way. A correct diagnosis matters because of what it does for the patient, so the measure that counts in the end is how patients do afterwards: over the following days in an emergency department, over months or years in primary care, depending on the specialty. We keep that in mind when we design the assistant and its evaluations. We have a first piece of evidence that this approach can work, and a clinical trial underway to see whether it holds up. I’ll share what we find.

Manuel Pinilla is co-founder of Medsys AI, a Madrid-based startup building real-time AI support for physicians during the consultation, and a co-author of the study. The paper with the preclinical results.

By MIKE MAGEE

As we entered the 2026 Mid-terms, twenty physicians held seats in Congress, four in the Senate and 16 in the House, a record for physicians in the highest offices of the land. (I provide a list below.)

In the wake of Tuesday’s remarkable display of democracy, many have shined a light on several new elected physicians, including Sen. Abdul El-Sayed (D-Mich.), Sen. Annie Andrews (D-S.C.), and notably the election of a woman physician as governor of Ohio, Gov. Amy Acton (D-Ohio).

They will join, come January, two standing physician senators – John Barrasso (R-WY) and Rand Paul (R-KY), and leave behind two other Republican Senators, Bill Cassidy (R-LA) and Roger Marshall (R-KS), both mired in controversy.

Physician leaders. Is this a good thing for our nation, struggling with an inequitable and prohibitively expensive health care system, a frayed and collapsing public health infrastructure, and a reckless vaccine policy launched by a public official whose own family members describe him as, well, “crazy.”

My short answer is “On the whole with notable exceptions, yes.” In search of an answer why, I go back to a speech I gave, now two decades ago, on March 12, 2006 at the AMA President’s Forum in Washington, D.C.

In that speech to the national leadership of the AMA’s “House of Medicine,” I provided three reasons for my faith in entrusting physician leaders with the heavy mantle of leadership:

“The world is a better place today for having developed and supported strong, enduring patient-physician relationships. This is not simply a function of the nuts and bolts health care that this relationship provides, in full vision day in and day out. But also for those things it has reliably delivered over many decades around the world, quietly below the radar screen.

1. The first of these, as a collective, is the management and processing of a population’s fear and worry, which, absent our efforts, would accumulate to destabilizing effect.

2. The second is a subtle and rational reinforcement of important societal bonds as an outflow of personalization of care. The importance of this second contribution was best highlighted by the great Cardinal Bernardin, who shortly before he died said to several thousand AMA physicians, ‘There are four words in the English language that have common English roots. They are Heal, Health, Whole and Holy. I tell you this today to remind each of you that to heal in a modern world you must provide health; and to provide health you must keep the individual, the family, the community, and society whole. And if you can do all that, why that is a holy thing.’

3. The third mostly invisible contribution that the patient-physician relationship makes to society is hopefulness; a quiet confidence that most problems can be managed, most mysteries solved, and that, absent a cure today, there remains hope for a discovery tomorrow. That deep seated optimism, embodied in an individual whom you trust, have confidence in, and believe in, is transferable.

As a collective each day, and every day, we help create a reservoir of good will, a belief that life is worth living, that risks are worth taking, that loving and being loved is possible, that dreaming, imagining, purchasing a home, getting married, having children are all worth the effort.

So it is my firm belief that if tomorrow all patient-physician relationships were to disappear, stable civil societies would immediately notice the difference. Our populations would be more fearful, less trusting, less tolerant, less connected, less compassionate, less productive, and less committed to the future.”

Two decades later, as I approach 80, do I still believe these words? On the whole, the answer is “Yes.” But high expectations leave a steep decline emotionally when our fellow human physicians disappoint.

Back then, I labelled this minority of physicians as “negative leaders”:

“Negative leaders view change with fear and leverage that fear as a currency to control a population in order to reinforce existing and past power silos. They attempt to segregate us one from another to maintain the status quo, and they deserve our contempt. They not only draw down the reservoir of good will locally and globally, but ensure our medium and long term failure. While their highly effective tactics can lock a quivering public in check, they cannot stop the world from changing around them. And so for negative leaders, success is short term, for they are rapidly overtaken by each other, and by environmental events.”

With youthful optimism back then, I offered this alternative. Choose “positive leaders” to lead. I said that day:

“Contrast these with positive leaders for whom visioning, rather than fear, is the currency. Where one excludes, the other includes. Where one reinforces the status quo, the other makes constant incremental adjustments. Where one plays by the old rules, the other makes the new rules. Where one segregates, the other congregates. Where one reinforces silos, the other builds islands of common stewardship.”

I am encouraged today, now 20 years later, with your newly elected physician leaders – Abdul El-Sayed, Annie Andrews (D-S.C.), and Amy Acton (D-Ohio).

Here is a current list of physicians serving in the U.S. Senate and House of Representatives:

Senate:

John Barrasso (R) WY Orthopedic surgery
Bill Cassidy (R) LA Gastroenterology/hepatology
Roger Marshall (R) KS Obstetrics/gynecology
Rand Paul (R) KY Ophthalmology

House of Representatives:

Ami Bera (D) CA Internal medicine
Herb Conaway (D) NJ Internal Medicine
Scott DesJarlais (R) TN General/family medicine
Maxine Dexter (D) OR Pulmonary/critical care
Neal Dunn (R) FL Urology
Andy Harris (R) MD Anesthesiology
Ronny Jackson (R) TX Emergency medicine
John Joyce (R) PA Dermatology
Mike Kennedy (R) UT Family medicine
Rich McCormick (R) GA Emergency medicine
Mariannette Miller-Meeks (R) IA Ophthalmology
Kelly Morrison (D) MN Obstetrics/gynecology
Greg Murphy (R) NC Urology
Bob Onder (R) MO Allergy/immunology
Raul Ruiz (D) CA Emergency Medicine
Kim Schrier (D) WA Pediatrics

Here is the list of physicians who ran for office and the results:

Rich McCormick (R-GA) (+)
Mariannette Miller-Meeks (R-IA) (-)
Roger Marshall (R-KS) (-)
Andy Harris (R-MD) (+)
Bob Onder (R-MO) (+)
Greg Murphy (R-NC) (?)
John Joyce (R-PA) (+)
Scott DesJarlais (R-TN) (?)
Ronny Jackson (R-TX) (+)
Mike Kennedy (R-UT) (+)
Ami Bera (D-CA) (+)
Raul Ruiz (D-CA) (+)
Kelly Morrison (D-MN) (+)
Herb Conaway (D-NJ) (+)
Maxine Dexter (D-OR) (+)
Kim Schrier (D-WA) (+)
Amish Shah (D-AZ) (+)
Richard Pan (D-CA) (+)
Abdul El-Sayed (D-MI) (+)
Adam Hamawy (D-NJ) (+)
Annie Andrews (D-SC) (+)
John Cowan (R-GA) (+)
Ralph Alvarado (R-KY) (-)

Mike Magee MD is a Medical Historian and regular contributor to THCB. He is the author of CODE BLUE: Inside America’s Medical Industrial Complex. Grove/2020

By HANA KAHLEOVA

Tuesday October 20th was World Osteoporosis Day, and one number deserves more attention than it gets: one in two women over 50 will break a bone because of osteoporosis. Bone loss speeds up as estrogen falls during menopause. Yet in the rush to treat midlife weight gain with the newest drugs, bones rarely come up.

I think of a patient who sat down and named the drug before she said anything else. She was in her early 50s, and she wanted a GLP-1, the weekly injection everyone seems to be on now. She had gained weight around her middle that nothing would touch, and she had not slept through the night in months. Two or three times a night she woke up soaked, kicked off the covers and lay awake. Two of her friends had started the drug that spring. She expected a prescription, not a conversation.

She is a composite of women I have treated, but her request is common. About one in five adults ages 50 to 64 have taken a GLP-1 drug, the highest share of any age group. Many are women looking for relief from the weight gain and broken sleep of the menopause transition.

They have good reason to want something that works. For generations, women’s menopause symptoms were minimized or left for them to endure. A drug that finally does something can feel like overdue respect. In a 15-minute visit, it is also the fastest thing a doctor can offer. That is not an unreasonable choice. The question is whether it should be the first one.

What changed my answer for this patient was a line already in her chart. A bone density scan the year before had shown osteopenia, the thinning that comes before osteoporosis. It had earned a note to keep an eye on it and then dropped out of the conversation, as these findings so often do.

Here is why it mattered. The weight people lose on GLP-1 drugs is not all fat. Across clinical trials of these medications, roughly a quarter of the weight lost was lean mass. In one retrospective study of patients taking semaglutide or tirzepatide, bone density at the hip fell by nearly 3% over about three years. For a woman already losing bone as estrogen declines, rapid weight loss lands on top of an existing deficit. And about half of patients stop semaglutide within a year, often regaining weight, so some pay that cost without keeping the benefit.

So I asked her a simple question: Of everything bothering you, what is the worst part? It was not the weight. It was the nights. She had stopped making morning plans because she never knew what kind of night she would have.

That answer gave me something better to offer. In a randomized trial my colleagues and I published in the journal Menopause, 84 postmenopausal women with frequent hot flashes either switched to a low-fat, plant-based diet with half a cup of cooked soybeans a day or made no changes. After 12 weeks, moderate-to-severe hot flashes fell 88% in the diet group, compared with 34% in the control group. Half the women who changed their diet had no moderate-to-severe hot flashes left at all. Many also lost weight without counting calories.

I told her all of it, including the caveat that results vary and the only way to know is to try. I suggested 10 weeks, with the GLP-1 still on the table if nothing changed. She was skeptical. She had tried eating better before. I explained that this was not vague advice. It was a specific swap: less meat and dairy, more beans and lentils, and a daily serving of soybeans, tofu or soy milk, backed by a trial that measured exactly the symptom she wanted fixed.

She came back at 10 weeks. The night sweats had not vanished, but they had gone from most nights to about twice a week, and they were mild enough that she usually fell back asleep within minutes. She had lost a little weight. She had started making weekend plans again.

GLP-1 drugs are a real advance, and I prescribe them. For many women, they are the right choice. But our health care system is built to make the prescription the default. Short visits reward it. Insurers pay for it more readily than for the time it takes to explore other options. Bone density results get filed and forgotten. None of that serves women in midlife well.

Women can push back. Before starting any weight-loss drug, ask what your last bone density scan showed, or whether you need one. Ask what else could address the symptom that bothers you most. Doctors can start by asking what is actually driving the visit. And insurers should cover nutrition counseling and follow-up as readily as they cover the drug, so the choice is real.

Now, when a patient asks me for a GLP-1, I ask first what is keeping her up at night. Women in midlife have waited long enough to be asked.

Hana Kahleova, MD, PhD, is a physician board-certified in internal medicine, endocrinology and lifestyle medicine, and director of clinical research at the Physicians Committee for Responsible Medicine, a nonprofit that advocates plant-based nutrition and funded the hot flash trial described here.

By JOHN SAMARAS

Noom Med prices a GLP-1 plan at $249 a month. The last two words on the line matter.

“$129 To get started. $249/mo after the first 4-week supply (billed quarterly).”

That’s Noom’s pricing page on November 3, 2026. If a quarterly charge covers three months at $249, it comes to $747.

In a Buzz Health survey of 256 pharmacists in July, 68 percent reported more cash-pay GLP-1 sales over the past 12 months. Pay cash and your health plan may never see the purchase.

The $149 price that adds up to $298

Hims’ weight-loss page, read the same day, lists the Wegovy pill “From $149/mo.” A small mark leads to a footnote. That price “includes medication only,” and the required membership is “$39 for the first month, auto-renews at $149/month thereafter.”

At the lowest printed price, drug plus membership comes to $298 a month after month one, billed separately.

Regulators have noticed

On December 3, 2025, the FTC gave final approval to an order against the telehealth company NextMed, which it had alleged “sold weight-loss programs with undisclosed costs and membership commitments.” NextMed and its principals must pay $150,000.

In July 2026 the FTC, Utah and California sued Hims & Hers, alleging among other things that the company fails to “clearly and conspicuously inform consumers when their prescriptions will be refilled each month.” That case is pending.

Both cases came after the fact. The FTC’s fee rule, in effect since May 2025, reaches only live-event tickets and short-term lodging. California goes wider: since July 1, 2024, most businesses there can’t list a price that leaves out required fees, other than certain government taxes and shipping costs.

Put the billing terms beside the price

Wherever a buyer picks a cash GLP-1 plan, print the amount of each charge and how often it lands beside the monthly price, along with any recurring fee. A buyer could then work out the first 12 months from one page.

Noom would print the quarterly charge. Hims would show the membership beside the drug price.

The FTC could write that into its next order. An employer that reimburses cash purchases could ask for it.

Collective prints it. Collective sells compounded semaglutide, which is made by a pharmacy and is not FDA-approved. Its page, read November 3, prints “$59/month + $199 annual membership,” the same price at any dose.

The membership is required, starts with a 21-day free trial and is billed annually. A buyer can work out $907 for the first 12 months from that one line.

See the bill before your card does

Paying cash? Before you type in a card number, compare every cash GLP-1 program side by side on the GLP Chart homepage, then find the billing line on the program’s own page.

John Samaras is the founder of GLP Chart, which checks the published cash prices of the US GLP-1 telehealth programs it tracks every Monday and earns a commission when a reader signs up with some programs, including Collective.

By KIM BELLARD

I once again have to thank The New York Times for introducing me to a new concept. Sarah Kessler wrote about the rise of “legal engineering” – for which, it turns out, one doesn’t have to be either a lawyer or an engineer.

Their role is to help integrate A.I. into the practice of law, whether at law firms or companies. That sure sounds like a prudent, if possibly belated, idea – and one suspects we’re going to see such “engineers” in other fields. I certainly hope that healthcare is one of them, sooner rather than later.

Irene Liu, the executive director of the A.I. initiative at Stanford Law School, explained the job to Ms. Kessler thusly: “If you were to just put it into an equation, it’s law plus technology plus business process.” That Venn diagram gets pretty small very quickly.

Mary O’Carroll, the C.E.O. of LegalEng Consulting Group, also explained the role to Ms. Kessler: “You don’t necessarily need a giant pyramid of lawyers anymore. People who have this kind of skill set don’t have to have a J.D., and they don’t have to be software engineers, but there’s business process re-engineering, or this M.B.A. type skill, that I think is really necessary.”

Ms. Kessler notes that they are sometimes referred to as “knowledge and innovation counsels,” “innovation attorneys,” or “legal applications engineers,” which may illustrate that the role is both new and evolving.

If you’re considering why this role is important, note that the 2026 Legal Industry Report from 8am found that 69% of legal professionals report personally using Generative AI (GenAI) tools such as ChatGPT, Gemini, or Claude for work-related purposes, up from 31% a year earlier.

Ms. Liu’s Law AI Initiative joined Stanford Law School’s Executive Education to launch the Stanford Legal Engineering Academy in September, a two-week residential program for intensive training in AI strategy, implementation, governance, and organizational transformation, starting in the summer of 2027. Adam Sterling, associate dean of executive education and strategic partnerships, said the program was “designed to prepare participants to act as chief AI leaders within their organizations, whether or not they hold that formal title.”

“Using an AI tool is not the same as understanding how to redesign work around it,” said Julian Nyarko, professor of law and co-director of the Stanford Law School AI Initiative. “The leaders who will be most effective in this transition will be able to assess where the technology adds value, recognize its limits, and translate that understanding into new workflows, services, and organizational structures.”

Howard Glazer, a longtime partner at Ropes & Gray who became its head of practice A.I. about a year ago, told Ms. Kessler: “In the old days, there was technology and there was law. And as long as my email worked, I didn’t have to talk to people who dealt with the technology. But now we actually provide our service and provide our product using the tools.” More to the point, he added: “‘Legal engineer’ kind of makes it sound like someone made it up. But it’s not. It’s a role which the current environment just screams for.”

Legal tech firm Legora , which describes itself as “the agentic operating system for legal work, supporting lawyers in research, review, and drafting across complex matters,” claims to have some 200 legal engineers. “We’re hiring hundreds, and competitors are hiring them, and our clients are hiring them,” David Himmel, Legora’s head of legal engineering for North America, told Ms. Kessler. “It’s a true profession.”

She reports that legal AI firm Harvey, which describes itself as “the operating system for legal and professional services,” has similarly grown its team of legal engineers to over 200.

Legora profiled Ricardo Anaya, whom it described as its founding legal engineer for the U.S. He recalled: “By the time I came to Legora, I’d been the lawyer, been the client, and worked on the product side. That’s basically the job description for a legal engineer, though nobody called it that yet.”

Mr. Anaya has found that there is often a gap between introducing an AI tool and its use. “The gap wasn’t knowledge,” he said. “It was that we were teaching the tool, not their work…Most adoption problems are diagnosis problems, and diagnosis doesn’t happen over video. That’s why we go to them. Meeting lawyers where they are is a critical part of the job.”

If your industry has complicated workflows, you’re going to use AI. If it has lots of arcane knowledge, you’re going to use AI. And, of course, if people in it care about making money, you’re going to use AI. If your industry doesn’t yet have its version of a legal engineer to help master this transition, it will.

Two years ago I wrote about the need for more engineers in healthcare, noting that most people who present themselves as “healthcare engineers” are focused on designing, building, and operating healthcare facilities, although some felt it was (or should be) “involved in all aspects of the prevention, diagnosis, treatment, and management of illness, as well as the preservation and improvement of physical and mental health and well-being, through the services offered to humans by the medical and allied health professions.”

If they were doing all that, healthcare wouldn’t be quite the mess it is today, with all parties seeing the system as dysfunctional. Healthcare needs its version of legal engineers – perhaps clinicians, perhaps not, but certainly people who understand how clinicians practice, how the system “works,” and enough about AI to see where and how it can add value.

Again, that Venn diagram gets small quickly.

I’d note that the American Medical Association (AMA) has launched the AMA Center for Digital Health and AI, and named the well-respected Jennifer Goldsack to head it up. Still, it is under the AMA’s umbrella, and its own description says it “educates and supports physicians, and gives them a voice in how technology is used to enhance the patient and clinician experience,” so if we’re looking for radical system reforms or even putting patients first, we may need to look elsewhere.

I remember when healthcare first started discovering the internet and “eHealth” became a thing. Then mobile came along and “mHealth” was the catchphrase. Smart people pointed out we’d know when healthcare had truly adopted to the new technologies when things just became “health” again. Same with AI.

We need legal engineers and their equivalents now to help us transition to the AI world, but we’re going to get to the point when AI is so baked into everything that we won’t have distinct roles like them.

Kim is a former emarketing exec at a major Blues plan, editor of the late & lamented Tincture.io, and now regular THCB contributor

By MIKE MAGEE

This week, New York Times political columnist, Linda Kindler reflected on “How Silicon Valley Stole the Future.” Signaling that this wasn’t simply another article on the binary choices of an AI world – or as she put it “either a world in which A.I. has eliminated all jobs, unleashed bioengineered viruses and enabled catastrophic warfare or a utopia in which the technology gives people ‘greater power to live life as they choose.’”

Instead, the AI discussion steered away from Fellow Times columnist, Thomas Friedman’s scary Halloween AI piece the day before (“We Just Set Off the Second Big Bang”) and focused instead on the work of Jonathan White, the Professor of Politics at the London School of Economics and Political Science where he teaches on democracy, political thought and political theory.

Professor White’s message was a clear warning not to abdicate control over AI to the Tech bro’s, in that “the capacity to imagine alternatives that may be improbable and to pursue them against the odds” is a “democratic resource.” In Kindler’s words, White argues, “When the future increasingly seems like something to avoid rather than something to look forward to, the very foundations of democracy are weakened. The future has been the primary terrain upon which all politics are practiced since the French Revolution challenged the divine right of kings.”

And Americans have made it pretty clear over the past two years that they reject Kings in the White House. Last week’s “No Kings” rally at thousands of sites across the nation were in fact a recurring event. Back in June 19, 2025, the Guardian took notice commenting that “The scale of last weekend’s “No Kings” protests is now becoming clearer, with one estimate suggesting that Saturday was among the biggest ever single-day protests in US history.”

Americans apparently want control of their own future destinies, AI and otherwise. With 5 days remaining before the 2026 Mid-Term elections, the levels of participation by the electorate are projected to set a record nationwide at 54% (past record was 49% in 2022) with a projected 135 million of the eligible 250+ million casting votes.

More than 40 million votes have already been cast with 5 days remaining. The demographic breakdown shows a sharp increase in Hispanic participation which makes up 16% of the electorate. Black participation, now 12% of eligible voters, is also historically high. Whites are 61% of the voters and Asian Americans represent 5% of likely voters.

Current polling from The RealClearPolitics and CNN polls averaging for the generic congressional vote show Democrats leading Republicans by roughly 8 to 9 points (averaging around 52% to 43%).

According to one forecaster, DDHQ, “Forecasters such as Decision Desk HQ and prediction markets (like Kalshi) indicate a strong statistical probability—ranging from an 80% to over 90% chance—of Democrats winning back a majority in the 435-seat House, with current model projections anticipating a pickup reaching into the vicinity of 238 seats.”

As for the Senate: “Current forecasting models lean slightly toward Democrats retaining or narrowly securing a 51-seat majority, giving them roughly a 58% probability of controlling the chamber, though outcomes remain closely tied to a handful of toss-up states.”

The US polling data site has these additional insights:

“Three structural factors push 2026 turnout above the midterm baseline. First, Democratic base enthusiasm in opposition to Trump’s second term mirrors 2018 conditions. Early indicators — special election overperformance, record small-dollar donations, grassroots volunteer sign-ups — all suggest elevated Democratic motivation. This is the single biggest driver.”

“Second, the expansion of mail and early voting infrastructure since 2020 means the structural floor for participation is higher. States that added permanent mail voting options between 2020 and 2024 saw turnout increases averaging 4–6 points. These structural changes do not disappear between elections. Third, Gen Z voters (now 18–27 in 2026) are adding roughly 3–4 million new eligible voters per election cycle, and this cohort has shown higher midterm participation rates than prior generations at the same age.”

What Professor White would suggest is that the future is what is really on the ballot, in the middle of a massively disruptive technologic revolution and “generational social change.” Who gets to decide the future. In a democracy, you do. VOTE!

Mike Magee MD is a Medical Historian and regular contributor to THCB. He is the author of CODE BLUE: Inside America’s Medical Industrial Complex. (Grove/2020)

By KIM BELLARD

Let’s say your favorite social media platform is bought by a right wing trillionaire, who espouses free speech, but only when it echoes his. You can switch to, say, Threads, but – whoops – that’s owned by Mark Zuckerberg, another of the super-rich who have turned to the right. There’s always Facebook but – damn! – Zucked again! So you migrate to Bluesky, where the discourse is more civil but the audiences admittedly smaller. At least you have some choices.

Not so much if, say, you are the President of Ukraine. Ukraine has unexpectedly held off the much bigger Russia for over four years now, in the process reinventing modern warfare by turning it from the traditional men, tanks, and artillery to a technology-powered war, especially with drones. To do so, it relied heavily on Starlink and its Pentagon-version Starshield. Unfortunately, Ukraine was blindsided a couple weeks ago when Starlink speeds were drastically lowered. Drones were literally falling out of the sky.

President Zelensky raised the alarm. Secretary of State Rubio said he thought software “glitches” were to blame, but most people suspected either Starlink’s owner Elon Musk and/or President Trump. Mr. Musk pointed fingers at the Pentagon, President Trump blamed Mr. Musk, and for several days Ukraine was hampered. Finally Mr. Musk admitted he’d ordered the slowdown, saying Ukraine had violated the “terms of service” of the service, using it “for offensive drone strikes to kill people.”

Note that Serhii Beskrestnov, a Ukrainian presidential adviser on military technology, wrote on social media “we are PHYSICALLY UNABLE TO USE COMMERCIAL STARLINK for deep strikes or medium-range strikes.”

The “loophole was closed last week,” according to both Mr. Musk and President Zelensky, with normal speeds resumed, but if you were President Zelensky you’d sure have to be looking for alternatives, as you’d never know when the equally mercurial Mr. Musk or President Trump might pull another Lucy on you (Gen Z: check your “Peanuts” references).

Unfortunately, there really aren’t any, and that’s a problem.

One might expect that another one of the newly MAGA super-rich who is into space – e.g., Jeff Bezos – might have filled this space (pun intended), and you’d be partly right. Amazon has been working on Project Kuiper – now Project Leo – for years now, and now his space company Blue Origin has launched TeraWave, offering “space-based communications.” Unfortunately, despite all his and Amazon’s money, neither is ready for primetime; Leo has fewer than two hundred satellites in orbit, and TeraWave won’t even launch until late 2027. So President Zelensky better be looking elsewhere.

Accordingly, Ukraine’s Ministry of Defense signed a letter of intent with Finnish companies Nokia and ICEYE “to develop new defense and satellite communication capabilities.” President Zelensky acknowledged: “Modern warfare has proven that reliable communications are a matter of national security. We must all rely on capabilities that we can control.”

Rafal Modrzewski, CEO and Co-founder of ICEYE, agrees: “Ukraine needs secure, sovereign and high-performance satellite communications systems that it controls. We have a proven track record of delivering sovereign missions at speed to help build that capability.”

President Zelensky has also met with leaders in Norway to develop new satellite communications capabilities, along with air defense systems. And he told reporters Ukraine has its own plans: “There are developments. We are working on the same [similar to] Starlink.” Ironically, though, it would require SpaceX to launch the satellites. That may be a big problem. Chad Anderson, founder and CEO of venture capital firm Space Capital, told CNBC: “Capital isn’t the constraint any more. Launch is.”

European leaders are watching the U.S. turn its back on allies with alarm, and it is no surprise that they realize they can’t be dependent on Starlink either. The European Union already has Govsatcom, which went live with initial services in January of this year and expects to offer commercial services early next year, potentially extending them to Ukraine.

Merien ten Houten writes in Innovation Origins that Europe is working on at least three alternatives to it: the IRIS² project, its own secure satellite network; Project Bromo, planned merger of the space and satellite divisions of Airbus, Thales and Leonardo to create a SpaceX competitor; a planned consortium of Deutsche Telekom, Orange, Telefónica and Vodafone to create a satellite-based network for mobile phones.

Obviously, none of these are anywhere near being ready to help Ukraine, or anyone else. IRIS² is furthest along, and it won’t launch until at least 2029.

Canada has its own concerns about dependencies on U.S. capabilities, and so it has awarded Toronto-based Telesat a contract to “strengthen secure and reliable military communications.” Telesat plans to have Telesat Lightspeed, a global low-Earth-orbit high-speed broadband network, operational by first quarter 2028, with a first launch later this year. The company told Cybernews that, although its primary customer is the Canadian Armed Forces, it expects to offer its services to NATO and other partners, including Ukraine.

All that sounds very promising, but it leaves President Zelensky knowing he is still dependent on the U.S. and Starlink for the foreseeable future.

I have to confess, I don’t quite understand how, in a nation, in a world, of supposed capitalists and rich venture capital firms, Starlink has such a monopoly. I understand even less how SpaceX has not only built such a lead in space launch capabilities but is also allowed to own a satellite service that depends on launch capabilities. Talk about building a moat to competitors. Talk about monopolistic practices.

Similarly, if Russia were to invade Netherlands, or China Taiwan, taking over, respectively, ASML and TSMC, the rest of the world’s chip-making abilities would fall back to the 20^th century.

And, gosh, last week we discovered that an AI-discovered flaw in a programming language most of us had never heard of – PHP – exposed as much as 70% of the web to attack, forcing many sites to shut down to fix the flaw before hackers exploit it. The point of the internet was to be resilient and redundant, so having 70% of its websites with the same vulnerability sure seems like a problem.

If you work in healthcare, think about what will happen when – not if – AI discovers a critical security flaw in Epic, or if – hopefully not when – Judy decides to slow speeds to restrict capabilities. Healthcare would grind to a halt.

The Starlink slowdown might seem like just a problem with Starlink and just for Ukraine, but it’s not. When a single company and, indeed, a single person, has such control over the lives of so many people, with no alternatives and no recourse, that’s a problem for us all.

Kim is a former emarketing exec at a major Blues plan, editor of the late & lamented Tincture.io, and now regular THCB contributor

By XAVIER WARTELLE

The Metric Healthcare Could Never Move

One statistic should capture the attention of every hospital executive. A peer-reviewed analysis published in Medical Care examined HCAHPS (Hospital Consumer Assessment of Healthcare Providers and Systems) performance across nearly 4,000 U.S. hospitals over an eleven-year period. Nearly every measured category improved meaningfully except one. Overall hospital ratings improved. Nurse communication improved. Discharge information improved.

Physician communication barely moved.

After more than a decade of public reporting, value-based purchasing incentives, star ratings, communication training, and patient experience initiatives, physician communication improved by just 0.8 percentage points, the smallest gain of any major HCAHPS domain. That finding should force healthcare leaders to ask an uncomfortable question: What if the industry’s approach to physician communication has reached its limits?

Many health systems have invested heavily in communication training, service excellence programs, scripting initiatives, and patient experience workshops. These efforts have value. But they largely focus on behavior. HCAHPS 2.0 suggests that behavior alone may no longer be enough. What if the real challenge was the antiquated tools provided to physicians that haven’t changed in nearly 50 years?

And HCAHPS 2.0 will put an even greater emphasis on this. For years, patient experience metrics largely focused on satisfaction. Did patients feel respected? Were they treated courteously? Did they have a positive overall impression of their hospital stay? Those questions remain important, but the updated framework moves beyond satisfaction and toward something far more consequential – understanding.

The new methodology places greater emphasis on care coordination, communication across care teams, family involvement, and post-discharge planning. At the same time, CMS is moving away from a scoring model that rewards only top-performing responses and toward one that recognizes incremental improvement across the entire patient population. Every improvement now matters. Every patient interaction counts. Every conversation has measurable value.

For healthcare executives, this should be viewed as welcome news because communication has always mattered. Healthcare is only now beginning to measure it differently.

The Exam Room Hasn’t Changed

Hospitals have transformed nearly every aspect of care delivery. The exam room, however, remains remarkably unchanged. Physicians still rely primarily on words to explain concepts that took them years of training to understand.

Over the past two decades, healthcare organizations have invested billions of dollars in digital transformation. Electronic health records, robotic surgery, artificial intelligence, advanced imaging, precision medicine, and predictive analytics have fundamentally changed how care is delivered. Yet the moment when a physician sits down with a patient to explain a diagnosis, discuss a treatment plan, or review the outcome of a procedure often looks much the same as it did a generation ago. This disconnect is becoming increasingly important.

Medicine has never been more sophisticated. At the same time, the gap between what clinicians know and what patients understand has never been wider. A neurosurgeon may spend years learning how to interpret complex imaging. A patient may be expected to understand the same anatomy, diagnosis, and treatment options during a twelve-minute consultation. It is hardly surprising that misunderstandings occur. What is surprising is how long healthcare has accepted this reality as unavoidable.

Why CMS Is Expanding the Definition of Patient Experience

The new HCAHPS questions tell an important story. Patients are now being asked whether doctors, nurses, and staff appeared informed and aligned. They are being asked whether care teams worked well together. They are being asked whether families and caregivers were involved in planning for recovery and discharge.

These questions recognize a reality every healthcare executive understands. Patients do not experience healthcare through organizational charts; they experience it as a journey. They don’t distinguish between departments, specialties, care settings, or reporting structures; they simply know whether they felt informed and whether they trusted what they were being told. They know whether they understood what was happening to them and what would happen next.

When communication breaks down, patients often perceive fragmentation, regardless of how coordinated care may be behind the scenes. Conversely, when they understand their diagnosis, treatment plan, and recovery process, they experience care as cohesive and coordinated. This is where HCAHPS 2.0 becomes particularly significant. More directly than before, it evaluates whether patients experienced care as a connected and understandable journey, rather than a series of isolated clinical encounters.

From Clinical Infrastructure to Precision Communication

Healthcare has spent decades advancing medicine. The next challenge is making it more understandable.

Precision medicine changed healthcare by shifting the focus from population averages to individual patients. Communication, however, has not undergone the same transformation. Most clinical conversations still rely on generalized tools: educational brochures, condition-specific videos, static images, and explanations that require patients to translate complex medical information into something personally meaningful. These tools describe disease in general. They rarely help patients understand what is happening inside their own bodies. The next frontier is precision communication: using patient-specific information to create patient-specific understanding. The practice of medicine, personalized.

This shift is already underway. Across healthcare, organizations are exploring visual communication tools, AI-assisted consultations, language access technologies, family engagement platforms, and new approaches to shared decision-making. Together, these innovations are creating a new layer of healthcare infrastructure focused not on delivering care, but on helping patients understand it.

The goal is not to replace physicians. It is to strengthen one of the most important moments in healthcare: the conversation between physician and patient. When communication becomes more personalized, patients become more engaged. Trust grows. Anxiety decreases. Families participate more meaningfully in care decisions. And clinicians can spend less time explaining and more time connecting. That is not simply a patient experience improvement. It is a strategic capability. And increasingly, under HCAHPS 2.0, it may become a competitive advantage.

Patient Understanding Is Becoming a Business Imperative

The implications extend far beyond survey scores. Healthcare leaders today face increasing pressure to improve outcomes while simultaneously strengthening financial performance. Patient understanding sits at the center of both objectives. Those who understand their diagnosis are more likely to adhere to treatment plans and proceed with recommended care. Those who understand what occurred during surgery are more likely to feel confident during recovery. And patients who understand their care journey are more likely to trust their providers and remain within the health system for the long term.

In an increasingly competitive healthcare landscape, that matters. Most discussions about patient experience focus on reimbursement and quality scores. Those are important considerations. But there is another dimension that deserves equal attention – loyalty. Today, patients have more choices than ever before. They compare providers online, review quality metrics, seek second opinions, and they travel for specialized care. The organizations that earn trust are often the organizations that earn repeat engagement.

Patients rarely leave a health system because they dislike clinical excellence. More often than not, they leave because there is a lack of communication and/or understanding. It is becoming clear that understanding is now a competitive advantage.

Recognizing that reality requires healthcare leaders to move from insight to action and to invest accordingly. That will not only improve patient satisfaction HCAHPS scores, but also improve top-line revenue.

What Leading Health Systems Will Do Next

The organizations that thrive under HCAHPS 2.0 will likely move beyond traditional patient experience initiatives. They will focus on making understanding a measurable outcome; not just an aspiration. They will make care coordination visible to patients, involve families more intentionally, and adopt tools that help clinicians explain complex information more clearly. They will embrace technologies that transform abstract medical concepts into experiences patients can see and comprehend. Visual communication tools that allow patients to see their own anatomy in three dimensions are one such example: rather than asking patients to imagine what a physician sees on a scan, providers can increasingly show them at the consultation, and again in the recovery room, when families are present, when the stakes of understanding are highest.

These technologies are not replacing the physician-patient relationship; rather, they are strengthening it. More importantly, this does not require sacrificing efficiency for empathy, as the best communication technologies improve both.

When clinicians can communicate more effectively, consultations become more productive, decision-making becomes more collaborative, and patient confidence increases. Everybody benefits.

The Future Belongs to Health Systems. Patients Understand.

For decades, healthcare has measured whether patients were satisfied; HCAHPS 2.0 begins measuring something far more important – whether they understood. And that distinction may ultimately reshape how hospitals think about patient experience, care coordination, physician communication, and clinical excellence itself.

Hospitals have spent decades building the infrastructure to deliver better care. The next decade will belong to the organizations that build the precision communication infrastructure to help patients understand their care. HCAHPS 2.0 may ultimately be remembered not as a survey update, but as the moment healthcare began measuring what matters most: whether patients truly understood their journey. Because the future of healthcare will not be defined solely by what physicians know, but instead, by what patients understand.

Xavier Wartelle is Co-Founder and CEO of Avatar Medical, a visual intelligence company building a scalable solution to bring patient-specific 3D imaging throughout the surgical care journey.