Shaping the Future of Healthcare with AI

December 13, 2023 Business Talent Group

In this episode of BTG Insights on Demand, Mal Postings—innovation expert and technology executive—joins BTG's Rachel Halversen to discuss the impact of artificial intelligence within the healthcare industry. Together, they'll explore use cases for AI in healthcare, such as personalized treatment plans, real-time monitoring, preventative care, digital user experiences, and more. They'll also discuss the benefits and challenges of AI implementation industry-wide along with emerging trends related to driving innovation and successful AI implementations that will shape the future of healthcare. Listen to the episode, read our lightly edited transcript, or jump right to a specific section of the chat below:

Interview Highlights:

Rachel Halversen (1:17)

Hi Mal, thank you so much for joining us today.

Mal Postings (1:20):

Hey, no worries. Looking forward to the conversation.

Rachel Halversen (1:24)

So to get started, do you mind giving us a little introduction about you and your background?

Mal Postings (1:30):

Yeah, sure... English guy, started off as an actuary in the UK, moved into IT at a very early age, did everything and everything in IT from, configuring mainframes and machine learning, coding. Then got headhunted into Capgemini, as we all know, it's a consulting organization. Ran the internet team back in the year 2000, seems a long time ago. Moved to Paris in 2001 and led the Capgemini partnership team, which is basically doing agreements with IBM, Microsoft, HP, Cisco—the usual suspects—and started to lead more of a CTO position on, new technology waves. In 2004, moved to the USA working in Boston, primarily focused on, again, CTO type work, looking at blockchain technology and RFID technology in anti-substitution in the supply chain for drugs. So basically run a program, program of pharmaceuticals coming in over six months. And then I moved to Detroit; Capgemini wanted me to run a deal with the General Motors, so got to know Ralph Szygenda very well, developed a team of Oceans 50 architects.

Did that for five years, and I was headhunted to be the CTO of Ernst & Young and moved to New York. So did that for five years, worked with Canadian government, worked with Disney, Susan O'Day, people like that. And then I thought, "I've been in consulting for 25 years and want to get back into industry." The consulting is good, but you do six months here, six months there, and I just wanted to be responsible for delivery. So I was headhunted to be the CTO of Quintiles in North Carolina, which is a leading clinical trial company in pharmaceutical. That then merged with IMS Health, it's now called IQVIA.

So was the lead digital, whatever you want to call it, there, introducing robotic process optimization, built the team of AI, machine learning people, and then after five years came back to New York working with startup companies. And then the last two years was offered an opportunity to lead digital transformation but in essence it is working with Google, GCP in AI, machine learning. So yeah, that's my history, and it's going really well. We're now developing a whole suite of things around generative AI and all the kind of things we're going to probably cover in the future questions.

Rachel Halversen (4:31):

How do you view the overall impact of AI on the healthcare industry?

Mal Postings (4:36):

Yeah, there's so much out there today. You can go on LinkedIn. Everybody is talking about AI in healthcare. I would like to split it in terms of… what are the use cases from a personal point of view? If I'm looking from a patient point of view, right, what are the use cases? So the first one is this virtual user experience, right, you know? Medical chatbots, telehealth, avatars, augmented reality, patient engagement, all of that kind of thing, and how we do authentication profiling of a person, emergency services. That is huge from a personalization point of view. Equally, we're going to get hit with, oh, I’ve got COPD, I’ve got mental behavioral health... We are moving to a world with more precision treatment plans for multiple issues, so in reality a person will have several treatment plans running in parallel from various and different specialists.  I’m a single person so what should I do, say, at 6 pm on a Friday afternoon when I’m getting conflicting advice. So this is collaborative planning and that orchestration of customized treatment plans is going to be really key.

Every healthcare company—well, certainly from a payer point of view, whether it's Aetna, Blue Cross Blue Shield, whatever—they have the care team. How do they orchestrate what they're doing with your doctors, with all your applications on your mobile devices? So that collaborative planning and orchestration is key. I think the other thing from a use case, from a personal point of view, is precision. We're now moving into genomics, proteomics, and the pathways. Today, insurance companies—payers—think in terms of cohorts, which means segments. Oh, if you're in this segment, you're part of 30,000 people who need to do this. Yeah, great. That's going to change. No, I'm a population of one, population of one. That precision pathway, recognizing that I have COPD, type two diabetes, I have this, who's actually looking after me as an individual, right? That's another thing.

You know, real-time monitoring, right? I came from a life of clinical trials with IQVIA, and we have pharmacovigilance, which is a requirement, and surveillance, but we all have our Fitbit devices on. Hey, your heart is racing Saturday afternoon between 3:00 and... Why is that happening? Right? More and more that real-time monitoring is reality right now. We can get new FHIR industry standard information that we can now persist (on platforms like Google’s Healthcare Data Engine). That medical internet of things is huge. We need to predict when the adverse event's going to happen before it happens, and that's the kind of things we need to do. It's pharmacovigilance, surveillance, real-time monitoring, whatever you want to call it. It's using AI.

But also then if we know that somebody is susceptible to some environmental, and we know that there's a heat wave going on, I'm now talking in futures, right? But how do we link that? And let's talk about medical adherence. You've got medication and you have to take this…how can we use AI to ensure that we're monitoring if a person is adhering to the medical adherence that they're being given by their PCP? I think that's part of the future.

And then preventative, this is where obviously everybody's going into wellness and rewards, not fixing your problems, it's saying get away from your problems. I mean, that's huge. Everybody's talking about what you should eat, what you should drink, what shapes you should have and all this kind of thing. But it's really important to monitor that, you know? And in IT, we can do that. The big key thing here will be immunotherapy. It's going to be huge in the next five years.

And then from a company point of view, it’s… the use cases are robotic process automation, clinical process optimization. The whole area where you go and see your doctor, Epic EMR system, they can do automation of your clinical notes. They can listen to you and do NLP (Natural Language Processing) on that. The scheduling optimization, right? Nurses are going crazy at the moment because they don't know if they're going to be there on a Friday night at 11:00 or not. So all this scheduling optimization, whether it's an operating room, nurse working patterns, forecasting, that is all AI related. And then more specifically, robotic surgery. We can train surgeons in robotic areas. Then regulatory affairs, a new regulation comes in, how can you then, through simulation, understand, well, this regulation affects area A, C, and D? So yeah, to answer your question, and I can go on and on, the use cases are just crazy.

Rachel Halversen (10:07):

What about some of the benefits that you haven't already covered today?

Mal Postings (10:11):

Well, the benefits are cost optimization, the personalization. But defining quantifiable benefits in a normal business case is obviously difficult. Take medical imaging, right? I can use AI, machine learning, or medical imaging. I can get a PNG file, and use AI to do near real-time diagnosis. Wow, that is huge saving, huge saving. The second one is the future. You know, personally, I'm just coming out of remission in oncology, and I'm just taking my regular blood draws every month to say, am I still okay? And so we need to look at ways of automating for a blood draw, going through the DNA processing, and analyze that real-time information to say,  is it advisory? I would say we’re never the golfer, we are the caddy. We're advising the primary care physician or the oncologist, but boy, boy, that is huge in what we can do going forward. And we're working on that today in ovarian cancer for screening. That can't be done without genomics. So that's a really exciting area.

Rachel Halversen (11:27):

Definitely. Thank you for sharing. In your opinion, what are some of the most significant challenges that healthcare organizations face when adopting AI technologies and how can they overcome those hurdles?

Mal Postings (11:39):

Well, when we talk AI, and we've got to be careful with AI and generative AI, which is obviously the big topic. I mean, obviously hallucinations where—basically the AI model hasn't been trained well enough, right, and therefore it's coming up with wrong answers. There's no excuse for that, but it's going to happen. I think privacy as well. We need to recognize that, again, the data we use to train the models is based on a population. How representative is a population across a number of diverse population states?

I think bias is another thing. All of these AI models that's going into automation in operational areas, we really need to think how we created those models and were they fully a hundred percent validated, put it this way, from different cultures. That is so, so important, and it's so difficult. I think the lack of trust as well. With generative AI, it's just like, "Oh, wow, we're seeing all this." But do you actually trust the information you're seeing?

The Large Language Models, whether it’s MedPalm or GPT4 from Microsoft...how can you trust that what you are hearing is true? And actually you can't, because it's not referenceable. You can't hit the button and say, "Okay, so tell me, this person I'm seeing now, you're suggesting to me the diagnosis is this." I want to press the button and say, "Give me the audit traceability of your suggestion." You can't do that today. Until that reference traceability back to the source and that, we still have a way to go. We're in a world now of AI and generative AI, which is still 20% into the tunnel, really rock and roll, good stuff. We still have like 50% to learn, but overall, it's not going to change. We're still going to be there.

Rachel Halversen (13:47):

So what, in your experience, in addition to those, are the key considerations that organizations should keep in mind when selecting or developing AI algorithms for healthcare application?

Mal Postings (13:57):

I think firstly just data availability, right? If you don't have quality data, you can't create algorithms. It's as simple as that. But we're still in that model of people still not recognizing executives understanding the investment. How many companies are saying generative AI, AI? We've got an advanced analytic team, right? Who? 20 people sat in a corner. In IQVIA, I tell you, no word of a lie, we recruited 2,000 people in 18 months, and Ari, who was the CEO of IQVIA, and he said, "Mal, I don't care the business case, I don't care. I'm going to hire 2,000 data scientists today." That is a smart move. How many companies would do that without, "Oh, we need the business case. We need exactly..." No, the CEO says, "This is the direction for the future." That is smart.

Rachel Halversen (15:02)

Do you have any advice for companies that are just beginning their AI journey?

Mal Postings (15:08):

You need sponsorship for sure. Without that, you're dead. The strategy, building a team of data scientists. But I think in the healthcare industry sector, we're at an advantage. The reason I say that is  when you look at the Generation X, Y, Z, whatever you want to call them, people now want to work with companies that are generally making a difference. Climatic change, whatever, right? Healthcare is one of those. So I would say from an organizational point of view, we as healthcare companies are saying, "This is the industry you want to be in." I mean, how many people coming out of a Harvard Business Review, degree, whatever, can say they're working on a program that's using generative AI linked with biomarkers from genomics that is now doing ovarian cancer screening, which hadn't been done before—which is basically what I'm doing today.

And I'm thinking, "I think a lot of people want to be involved in those kinds of programs, right?" The problem is it's not only the education within a company, but the communication. Because little cool projects are happening, but they're happening in isolation and there's not the awareness. And I think that's our challenge, that in any areas of new technology, there's always cool areas that are forging a new way, but the scalability of that is a key question right? How do you scale that to become—obviously as an architect, that's part of my job. But equally, I'm forced to focus on a single use case, which is ovarian cancer screening, which is good, which is rock and roll. We're going to do that. But then ensuring that's repeatable for other ailments, other therapies. But companies need to realize that.

And I think GSK, the big pharmaceutical companies are using AI for recognizing where are the gaps in treatments. So new drug development, they can analyze through AI, machine learning where, "Oh yeah, we're not really hitting the mark," or some of the drugs that they've developed aren't really doing what they thought they were doing. So that's a huge area. Clinical trials, problem there is matching people quickly to the drug protocol from the pharmaceutical company to individuals. That can be done very quickly with AI.

Rachel Halversen (17:58):

So interoperability remains a challenge in healthcare. How do you envision AI aiding in the seamless exchange of patient data across different systems and providers?

Mal Postings (18:08):

Yeah, you hit one of my key points, interoperability. I think the first thing that people don't understand is… do people really understand interoperability? They just say no, right? So let me explain more. You've got primary care physicians working in the EMR systems, Epic Cerner, whatever, and they'd live in their own world and always will do, right? Cool. You've got many payer companies having their own call center and care management team running in their own system. Great. Cool. You've then got specialist applications now on iPhones for Type 2 diabetes, for COPD, for whatever, right? All of these endpoints—let's call it endpoints—each have amazing intelligence at the endpoints. The problem is nobody's connecting with each other.

And the primary care physician is saying, "Hey, I just did a referral for you for Type 2 diabetes with this Onduo application." "Great, thank you." And then a primary care physician is saying, "I'm seeing no information back one month later, two months later, three months later." The integration, it's like... I talk about this like air traffic control. Everybody's working in a specialized area at the edge, which is perfect. The communication of each area's edge transactions, information is completely non-existent. So what we're doing today, my solution, obviously working with Google and healthcare data, first and foremost, the difficulty is data formats. So we have to ensure that somehow we get a common format. Once we get a common format in the center, like air traffic control, we can now run intelligence, AI, machine learning, but we can't do that when 20 companies, we're all talking different 20 languages.

That's exactly the world we're living in today, and that's what the next 18 months is about. It's all about interconnectivity. Every endpoint is good. Endpoints aren't going to go away. Specialization in the endpoints is absolutely key. It's the sharing of information across the different endpoints is what's missing today. And the only way you can do that is having a common information format, So we're moving in that direction. It's still going to be another two, three years, but that's a challenge.

Rachel Halversen (20:58):

So how does AI contribute to healthcare cost containment and resource optimizations?

Mal Postings (21:03):

Well, obviously automation in—We've got a thing called natural language processing, which is just basically looking at how we interpret data. But if a nurse goes to a clinical chart, some clinical charts have a hundred pages, right? So we've already deployed NLP, natural language processing—AI basically—to say, "Hey, we know this person. We know what you are looking at right now. We think you need to go to page 54, not 63, not 21, to page 54." We've implemented that right now. So that way of automating existing processes using intelligence is absolutely there. Customer service operations, contact centers, that's another way of cost containment. I think using predictive analytics to say, "We can predict what the customer needs are before they're going to come to us with it." And of course, that's going to save costs. I think monitoring behavior patterns so we can react quicker than the normal way of reacting, which is going to be completely lengthier processes. So cost containment, I think it's automation, I think it's customer service, it's predictive analytics, it's real-time monitoring. These are all cost containment ways of using AI.

Rachel Halversen (22:38)

Do you have any specific examples of AI projects in healthcare where you've seen a particularly profound impact, either in terms of patient outcomes or operational efficiency?

Mal Postings (22:49)

Wow, that's a great question. Profound impacts, okay. I would say the first one would be, let's call it next best actions, right? And I'm sure you've heard that before. We can look at... Say me, Mal Postings, I just got a new, I don't know, pharmaceutical, went to the doctor, got a new prescription, went to, I don't know, Walgreens, CVS, whatever. It's now being recognized in the longitudinal record that Mal has some new medication. That new medication now is saying, "Holy crap, Mal, you've just got this new medication, but we know you are already a critical heart failure patient. I'm going to send you a next best action saying stop taking that medication right now." The whole area of next best actions is key. Or let's say it's a Fitbit device. I'm at home, I fell down the stairs, nobody's with me, but my Fitbit device recognizes I fell down the stairs. It's sending out a warning message saying, "Mal just fell down the stairs," right? So this whole area, you're talking about profound impact programs, next best actions is key. It's based on near real-time data. And again, it goes back to this networking of real-time, air traffic control, messages always flying around saying, "Hey, I've just seen you do this. We need to do this now."

Give you another example in another industry. I don't know if you're into Formula 1 racing—I love Formula 1 racing—but cars are going around the circuit at 200 mile an hour, say, in China, the information from the car is going back to a UK working establishment, and they're analyzing information real time. So you're a driver, right, 200 miles an hour, they're giving you a message back from your last corner before the next corner. Now that is happening today, From China to the UK, back to the driver in milliseconds. Now, apply that to healthcare. Yeah, we don't need that millisecond, but I'm walking down the street or... So profound impact is real time.

The second one is, I think, collaboration. How do we ensure everybody's aligned on my care plan? I've got three care plans. What is that collaboration? That's a profound impact to say, "Actually as an individual, I love you, my COPD provider, I love you, my, wellness and rewards coach, but who's actually looking after me?" The answer is nobody, because everybody is focused on different verticals. So I think that is the second one.

And then the third one I think is precision. We've got genomic profiling. We're going into proteomic profiling. Why do I go in to go see my primary care physician? Yeah, whatever. Fine, right? And I go in and they take my vital signs, whatever, and they treat me like 101 other people, right? No, I'm not like 101 other people. Where is that focused precision medicine type thinking? You know my genomic profile, you know my proteomic profile, you know my resistance to drugs. It's just like, why do I need to go through processes that are just boring that I don't need to go through?

So the future is going to be like, "Oh, my gosh, I know you Mal. Whoever you are, I know you, already know all your vital signs and this and that, and I know your DNA, I know your genomics. I know your proteomics." And I think people are reading that now on the internet, and they're beginning to expect that. Digital Twin technology will have an impact on this area. And the expectations versus reality, there's a big gap there. I think the future is three things. It's real time, it's precision, and it's collaboration.

Rachel Halversen (27:32):

I just hope we get to a point where I don't have to fill out the same form every time I go to a doctor.

Mal Postings (27:37):

Well, exactly. Exactly. It's just we're 2023, and you've just said something which is just crazy.

Rachel Halversen (27:44)

So how can AI contribute to reducing healthcare disparities and improving access to quality care for underserved populations?

Mal Postings (27:52):

Yeah. Well, first and foremost, AI can identify the health disparities. And I think a lot of that is allocation of resources, right? It's an underserved population. What are the decisions in terms of resource allocation? Where do you put the water? Where do you put the emergency aid? AI can help in all that more than any human. Rehabilitation, identifying people that need that and are getting that. But I think the most important area here is education. It's the level of education and the distribution channels of the education and monitoring that. I think the willingness is there. It's huge opportunity though.

Rachel Halversen (28:44)

So we talked a little bit earlier about how AI and machine learning rely heavily on data. So what strategies should healthcare organizations use to ensure data quality and privacy in AI-driven initiatives?

Mal Postings (28:54):

Yeah, I think from a quality point of view, it's around obviously understanding the data source and the curation, the common... Obviously in healthcare, we're talking about FHIR format of data. But more importantly, where are the gaps in the data? AI can actually fill the gaps in data, which sounds a bit scary, but it's a way of doing that. Looking at when we have three sources of data from three different companies all saying that, "Oh, this is a procedure information," which means actually a surgery record. That's going to come in three different formats from three different companies. And we want to now centrally say, "Yeah, somebody performed the surgery." The surgery was oncology, cancer, whatever. Who performed it? What was the date? All that kind of thing is easy, right? What's problematic is the outcome, because that's more subjective.

So I think with data and data quality, there's certain things which are ground rules that we say we have to do, and it's very easy. The complexity is the 20% of those fields that we've defined like outcome from surgery, we've never been specific enough to say what outcome from surgery really means. So what we're going to get is three paragraphs from the surgeon on the outcome, which is really good. But then we have to run natural language processing from an AI machine learning point of view to really say what really was that outcome?

And we need to better define, I think, certain parts of AI, machine learning data sources, to be—I can't say repeatable. I don't want everything to be robotic. I don't want to say, "Hey, you've just performed a surgery on ovarian cancer. Was it A, B or C?" No, I'd love the idea of somebody just giving free text what happened. We should never stop that. But equally, we have to look at how we record that for analytics. That is a huge challenge for those areas. It's the 20% areas which are difficult. The 80%, I think we should just normalize and recognize that those 20% value-add areas, it's very difficult to capture that in a field in the database. It's more subjective, but it's extremely valuable information.

Rachel Halversen (31:30):

That makes sense. Could you discuss the importance of collaboration between AI experts, healthcare professionals, and other stakeholders in driving innovation and successful AI implementations in healthcare?

Mal Postings (31:43):

Yeah. again, personal experience is that... To be honest, it's like any area of IT, innovation, connecting all the stakeholders, it's really difficult. First and foremost, you need money. If you don't have the money, you don't have the budget, you can't do anything. And that will... So you need to agree who is the sponsor, who is passionate, who's going to lead the charge, and agree that sponsorship. Then you need to, for each of those business units that are involved, I think agree, the—I call it workflow—the common workflow that where are the roles and responsibilities at each stage and get people to sign up for that. That's very basic stuff. So I think first and foremost, there's also the education. This is a big thing. Again, I spoke to leadership a couple of days ago, and they were like, "Yeah, we can give money if the business asks for this." And I'm saying, "Well, the business ain't going to ask for it because they're not educated enough."

There's a huge gap there because IT is saying, "We've got money if the business asks for it." And the business is saying, "We're running our day-to-day jobs at the frontline. We don't know, what is digital twins?" Have no clue. And then IT people say, "Oh, digital twins can save me 20% a year." And here's the gap. So I think we need to work on the education between IT and the business to say, "Here's the art of the possible." Yeah, certain things we can do, certain things we can't do. But at the moment, the governance is the key thing. The governance is expecting ROI clean models to say everything we do as a program is accountable and we tick in the box, we're going to get a return on investment.

Sorry, you don't really understand the art of the possible from new areas of AI, which is growing incrementally like crazy, to educate the business on how you do things differently. And the business is like, "Holy crap." I'm reading this article in Harvard Business Review or on my plane going somewhere, but it's just like, "Are we as a company?" That is the problem. The connection of innovation, IT, ideas, and business. Because at the moment we're set up for two different organizations. Business are leading the charge. They have the money, and they throw it over the fence for IT to implement solutions. The world we're moving into is IT is now leading the charge to business saying, "Here's the innovative opportunities." So that model is changing from IT being a servant of the business to IT driving the business.

Rachel Halversen (34:48):

Thank you, and thank you so much for making the time. That's all of the questions I had, so thank you for joining.

Mal Postings (34:55)

Take care. Thank you.

Rachel Halversen (34:57):

As a reminder, our guest today has been Mal Postings, an innovation expert and technology executive focused on helping healthcare and life science companies achieve digital transformation. And I'm Rachel Halversen for Business Talent Group.

To start a project with Mal or thousands of our other highly skilled independent consultants, visit businesstalentgroup.com. Or subscribe for more of our conversations with on-demand experts and future of work thought leaders wherever you find your podcasts. Thanks for listening.

About the Author

Business Talent Group (BTG) is the leading talent platform that connects top independent management consultants, subject matter experts, project managers, and interim executives with the world's best companies. BTG provides just the right on-demand talent needed: remote or on-site, part-time or full-time, individuals or teams, all around the world. That’s why more than 50% of the F100 and hundreds of other leading companies trust BTG to curate, vet, and compliantly deliver talent who fuel growth, innovation, and performance improvement. Find in-demand skills and expertise at businesstalentgroup.com

Follow on Twitter Follow on Linkedin Visit Website More Content by Business Talent Group
Previous Article
The Essentials of Effective Organization Design
The Essentials of Effective Organization Design

Herman Vantrappen—strategy advisor and co-author of "The Organization Design Guide"—joins us to discuss the...

Next Article
The Art of Intelligent Automation
The Art of Intelligent Automation

Vivek Garg—an intelligent automation and AI expert—joins us to discuss the art of intelligent automation an...

Get our free guide to working with on-demand talent.

Read Now