Latest / Make It Real / The Data Problem Holding Back Healthcare AI with Arun Rangamani | EXL
Transcript
- Arun Rangamani (00:00): There is a perception that there is a AI magic wand which can be waved across the data and everything falls in order. The challenge is these AI models are very good for compute and processing power and ability to run things much faster. But without answering the question, the prompts, what questions are they answering and how are they actually solving for a problem and what problem are they trying to solve? Shirley Macbeth (00:26): You're listening to Make It Real brought to you by EXL. I'm your host, Shirley Macbeth, and on this show we're exploring how artificial intelligence is reshaping workflows, industries, and the way real work gets done. And yes, we're going to make it real. Did you know that healthcare is the biggest industry in the US and it generates far more data than any other industry out there? And yet AI is still in its infancy when it comes to healthcare. And I'm really excited about today's guest because he's going to help us unpack that. Today's guest is Arun Rangamani and he's the SVP of healthcare payment services at EXL. And Arun brings such a unique perspective to this conversation. He's had so many years of healthcare experiences, helping healthcare organizations better support their members, fight leakage, adopt, and adapt to increasingly stringent goals around governance and compliance. So I'm really excited, Arun, to have you here and to have you break this industry down, talk about the opportunities of AI and how companies can get started. Arun Rangamani (01:35): Thank you, Shirley. Pleasure to be here with you to talk about the most challenging problems in the US. Shirley Macbeth (01:40): Yeah. I know, and you've dedicated decades to this industry and learning and helping organizations thrive in this era. So tell me a little bit about your role at EXL and what you're seeing for AI adoption both across healthcare and across other industries. Arun Rangamani (01:57): I'm Arun Rangamani and I'm part of the healthcare leadership team and also an operating committee member for the EXL, my entire executive team. I actually help healthcare organizations fight fraud-based and leakage, large healthcare organization, middle or healthcare organization, small. And EXL has been at the forefront of this the last couple of decades. I actually lead payment integrity, revenue cycle management and risk adjustment business for EXL Healthcare. And last year we saved probably $5 billion plus for our clients. And the industry itself is in an inflection point because of AI adoption right now. Like any other industry, healthcare is also very vested in trying to find out what AI can do to transform the way the processes and the workflows go. But healthcare has been structurally very complex despite it being the largest industry and probably the biggest employer in the country and probably generates, as you mentioned earlier, far more data than any other industry. (02:59): The application on the data or deployment of AI or even technology modules or even transforming the workflows does take considerable effort because of the way the data is stored in silos or the processes are complex. These are really unique workflows which differ by each large healthcare organizations who all build their legacies. So it's been a difficult structure to unpack in the past and the hope is AI can simplify this and as EXL and what we do is to understand these problems better, get the context right and provide those solutions for customers. Shirley Macbeth (03:36): Well, nobody better than you and the teams that you've built who have been studying this problem and working on this problem for decades. I'm really interested in where you see a starting point. I think there's no lack of interest and excitement around how can we change an industry? How can we transform? How can we save costs? How can we fight fraud like you've talked about? And ultimately, how can we provide better care? But how do you get started? Where are you finding and counseling organizations to get started with AI? Arun Rangamani (04:06): It all starts with data. We kind of assume or we have data to prove 50% of the data generated by healthcare is still not touched. Imagine that in a bank or in a financial services industry, a consumer's 50% data is not touched is unheard of. But in healthcare, my industry expert, the CTOs of large healthcare organizations already confessed 45, 50% data never get used for any kind of insights because it's not possible to thread them together. So for us to have any kind of AI-based impact on the industry, it starts with getting the data right, which means a lot of effort that needs to go into preparing the data, making it ready. So data prep will be the primary factor to decide whether you're ready for any kind of transformation. The second important aspect is all the complex workflows. So target those workflows which are most important and prepare the data around it so that the context and the use cases for AI-based transformation is real. (05:09): So those two are the most important primary factors in preparing for a transformation journey. Shirley Macbeth (05:14): So get your data house in order essentially and really find and focus initially on some of the use cases that are ripe for the most impact and expand from there. That makes a lot of sense. Maybe you could give us some concrete examples of the types of areas when you talk about a use case that you see that are so ready for AI transformation. Arun Rangamani (05:34): I have a few of them, but the most important one is the encounters which happen in patients visit hospitals or physician offices. Typically and traditionally last three or four decades, the data which gets generated is typically a claim to the structured information set. It probably will have about 200 to 300 attributes in it, but it does not truly contain all the clinical medical information set. There is another set of data which is called medical records. There are other unstructured data formats available which were not available for analysis or insights earlier because it took a lot of effort to translate the data because those technologies were not available a few years back. But currently with the advent of AI with data transformation techniques and how to actually pars through those medical records and medical records are typically 600, 700 pages long and able to integrate that and use it for analysis, basically unstructured data to structured data to analysis is available now. (06:35): So I feel that that'll be the big inflection point positively for the industry, particularly understanding the clinical decisions made in a process because it has both a financial and a clinical impact to the organization, both the providers or the payer to have that information set from the structured and unstructured data will be so useful to control waste and abuse. Because keep in mind healthcare, as you're mentioning earlier, is a $5.3 trillion industry and they say or we assume three to 10% of that goes to waste and abuse. So imagine those hundreds of billion dollars available for you just by integrating the data together and using the right data set is phenomenal. That's what I feel is most transformational. Shirley Macbeth (07:23): No, it's kind of mind-boggling what you say there when you think about, I think two sides of it. One is we're all patients at all times of healthcare and our medical records and to think that those records aren't necessarily talking to each other, you want the best possible care. So there's the human aspect and there's also this incredible fraud and waste, and it's incredible to be able to think about how we can save those dollars and invest them better across the industry for better outcomes. So massive potential, very exciting potential. So you had mentioned a lot of this data doesn't talk to each other or it's in different places, and a lot of this goes around to strict privacy, data governance, things are sitting in different areas, and healthcare is probably one of the most regulated industries out there. What can we learn from how you are approaching this and some of the clients you're working with that may apply to other industries as Arun Rangamani (08:15): Well? We learned both ways from the other industries to how healthcare can learn from other industries like financial services and what healthcare can teach or help the other industries. So I'll go with the first healthcare, how the data restrictions and the privacy, which is there in healthcare today, how it actually beneficial. We can flip it because what can be a possible impediment to data integration because of privacy and usability also means it is a very structured framework how PHA data and how exclusivity and privacy is being considered. It is not a random thing. There are very clear HIPAA rules on how data should be considered, how data should be consolidated, how should be aggregated. Now with AI and models, these are rules to be fed into the AI model, which means to consolidate and aggregate and give broad trends, at least healthcare has clear defined roles on how data is excluded or included. (09:07): It can be used with AI models for learning and then you apply context to it. Context is the most important thing in healthcare, the clinical context to it, the medical context to it, which is again well defined by both the federal government and by the payers to the policies and procedures. No other industries has such regulation, which means if they'll document it. So the well-documented aspect of it actually applies in benefit to AI models because they're not hallucinating anymore. The defined direction gets, so that is what healthcare can teach with other industries by having regulations on how data can be used. That means you've got guardrails on AI's privacy is protect. That's what a bank or an insurance company can utilize from healthcare. On the other side, how can a financial industry or a credit card industry teach, for example, healthcare? The time it takes for a financial services industry on to determine something's a waste and abuse transaction is much shorter because they kind of optimized what they should look for. (10:02): In those industries, they don't fish for a wide net. They actually know where to fish for information and identify waste and abuse. Healthcare can learn a lot by actually identifying those key attributes so that they can dive deep. So the search or the algorithms go after waste and abuse mile deep in inch wide rather than casting a wide net, which always results in more base dash. Shirley Macbeth (10:24): We started with data and then you mentioned context, and I think that is so important in healthcare. Tell me what that means to these very specific use cases and the importance of context now as part of the overall solution. Arun Rangamani (10:38): Yeah, I think there is a perception that there is a AI magic wand which can be waved across the data and everything falls in order. The challenge is these AI models are very good for compute and processing power and ability to run things much faster. But without answering the question, the prompts, what questions are they answering and how are they actually solving for a problem and what problem are they trying to solve is all defined by that individual who's working in the medical field or in a hospital system or in an administration which controls this hospital workflows or health plan workflows. Without the input to that AI model, you can have the most powerful AI models, but that context becomes so important on what they're trying to solve for. The challenge in healthcare has been healthcare industry is trying to copy some of the methods, pathways from other industries. (11:32): There is also a problem with that because typically these industries go for point solutions, whereas healthcare point solutions may not be applicable because it's a very integrated workflow between a provider, a provider, sometimes an employer, the member, patient pathways, care coordination, utilization. It's much more complex, which is why it is a biggest industry too. So to thread them all together and thread all the data fields together, even if you bring the data together without understanding the context, it'll not go anywhere. So bringing the clinical domain expertise is so important. So what it also means is the healthcare worker, in my opinion at least, in healthcare, AI will always augment what a worker does and the data and the AI prep and the models will make them more productive rather than replacing them as we think about in other industries, "Hey, can you have a board do everything?" Probably in other industries. (12:27): In healthcare, the context is specific. So at least for the next decade or two, AI is going to help the worker get better. In fact, it's good for the American industry because it is the largest employer by far in the States. So we have to be also looking at an angle and how AI can help them. So thread them all together to get the right use cases. Context is important, and that also augments the skill and efficiencies. Shirley Macbeth (12:49): You said so many good things there, and I think the people piece of it is something I want to talk about for a second. There's a fear always, "Oh, AI is going to replace what I do and who wants to talk to a robot if I'm getting medical care?" And you said something quite the opposite, that you're saying that this is augmenting and helping medical professionals provide better care and really help grow the overall industry. And so that's going to take a lot of knowledge management and transformation really of how the types of roles and the workers, if now AI is providing a leg up, if you will, to provide even better care. Can you comment maybe on some examples of that? Arun Rangamani (13:28): I think before even that, I want to set some context on why healthcare is very important because it's very personal. It is not a nice to have, it's a must have and you are to get treated. So that's why it is the largest industry because people do not want to get the right outcome. They want the quality of care to be the best and they don't shy away or industries don't shy away to spend money to provide the right care, particularly in a developed country like the US. (13:53): The idea here is the healthcare worker today is dealing with the legacy systems, is working on archaic information sets, but they're expected to provide the best of outcome to a patient. And then of course the healthcare industry as a whole gets blamed for you got legacy, you're not getting the right outcomes, and it's always in the news for the wrong reasons. But the matter of fact is if we focus on the worker not to replace them, but make them feel better because they have industry knowledge of 30 years, they have context, nuanced context on how each cohort of patients should be treated, how the outcomes should be informed. Only with that, making them feel better and how they actually perform their jobs better is so important. For example, in home health, home is a big use case where many after COVID people got to get treated at home. (14:44): Now what treatments happen at the home or at home and how are they getting the drugs or how are they actually getting the physiotherapy treatments or they get discharged from hospital? How do you avoid them getting readmitted for the same reason? All of this has to be care coordinated. Today, the systems are talking to each other in faxes, in phones, in emails, and there's no single place where you can get it. And it's not just data consolidation, you can consolidate the data, but who's going to provide the insights to you? The care coordinator and worker should get the insights in a simple format so they can then decide the next best action for the quality outcome the patient deserves. Now AI and digital transformation, if we want to consider that, should be talking and specifically using that use case and solve for it, not looking at as a technology enhancement or a system upgrade, which is probably easier problem solve, but essentially looking for outcomes which are important. (15:39): So when you define an outcome like that to a system, then it easy to measure and it keeps improving constantly. So that's where I think AI in healthcare is going to improve how we deliver care through a human being, at least for the next imaginable future. Shirley Macbeth (15:55): It's very exciting and we're seeing so much speed and adoption and interest in this for that very reason because it makes so much sense in providing better care at that moment than it needs to happen. I think that's the north star that everybody would want better outcomes delivered in that way. I want to go back for a second because you talked about some healthcare is learning from other industries and sometimes that's good and sometimes not. And you had said sometimes there's the idea of just bolting on AI to something that's already happening. And you had spoken about healthcare is so broad and it needs that coordination and orchestration of all that data. How do you think about that? It sounds like you have to think more broadly as a transformation rather than just adding AI here and there. Arun Rangamani (16:39): It's like a coral reef in a way I feel. You touch one part, people don't realize you're disturbing something else. That is very, very appropriate in healthcare ecosystems. And it's very unique to the US ecosystem too because we are private funded, we have payers, we are not a single payer country, which means the multiple contracts, multiple commercial and quality contracts between the government, between the employers, health plans, providers. If you want to just change one aspect of it by providing a point solution to a hospital system, it is not going to help the other systems in that loop. So unless you benchmark baseline, understand what workflow you want to touch and what data sets exist, that's why I keep repeating context and domain expertise. And this is right now sitting mostly in the healthcare workers' brain. They're the experts there. That is how you kind of track it, preparation of data, understanding the workflows, identifying the right use cases, and then applying the transformation strategy, which should be across the board. (17:39): So it should be a top-down approach, not a bottoms-up approach, not bolting onto a point solution and getting an upgrade on, "Hey, I got 50 people in a call center. Can I use an AI software to automate it?" That's going to just solve or put a bandage on something which is existing. It's not going to solve for the root cause. The root cause needs a deeper analysis. And I'm not trying to complicate it. All I'm saying is sensitizing on the complexity of these contracts between the multiple entities and with AI and the ability to process things at speed, that's not rocket science, that's not very complicated these days. So that's what the AI companies, companies like EXL, company healthcare organization should focus on, not fall for immediate fixes. Can it automate something? Can it apply an AI model somewhere? That is the pitfall if they just adapt what is being done in another industry because healthcare is very unique, particularly in US where it's commercial and clinical at the same time. Shirley Macbeth (18:36): Well, Aram, with what you've talked about, it's just massive transformation, massive opportunity. So I'm going to ask you to look in your crystal ball just a little bit ahead, two, three, five years. Where do you predict and where would you like to see the industry moving with the adoption of AI? Arun Rangamani (18:51): The first element I want to touch upon is as we look at this journey, I like to give an anecdote of how do you eat an elephant? What do you want to eat an elephant without an elephant? One bite at a time, which means the journey starts somewhere, but we have to be real about how do you measure success every day, every week, every month as we go through this. The vision, of course, is to transform the way patients, members get care delivered the right way. The quality of care goes up. That's the outcome that's required in healthcare. Our costs are optimized. Avoid waste is the other angle to this. The key thing that AI can do is understand from baseline what we have from a data and a workflow perspective, that could be the first thing I would expect the companies to invest in, not go for point solutions, not go for shortcuts. (19:40): And I think the industry is looking at that from a transformative experience. The two to three year journey, I think there'll be massive unstructured data infusion into structured data as an example of medical records, hospital bills integrated into claims. So some of the decisions being taken by providers on a claim, some of the decisions being taken by the health systems on utilization or prior authorization, all of them we get completely transformed. The inefficiencies in the administration would possibly go. That'd be the first immediate benefit of deploying digital AI. Longer term benefit would be we should be able to predict based on information sets and data available. We should be able to predict and manage diseases and chronic conditions much better because we're getting an older population, so there'll be more care. And can that be a self-sufficient care? Can we avoid a patient getting into a hospital system, not just for cost avoidance, that is from a cost angle, but even otherwise, how do you keep a healthier population and manage a disease? (20:44): So that's where this is all going towards. And this has been there for 20, 30 years, not a new thing. But with the advent of AI, this is going to be definitely possible. I'm extremely positive and an optimist, and with the kind of leverage we have around AI in this country, we should be able to go there. Shirley Macbeth (21:01): Well, it's a bright future and it's exciting. I think there's this catalyst to accelerate what we've wanted to do as a healthcare industry for many, many years. It sounds like we're on the cusp of being able to really bring that transformation. And obviously it's a journey, like you said, with the elephant one bite at a time, but you can see where the speed will accelerate and pick up over time. So we've talked about so many amazing things here. I'm going to try to recap our conversation and you can let me know how we did, but some of the key takeaways of what you said, number one for healthcare being the massive explosion of data, and the first thing that matters is to get your data house in order, really bring those data sets together. I heard you say start now, think of an important use case, but have that broad purview of the broader effect. (21:46): You said also don't bolt on. You said think more holistically. And we also talked a lot about the importance of context in addition to data, but that context and why that matters so much in both the people care, the person delivering it, and then the data that's being trained and to help provide that great care on the spot. How did I do? Did I miss anything, Arun, that you wanted to add? Arun Rangamani (22:09): No, you covered it well. You're a healthcare expert now. Surely thank you for the opportunity. Shirley Macbeth (22:15): Arun, I'll do a few rapid fire questions with you. So first, start fast with AI or get it right? Arun Rangamani (22:22): Get it right. Shirley Macbeth (22:23): All right. What do you think about get your data ready for AI or get your AI ready for data? Arun Rangamani (22:31): Get your data right for AI. Shirley Macbeth (22:34): Okay. When you think of the benefits of AI, is quality of care more important or cost of care? Arun Rangamani (22:41): Quality of care. It'll take care of cost of care eventually. Shirley Macbeth (22:44): And when you think of payment integrity prepay, what's more important? Speed or accuracy? Arun Rangamani (22:50): Accuracy. Shirley Macbeth (22:52): Accuracy. And when you're looking to implement AI, do you think about building or buying? Arun Rangamani (22:59): Building. Shirley Macbeth (23:00): All right. And then finally, when we are transforming for AI, upskilling existing talent or hiring new talent? Arun Rangamani (23:09): Mix of both. Shirley Macbeth (23:10): Excellent. Well, this has been fascinating. I think there's so much that other industries can learn from healthcare and the massive potential here. I'm excited to watch on the sidelines, both as a AI professional and also a patient to see how quickly this can help transform our day-to-day interactions and overall help improve and save lives. So thank you so much, Arun, for joining us today on this podcast. Arun Rangamani (23:36): Thank you for the opportunities, Marie. Thank you. Shirley Macbeth (23:40): Thanks for listening to Make It Real. We hope today's conversation gave you ideas, insights, and inspiration to help bring AI to life in your organization. Remember, big ideas don't drive change, people do. Keep learning, keep experimenting, and keep embedding AI where it matters most. Follow along so you never miss an episode.