a16z Podcast - a16z Podcast: The Rise of the Digital 'Pill'

Episode Date: April 28, 2017

The first thing that comes to mind when treating health problems is the need to take a pill (or other pharmaceutical) of some kind. But could a digital therapeutic -- a software-based intervention -- not only complement, but possibly even replace pills? In this episode of the a16z Podcast, CEO of Omada Health Sean Duffy and a16z bio fund general partner Vijay Pande (in conversation with Malinka Walaliyadde) discuss the potential of digital therapeutics, which use software, design, and other carefully orchestrated elements to change behavior. (Because what is software, really, observes Duffy, but a way of changing behavior?) Of course, digital therapeutics can augment medical treatment and make doctors better -- but what advantages do such methods have over pills? How do we know it’s really working? And what role does digital health have in the continuing push towards value-based care?

Hi, I'm Hannah and welcome to the A16Z podcast. Our biofund team often wonders if one day in the future we might be shocked that we once relied only on something as backwards and often toxic as pills for diseases like diabetes, depression, PTSD. So then what's the alternative? Could it be digital therapeutics? This episode on Digital Health includes Omata Health CEO Sean Duffy, A16Z partner Vijay Pande, and is moderated by our own Malinka Wallaliate. Maybe we'll start with what is digital health? It's a pretty overuse term. It means different things for different people.

So when people say it, what do we mean? Yeah, you know, it's interesting. If you look at what software that had health care implications was called before, it was usually some amalgam of e-health or M-health or health IT. And I think at large software entrepreneurs started to latch on to digital health because it felt fresher, frankly. It captured more. Well, especially there was something different than what we were.

we were talking about before. And so that's why we needed a term. Most of the time people think of digital health where it's software driving, right? Software is leading the strategy, not a secondary, right? That's really the next generation of software in healthcare, where software has to be a solution to make the big change versus kind of a second order of operations to make something real. Yeah, the way I see it especially is that software allows us to do something new and at scale. What's a good example of that? I think there's a whole host of companies that are kind of capitalizing, on this next era of how software can empower outcomes in a whole host of new ways. We're really shooting to deliver a clinical outcome through software.

You can think about the traditional breakdown for how we think about health care as being DX and RX. You know, and so the RX here are digital therapeutics, but there's also, in a sense, digital diagnostics, the ability to be able to use machine learning and other areas to really propel diagnostics in new ways as well. And digital adjuncts, where it's digital is a complement to something that could be a, you know, a drug, it amplifies the outcome that you're working to achieve. I think historically when people thought about, you know, the therapeutics or the devices

or the diagnostic industries, it was very physical, right? It's the small molecule that gets the outcome, right? And you, you know, you figure that out, you test it, it works, you lock it in, and that's what you use. And now, because so many people are using software in brand new ways and you can learn how to guide therapies and guide people with software, that's core to the outcome, which is a whole, that's a brand new thing for the industry at large. The interesting thing here is that, you know, we've seen this before in a sense and that the in the 80s with the origin of biotech and protein biologics, we already went through, oh, you know, I know what a drug is. A drug is a small molecule.

This protein thing, that's weird. I mean, the protein thing, that'll never be a drug, and you can list all these reasons why it'll be a problem. This is some degree there's truth there, but actually protein biologics are a huge part of how people live these days. Oh, for sure. And it took, you know, it took the industry while to evolve to accommodate that. There were a whole host of new, you know, regulatory pathways that had to be created. There was, you know, commercialization pathways, manufacturing pathways, ways to inject and administer the medications. There's this entire surrounding ecosystem that had to emerge to commercialize biologics. So this is really the third wave. There's small molecules, biologics.

This is the new format of therapeutics. So let's talk about, are we actually talking about versus small molecules and biologics? Are we talking about replacing some of these builds? Do some of these therapies work together? What sort of diseases do they work best in? You know, when we think about this space, there are some things that I really imagine digital therapeutics making a huge impact in, anything that's behavioral in nature, that when you talk about PTSD, depression, anxiety, sleeping issues, and especially something like type

two diabetes, these are behavioral problems for which a behavioral solution is very natural. You know, I can imagine that there will be things for which digital therapists just don't make any sense. Like, it's hard for me imagine what an antibiotic looks like. But the reason why is that antibiotics fit the standard Western model really well. Bacterian mice is very similar, bacteria in people. Antibiot kills a bacteria. It all just makes sense. Whereas PTSD in mice versus humans, it's hard for me to know whether actually that makes even any sense to be talking about that. And so for behavioral problems, a behavioral solution is very natural. Yeah, I completely agree. And it's, you know, I think it's very hard to find someone who doesn't

agree that software experiences move behavior. I mean, that's the, you know, any good product designer at their core is kind of a behavioralist. Now software and using a combination of social experiences, people, you know, devices, curriculum, content, you know, timelines, goals, all of that allows these approaches to actually scale in a way that you might imagine a traditional drug could scale where it's reliable, it's reproducible, you can deploy overnight anywhere. How do you know it's effective? There's a reason why digital therapeutics has a right to the term therapeutic.

And in my mind, it's because one can assess its efficacy in exactly the same way you would have drug. So you could take a drug for a given behavioral disorder, and then you can compare that in efficacy to a digital therapeutic. And in a sense, if no one told the doctors which was which, the rest would look the same. It's just you're assessing the disorder over time. And so that's the part I think that's particularly intriguing is that we can assess efficacy using basically identical metrics. Have they been head-to-head comparisons against existing? pharmaceutical products. In the original diabetes prevention program trial, they actually looked to see if these programs delivered in person would be better than drugs that they

hypothesized to work. And they were. You know, a program delivered far more efficacy in diabetes risk reduction than an actual drug. The final thing is I would have thought that digital therapeutic would be maybe just a little worse than a drug. Like you think of a drug is like a really powerful thing and then compare it to software. But the irony is that I can imagine 10 years from now, people would be like, well, what'd you expect? I mean, you wouldn't expect this small molecule that we didn't have like this sense for how all of the human brain works and understanding behavior from a fomental molecular biological sense. How could a small molecule do as well compared to something that directly gets at the behavioral problem? Oh, for sure.

And you might even imagine that when drugs are commercialized, if there's nuances to the medication that require behavioral support, you know, maybe the pill's really big and it tastes bad. Maybe you have to eat it with, you know, something that's high in fats. Maybe it has to be exact time of day or it doesn't work, you know, any element that behavior is required as part of the drug therapy, you could imagine that it would be viewed as laughable to not have a digital companion experience with a medicine. Well, that goes to a question that I think in Malinka was getting at earlier, which is, are these replacements or are they supplements? In some cases, there may not be a behavioral drug that has any sort of efficacy. And there's certainly

examples of that, or the efficacy is fairly minimal. And then some cases where I could imagine synergies. Oh, they're like antibiotics. You brought up early. That's a classic example where a lot of people don't comply in the right way with antibiotic medicine. What's changed in the last few years to make all of this possible? Yeah, I think one is just the spread of technology, right? Some of the biggest levers you can pull on behavioral elements have to do with, you know, social interactions, you know, goal setting, peer support, and, you know, people are more comfortable engaging in different and new ways through web and mobile.

I mean, that's kind of how people are living their lives now in front of their screens. So you can meet them right where they are. I think those two elements and the ability to measure and use data science to improve are really enabling this transformative new wave in behavioral medicine. It's also intriguing. I think culturally, people are very comfortable texting, and you actually might be much more comfortable texting a stranger than you would be talking to them face-to-face. Those technological changes have engendered cultural changes as well. You can measure the inputs and the outputs at the same time, which is traditionally very, very hard in the healthcare system. What does that allow you to do? Well, I was you to improve, and that's one important nuance in this space.

If you're building a cardiometabolic med, you've got it, you lock it down. It's a small molecule, like done, right? Cool. In digital, that's not the case. You can constantly evolve, personalized, Taylor, improve, you know, create kind of the next chapter of how you create really almost like, you know, personalized behavioral medicine. And you can do that through measurement. You know, I mean, you look at what we've done, we've amassed about 11.5 million weight readings.

It's the largest longitudinal data set of these sorts of programs in the world. And if you make sure that every single user in your program you're learning from to make it better for the next, you actually increase the innovation velocity as you grow as a company. Now, you wouldn't have that in traditional kind of classics, classical, you know, drug device or men. And you have an interesting barrier to entry that is far superior than something like patents. Something you put in the oven, you know, you set a timer, you're here, ding, and like, oh, great, I'm done. That's not, that's just not the reality of the space. It needs to constantly improve.

And the barrier and the competitive differentiation is in all the millions of nuanced. details and how you deploy. I actually remember you're telling me when you're able to change the notification frequency and how that affected health outcomes. Our program occurs on a timeline and that gives you every single week a new chance to try something new and different. And you can really adjust it appropriately on demographic considerations and know if it's working and see who it's working for who. Think of them as internal randomized controlled trials where we reduced the latency with which we gave people feedback on the food after they entered it in that section of the program where that's most relevant. And just that latency reduction, you know, you see a

lift in the engagement. And then at 16 weeks, we saw like a 0.34% improvement in weight loss outcomes. What are some of the more surprising things that you found through some of these efforts? One is kind of interesting is the question is, well, can people who are viewed as less tech savvy engage? Would this work for older people? I was a question we got a lot early on. And until we'd done work in 65 plus, it was kind of a hypothesis for us too. We weren't sure. It turns out that age, the older you are, the better you actually do, which is a little bit counterintuitive. You know, I think it would be easy to think, well, digital, of course, young people are going to do. Young people are going to do better.

But people 65 plus, they engage more, you know, with their peers and the groups. They're, you know, they're more dialed into the program. It's, you know, it's really fascinating. And, you know, and we think that that's a combination of two reasons. One is building tech that's easy to use, right, mixed with the reality that, you know, I think as you age in life, you're just more in tuned with your health and kind of your healthcare goals. So that was one that was really counterintuitive.

But it's still very new. it's difficult for those in the traditional healthcare system to wrap their mind around this. They don't completely get it yet. What are the worries that you've seen from people in the system? I remember reading about the American Medical Association. The head of the AMA put out a statement calling digital health technologies digital snake oil,

which I thought was a fairly interesting statement. What do you think about that? Yeah, I mean, that's a fascinating example because that became a talking point quite quickly and was picked up by media. The AMA views digital health, the digital snake oil. But if you actually look at that in context of what they were saying, it's very sound. And AMA is not against, you know, digital health.

Digital health that's going to and be embraced by the digital, by the enterprise healthcare community at large. It needs to publish evidence. You need to show the effectiveness. Otherwise, it, you know, it could be snake oil, right? It could not work. It needs to abide by guidelines. You know, what makes sense clinically for the person.

It just needs to be done in a way that actually delivers a result, which increasingly companies are doing. Traditional medicine had its long history of dealing with snake oil on its own. In traditional medicine, you have to go through clinical trials. It's hard to have large-end clinical trials. That's what makes drugs so expensive in part. And instead, on the digital therapeutic side, you can be running the equivalent RCTs of these randomized clinical trials all the time with super huge end.

But that, of course, then puts the onus on those having digital therapeutics to make the case. In the consumer world, I mean, we have the equivalent of snake oil everywhere. I was just Ottawa for a wedding was at this amazing spas and, you know, walk in and they're like, oh, this pool, it's going to take all the toxins out of your body. There will continue to be a need for evidence for evidence generation for digital health companies in a way that has traditionally been viewed as acceptable, which is peer-reviewed publications. I mean, that's kind of the currency du jour of gaining clinical acceptance. And it's, you know, it's not one. It's many. Digital has to do that as well.

So the traditional healthcare system is entirely based on face-to-face interactions with PCPs, between PCBs and patients? What do you think about the concern that we're losing something here by reducing the amount of that face-to-face interaction? Well, I think, you know, face-to-face is a proxy for feeling supported and cared for by a person, which with modern technology can be done on the other end of an internet connection. I think the systems where people are included as part of the core offering

are scaling in better ways than people believe they could 10 years ago. And I think the next era of how individuals can interact with the health care system at our level or at their primary care level will be through feeling cared for and feeling that there's people kind of rooting for them and looking out for them doesn't have to be in person. Also, I think whenever anything is new that's different than the way we do things, we tend to concentrate on what is lost in the difference rather than what's gained. There are people who actually would prefer this type of interaction. Yeah, I agree. In addition to creating really rich feelings for users, you can actually add efficiency and efficacy because you can use data to help the people that are part of it, right? So suggestions based on data analytics and data science of what the behaviors of our participants are expressing.

So your team of, you know, people, you make them smart, right? You know, give them superpowers with software. And that actually enables a better experience. It's kind of what you gain from it. Well, you know, I'm curious how far we can push software. So for some companies in this space have human coaches or human participants and software really scales. Another approach would be to have pure software where even the coaching is done algorithmically. And with all the excitement about AI, it's intriguing to think how far we can go.

Yeah, it's funny. I mean, in our space, the answer I'd give would be not too dissimilar to what Mark Andreessen said on a panel at Stanford with AI. Or it's like, you know, someone in the audience act, hey, Mark, you know, if AI is coming up, why isn't there an AI investor? And Mark's response was, if that does happen, I want Andresen to be the entity and the firm that figures it out first. Yes. Yet when you look at the majority of the complexity in scaling and all the challenges and building a business, a lot of it's very, very people forward and relationship forward. And it's all those kind of softer, kind of tougher elements that right now the world of AI is not equipped to address. And I feel that same way with us.

Like if we can, if there were ever a world where, you know, a software-only, an AI-only approach could coach you in the right way toward positive health, I want Omada to be the company that figures that out. But right now, you know, color me a little skeptical that would, that the world's ready for that. Do you think AI will be, will make a difference in this digital health world? Where do you think it would have the most impact? Yeah, I mean, we talk a lot about what that might look like for our, you know, our company. You might imagine a couple of permutation. One is like, you know, you could personify it, like, almost like a little robot that's a group member. right you know make it kind of cute and it contributes and sometimes when it's you know so when it's

when it's off it doesn't cross the uncanny value you're pretending that it's something it's not it's closer to r2d2 than yeah exactly closer to r2d2 because you can't you've got to like it's first principles of like social interaction yeah or eliza like how does that make you feel yeah exactly so let's go back to humans then for a little bit because i actually fundamentally do believe that in order for us to have better health care for a larger set of people we absolutely do need to scale up humans, providers, doctors, human doctors with software. We have a shortage. It's very clear. So what are the challenges you've seen in terms of scaling up humans using software? It's interesting. Primary care of tomorrow will be so different than primary care of today.

You know, it's very tough to take an existing healthcare system that's incredibly people-based and start from scratch with it. If you design with a vision that this is how you want to use this confluence of tech and software from ground zero, it makes a little easier than on the other side trying to retrofit like a 30,000 person, like health system to do that. Actually, this speaks of a really interesting point. What do you think about startups innovating within the existing healthcare system versus being completely outside?

It's funny. I mean, I would actually say that we sit, we deliberately have pointed ourselves toward fitting in with the existing healthcare system, just in a digital way, right? We want referrals. So we want the existing healthcare system to send us participants or to get participants from employers, plans, understanding what is the current reality of the healthcare system

and knowing how you can latch onto that is really important for entrepreneurs in this space. And you have to take an approach of ramping the learning curve and embracing the complexity, even though there is a lot of it. Yeah, no, it's interesting to think about what was coming. Computation will scale existing things. And we'll see a lot of that. You could imagine how machine learning applied to radiology

could be almost like a grammar speller checker for radiologists and how they can see so many more images they could see before. And I think before we talk about replacing doctors at all, that does make any sense when we can be accelerating them and having them do so much more. And I'm pretty sure we'll see that first, just like you don't think a word processor with a grammar checker is going to replace a writer anytime soon. On the other hand, it could be that the real transformative changes are in areas where human being can't even do that at all right now. So, My very example is looking at a genome. Like a human being can't look at all the different snips or something like that

and just say, aha, that's a T and 7A. Therefore, we're going to give you this drug instead of that drug. You know, that's not something human beings really are good at. And they may never be good at it, whereas competition's perfect for it. I think people who are really future forward are not thinking of primary care as a discipline anymore. It's more of a concept. Right? It's, you know, like all docs used to be primary care docs, 150 years ago.

then you got some specialties and, you know, the amount of information that medicine had so quickly outpaced the ability of any one person's brain to absorb it. And software is really good at knowing everything. So, you know, you just have to connect it with the person and help guide their judgment and layer on the human instinct. Well, and on top of that, we've been talking, I think, implicitly about a sort of a U.S. or at least a first world approach to health care. And there's a lot more of the globe than just the U.S. And, you know, when I think about other countries like India, India, you know, has billion-ish people and apparently half a million doctors. Mm-hmm.

So, you know, that ratio is pretty scary. And a lot of doctors are primary care doctors. And I can imagine for them having a computational radiologist on call could be something that could really further propel them to be able to help more people. And especially the intriguing thing about computation here is that it's not just an average radiologist. It could be the best radiologist. given. And so what you could see in the third world is something very much analogous to what happened with telephones. I can imagine leapfrogging in countries like this that don't have the infrastructure we have now and that they would have an even

greater benefit to moving quickly into this future. Yeah, no, I mean, I agree. I think that'll start to happen in certain kind of pockets, especially pockets where the current existing healthcare system in the U.S. might be just slightly uncomfortable with like how different something is. I mean, you see that drones, right? You see, you know, you see countries across the world who recognize challenges and design kind of regulatory frameworks to better support kind of new and different things and kind of leaprogate. We often think about human nature as being slow to adopt change, but that medicine in general has that reputation because, you know, it's charged with human life and you don't want to

do any harm. But I could see this approach of just enhancing doctors is a way to sort of naturally work in and in time. It would be one of these things where I think we wake up 10 years later and it's just all there. You know, as a practitioner, I think that when done right, man, this makes you feel awesome about being a doctor. Yeah. We're in a primary care crisis in the health care system. Like, there's not enough to go around.

So much of what you do is, like, maybe not the best use of your brain. And if all of a sudden, like, you're surfaced with the areas where you just feel on a day-to-day basis that you can make the most impact, you have a much more fulfilling profession. Yeah, absolutely. My friends that go into the profession, they go there to save people. Yeah. And not to do paperwork or not to... That totally makes sense. So maybe let's shift over a little bit to the commercialization aspects.

What have been the major challenges that you think companies in this space would face? Yeah, I mean, I guess not too dissimilar to the reference point in biologics, it's just new. We always tried to get reviewed and have discussions with the P&T committees at, you know, health tons of pharmacy and therapeutics committees or the people who evaluate kind of new services for clinical benefit. You know, sometimes the entities can think of you as software and like, well, this is a, you know, you're a vendor, you're a vendor of software for us. Like, no, no, no, it's not, you know, it's not quite like that. And so that's, you know, that's been a constant bit of evangelism where it's like, here's how to think of us, right? So that's kind of

a classic thing. And then, you know, the other, there's just a lot of brass tax, like, you know, tactical considerations of like how to even operate and work with each other. So you've got to got to get in a room together and figure out how to use their existing infrastructure to do this in a different way. As a, for instance, prices on outcomes. I think digital companies are better equipped to do that because they're better able to measure their impact. In order to do that using the existing health care system, you've got to set it up in an interesting way and use fee-for-service billing, but done in an outcomes-based way. And you find the early customers that you figured out with them. And then they're on the next calls with the other health plans.

So here's how we set up. And then it just, you know, it starts to spread. So it happens, it happens organically, but it is wind in your face when you're so unique. Like, you don't really fit into any box. It's tempting to sort of put a lot of effort and thinking about does it work and the efficacy, and those are all important things. But I think the part that often people don't think enough about is also, is there going to be a huge ROI for your customers? And so, you know, you want to be able to ideally go after not just the wellness space, but something where you're having a really material impact on the cost of health care or the quality of care or ideally. Yeah, I mean, early on, in enterprise healthcare, it can be tough because it's a very appropriately

risk-over-spying market. I mean, you know, they want to make sure to do right by their member their patients, et cetera. So the areas where we found our early adopters were just clinical evangelists. I mean, we've got a reality across the globe where preventable chronic diseases killing more people worldwide than infectious. The epidemiologist is just frankly scary. And you'll find evangelists, medical directors at health plans, you know, in providers that have such conviction that we're heading in the wrong direction clinically for their patients that they're willing to make a bet on something, you know, if it's new or neat. So you just, you know, you kind of have to find them and then learn with them before you can bring, you know, what you've got.

to a more broad set of customers. Actually, you brought up prevention briefly, and this is something I want to dig into a little bit. People often talk about healthcare today as sick care. We take care of people when they're sick. We don't try and keep them healthy. And something of a general trend of notice with some of these software companies in the healthcare space

are that they really enable a prevention-first healthcare system. Is that something you've seen as well? What do you think about that? Yeah, no, it's funny. A piece of me I always think that's a little bit of an oversimplification. I mean, vaccines are preventive. You know, the U.S. has something called the U.S. preventive services task force, which is one of the most illustrious, like, positions as an

internal med or primary care med, like doctor to sit on that evaluates preventive services, you know, mammography's preventive, screenings, preventive. We don't see all the lives that were saved from vaccines. You don't see it. It is true that the U.S. health care system is a little bit more like reactive. I mean, we're built to address kind of acute problems and reimburse for acute problems. You prevent the progression and slow the progression. You know, it's not like we enroll people with no risk factors into our program.

I mean, we, you know, we sometimes call it the tipping point population where they're already warning signs. Like, this is the right moment clinically to intervene versus earlier. But I think, you know, I think software and data helps you do that. You just have to fit in the same sort of framework that the rest of prevention does in the U.S. Also, I think there's a big opportunity for changes that reflect the desire for value-based versus fee for service. If you think about capitalism, you're thinking about psychology, human nature, people will optimize within the rules you give them. And if you're paying for services, you probably get a lot of money. services. If you're paying for value, that's the best chance to get value. If there's any one thing

that I would like to see is to preserve the value-based approach, this would have the greatest impact on changing things, both in terms of quality care and decreasing cost. Decreasing cost is such a nightmare. And obviously, we all care about value. We're still very early on at this intersection of health care, bio, and software. What do you think the next few years look like? It'll be a lot of infrastructure building. I think more companies are going to continue to realize that evidence is going to be important in this space. I think you'll continue to see a real leverage of using data signs to adapt and personalize as companies get volume. We will continue to see a push for Alchem's-based pricing. You've got plans across the country of asking drug companies

to do it, asking device companies to do it, asking their providers to do it. I think that'll continue. And then the next leap that I think really is the transition even closer to clinical settings for digital, right, where it's not often that a clinician or a practitioner thinks I'm going to prescribe or refer to digital. I think that's changing. I would add one thing, which is that I think about where we're going to be in 10 years. It's something where machine learning and AI will have crept in so much that we wouldn't even have thought about. So if you think about even just where we were 10 years ago and you could bring someone straight into the future, you show them the apps on your phone with automatic face recognition and

the huge possibilities that we have that are just things that we just routinely think about. I think those will be deeply embedded in all these different spaces because it's basically coming along the ride with all the other technological advances that we're having. And so the real question is, how can we sort of accelerate that change and how can we get there faster? And I think one of the key things that we're seeing is the emergent properties of having software pushing this is the driving feature. And that is, as Sean said, I think that's what's new.

and I think that's what's important. All right. Thank you very much, everyone. Thank you.