In this Becker's Healthcare webinar sponsored by Charta Health, Brian Zimmerman of Becker's talks with Koh Kerdsri, chief information officer of CloverCare Health Plan, a Medicare Advantage plan in Southern California; Aleesa Jansick, senior director of government programs at Sanford Health Plan; and Nicole LaBarbera, vice president for North America at Charta Health, about why Medicare Advantage risk adjustment still runs as a seasonal sprint and how to make it a continuous operation. The panelists trace the inefficiency to siloed ownership, data lag, and redundant chart work, and agree the fix is proactive, cross-functional operations supported by AI and strong governance.
Key takeaways
- Aleesa found risk adjustment work siloed across departments that each held a piece of it, with no single area overseeing it in its entirety.
- In California's delegated model, Koh says encounter data passes through clearinghouses and IPAs before reaching the health plan, so it can arrive 60 or 90 days after the service, and not all of it arrives.
- Nicole says quality abstraction and HCC coding often have two teams read the same charts at different times, generating redundant retrieval costs, and that LLMs can extract both kinds of evidence in one automated pass.
- Koh's plan promotes annual wellness visits and aligns incentives in the first half of the year rather than waiting until the third or fourth quarter.
- Before adding FTEs, Aleesa says leaders need to understand what each individual is producing, which is what makes dashboards and metrics so important.
- Koh says many risk adjustment gaps are data gaps, not clinical gaps: the information is in the medical record, but the code isn't translated or submitted to the plan.
Introducing the session
Brian: Hello, everyone. This is Brian Zimmerman with Becker's Healthcare. Thank you for joining us for today's webinar, Risk Adjustment Without the Sprint: What Outside Leaders See That [Specialists] Miss. Before we begin, I'm going to cover a few quick housekeeping instructions. At any time you have questions for our speakers throughout the webinar, please feel free to type them into the Q&A box you see on your screen. We will follow up and get you an answer. Today's session is being recorded and will be available after the event. You can use the same link you used to log into today's webinar to access the recording. If at any time you have issues with audio or visuals, try refreshing your browser. That often resolves those issues. You can also submit any technical questions into the Q&A box as well. We've got folks on the back end who can help you out.
So with that, let's jump in and have our speakers introduce themselves. Can each of you share a little bit about your professional background and your work in the payer space in your organization? Koh, maybe we can begin with you.
Koh: Thank you, Brian, and thank you for the opportunity to join the panel today. My name is Koh Kerdsri. I'm the chief information officer for CloverCare Health Plan, which is a Medicare Advantage health plan serving members in Southern California. I have responsibility for technology and data services, along with cybersecurity and AI, within the organization. At the same time, I also have Star quality and risk adjustment for the organization. My role is helping the organization use data more effectively to help improve member outcomes, [provider] performance, and operational efficiency.
Brian: Excellent. Wonderful to have you, Koh. Looking forward to hearing more about your experience as we move through the conversation today. Nicole, let's meet you.
Nicole: Absolutely. I'm Nicole LaBarbera, vice president for North America at Charta Health. Quick background on us: Charta builds AI-powered solutions for provider teams and payer organizations. The foundation of our business is that clinical charts contain a wealth of valuable data, but historically, that wealth has gone largely untapped due to the difficulty of reading and interpreting clinical documentation at scale. AI has fundamentally changed that. Large language models, LLMs, the same technology behind ChatGPT, Gemini, and Claude, can read and perform complex analysis of clinical documentation across hundreds of thousands or millions of charts in a very small amount of time.
The application of this technology looks different whether you're a provider or a payer organization. For providers, we build custom EHR integrations to review 100% of encounter notes at the pre-billing stage to ensure accurate, compliant billing and coding, and we can analyze and report on quality measures. If you can train a human to look for it, we can train the AI to look for it.
For payers and risk-bearing provider organizations, the same LLM technology enables incredible efficiencies in risk adjustment operations. Our models are trained in HCC coding and CMS guidelines. They can ingest charts, parse them by dates of service, and pinpoint exactly where there's evidence for incremental coding opportunities. With the same inputs, we can also tell you which claims lack linked face-to-face encounters and codable evidence, and that's a huge leap forward from a [mock RADV].
But the efficiencies go beyond the scope of what's traditionally covered in a retrospective chart review for risk adjustment. We're able to take a [tranche] of clinical charts, read them at scale, and in a single scan, identify all the valuable clinical evidence, like HCC evidence or quality gap closures. And [we] organize that data in a way that makes it very fast and easy for teams to consult the charts. Instead of going page by page, they can click on the evidence and go directly to it. We call this a clinical intelligence layer for risk adjustment, and it's saving teams a ton of time and money.
Brian: Thank you, Nicole. We'll learn more about that, and I appreciate you bringing that perspective. There's a real opportunity here with the new technology to do more with the information that healthcare has. Aleesa, let's meet you now.
Aleesa: Good afternoon. I'm Aleesa Jansick, senior director of government programs at Sanford Health Plan. We're part of an integrated health system in the Upper Midwest. My work has spanned regulatory strategy, implementation, and operations across several lines of business, including previously overseeing risk adjustment for ACA and Medicare Advantage, and I appreciate the opportunity to join you this afternoon.
Why does risk adjustment run in silos at health plans?
In brief: Because the work is split across departments that each hold a piece of it but can't see the whole picture, Aleesa says, which makes it hard to set strategy around upstream and downstream impacts. At Sanford, which launched its ACA business before Medicare Advantage, she says risk adjustment was "done off the side of people's desks," with each function owning its own chunk.
Brian: Wonderful to have you. Let's dive in with our first question, and, Aleesa, we'll begin with you. When you first looked closely at how risk adjustment runs day to day, what stood out to you as inefficient or out of step with how the rest of the business operates?
Aleesa: Two things. At my organization in particular, I was surprised by how siloed the work is and how many departments had a piece of it but couldn't see the entirety of the picture. Pieces and parts live in the provider space, the clinical space, and the data technology space, but it didn't seem like one area was overseeing it in its entirety. That made it difficult to look at the strategy to be able to see the whole and make decisions based on the upstream and downstream impacts. The second piece was the fascination with how far backward you went and how many different plan years you were living in at the same time. The longevity in that aspect.
Brian: Can you share a little bit of a diagnosis of why some of these inefficiencies have come to exist, in your opinion?
Aleesa: For my organization in particular, we started as an ACA org before we launched MA, which might be a little bit backwards from other organizations. It became a little bit that risk adjustment was done off the side of people's desks, and so the function lived within the functional area. Charts would come in, and they would live within [clinical], because it is clinical, and [clinical] would reach out to their counterpart to do that chart retrieval. Then the coding piece: where do we have a coder? That will go to that coding function. And so you end up with little chunked-out portions of it. Then at the end, you look at it, and you're wondering, why do I have a pay-in of $6,000,000 for this line? Because no one's looked at it in its totality in that way. So it requires looking at it holistically and taking a step back and saying, where do all these bits and pieces and functions feed into? I wouldn't say that nobody cares until money is involved, but a little bit.
How does data lag affect Medicare Advantage risk adjustment?
In brief: In California's delegated model, Koh says encounter data travels from the PCP through clearinghouses and IPAs before reaching the health plan, so the plan may receive it 60 or 90 days after the service. Data can also drop at each stop, so a gap the plan pushes to a provider may already have been closed two months earlier.
Brian: Right. You're segmenting by expertise. Is that fair as a description of what's been done in the past? Thank you, Aleesa. Koh, coming to you now. When you get under the hood of risk adjustment, what stands out to you as inefficient or an outlier compared with how other business functions run?
Koh: For us in California in particular, given the delegated model with IPA and MSO arrangements, when the PCP sees the patient, they have to submit the data through their clearinghouse, the clearinghouse to the delegate, the IPA, and the IPA through the clearinghouse to the health plan. Because this data is submitted from one partner to another through the [trading] partner, the data itself gets delayed. By the time the health plan gets the data, the service was done 60 days or 90 days ago. With that delay in the data, we don't have closer-to-real-time monitoring of the patient. So when we push out a gap to the provider to work on, the provider may say, "I already closed the gap two months ago," but we may not have the up-to-date data because of the additional delay.
And we may not get 100% of the data from the private provider office, because when you drop to a different stop, you drop a percent here, a percent there, so we might [not] get a full universe of the data. That's the challenge. So we have to work closely with our delegates and find out which particular providers happen to be outliers. Are we not getting the data in a timely manner?
Brian: There's a delay here that keeps you from having the most accurate picture. Is that fair?
Koh: Yeah. Correct.
Why hasn't technology fixed risk adjustment inefficiency?
In brief: Nicole says it comes down to two things: risk adjustment has always been done this way, and until LLMs, no technology lived up to the promises made to risk adjustment leaders. She says many teams still live in spreadsheets despite having technology budgets, and that LLMs can now pull quality gap evidence and HCC evidence from the same charts in one automated pass.
Brian: Nicole, I want to come back to you now and get your reactions to what you've heard from Aleesa and Koh, and share if that's what you've seen at a market level. You're working with a lot of different organizations. Does this all resonate with you?
Nicole: Absolutely. I have a pretty extensive background across different technology organizations, but it wasn't until I started working with revenue and risk adjustment teams that I started to understand the magnitude of the problem facing US healthcare. This is the problem that so many point solutions were built to solve. There's so much effort put into all of these intricate processes conveying what happens in a clinical setting to a payer, whether that's CPT coding or HCC coding. It's time-consuming, it's manual, and it's ultimately very, very expensive work. Nowhere is that more true than in risk adjustment. Retro coding teams spend hours and hours [poring] over charts of MA patients looking for [missed] diagnosis codes. This is skilled labor, and it takes training. So it's very expensive whether those services are performed here in the US or overseas.
Over the last 20 years, technology has evolved way past spreadsheets and manual chart pulls. But overall, in the conversations I have with risk adjustment leaders, even though many of them have budget for technology, we're still living in spreadsheets or seeing these inefficiencies in our processes. It's not because other vendors haven't tried to deliver solutions. They have. It comes down to two things. First, this is the way that things in risk adjustment have always been done, so it's the accepted wisdom for how to operate. And two, no technology until now has lived up to the promises that are constantly being made to risk adjustment leaders.
I say until now because that comes down to LLM technology. These large language models have been a breakthrough for risk adjustment teams. I'll give you an example. Doing a single pass for quality abstraction and HCC coding is the aspiration we hear from a lot of risk adjustment teams. But these are two very different workflows, and they happen on different timelines. They require different types of expertise. For most teams I work with, these two workflows create redundant labor and costs. The same charts are read by two different teams at two different times for two different reasons. And due to a lack of coordination, they often generate two separate but redundant costs for retrievals.
But LLMs like the ones Charta has trained for risk adjustment can extract all of this information in one automated pass. It can identify evidence of quality gap closure in a chart just as it can identify evidence for an HCC coding opportunity. We can then structure this data and store it in a way that makes the entire chart easily accessible and easy to query. You can click on the HCC code and go straight to the page in the chart that contains the evidence, reducing a ton of labor and cost. It's been a game changer in terms of technology delivery for the market.
How can risk adjustment move from a seasonal sprint to a continuous operation?
In brief: By starting earlier and keeping the work visible all year. Koh's plan promotes annual wellness visits and aligns incentives in the first half of the year, and Aleesa says risk adjustment should stay part of the health plan's regular "heartbeat," not a year-end push. Nicole says vendors also need to learn a plan's operations, starting with friction like redundant chart retrieval.
Brian: We diagnosed the problem with some early answers, and, Nicole, your comments point at how this is changing and becoming something different. Can you unpack that even further: risk adjustment moving away from a seasonal sprint, so to speak, and becoming a continuous operation? Can you talk more about the driving forces behind that? The technology available is a huge component, as you've laid out. But what are any other driving forces, and what do organizations need to do to make this transition successful?
Nicole: I immediately go to redundant chart retrieval because it's such an obvious example of costly friction in risk adjustment, and one that all the people who are watching can relate to. A lot of teams are crying out for a solution, but you won't hear those cries unless you're listening. We heard that pain because we spend so much time listening to risk adjustment leaders talk about their work through an engineering lens. So [as a forward-deployed team], we take everything that we hear in these conversations, and then we build tailored solutions based on that.
For example, we work with a top 10 payer. They're a competitive shop, and they should be. We had to work to gain their trust, and we didn't do it by making a bunch of promises. Our team sat down with them, modeled out a solution, and listened to how their operations work as an engineer would, asking curious questions. This opened up all the inefficiencies around their RADV processes and how painful it was going to be in the next year and a half as all of these audits start coming down the pipe. They asked us where we could help. So in addition to this redundant chart retrieval, I would say RADV, and a few months later, we launched this HCC validation platform that was designed to payer specifications to make their operations and their lives a whole lot easier.
Brian: So if you're going to leverage technology to change this, whoever's supplying that technology has to be able to get in the weeds with you and understand your organization. Correct?
Nicole: Exactly.
Brian: Koh, coming to you now with the same question. If you're going to change how risk adjustment has been conducted traditionally and advance it toward this more continuous operation, what do you have to get right to make that a success?
Koh: Our organization tries to push promoting the annual wellness visit, aligning the incentive program and all of that, in the first half of the year rather than waiting until the third quarter or the final quarter. We see a lot of success by closing the gap earlier in the year. Also, towards the end of the year, everyone's busy with their seasonal scheduled vacations. So pushing to get it in, doing preventative screening, getting it screened, and if they need to see a specialist, promoting that, we see better success. And at the end, it improves the member outcome, pushing that year-round and earlier in the year.
Brian: Being more proactive about getting moving.
Koh: Rather than reactive. Yes.
Brian: Aleesa, coming to you, what would you say folks need to get right to change this process?
Aleesa: From a payer perspective, making sure that risk adjustment stays a part of the general cycle and heartbeat of the health plan, making sure it doesn't stay just that year-end push, and that it stays constant and visible and not purely outcome-based. And from the payer-provider perspective, the benefit of being part of an integrated system is making sure that we are working with our health care system in general to have some of those earlier scheduled visits or earlier partnerships, getting upstream for all of those interventions.
Is risk adjustment understaffed, or is it a workflow problem?
In brief: Often a workflow problem. Aleesa says the answer is always "we need more hands," but adding bodies doesn't fix inefficiency unless you know what each individual is producing, which is why dashboards and metrics matter. Koh, whose plan added nearly 15,000 members in one enrollment period, frames it as people, process, and technology, backed by internal KPIs, vendor SLAs, and AI that reduces manual work.
Brian: Excellent. I want to talk now about the workforce issues that can come into play here. Aleesa, maybe we can stay with you. Have you spotted anything that you initially thought was related to staffing, to having enough folks doing the work, that surfaced as, no, this is a workflow problem that we can resolve? Does anything come to mind, Aleesa?
Aleesa: That's always the question. One of the things that we've had a unique opportunity to do recently was, we merged between [Sanford] Health Plan and Security Health Plan, and we've done a lot of vendor comparison, which is even a philosophical comparison: do we insource, or do we outsource, or what do we do? And looking at these lack of efficiencies, even if you throw more bodies at this, what are those bodies doing? So having a real understanding of what each individual is doing for the output, which is what makes those dashboards and metrics so important. The answer is always, we need more hands, we need more FTEs for that output, but [what matters is] having an understanding of what each individual is producing.
Brian: You have to understand it at that granular level to be able to pinpoint a workflow issue that could be easily resolved. Is that fair?
Aleesa: Fair.
Brian: Koh, coming to you again. Correct me if I'm wrong here, but you added nearly 15,000 members in one enrollment period. Is that correct?
Koh: Correct. Yeah.
Brian: Did that growth expose any workflow problems that were maybe masked as staffing challenges? I'm curious to hear what your thoughts might be there.
Koh: Yes. We have a pretty good forecast, so we foresee [growth] based on how competitive the benefits are in our market and how the organization advertises itself. We have to be tight on the budget, but are we exactly on point with what we have? No. At the end of the day, it's the people, the technology, and the process. Is it the technology that needs to be better, or the process that needs to be better? We use a lot of vendors out there, so how do you monitor the vendors' performance? That's one thing: having an internal KPI dashboard, and the same with the SLAs with our vendors and how they're performing.
And from the IT technology perspective, it's how to leverage AI to improve the workflow in risk adjustment and chart review. So we implemented that and educated the rest of the team. It has been helping the team leverage the technology and process improvement. And we're always looking for opportunities to streamline the process and get rid of or reduce the manual work. That's something we always explore and look into improving.
How do health plans manage change and AI governance in risk adjustment?
In brief: With alignment, clear purpose, and proof. Koh says everyone from leadership to line staff needs to share the same strategy, backed by AI governance, signed BAAs before any AI-enabled tool is explored, blocked unapproved tools, prompt rules staff attest to, and human review. Aleesa centers change on what risk adjustment is for, and Nicole says a proven business case and an internal champion move teams to change.
Where internal objections come from
Brian: That's exciting to a lot of folks, Koh. But what we're talking about here is a pretty significant change, and significant change is always challenging, even if it's also exciting in some ways. As you're making these changes, where are the hardest internal objections coming from? Whether it's staff doing the work, or the executive level being skeptical of certain things, where are you finding objections, and how are you overcoming them?
Koh: At the end of the day, from the top down to the line-level staff, everyone needs to be on the same page. We have full transparency, as much as we can, from the leaders down to the line workers, to understand the strategy for the organization one year, two years, three years. So we're on the same page. And because I'm sitting in the AI governance, I promote the use of AI to streamline. We have a lot of use cases coming in, and then we present them to the governance team and also at the town hall so that all the employees understand what we have and what we block. At the same time, we promote its use. We monitor that closely. And at the end of the day, human in the loop, given that we are a highly regulated organization here in healthcare, with HIPAA and PHI. So we have guardrails on that, so that people understand what they can do and what they cannot do. That's important.
Balancing speed and rigor with AI
Brian: One more follow-up for you there, and then, Aleesa, I'm going to come to you. The stakes are so high in healthcare in terms of data and security and things like that. I imagine there's a tension between wanting to innovate and move fast, but also taking the time to be rigorous and make sure you get all your ducks in a row. How are you navigating what is probably a natural tension for this industry?
Koh: Our vision: stay curious, [think] outside the box, but at the same time, be cautious about what you do. So for us, any tool the organization is approving that happens to be AI-enabled goes through a strong process for the vendor assessment and business agreement. The BAA needs to be signed before we can explore. Anything that is not approved within our organization, we block. And we have [tight] prompt governance: what you can type in the prompt, what is allowed and what is not allowed. The employees understand that and sign the attestation. Providing appropriate training to the staff is also key, so they know what's available to them, so they can leverage and learn from the tool too.
Brian: Resources, and everybody's aligned. Get everybody on the same page. Correct?
Koh: Right. Yes.
Keeping the organization centered on purpose
Brian: Aleesa, coming to you now, thinking about this from the change management perspective as well. As you're trying to leverage more technology and change the way you're doing things, where are you meeting resistance, and how are you overcoming that resistance?
Aleesa: Meaning resistance with the [adaptation] of technology, or meeting change in particular?
Brian: Either. Whichever one most resonates with your experience.
Aleesa: Everyone is always excited about improvement until it means changing something. As we've grown in ACA and MA, started to wrap our arms around the complexity of risk adjustment, and matured in that over the past few years, we've grown in leaps and bounds in our understanding and in the integrated model we've had with our system. That has been a lot of change. One of the things that we've had to consistently go back to is: what is the purpose of risk adjustment, and what are we trying to accomplish here? That's one of the core components anytime you're trying to do change management: focusing on what the purpose is, what success means now, what it means in six months, and what it means a year out. That goes back to continuous improvement, and coming back to what that definition means.
The integration of technology into that can be a little bit difficult because it speeds up the conversation. Our organization, too, has a very rigorous definition that goes back and forth between AI governance and cybersecurity and leans on the conservative side, because we are dealing with patient data and members. It's an interesting time to be on that balance of impact and [patience].
Brian: I imagine getting people focused on the core, what risk adjustment is, why you're doing this, what success looks like, requires some pretty disciplined, consistent communication. Can you unpack for attendees how you continue to center the organization around that mission? It probably sounds a little bit easier said than done.
Aleesa: Yes, it does. Risk adjustment can sometimes get some bad press, too, even if you Google, "What is risk adjustment?" So at the end of the day, it's going back to its core and saying: risk adjustment, from the payer side, is making sure that we are appropriately compensated to take care of our members. That is what it is, in some ways. And then: how do we work with the payers and the provider side to make sure that we're capturing what that means in the right way, so that it can be translated all the way through? Now, that is boiling the ocean into a very simplistic way, and it does require campaigns and attention and follow-through.
Again, we have the benefit of being part of an integrated system, because we have that touchpoint baked in. We do have a sharing of data within ourselves, so we can see a little bit more, and we can have more of that vested interest in some of those things. But it's getting people constantly on board with that same thing and saying, we have the same end goal. We both want to take care of this patient member. When all else fails, coming back to that.
Building the business case for change
Brian: Helping them understand what it's truly about, so you don't get lost in the difficulty of the work and you're focused on that end goal. Thank you, Aleesa. Nicole, I want to come back to you now and get your reactions to Koh and Aleesa. In your experience, when you're working with folks to get new technology up and running and change processes, where is the pushback coming from, and how are folks overcoming that?
Nicole: Both of them put it well. Change is hard. I love the framing around people, process, and technology. We believe at Charta that technology alone will not change somebody's behavior. People have to see the efficiencies that the technology can create in their processes, and this might already be a process that they trust.
I'll give you an example of a customer that we work with, a very large capitation provider organization. It wasn't enough for us to show them the technology, nor should it be. We had to get much deeper into the processes they already had in place, and they didn't understand the true business case for changing the way they operated. So we took a very simple approach, and we proved it out for them. We took a sample of their charts and demonstrated a different process that could lead to different and substantially better financial outcomes. But even the data on its own wasn't enough. You have to make sure the story that you're telling with the data is going to reach the right ears. You need the right person who's going to champion that change. Change management is a whole category on its own. But essentially, that's what we did. We co-built this business case together, and we were able to tell the story that made a meaningful shift in their operations and their business overall.
Brian: Can you talk a little bit about what that story was? What story were you telling?
Nicole: At the end of the day, it was around the financial impact of changing how they were doing risk adjustment. A lot of it was giving them more transparency, supplementing their manual team, and ultimately bringing that real data and proof point into the organization. What AI is best at is focusing human attention on the clinical nuance, getting them out of those manual processes, those day-to-day things that suck all of our energy away. Instead of showing up with a solution and data, we told the story around the financial impact, the overall operational benefit to the entire organization, and how it was going to uplevel their organization as a whole, as well as their members.
Brian: Were there any particular numbers or hypothetical metrics or improvements that carried the day or resonated with or excited people?
Nicole: It was a very significant RAF lift.
Brian: Gotcha. I imagine that was exciting.
What lessons from technology, IT, and policy carry over to risk adjustment?
In brief: Nicole says the most successful technology partners pair outside engineering perspectives with deep subject matter expertise. Koh says many problems are "not a clinical gap, but it is a data gap": the information is in the record, but the code isn't submitted. Aleesa says her risk and governance background helped stand up the program, but day-to-day operations needed more technical oversight, so risk adjustment moved under finance.
Brian: Thinking about your broader professional backgrounds, I'll come to each of you now. Nicole, you've touched on your experience with technology. What did you learn from your past experience that you've brought into risk adjustment that has translated well? And maybe something that didn't translate as well, or where you had to learn something different. Nicole, I'll bring that one to you first.
Nicole: Absolutely. I've been very, very fortunate to be part of teams that have built successful enterprise platforms that fundamentally changed the way organizations operate across multiple industries. The most successful technology partners typically combine outside perspectives from top engineering talent with deep subject matter expertise, and that's exactly what we've done with Charta. The engineering lens is very valuable, and we stay curious, and we talk to our customers and deeply understand their problems. We also bring in true industry experts to help formulate our overall hypotheses around how we can help the industry. I've done that at past organizations. That's what we're doing here at Charta.
We are building out an advisory team with the most diverse array of risk adjustment leaders that we could find. We have a physician, an actuary, an accountant, and a career risk adjustment lead. They all came up through the ranks, and they're from big Blues, provider groups, and smaller regional plans. They all operate differently, and we're taking all of that rich insight to think strategically about how we are going to bring solutions that, like I said, are going to fundamentally change how these organizations operate. There are things that they all do the same; there are common problems, and those are the problems that we're starting to solve. So if anyone is interested, DM me. We're always looking for input. That's how we're building our platform, with direct feedback from the market.
Brian: Getting the right folks around the table, hearing from a broad range of perspectives, and building the right expertise in teams. Thank you, Nicole. Koh, thinking about your background in IT systems design and things like that, how do those lessons translate, or not translate, to risk adjustment work?
Koh: I have a technical background in engineering and also from the IT perspective. And going to the front line and talking from the business perspective and operational perspective on risk adjustment with providers gives me an end-to-end connection. I'm able to identify potential issues and troubleshoot them. What I learned is that a lot of these is not a clinical gap, but it is a data gap. We have a lot of this information in the medical record, but somehow the code is not being translated or submitted to the organization. That's why many organizations spend a lot of time doing retrospective chart review. That, to me, is a big finding and learning opportunity.
In addition to the data lag, we don't get the data in real time. That's also one of the issues we learned: promoting real-time or closer-to-real-time submission from the provider, so that we can get more real-time data from the provider to begin with. That's also something we monitor, and then we provide education. If certain things happen to be missing from the data gap perspective, we provide that feedback so that the provider can submit more complete and accurate information. That has been the education piece that we provide on top of the findings.
Brian: You're able to find those gaps, and maybe I'll call it an opportunity. You're finding opportunities and places to make connections. Is that how you think about it?
Koh: Correct. Yes. It's always an opportunity when we conduct the [chart review] with the provider. It's in there, but we're not receiving it. So we're able to identify which provider, which group. Is it a data gap, or is it a limitation of the technology? Because sometimes the technology is limited to only eight ICD [codes] versus 12, because of the modification. And for the many organizations that have more than that, how [are they] submitting the diagnosis codes?
Brian: Thank you, Koh. I appreciate the example. Aleesa, coming to you now, thinking about your experience in law, policy, and performance excellence, whatever lessons you think are most applicable here. What about that background has served you well in your work with risk adjustment, and maybe what hasn't?
Aleesa: The portions that have served it well were the ability to distinguish risk from noise, especially when trying to get executive attention and saying, no, this needs to move, this needs to be appropriately resourced, especially as we were attempting to get some of the earlier programming off the ground and build up the more basic risk adjustment program as it was probably five or six years ago. The same things, however, are what have been a bit of an inhibitor as you're trying to procure it, which is why, appropriately so, with my support, we've moved it back under finance. I still partner very closely with it, but the day-to-day operations need more technical oversight. The questions of "What's the risk? Are we going to get audited? Who's looking at it? Who owns this?" are good governance questions, and they're good risk control questions. But as we're looking at continuous improvement, [we] need that day-to-day ownership and oversight [to get] to the next level.
Brian: Solving those challenges around who owns this, what we're doing, who's accountable. If you're willing, can you get even more specific about how that's played out? Any specific examples of when you've made progress there?
Aleesa: I made progress very well early on because, again, there was no ownership. I think it was partially moved under me because they were like, well, this has risk, and you oversee risk, so it's risk. And I'm like, sure. So it was a bad stand-up of oversight and of learning and triage. But as it has matured here and grown, especially with the recent merger and acquirement of people with a lot more specific skill sets, the partnership we've been able to have cross-functionally with tech teams and with our data analytics, and the ability from my background to establish more of a cross-collaborative governance structure that's set up for that cross-accountability, has served a strong purpose. So we still have the oversight. One of the things that, from a risk adjustment perspective, I want other areas to learn from is that risk adjustment has a downstream impact from every other place. Everything impacts risk adjustment: data, clinical operations, and everything else. And it won't function well unless it has, not to overuse this word, that cross-collaboration from every single other department.
What will high-performing risk adjustment operations look like?
In brief: More proactive and more unified. Koh expects plans to move from reactive to proactive operations built on accurate real-time data, predictive analytics, and risk stratification. Nicole says leaders should ask why they do things the way they do and what else they want to learn from their charts. Aleesa suggests asking IT what data operations won't see for six months.
Brian: Yeah, because so much is influencing it, you've got to have that cross-collaboration. Thank you, Aleesa. Nicole, coming to you, we've got a couple of questions left here, and I want to pose this one as a forward-looking one. From your perspective, where will high-performing risk adjustment operations look most different from today in the future? And as leaders prepare for that future, what's one question they should be asking their actuarial or IT counterparts now? What's it going to look like, and what questions should folks be asking?
Nicole: Aleesa brought a good example. Sometimes mergers force us to look at processes and expertise overall. But taking a moment to pause and ask, "Why are we doing things this way?" can be very eye-opening. If the answer is something along the lines of, "This is the way we've always done it," it might be time to take a look at, like we said, those people, the processes, and the technology. Take a look at the market. There are teams out there who are ready to make recommendations. There are teams like ours here at Charta who are ready to put engineers to work solving your problems.
The second question would probably be, what else do we want to know about our members? What else can we get from these charts? Our customers bring us those questions all the time. So bring us that wish list. I don't want to say the sky's the limit and overpromise, but over the last three years, LLMs have been making the advances that I described, and the possibilities are so much broader than even two years ago, when NLP was the prominent technology. Those would be my suggestions. I do believe that LLMs are offering a different way forward around unified clinical intelligence that can be leveraged by teams across the entire organization, to your point, Aleesa, with risk adjustment meeting all of those impacts downstream. So I'm excited about what's ahead.
Brian: A lot can change in a couple of years, that is for certain. Koh, let's get your take now in terms of where you think this is going to transform the most in the coming future, and what question folks should be asking themselves.
Koh: Tied to the earlier comment: moving from reactive mode to become proactive. Leverage the technology, and make sure that we get all the data in real time. We get accurate data in real time up front, and at the same time, leverage technology to predict, including predictive analytics, to prevent hospitalization, and at the same time, have an accurate [risk stratification]: which members happen to be high, medium, and low, and pay attention to those so that we can predict when we believe the member will be in the hospital, will be readmitted, so that we can take a proactive approach rather than a reactive approach. That would be what I'm thinking.
Brian: That would be hugely transformative, to be able to get better for the patient and lower costs. The opportunities are huge. Is that fair, Koh?
Koh: Yes. And we have a lot of segmentation of the data, so being able to integrate and look at the entire 360 for the member, so that we have a comprehensive data view per member. That would be best, so that we can talk to the member, improve their health outcomes, and take that proactive approach.
Brian: Thank you. Aleesa, coming to you now, how do you anticipate this is going to change? What would you add or double down on from what you've heard so far, and can you offer up a question for folks to ask the counterparts who are trying to help them?
Aleesa: Yes. Proactivity is key. I would specifically ask IT teams, "What information do you have today that my operations won't see for another six months?" And for actuarial, "What gaps in our data or processes create the most uncertainty in the forecast, and how early can we identify them?" It's the same thing for proactivity, but we can get a little bit more precise in those questions.
Where should risk adjustment leaders start?
In brief: Aleesa says the first step toward continuous improvement is to look back at the previous year, see what didn't work, and take the lessons learned. Koh adds that leaders should look for opportunities to apply fast-moving AI to process improvement and earlier issue detection, and Nicole recommends auditing your processes, talent, and technology.
Brian: Excellent. We will close with one final question. If folks watching this remember just one thing from our conversation today, though hopefully they remember more than one thing, what should that be? Aleesa, let's start with you.
Aleesa: For a successful risk adjustment process, the first thing that everyone should try to do to move forward for continuous improvement is to look backward at the previous year and try to see what didn't work and take the lessons learned. That's where I would start.
Brian: Thank you, Aleesa. Koh, same question to you.
Koh: I would say something similar to Aleesa, but I want to add: look for opportunities where we can leverage the technology with the fast-moving AI. What can we leverage the technology for to help with process improvement and to help identify any issues or improve the process earlier, and then, again, move into a proactive mode rather than a reactive mode? That would be my comment.
Brian: Definitely a running [theme] through our conversation today. Nicole, the final word to you. What do you want to leave attendees with?
Nicole: There's a tremendous opportunity to audit everything, from the processes that are in place to the talent that you have and, of course, the technology, and to think strategically about how you can bring the best, the brightest, something new and innovative to the organization to drive tremendous outcomes. There are tremendous technologies and engineers and thought leaders out there to help everyone do just that. And I hope that's something that everyone will take on.
Brian: Thank you, Nicole. Thank you, Aleesa. Thank you, Koh, for making time to be a part of this conversation today. I appreciate it. On behalf of Becker's, thank you as well to Charta Health for sponsoring today's webinar and helping us put this content together. Thank you for joining us today. For everyone who's watching, we hope you have a wonderful rest of your day. Thank you so much.
Koh: Thank you.
Aleesa: Thank you.
Frequently asked questions
Why is Medicare Advantage risk adjustment so often inefficient?
The panelists point to fragmentation. Aleesa says the work is split across clinical, provider, data, and coding functions with no single owner, Koh says encounter data can reach the plan 60 to 90 days late and incomplete in a delegated model, and Nicole says retro coding remains manual, skilled, and expensive.
How can a health plan avoid a year-end risk adjustment sprint?
Start earlier and keep the work visible. Koh's plan promotes annual wellness visits and aligns incentive programs in the first half of the year, and Aleesa says risk adjustment should stay part of the plan's regular cycle, with earlier scheduled visits and partnerships with providers.
Can quality abstraction and HCC coding happen in one chart review?
Nicole says yes. The two are usually separate workflows on different timelines that often have two teams reading the same charts and paying twice for retrieval, but LLMs can identify quality gap closure evidence and HCC evidence in one automated pass.
Should a risk adjustment team hire more staff or fix its workflows first?
Aleesa says to understand what each individual is producing before adding FTEs, which is what makes dashboards and metrics important. Koh frames it as people, process, and technology, monitored through internal KPIs and vendor SLAs.
How should health plans govern AI tools used in risk adjustment?
Koh's plan assesses every AI-enabled vendor and requires a signed BAA before exploring a tool, blocks unapproved tools, sets prompt rules that employees sign an attestation for, trains staff, and keeps a human in the loop because of HIPAA and PHI.
What should risk adjustment leaders ask their IT and actuarial teams?
Aleesa suggests asking IT, "What information do you have today that my operations won't see for another six months?" and asking actuarial, "What gaps in our data or processes create the most uncertainty in the forecast, and how early can we identify them?"
