Can AI Drive Authentic Patient Experiences?

Chris Hemphill (00:02):
Consumer experiences, major disruptors in AI tech are shaping healthcare for years to come. On hello healthcare we dive deep on these issues with leaders who are driving change. I’m Chris Hemphill, VP of applied AI at Actium Health and we hope that these stories will help you to create or demand a better future in healthcare. Artificial intelligence, AI, we’ve all heard the term, maybe you even heard it in my job title on the podcast, perhaps it inspires wild sci-fi fantasies of learning machines, conscious computers and evil robots seeking to take over the planet, ALA the Terminator or Matrix movie franchises.
 
Chris Hemphill (00:45):
But what is AI really in? What can it do for us? Outside of Hollywood, intelligent machines are not taking over the world, but in fact AI is already well integrated into many aspects of our daily lives. Most of us are probably unaware of the influence that it has in some of the most important parts of our world. Artificial intelligence has infiltrated virtually every industry and become a major component of things like online shopping, commercial advertising, online search engines, your social media accounts, banking, smart home devices, some of the systems in your car and of course we can’t forget the phone that you’re listening to this on. Virtually every business of the future will need to rely on AI in some form or another to stay competitive. According to IBM’s 2021 global AI adoption index, nearly three quarters of companies are now using or exploring the use of AI. And that pace is still accelerating.
 
Chris Hemphill (01:45):
Nearly half of all companies are now reporting accelerated AI rollouts due to the pandemic. AI has allowed us to move faster, smarter, and at a scale that was previously thought impossible, but we are focused on healthcare, healthcare is supposed to be deeply personal and meaningful. Is there any room for artificial intelligence to play a role in that? Well, let’s find out.
 
Sheetal Shah (02:09):
Maybe I’ll start with what artificial intelligence isn’t in healthcare in terms of what we’re talking about today, right?
 
Chris Hemphill (02:14):
That’s Sheetal Shah, digital health executive at Woebot and board member here at Actium. Sheetal has a wealth of experience leading digital transformation at top 10 health systems.
 
Sheetal Shah (02:26):
We’re not here to talk about replacing physicians, we’re talking about how do we use massive data sets though to find insights and make decisions on what would take humans years to do by themselves? In our case we’re talking about outreach, right? And engagement. So what I wanted to share was let’s establish the role for AI in healthcare, right? And if you look at all the research that’s out there that’s being published at a feverish rate, there’s also a flip side of not only how healthcare models are changing and the requirements we have on why we engage the community, patients, families and caregivers, but also the provider side and how much burden we’re shifting to them to take on the role of engaging the right people at the right time for the right things.
 
Sheetal Shah (03:13):
So if we looked at, in the US, they say the typical primary care at the average panels, 2300, I’ve seen it go as high as 2,500 patients per PCP. If you just looked at that and you asked a PCP to basically provide all the recommended acute, chronic and preventative care itself, but then also the engagement to get the individual in the door, they would be spending 21.7 hours a day working. Let’s settle in for a moment, you basically are asking an individual to provide the acute, the chronic, the preventative care, the anticipatory care ahead of time and getting that’s just obviously this type of engagement just does not scale.
 
Sheetal Shah (03:58):
We’re sitting on a treasure trove of data that allows us to do things that are scale. And I think that’s the piece that when we say intelligence, it’s at scale, it’s at scale on how we personalize the communication that it has purpose for both the patient, it’s relatable to the individual that they actually will open it and read it, right?
 
Chris Hemphill (04:18):
Make note of that, Sheetal, focused on AI’s role in communication and outreach.
 
Sheetal Shah (04:24):
It’s proactive, it’s about forming a relationship. We may use your data to inform what we tell you and by the way we’re going to reach out to you by email or phone or whatever preferred method you have. So, if you look at it, we know the ability for providers in a typical model of healthcare to scale engagement is not possible, they’re already slammed with that 21 hours a day if they wanted to do everything for everyone, we know that today it’s not personalized, or if anything that engagement is absent.
 
Sheetal Shah (05:00):
And then you look from the periphery and you’re starting to see organizations look at, Hey, we found the hole that we’re going to plug, we’re going to form a strong relationship.
 
Chris Hemphill (05:09):
Communication and outreach. While we constantly hear about AI to scan images or replace doctors, there’s remarkably little talk about its role in influencing patient relationships, maybe communicator 2 wasn’t as catchy a title as Terminator 2. That’s a shame because there are people actually doing this in the healthcare space. I spoke with Dave Pavaletz, who at the time led AI based outreach efforts at Virtua Health, which is a $2 billion health system in New Jersey, since his hands who are on AI based tools every day, it’s interesting to hear his thoughts about how AI and healthcare consumerism combine.
 
Dave Pavaletz (05:52):
In its simplest form healthcare is really not much different than retail, everybody always is like, oh, it’s so much different, I’ve been outside the healthcare industry, it really isn’t that much different. Everyone has a need or a want that they want to research and then they select what they want or who they want to go with. So this is where AI and healthcare marketing has really entered the playing field because consumers have evolved to really expect more. And this is really that Amazon effect that everybody talks about crossing across multiple industries.
 
Dave Pavaletz (06:21):
So it leads to the challenges that can be addressed, so the classic model for direct marketing and healthcare has always been, what’s good for the goose is good for the gander. And the shift has to be made in healthcare to become more personalized, especially in an industry that for the most part is one of the most empathetic industries out there. So, we know how to treat our patients well, but do we really listen to them and get to understand them and put this persona to them. So what does the day in their life look like? What are their pain points? What do they value most? What are their goals and especially now in this new unprecedented daily routine is how do we keep them safe when they come back?
 
Chris Hemphill (06:58):
Now that’s a lot of moving parts.
 
Dave Pavaletz (07:01):
So what AI does is it enables that personalized targeting, it really stops us to stop this from generalizing and puts a space to that persona, artificial intelligence really takes the data that you have internally and combines it with really any outside data sources that you have, whether it’s demographic, internal data, third party data. And what it does is it produces this target that you’re looking for. And it could be multiple targets for the sake of this, I just put one here. What it’ll do is throw it out across your population. So it says, okay, here’s this persona, here’s who we should be marketing to, they’re most likely to utilize our services. And what this does is it stratifies it across there, really does the work for the marketing team to say, this is a segmentation and this is the persona.
 
Dave Pavaletz (07:45):
And it really changes the way we market, so it takes me over to, this is what the old patient journey looks like, it’s very transactional, impersonal. Now imagine how this feels to a patient, right? They can be anywhere between 24 hours to six weeks to academic medical centers that take up to six months in between these touch points. And what you do now is you would take that persona that you’ve done and you’ve utilized with artificial intelligence. And what you do is you drop that impersonal feeling there, that detachment, you drop that time limit between there. And what you do is you start building this relationship.
 
Chris Hemphill (08:22):
That’s a really interesting transition, terms like AI and machine learning sound very cold and impersonal, right? But Dave’s focus here is how to use it to be warmer, more personal and more engaging to patients. But how do we know that this tactic is working?
 
Dave Pavaletz (08:41):
So the smart thing to understand with AI, there’s both hard and soft benefits, those soft benefits are you’re understanding your consumer and your patient’s intent, their interest, their engagement levels and then we start getting into those hard benefits of conversions and revenue. And then AI helps us with these personas to be able to target these patients. And the number one success is that we’re able to bring them back as a customer again. So what AI is constantly doing to engage that is understanding the content you’re sending out, understanding the emails you’re sending, understanding who you’re targeting, the keywords that they’re searching for, the analytics behind what we’re sending to them. AI is compiling all that and it’s constantly learning to retarget and continue to target those patients to develop better personalized content, to be able to understand engagement and how people are engaging with your health system.
 
Dave Pavaletz (09:28):
And ultimately for Virtua what success was for our breast health campaign that we launched in January is we had a 350% increase in mammogram orders versus our control group internally. We see [inaudible 00:09:40] the industry benchmarks by close to if not over some 300% in some areas of open rate click through rates and form conversions. And this is what I was talking about in the realm of those soft benefits, you’re understanding how these certain patients are engaging with you and assuming what you’re sending out.
 
Dave Pavaletz (09:57):
But the most important thing is we’re gaining new insights daily, this was brand new for virtual when we started. So for us, it’s understanding that intent, that interest and engagement and constantly learning and being able to utilize that for marketing campaigns in the future. We’re still and always learning, so we’re looking at the voice of the customer and what they’re providing to us, we’re looking at the analytics and the results in the background and the important thing with AI is that you can always pivot on a dime. So it is always great to hear those hard benefits of these increased orders, the increased revenues, the increased conversions, but it is important to always stress across the organization, that there are soft benefits that people may not notice, but keep you connected in the long run.
 
Chris Hemphill (10:35):
To hear Dave tell it, machine learning can play a powerful role in the nature and quality of patient outreach. In previous episodes, we’ve explored how important that quality is. And the marketer’s role in delivering care. To explore this a little bit deeper, let’s bring in Christopher Penn. Christopher is the founder of Trust Insights with delivers data driven marketing experiences for their clients. He’s chief data scientists as well and he’s also a great communicator. Check out his podcast, Marketing Over Coffee and hear what he had to say on the concept of AI in patient outreach.
 
Christopher Penn (11:13):
Data plays a leading role in helping marketing leaders, or in a seat at the table, but to guide the direction of the company, when you think about how you market your organization, if you don’t know what’s working, you can’t fix what’s wrong, the old adage you can’t manage what you don’t measure is still true, right? No amount of cool technology has fixed that simple basic fact. And when we look at the role that data plays in marketing of telling us, who’s interested in our stuff, how are people engaging with us? What are the things that they’re telling us through their words, through their actions, through their interactions with us, it gives us tremendous insights into how we should be changing and pivoting.
 
Christopher Penn (11:52):
So a real simple example, if you think about the way we look at the customer journey, life cycle in healthcare, customer lifetime value actually means the lifetime of that customer, right? That’s one of the few industries where, yes, you will literally be with us until you die. And there’s so many interactions and data points that are captured that are simply not in intelligently used, think about, even if you remove it, you anonymize it, you de-identify it completely, think about just the sheer number of data points in an EHR, right? How many interactions has this person had with a provider?
 
Chris Hemphill (12:25):
Remember when Sheetal said that healthcare data is a treasure trove.
 
Christopher Penn (12:29):
All that data could be part of your marketing if you are using it well, I mean you have permission to use it and you’re careful with it to get a pulse for what’s happening with your audience, what’s happening with your patients in their lives. And one of the really unique things about healthcare marketing is that you can use this data responsibly to guide outcomes, right? You can say, hey, we’re noticing there’s a substantial number of people who are buying this garbage fast food and it’s going up in this region. We know in 20 years time, we’re going to have a major cardiovascular problem in this area if this does not change, if these behaviors don’t change. So if you have that data, you can start to help guide and say, Hey, you know what? This looks like a food desert, we need to fix this. How can we use marketing’s power to communicate with people? These are better choices or worse choices you could make if you want to see your grandkids graduate college.
 
Chris Hemphill (13:20):
So we understand the data space and we could admire the data and it’s super interesting insights, but interesting isn’t good enough in healthcare. How do we get to a point where this data drives actions?
 
Christopher Penn (13:34):
It depends on your business models, because one of the things that is changing and is on people’s minds in healthcare is how do we get away from treating illness to treating wellness, right? How do we apply more preventative measures so that we’re seeing fewer things like massive cardiac events and strokes and stuff. And that requires realigning your data to understand and be able to forecast and predict beforehand, Hey, these are some likely things. If you do basic propensity score modeling on data around the top 20 conditions, you can get a pretty good predictive perspective like, Hey, we know that obesity is a pretty big problem, not just in America, but planet-wide, we know cardiac disease kills almost more people than I think almost anything else. We know that COVID-19 is the number three killer in the United States, right?
 
Christopher Penn (14:23):
Marketing technology has the ability to offer some benefits to say, here let’s help you build systems that can process the data, let’s create storage engines that can analyze it, let’s create machine learning models that can get through the data faster so that you can start looking at the mathematics underneath it and figure out, well, these are the trends, these are the patterns. Right now, marketers don’t have that. Right now marketers have big piles of data and no way to handle it.
 
Christopher Penn (14:49):
So there’s a bunch of different things and over the next year to two, we’re going to see a lot more of that data get more used, better used and the technology supporting it mature.
 
Chris Hemphill (14:58):
This is where Fun stuff comes in, the ability for AI to write copy for you.
 
Christopher Penn (15:04):
The generation of text comes from open AI’s GPT models, which is a pre-trained transformers, these are big, big, big, big language models that create language. One of the most powerful uses of them right now is in question and answer scenarios. So right now a lot of companies have tried with differing degrees of success to create conversational chatbots, things that offer customer service experiences and things on websites. And they have uniformly been lackluster, some of the newer ones that are using GPT-2 and GPT-3 are creating legitimately lifelike interaction capabilities, ones that you would be hard pressed after some time interacting with them to not know that they are … Isn’t a human actually behind the scenes guiding it, they’re trained amazingly.
 
Christopher Penn (15:48):
There’s some fascinating applications for this for things like behavioral health particularly in a pandemic we have a lot of people who are suffering from significant mental health impairments from what’s happening, what’s necessary to keep people alive. These technologies are really, really interesting.
 
Chris Hemphill (16:03):
We’re not kidding here, we grew up told that AI probably wouldn’t have much use in fields that required creativity, fields like writing and music, but right now we’re seeing that being proven wrong. Chris dug deeper on models and tools that are available right now and assist in the generation of creative itself.
 
Christopher Penn (16:23):
If you think about it, you could give you substantially better marketing content because now instead of having to get subject matter experts to digest down a 20 page paper, you can have a machine give you the one paragraph summary. And that’s what goes in your hospital newsletter, right? Your hospital newsletter can be a Roundup of the very best content scored by whatever metrics you have that you use for it, but summarized by machines to be relevant so that your stuff that you’re publishing all the time is always the best of what’s out there. That is a substantial advantage.
 
Christopher Penn (16:54):
We have a project right now for one of the hospitals in the nation where we’re curating content for them to help them assemble their newsletter and their social media content and we ship them files and here’s the best of the best that’s out there that we’ve been able to identify with machine learning. And A, it saves them from spending 40 to 80 hours a month curating this content, B, it’s the stuff that performed well already. So we know this is the good stuff and now it’s just a question getting it packaged up for them.
 
Christopher Penn (17:21):
These types of technologies, this is not theoretical, this is something that you can go and try right now on your own and play with it, play with all the different settings. So what’s happening is that these technologies keep getting better and better and better until you get to a point where the machines will be able to generate credible content. One of the challenges that is going to be a big problem for many marketers is there’ll be a huge Gulf between the haves and the have nots and not in terms of finances, but in terms of who can use technology and who can’t, because if you are a three person marketing team in a market and you’ve got another three person marketing team that has AI capabilities, they’re going to create a hundred, 200,000 times more content than you will. They will be able to just saturate an area, a field and you won’t be able to deal with that.
 
Chris Hemphill (18:06):
If you’re a healthcare marketer or in marketing or in engagement in general, keep an eye on content generation, it might not be much right now, but over time you’re going to be able to create content and connect with your audiences much faster. However, when it comes to engaging your patients, the outreach needs to be smarter too.
 
Dr. John Glaser (18:26):
Sometimes think when we use the word patient engagement, what we really mean is you do what I tell you to do.
 
Chris Hemphill (18:30):
You’re hearing from Dr. John Glaser. Glaser is an EMR pioneer who’s been CEO of Siemens Healthcare, CIO at Partners Health and is currently executive in residence at Harvard Medical School. I recently spoke with Dr. Glasser about the need for intelligence oriented experiences within healthcare, instead of our typical transaction based.
 
Dr. John Glaser (18:53):
If you don’t do what I tell you you’re not engaged, said, man, I’m fine being overweight, okay. You can tell me all you want. What up? Well, you’re not engaged. I’m perfectly engaged, man, I just don’t want to do what you’re telling me to do. So I think there is the learning what will work and not. And it’s like it can’t, are you raising a teenager? There’s only so much they’ll do at a moment in time, they got to be ready for certain messages, et cetera.
 
Dr. John Glaser (19:12):
So you do what you can and sometimes it’s because they’re not motivated, sometimes they’re motivated, but they just can’t, they’re poor, they don’t have the copay. They say, you want to walk 10,000 steps, yeah, but I’m in a high crime neighborhood, I’m afraid. We continue to view, your and my health as a health, it’s a medical care problem rather than a healthcare problem or health. And so we talk about, geez, let’s keep you healthy and your weight in the right place and if you got diabetes managing, et cetera, it’s interesting to me, as a country per capita, we spend more on medical care than any other developed country.
 
Dr. John Glaser (19:44):
On the other hand we spend less per capita than any other developed country on what we call social care, make sure people are housed and et cetera, et cetera. And when you add them up together, we’re right, you add the social and the medical care we’re right in the middle of all the other countries. So we are still way skewed in terms of medical care relative and hence the sort of interest in social determinants also. But anyway, I think Chris, there’s plenty of bad habits, payment is a bad habit, the sort of view of health is being AKA medical care, which it isn’t. We still manage to screw up implementations and we still manage to pave [CalPaz 00:20:13]. And we really ought be taking advantage of the technology to.
 
Chris Hemphill (20:16):
So what needs to happen to create better consumer and patient experiences?
 
Dr. John Glaser (20:21):
If you look at healthcare, the old fee for service fragmented healthcare system, we had this called dominant design called the electronic health record, it reflected the [inaudible 00:20:29]. Let’s get the billing done and let’s make sure we can do a variety of transactions, this, that and the other, but it reflected the moment.
 
Dr. John Glaser (20:36):
But if we’re shifting to this new business model, this model of value based care and where we keep you healthy and where we manage you across the continuum, et cetera, then you say the dominant design of the core IT systems has to change as well. And they said, well, okay, how does the EHR have to change then if we’re going into a new dominant design, what does it look like?
 
Dr. John Glaser (20:55):
Well, I mean, there’s probably two ways, one is, since I’ve been in this field for 35 years, it’s always been the record, automated medical record, computerized patient record, the electronic medical record, the electronic health record, personal health … We always record was the now, that we had of these kinds of things here. Well, in the years to come, you can have the best record on the planet, but that doesn’t mean you’re going to do well in value based care, does it mean your care is consistent, does it mean it’s any good, does it mean it’s safe at all? Doesn’t mean any of that stuff, what it will matter is whether I’ve got a plan, I’ve got a plan to keep Chris health.
 
Chris Hemphill (21:22):
John continued by sharing what this plan may look like.
 
Dr. John Glaser (21:27):
I got a plan to keep John healthy, and if John needs his hip replaced, I got a plan to handle that. And if John gets diabetes, I got a plan. And if I can get a plan for John and manage to that plan and get the outcomes, that will help me. So they focus, we always have a record, we always need that, has to shift from automating a record to automating plans, because you’ll have to with a new business model.
 
Dr. John Glaser (21:48):
And then the second part of that was well, when I was a early CIO, what we were doing introducing the EHR was automating paper transactions. So, you retrieve results, you wrote orders, you documented and we took paper away, we gave you a computer and said, that’s great. And we still have work to do there, because it’s too many clicks and not usable, et cetera.
 
Dr. John Glaser (22:06):
But even if we were flawless at that, that doesn’t mean you’re going to do well in this new business model, you got to have intelligence, which is guiding the clinician’s decision and the patient’s decision. And also looking at the data, say, golly, treatment A is better than treatment B or the drug is hurting people. And even intelligence that shapes the interaction, where the machine says, I know who you are, I know what you’re trying to do, I’m going to present data to you that is most relevant here.
 
Dr. John Glaser (22:28):
So moving from a transaction to an intelligence orientation, et cetera, still have transactions, et cetera. So those two fundamental changes will need to occur, because of the shift in the business model, the new dominant design has to have those.
 
Chris Hemphill (22:41):
So there’s a clear need for more intelligent systems for systems to better understand patients, but what’s next is that action space I was talking about earlier. To dig in to how close we can get to precision medicine, I spoke with Dr. Bill Hersh. Bill is the professor and chair of the department of medical informatics and clinical epidemiology at OHSU. In short, this focuses heavily on driving actions and finding information using data science.
 
Dr. William Hersh (23:13):
I think a challenge sometimes is to motivate people to take care of their health, especially when they’re still healthy and haven’t developed any illness. I think probably looking for innovative ways of communicating to get people to change behavior, simple things just eating a reasonably healthy diet, exercise, those sorts of things that help to maintain their health. But people will get sick either because unfortunately the human body, especially as one gets older will develop things, but getting people to especially illnesses that have a lifestyle or behavioral component of getting information to people so they can undertake changes that will be beneficial to their health.
 
Dr. William Hersh (24:01):
I’m not an expert in marketing in the best ways to communicate with people, but certainly having information, so knowing people who might … Identifying who might be amenable to good sorts of messaging around improving their health is an important thing. I think probably the lowest hanging fruit is people who have chronic illnesses that have a behavioral component. And by behavioral, things like diabetes, hypertension, things that can be treated, they may still need medical treatment, but we certainly know that for example a person who has type two diabetes, because they’re overweight, you can make a lot of head way by them losing weight.
 
Dr. William Hersh (24:44):
The diabetes may or may not completely go away, but it will probably be better for their health and things like diet, exercise, quitting smoking, things like that are important things. As time goes on and we learn this whole notion of precision medicine as we learn about genetic variations and so forth, that may predispose people to higher risk of certain diseases being able to find that information to help people out.
 
Dr. William Hersh (25:14):
That does though, you mentioned the word data governance and even some of the larger concerns about data and privacy and so forth, what are you going to do with that information? You know that I’m predisposed to some rare disease or something like that. I think we need to be careful and I think we need to be thoughtful in how we use data, make sure that doesn’t have unintended consequences of learning that someone might be at risk for this illness and they may have more difficulty getting health insurance or something like that.
 
Chris Hemphill (25:44):
Bill then went in to focus on how other industries use AI to drive intended actions.
 
Dr. William Hersh (25:51):
I think a lot of the technology companies, I’m sure that I’ve ended up buying things on Amazon that I didn’t intend to buy because they suggested it would be a good idea. I’m sure there’s some research going on behind the scenes that, do we show them this product or that product and healthcare has been slow to adopt that. I mean, healthcare, maybe unfortunately has had this, if we build it they will come attitude, but not looking at what are the best ways to get everyone who should have a flu shot to actually show up and get one? For managing chronic illnesses. What are the best approaches to keep someone’s blood sugar, blood pressure in the normal range, their weight in the normal range and testing different approaches.
 
Dr. William Hersh (26:38):
I think probably the main thing would be the definition of informatics. Informatics is about information technology, but it’s more about the information and the technology and making sure that what we do with the information benefits people. And so I think that’s the focus and you certainly need to know a lot about the technology and the science behind informatics, but at the end of the day the focus of it should be the benefit that it provides to people, whether it’s patients, healthcare providers, healthcare systems and so forth.
 
Chris Hemphill (27:13):
So let’s dig even deeper, what does healthcare informatics and AI have in common with the way it’s being done in other industries?
 
Dr. William Hersh (27:21):
In the informatics field we sometimes talk about banking, back in the day when I used to travel to far flung places, hopefully will again in the future, people always amaze, you land in Singapore and the first thing you do is you take your ATM card from your local bank and stick it into an ATM and pull out Singapore dollars. Part of the reason why that system is successful, I mean, by one, I think they can overdo the analogy too much because healthcare is actually more, the data of healthcare is a lot more complicated than banking, banking is mostly numbers.
 
Dr. William Hersh (27:55):
I think the reason why banking works really well is everyone makes a little money along the supply chain, when I plug my local banks card into the machine, that bank that owns that machine makes a little money, my bank makes a little money and it’s convenience for me. And again healthcare sometimes has a different model than that. So I think we can hopefully as we move more into population health and paying for healthcare with some component of doing the right thing, immunizing everyone, patients with chronic illness getting things under control will help. And it might be more function of the business model than the technology, I mean, in many ways healthcare has been slow to take up information technology, but also because I don’t know that the financial incentives there.
 
Dr. William Hersh (28:48):
Healthcare is not slow to take up technology, there’s a ton of technology in healthcare. And in fact sometimes there is overuse of medical technology, because there’s incentive to use it and so forth. So I think it too depends on the business model.
 
Chris Hemphill (29:01):
One of the most powerful points Dr. Hersh makes, no matter what technologies is available, healthcare has an incentive problem. When I say the word incentive, who do you think of? Well, incentive is a behavioral economist favorite word, so we brought in healthcare, behavioral economist, Dr. Matt Cybulsky, his background sits at a unique intersect, behavioral economics, healthcare finance and healthcare administration. Plus he’s done work with companies like Deloitte and Optum. If we’re discussing AI and healthcare patient engagement, let’s talk about the bigger picture on where incentives align and what outcomes to expect.
 
Dr. Matt Cybulsky (29:38):
Machine learning and AI looks like a black box to people on the outside, we’re always on this sort of euphemism treadmill of tech, oh, big data, that’s our savior. So when you think about something like AI and machine learning that’s hard for a potential client or a consumer to even understand, you have to explain why it’s even needed, what’s the problem that it’s trying to solve? Well, the problem is, especially in healthcare with remote devices, with hearables, with social interaction that we have, with now secure text messaging that’s ongoing and more, there is an incomprehensible amount of data not only from the consumer, or from laboratory devices, inpatient use case care, even glucometers from Livongo and Teladoc that has to be understood.
 
Dr. Matt Cybulsky (30:23):
And why does it have to be understood? Well, the more of this sort of raw data that we get and the fewer humans we have to process it, the more questions we don’t have answered. A media case in point, Walmart, we’ve all been there. Walmart figured out a long time ago and by the way, they have a huge data center, in fact I don’t know if it still is the case, but it’s like 11 times larger than the IRS’s data center. It’s always processing information about consumers and pricing, where things are in their stores, they even go as far as to say something, a discovery they figured out using machine learning and AI, hurricane in the Gulf at their Southeastern Walmarts, if they ship frosted strawberry pop-tarts and EverReady flashlights or flashlights with batteries in them, they’ll liquidate each of those cases once a week with those, as long as there’s a hurricane in the Gulf. They didn’t figure that out without taking tons and tons of data and trying to understand what is all this noise telling?
 
Dr. Matt Cybulsky (31:14):
So healthcare is in that place now more than we’ve ever been before, because we don’t just have a one way data exchange between EMR and Ensure, you’ve got third party data exchanges happening through an EMR that are incorporating devices in the hospital, devices at home patient interaction with tools, text messages all those things, you’re looking at maybe nine, 10 times amount of data from search alone that something like voice first technologies are adding to the mix of this, so a tool like what you all are creating has to exist to understand what the needs of the patients are or what the patients are doing when they’re interacting with their own care.
Chris Hemphill (31:52):
Let’s dig a little bit deeper into the mind of a healthcare behavioral economist, what technologies have Matt excited?
 
Dr. Matt Cybulsky (31:59):
A remote connection. So this idea that with consent from a patient that I will have these nodes of access that are all the time on. And I know for the people who are really focused on privacy out there that is absolutely nightmare fuel to hear, that something would be connected to you all the time. But if you look at things like, I think it’s called [poop 00:32:21], you have Amazon’s Halo, which is a wearable. And even just the smart speaker at home and voice 10 interactions has got me really, really excited, because it’s something that can happen all the time and why am I even paying attention to it?
 
Dr. Matt Cybulsky (32:32):
And so we can remove this effect of observation, I’m not changing my behavior anymore because someone’s observing me or I know I’m being heard, I’ve been habituating, I’ve got the gateway drug and a smart speaker in my house that eventually makes me focus on my mobile as my primary device or my wearable if I have it on, I don’t have it on today, where I’m interacting with things like my vital signs if I’m taking my meds or not, my sleep, my weight, my blood pressure.
 
Dr. Matt Cybulsky (33:01):
I mean, we could go on and on and on, these passive observational tool sets is really important. One of the most exciting things that I saw was at Carnegie Mellon, a mentor of mine there, Joe Marks, who’s just a brilliant mind, he’s a professor there at the machine learning department, I think he’s chair if I’m not mistaken. He and I were talk at Carnegie Mellon where there was a grad student talking about having a audio listening device in a home, similar that you could put into a smart speaker, it’s a piece of tech that could tell you within really, really close proximity, what clients was running in your house, what model it was and maybe how old it was.
 
Dr. Matt Cybulsky (33:34):
It also knew, it could listen closely and know if a toilet was flushed, [inaudible 00:33:37] was flushed and knew the difference between a chips bag being opened and a homeless bag being opened. So consider that technology along with the other passive sort of parables, wearable smart speakers that are in a home, that’s got me super duper excited, why? If someone has hypertension, you know how much they’re urinating in a day, I know it’s clinical to talk about that here. You know if someone’s eating in the middle of night or when they’re finishing dinner, you know when the stove is on or off, you know when they have water or they’re opening up a pot-bottle or a carbonated drink and in some case that’s really important. Or more importantly you can hear and listen, if someone is taking their medicine or not, even if they don’t interact with a message that says, Hey, you take your medicine today.
 
Dr. Matt Cybulsky (34:16):
That, because of the law of least effort can give patients the sense, because of the interaction we can give them based on tools that you all are creating a sense that they’re being really cared for well, now that placebo effect alone of someone cares about me, even if it’s done through machine learning and AI and all kinds of relational data sets on the outside our lives can give someone a lot of hope a lot of optimism, but not only that, it can reduce the variance of exacerbation and chronic conditions in long term care.
 
Dr. Matt Cybulsky (34:46):
Now imagine that, the quality of life you get from that, the cost savings from that, the longitudinal life expectancy you see from that, that keeping people at work and keeping people contributing to their communities, which in aggregate makes the United States, the world a much better healthier place to live using that long tail of supply that we can all get healthcare without it being a big, huge burden, because we have so many more people in the market all of a sudden engaging it. And we have machines and we have data science helping them live.
 
Chris Hemphill (35:16):
So that’s a fascinating look at all different things that remote technology and wearables will help us do. But at the same time with great technology comes great responsibility. I think that was the quote from uncle Ben. But we are trying to use algorithms to help people make better decisions about their lives, but at the same time there’s a potential for data science to be exploitative, we’ve seen it play out in other industries. How can healthcare avoid these types of trappings?
 
Dr. Matt Cybulsky (35:45):
When you look at new tech, there’s always this sort of ethical roar that starts to rise off in the distance about what’s going to happen, I mean, we’ve seen that in the history of science, I mean from laboratory biomedical sciences, IVF, there is all kinds … Even stem cells, utilizing stem cells. I mean there’s some hub even right now about Regeneron’s tool, right? And how they’ve developed that.
 
Dr. Matt Cybulsky (36:07):
So I think there’s always going to be sort of this ethical roar you hear when it comes to new tech, even if it’s non laboratory of science and medical science, I mean, even technology that are optimizing social connectedness obviously have been part of large controversies lately, and they should have a spectrum of experts and professionals and legislative bodies monitoring what that sort of impact is on the human population, the social population, because it can cause some problems. But it also, I think when I weigh that is the benefit you get out of it. Healthcare is no stranger to controversies when it comes to ethics, there’s an entire body of faculty that study it alone. I have a master’s in biomedical ethics and did that on our way through other training was a fascinating time, but what I’ll tell you is that medicine and healthcare I think amongst many industries really does look heavily upon doing the right thing, doing what’s good. The chief aim of medicine is doing good, right? That’s the bottom line.
 
Dr. Matt Cybulsky (37:08):
So as much as this could be used in a sort of negative nasty way, my opinion is it won’t win, moral constructs that are the norm. Healthcare has that a [priora 00:37:18], before you introduce new tech in the healthcare, there is an ethical concept that exists already. And I think that in cases like new tech, voice tech, AI, machine learning, taking private data sets to learn about human behavior, we take great links to protect that in healthcare. And if you didn’t, it would really kill the market.
 
Chris Hemphill (37:40):
And what should organizations and healthcare leaders think about when it comes to how to make sure that the new technology actually gets the adoption that it needs, actually gets people to use it long term?
 
Dr. Matt Cybulsky (37:52):
This is a great question, I get this one in variance a lot Chris, I’m glad you’re asking me, but the two things that come to mind here are convenience and lowering barriers. If you’re in a position as a patient and your health delivery system ensure whatever entity you’re interacting with offers you something that saves energy or reduces your effort, you’re likely to adopt it. Humans love convenience, American love convenience, let’s look at our packaging system at the grocery store for example, or even just fast food, that’s what health systems need to look at, is this convenient for the patient or for us? Does it lower barriers to action? Those two things go a very, very long way and I would say evaluating any new tech, any adoption, any new relationship with a tech firm, those are the two things you have to look at alone, because without that, it’s no good, it’s like having the pill that stays in the pill cabinet, you don’t use it. And these things have to be utilized.
 
Dr. Matt Cybulsky (38:45):
A lot of tech firms struggle with activation of a product when someone has it, so if someone gets your product and they don’t use it and it’s not active, that’s a problem. That’s not the product problem, it’s the way you designed it. Because you haven’t designed it in a way that reduces the barrier to using it, this is why we have the advent of badges, for example, on the apple iPhone, or you can even interact now inside and notification instead of going to the app itself, these are all examples of making something very easy to use and reducing the human effort. Remember one of the principles of behavioral economics is the law of least effort.
 
Dr. Matt Cybulsky (39:20):
So if you can look at it that way, whether it be an organization as a person who’s purchasing this use case, or whether it be a patient where you’re trying to get compliance out of for either a pill, an exercise or even just a sleep schedule, you have to look at the [inaudible 00:39:35] separate.
 
Chris Hemphill (39:36):
Fascinating overview for Matt. Matt actually heads up the Voice of Healthcare podcast and we’ll link to that in the show notes if you’re looking for more. So when exploring AI and patient engagement, we wanted to look at what’s available today? What use cases are being used, but underappreciated in general? But it still begs the question of what does the future hold and ultimately, where are things going? Let’s take it back to Sheetal Shah.
 
Sheetal Shah (40:04):
I think if you’re in the business today of patient engagement and outreach, it’s the best time to be in this, because if you look at the last decade, right? We spent the last decade on getting data digital, right? Going from paper charts to EHR through meaningful use, et cetera, whole struggle there on what it introduced into the environment, but we’re here. We have enough context about an individual, now we went through this sort of second phase I would say, which is a lot of platforms, spewed the 360, right? Let’s get all the data in one spot so now we can do something with it.
 
Sheetal Shah (40:38):
I think we’re at this third phase now, which is the action side, the patient engagement side and it is now data informed, proactive, personalized outreach. I think the struggle is going to be, how do you take that 360 of that consumer or that multiple 360 of a household and how do you distill it down to what is the right thing to engage them on? And what is the threshold to start acting? What is the value of that action?
 
Sheetal Shah (41:04):
I think it requires an organization to start looking at values of opportunities, it almost requires you to have in healthcare a product mindset, right? How many times are we on Amazon or Netflix or the like and them suggesting something or nudging us? I think the world of patient outreach and patient engagement is going to now be … Is basically, they will be the orchestration engine for all the siloed needs that different teams have and sort of be the conduit to engage members or patients on. But it does require a rethink about who is the, I hate to use the word owner of the relationship, but who is the team and what are the technology elements that help drive the relationship.
 
Sheetal Shah (41:47):
Because I don’t think lifetime value is sort of a one time thing that you achieve once the first encounter happens. You got to put something into the relationship and that’s to get that lifetime value. And I actually think that’s where AI has its greatest potential because instead of being another tool that provides burnout that just burns teams out, AI is informing us what to do when from a relationship perspective and I think that is probably what organizations need to start thinking about now.
 
Chris Hemphill (42:14):
As we’ve heard on today’s episode, healthcare is using AI and machine learning, many other technologies to help improve existing processes and create nuances as well from analyzing massive amounts of data and finding hidden patterns to determining actions and journey steps for patients. As healthcare leaders explore and examine these use cases and what makes a best fit for their health system, they’ll need new skills and new ways of understanding some of the pitfalls and challenges of bad AI strategy. We’ll link to some helpers for that in the show notes.
 
Chris Hemphill (42:50):
If there’s anything to take away from this, you’ve heard from people who develop AI algorithms, from people who are the end users of those and from other folks who employ them for strategic ends. We hope that from hearing from all of these people, we dissect that word, artificial intelligence. Remember the root word of artificial is art and art requires human involvement, attention and care. That care comes from multiple sides, it needs to come from the end users and also the creators of the algorithm. They need to understand the human context of what decisions will be made. That same care also needs to come from the healthcare leaders who are employing AI to help make patient experiences better. Whether you’re creating AI, making decisions about it or influencing decisions about it. The future of AI depends on you.
 
Chris Hemphill (43:49):
Thanks again for tuning into hello healthcare. If you like what you heard, we appreciate a review on Apple, Spotify or wherever you’re listening. You and your feedback fuel us. This conversation is brought to you by Actium Health, to get the latest on what these healthcare leaders are saying, subscribe to our newsletter on hellohealthcare.com or join us for our weekly sessions on LinkedIn. Thanks and when we see you next time, hello.