Perspectives on Leveraging Data and AI in Payer and Provider Space with Frank Webster, the Chief Medical Officer of Behavioral Health at Health Care Service Corporation and Amir Azarbad, the Managing Director of 10Pearls

[00:00:08] Hey, everybody. Welcome back to the Health Matters podcast. Saul Marquez here hosting

[00:00:13] today's episode. I have the privilege of hosting two distinguished health leaders, and I want

[00:00:19] to introduce them to you today. First, I want to introduce you to Frank Webster. He is the

[00:00:24] Chief Medical Officer of Behavioral Health at HCSC. I also have Amir Azarbagh with us.

[00:00:32] He is the Managing Director of 10 Pearls Healthcare Practice, focusing on strategic and technological

[00:00:39] solutions. Welcome to the podcast, gentlemen.

[00:00:41] Frank Webster, Chief Medical Officer, 10 Pearls Healthcare Practice, Healthcare Solutions

[00:00:42] Program, HCSC

[00:00:43] Thank you very much.

[00:00:44] Amir Azarbagh, Chief Medical Officer, 10 Pearls Healthcare Practice, Healthcare Solutions Program

[00:00:45] I would love to begin the podcast by really just asking you guys, what is it that inspires

[00:00:47] your work in healthcare?

[00:00:48] Frank Webster, Chief Medical Officer, 10 Pearls Healthcare Practice, Healthcare Solutions

[00:00:49] Program Sure. I'm a psychiatrist by training, run

[00:00:52] emergency psychiatric services, doing two at the same time, which is really stupid.

[00:00:57] I don't recommend anyone try to do that. One of the things that you do when you see that

[00:01:01] and also just working in the managed care side of things, you begin to see where the

[00:01:05] holes are in the system. Those are not just academic concerns that, oh, the system could

[00:01:10] be better. Those are people that aren't getting services that they really need to get to be

[00:01:15] stable. It really began to think systematically. I like to say I have an engineer brain and

[00:01:21] a psychiatrist body. My dad was an electrical engineer. That's how I think. When I start

[00:01:26] seeing systems and things not working correctly, it just drives me up the wall. That motivation

[00:01:32] to really address things in the healthcare system, especially around behavioral healthcare

[00:01:37] that aren't working well, is really what has been driving me and probably to the point

[00:01:42] of distraction now for several years.

[00:01:43] I love it, Frank. Thanks. Your motivation is palpable.

[00:01:46] I've been told that before maybe.

[00:01:49] Amir, how about you?

[00:01:50] Amir Shukri, Chief Medical Officer, 10 Pearls Healthcare Practice, Healthcare Solutions Program

[00:01:51] Yeah, I think so. For me, it's been just such a... The healthcare industry in general, just

[00:01:56] as part of GPD span has been such a... It's a broken system and there's so much opportunity

[00:02:02] to fix it. There's so much opportunity to get in and make a change. I've been fortunate

[00:02:06] the last 20 years of my career, all of my career, I should say, outside of a small stint

[00:02:11] at a bank, has been focused on healthcare and trying to figure out how do you move the needle

[00:02:15] in healthcare? And it's very interesting being at conferences like this and doing other things

[00:02:20] and hearing all the buzzwords and all the different new technologies and new solutions.

[00:02:24] But at the end of the day, it's still an old sort of system that needs to keep moving

[00:02:28] forward that's not really focused on buzzwords, etc. So it's just been interesting trying

[00:02:33] to figure out how do you move the needle slowly, practically so that things get done versus

[00:02:39] dreaming pie in the sky and hoping that it happens. And then every five years, re-strategizing.

[00:02:44] And so I'm excited. Healthcare is a place that is in desperate need of innovation. And it's

[00:02:49] exciting to be a part of it.

[00:02:51] That's great. Yeah, I appreciate that, Amir. And it's certainly great to have that commitment,

[00:02:55] right? It happens over a long span and you got to take bites of the elephant. Okay, great,

[00:03:01] guys. Thank you for getting to know you a little bit better. And so maybe let's kick

[00:03:05] it off with Frank. As a health plan, what are some of the unique challenges you face

[00:03:09] as one of the largest customer-owned health insurers in the US?

[00:03:12] HCSC is a mutual legal reserve company, which means it's member or customer owned. And it's

[00:03:19] not-for-profit. Now, there are clear advantages of being not-for-profit health care. You don't

[00:03:23] have to worry about quarterly earnings reports. You don't have to worry about layoffs if somebody

[00:03:27] were underperforming by 2% points off our day, either revenue or margin targets. But

[00:03:32] the disadvantage is sometimes they don't pay as much attention to the financials as they

[00:03:37] probably ought to. And it's difficult to raise capital. You can either borrow money or you

[00:03:43] can make money off of operations. There is no stock. So unlike some of our competition,

[00:03:48] you can't issue 5 million more shares and raise capital immediately on that. It has

[00:03:53] to be done. It takes us years if we're not borrowing it to raise enough capital to actually

[00:04:00] make strategic investments in health care infrastructure, health care, strategic investments

[00:04:05] and partnerships. Those are expensive. And it takes five years to get to a point where

[00:04:09] you can actually really start exercising some of those options that our competition can

[00:04:13] just issue stock on and do very quickly.

[00:04:16] Yeah, it's definitely a challenge. There's many companies that consider themselves health

[00:04:21] care solutions providers. And so what would you say sets Ten Pearls apart from the rest?

[00:04:27] I think it comes back to being a practical, pragmatic sort of outlook on it. It's not...

[00:04:33] We do a lot of work across the payer space and the provider space, and it has to do with

[00:04:38] us going to them with practical strategies like achievable. Here's the couple of steps

[00:04:43] you got to take. Here are the things you're going to get done. It's easy to go in like

[00:04:46] our competitors and come up with a hundred page deck and put it in front of the executives

[00:04:51] and say, this is the dream. This is where you should go. The alternative and what's

[00:04:55] worked really well for us and what sets us apart is our ability to actually be very sort

[00:05:01] of step by step focused on delivery of a solution. Have the conversations, people involved, not

[00:05:07] try to take a big bite out of the pie, but slowly get yourself towards that. And so as

[00:05:12] our clients have started to see true deliverables and true go to market and frankly, market

[00:05:17] differentiator products. That's really what sets us apart. We're able to say we're going

[00:05:22] to do something and do it slowly and get there versus here's a hundred page deck, go get

[00:05:27] X millions of dollars. It's going to take you five years. As Frank can attest, we've

[00:05:31] been delivering products every four weeks to production. That's a whole different shift

[00:05:36] in mindset within the payer space specifically because everyone's so used to, well, here's

[00:05:41] the money. We'll see it at the end of the year, hopefully, oh, you missed the target.

[00:05:46] Let's put another business case, go forward. So from a 10 Pearls health perspective, we

[00:05:50] are truly good steward of the dollars for our clients and want to make sure results

[00:05:54] in an agile way in a industry that's really not agile.

[00:05:58] Yeah. And thank you for that Amir. So Frank is do these types of capital challenges, partnerships

[00:06:04] like the one with 10 Pearl help?

[00:06:07] We pay about 3.8 billion in behavioral health claims every year alone. And that's not counting

[00:06:11] our Medicare, Medicaid business. We have to get and calculate that slightly differently.

[00:06:15] But the company itself, I mean about $54 billion a year of revenue, 27,000 employees, about

[00:06:21] 150 million a year of claims paid. So it's a huge company. And like all huge companies,

[00:06:29] especially in something that's as complex as healthcare, there is no one person who knows

[00:06:33] everything that goes on in the company. It's just, it's too large, too complex. That's

[00:06:36] just human nature. So you have to manage through kind of cooperation and collaboration.

[00:06:42] And you tend to get things siloed because we silo by state and we have different lines

[00:06:45] of business. So behavioral health, we're in an advantage because we're an enterprise shop.

[00:06:50] So I'm over all five states, I'm over all of our lines of business to some extent. And

[00:06:55] so when I think about how to get something done, I'm thinking about a solution for the

[00:07:00] company. When our Texas team or Illinois team or whatever, they may think, be thinking I

[00:07:05] need a solution for our area. So when we started talking with Amir and working on this, it

[00:07:10] was a clinical, we built a clinical data aggregation tool. All states, all lines of business, not

[00:07:15] behavioral health, not medical data, everything. Most of our 800 or 900 users now are on the

[00:07:21] medical side of the house. About 20% of them are on the behavioral health side of the house.

[00:07:25] Yet even today, and I've been doing this for three or four years now, still get asked,

[00:07:29] this is a behavioral health tool? Or is this for Texas? Or is this for, does it include

[00:07:34] Medicare or Medicaid? And my answer is yes, it includes everything. And still people are

[00:07:39] just, they get confused.

[00:07:40] Like amazed that it works that way, that seamlessly?

[00:07:43] Because other parts of the company would have. They didn't, Texas doesn't want to pay for

[00:07:46] Illinois. Totally understandable because of the structure of the company. This is normal.

[00:07:50] This is true in all blues, all health insurance entities. And if you want to try to get

[00:07:54] everybody to cooperate, they will, but that can take, that can add a ear to the conversation.

[00:08:00] We were just very blessed to be able to work with Amir, work, find some dedicated people

[00:08:04] internally who are working because we needed those internal IT resources. Amir's resources

[00:08:09] were very reasonable and we got users who are physicians and other clinicians. Hey guys,

[00:08:15] these are the clinical data that we have and this is what's in the pipeline to get. What do

[00:08:19] you want? How do you want to see it? And we let the users dictate what they needed to see.

[00:08:24] And that's why it's been a popular and highly used tool versus historically we would see

[00:08:30] might put in a business case, it would probably, they would actually do extra work to filter

[00:08:35] data for maybe certain lines of business or states. Because to me, that's actually harder

[00:08:39] to not collect the information than it is if it's just there, pull it. And so we really had

[00:08:46] to work in a different way than our company is normally used to working. And I think we

[00:08:50] got this done way under what we would have done internally and much faster. And the fact

[00:08:54] that we can just, we can make a change in almost immediately, which would be very difficult

[00:08:59] to do because we have a very large structure that is just, it's just baked into the structure

[00:09:05] of the company. So the fact that we were able to do this in such an agile fashion, it's

[00:09:11] hard to do that in a company our size. It's not because it's bad, it's just because our

[00:09:15] company is better with very large projects. Something this small almost doesn't even

[00:09:21] show up in the radar, yet the value of the data that we have in there is actually really

[00:09:25] critical because that's if we want to be a clinical company that's working off clinical

[00:09:29] data, we need to be able to display that stuff quickly and easily.

[00:09:33] That's great, Frank. Thank you for highlighting that. And it sounds like a win. I mean, it

[00:09:37] sounds like a win to speak about.

[00:09:38] We're thrilled. My boss loves it, who is the chief clinical officer for the entire company

[00:09:42] and is a family practitioner. And she's like, this is great. So yeah.

[00:09:45] Love it. That's great. What a great story. And I appreciate you sharing that. So look,

[00:09:49] we're here at Health 2023 and AI has definitely been the buzz here and a lot of conferences

[00:09:56] and everywhere. Where do you see AI having the most significant impact on payers?

[00:10:02] Yes, AI is the big buzzword, just like digital transformation was three years ago. And I

[00:10:06] don't know what was before that. There is a real opportunity for AI done right. And being

[00:10:11] practical about it to really transform the payers. And it's not about trying to completely

[00:10:17] remove everybody from the equation. AI is not going to all of a sudden magically reduce

[00:10:22] everything, your administrative costs, your people cost, your claim cost. It's not going

[00:10:26] to just happen overnight. But if you think about AI in the most pragmatic way, what it

[00:10:32] will enable you to do is make your people more efficient. When you allow your people

[00:10:36] to be more efficient at their job versus having to go through 10 systems to collect

[00:10:41] data or look through a medical record notes to try to figure out what's going on. If you

[00:10:46] can use AI in more practical ways, then your people are going to be more efficient, which

[00:10:51] ultimately is as a member owned, for example, with HCSC. That's good for the members. That

[00:10:57] helps with premium costs. That helps with all those things. AI from a practical sense. And

[00:11:02] there's a whole authorization piece of this that AI could help with. There's a lot of

[00:11:06] machine learning and algorithms that can manage the members risk continuum and be able

[00:11:11] to predict if someone's going to be moving from one level to another and those things

[00:11:16] based on prior data. The reality of AI is there's a whole everyone wants to talk about

[00:11:22] and say everything's driven by AI. Then there's a reality within the health care space, which

[00:11:27] I'll let Frank talk through. AI needs good data and it needs to have access to that data.

[00:11:34] If you have garbage data, AI is going to give you garbage results. And there's a real discussion

[00:11:38] within the health care industry around who has the data? How do we aggregate all the data?

[00:11:43] How do you get the provider data out of the EMRs? What do you want to do with the payer

[00:11:46] data? What about the consumer data? All the wearable information? How do you bring all

[00:11:50] that together? Then AI will make a significant difference. But if it's segmented data, AI

[00:11:56] is only going to be as good as that segmented data is.

[00:11:59] I would agree. And I think the biggest problem in health care and frequently by the people

[00:12:05] advocating an AI product is basically they're assuming AI is equivalent to magic. It's not.

[00:12:13] It's basically a fancy algorithm, maybe with self-coding, self-training capability, which

[00:12:19] is still a really good algorithm. What are algorithms useful for? Pattern recognition,

[00:12:24] pattern matching, right? But you have to have the right data. So what do most insurance

[00:12:29] companies have? We have claims data. Claims tells you what the diagnosis was, what procedure

[00:12:34] was performed, who the provider was, who the patient was, how much it cost, what date did

[00:12:38] it happen on. It doesn't tell you anything about why. If you're going to make decisions

[00:12:41] in health care, well, why is a decision was made or not made is critical. So you have

[00:12:48] to go beyond claims data. You have to go to something that actually is reflective of somebody's

[00:12:54] clinical condition. And that could be something as simple as a PHQ-9 standardized nine item

[00:12:59] rating scale for depression. That's direct clinical data. That's not a claim. It's also

[00:13:04] real time instead of frequently 90 days in arrears. So the type of data that you're getting

[00:13:08] direct access to clinical data or electronic health records, that's where there's a possibility

[00:13:13] through. And I don't even like to use AI because people misuse it all the time. So I would

[00:13:19] say things like natural language processing, where you can optical character recognition,

[00:13:23] charts, create constructs in the language of what certain conditions may be assessed

[00:13:28] as and then actually really begin to work on those constructs, train them, make sure

[00:13:32] they're just seeing what you're seeing. There is also a lot of variability, especially in

[00:13:37] behavioral health care. The diagnosis itself does not always drive the treatment. People

[00:13:42] does more so a little bit in medical side, but there's behavioral health that can, you

[00:13:46] can get a diagnosis of depression and people might do multiple different types of psychotherapy

[00:13:52] on it. They may or may not refer for medications, et cetera. So understanding that there's a

[00:13:57] lot of unexplained variability, some of it is probably a problem. A lot of it isn't. But

[00:14:04] this is where AI is going to really struggle because when you start looking at how people

[00:14:09] receive care in the behavioral health system, where AI is going to be useful once we have

[00:14:14] more information and we can begin to automate certain routine decisions, but it's going

[00:14:20] to really struggle because the patterns actually aren't going to look like they would in the

[00:14:26] medical side of the house. And one of the huge things that we see in any large healthcare

[00:14:31] entity that has behavioral health, there's an assumption that medical looks just, or

[00:14:35] behavioral health looks just like medical. It's completely different. But if you looking

[00:14:40] at the data with the assumptions, what you need is a diagnosis and a procedure code and

[00:14:44] a medical policy, and then you'll be able to do it. No, that doesn't work for behavioral

[00:14:48] health. Everything's subjective because behavioral health decisions are made on subjective level

[00:14:53] of impairment, not the diagnosis, the procedure code. It's the level of impairment that drives

[00:14:59] whether somebody needs a certain level of care or not. That's very subjective and can be

[00:15:05] reported quite variably. So we're a long way away from major impact in behavioral health

[00:15:12] space, a little closer in medical, I think. But again, it's really best for routine things.

[00:15:17] You want your clinicians to think clinically because right now when things aren't automated,

[00:15:22] your clinicians are doing routine, repetitive, mundane tasks that could easily be done by

[00:15:27] a machine and then freeing up that time to think critically. What I want is for our clinicians

[00:15:32] to spend a little time on routine stuff and more time than they're spending the day,

[00:15:37] which freaks out people when they're thinking about admin cost reduction. I want them to

[00:15:41] spend more time per case on more complex things because that'll actually help us make better

[00:15:45] decisions. So I view the AI's best role in the next five years as in freeing up clinical

[00:15:50] resources to focus on more complex scenarios. Now, not everyone agrees with me on that.

[00:15:58] There are a lot of interesting conversations, but I'm happy to be proven wrong. I love being

[00:16:02] wrong because when somebody explains why I'm wrong, I get smarter.

[00:16:05] You get to learn.

[00:16:06] It's so cool. It's so much less boring than if I write on something because that is the

[00:16:10] most boring outcome of any conversation I can imagine.

[00:16:13] I love that, Frank. You're such an engineer.

[00:16:15] I might have heard that comment a few times before.

[00:16:18] You don't know how true that statement is.

[00:16:21] I love that. No, that's great. And then the bottom line, thank you for that, right? You

[00:16:25] have to be objective. You have to understand the applications. And really, it's not a catch

[00:16:30] all. Great feedback there, Frank. And Amir, it's about clean data, right? It's about making

[00:16:34] sure that the data that's being leveraged for the areas that it can be leveraged in

[00:16:39] is clean. It's well organized. You guys, this has been a really great discussion. We

[00:16:44] could honestly do a part two on this. It's been so much fun. But we're here at the end.

[00:16:48] So what I'd like to do is ask you both for a call to action for the listeners and then

[00:16:53] we'll conclude.

[00:16:54] Yeah, I teach health informatics too, so this is a sledgehammering the same thing as I say

[00:16:58] to everybody. You need to understand what your data can tell you and what it can't,

[00:17:04] because the minute you're asking questions of your data that it is not able to answer. But

[00:17:11] the problem is, is most people don't know that there are limitations to the data or what

[00:17:14] those are. You've got to understand the implications of that. Because when you are

[00:17:18] getting recommended, if you're training for AI or decision making, that training set is

[00:17:23] actually not the set that you need. You're going to make really lousy decisions. And

[00:17:28] there are definitely negative implications of that. It's like you need to pay attention to

[00:17:32] details. Nobody in our country likes details. They bore me too and I get it. But that is

[00:17:37] actually that ability to critically think and to critically not accept what you're

[00:17:43] handed with AI or about your data and actually understand what it's telling you, what it

[00:17:47] isn't. That's going to differentiate people or organizations that are successfully

[00:17:52] executing on this and from the ones that aren't.

[00:17:55] Thank you, Frank. Appreciate that.

[00:17:57] Yeah, I think feeds in really well into sort of our view of from a 10 pearls perspective. We

[00:18:03] always want to start with the end in mind and then work our way backwards and ask some

[00:18:06] of those challenging questions up front. Because I think it helps everybody think what is

[00:18:11] it that you're really trying to do versus did someone come up with a shiny object and

[00:18:16] waved it in front of you? This is the greatest idea ever, right? We love to work through

[00:18:21] that with our clients. What we've done with Frank and with HCSC has been a very good

[00:18:26] example. What I would say to folks that are listening is get in touch with us.

[00:18:30] Don't just settle for the traditional way of doing things.

[00:18:34] Health care is ripe for change.

[00:18:35] There's always opportunities now with AI and there are other tools, but it needs to be

[00:18:40] practical. And that's, I think, what sets us apart from a 10 pearls perspective.

[00:18:44] We'll put the path forward.

[00:18:46] We'll have the right conversations.

[00:18:48] We will challenge.

[00:18:49] We will give you point of views.

[00:18:51] And then ultimately we all agree.

[00:18:52] And then it's about incremental value for our clients, delivery and success versus the

[00:18:58] all right, I'll see you in a year and hopefully this is what you wanted because you can

[00:19:02] see how fast technology changes.

[00:19:04] I look forward to continuing this discussion.

[00:19:07] And for folks who are listening to reach out to us and let's have a very practical

[00:19:11] discussion of what we can do to help you going forward.

[00:19:13] That's fantastic. Thank you for that, Amir.

[00:19:15] And thank you, Frank.

[00:19:17] And folks, thanks for tuning into this podcast.

[00:19:20] For quick ways to get in touch with Amir or Frank, you'll see it in the show notes.

[00:19:24] So make sure you check them out, reach out to them and take action on the very practical

[00:19:29] advice that we got from both of them today.

[00:19:32] Thank you both for being with us.

[00:19:33] Thank you. Appreciate it.