Key takeaways
- CMS has expanded RADV audit scale drastically, with all 550 active Medicare Advantage contracts now targeted for audits. The expansion includes overlapping timelines for RADV audit response that will create significant operational complexity for many MA organizations.
- Minimizing provider abrasion begins with cleaning up internal data, consolidating provider/TIN contacts, and leveraging EMR APIs or record repositories before reaching out to doctors.
- Implementing a robust chart triage process to confirm complete dates of service early —rather than discovering missing records late in the coding stage aids—reduces last-minute scrambles for charts.
- Treating every retrieved chart as a persistent data asset—by indexing dates of service, HCCs, and HEDIS measures to reuse across future audit cycles and risk adjustment programs—is the future of risk adjustment operations.
Introduction
Rachel Lee: Hi, everyone. We’re excited to have you join the first of our three-part series on best practices for RADV in 2026. My name is Rachel Lee. I'm head of payer partnerships here at Charta Health. Charta partners with health plans and risk adjustment organizations to support retrospective coding, HCC validation, and RADV audit operations. We work with a ton of payer validation teams, retrieval vendors, coders, audit leaders—you name it. And we are excited to share what we've seen work and what doesn't work when it comes to managing complex RADV programs.
Here with me is my colleague, Tam Pham, VP of health care solutions for value-based care. She's been on both sides of the table—payer and provider—and really knows the RADV process better than anyone else here. So, Tam, I'll toss it over to you for a quick intro.
Tam Pham: Thanks, Rachel. Hi, everyone. I'm Tam Pham. I'm the VP of health care solutions here at Charta. I spent many years in health care operations working across payer, provider, and now on the vendor side. And much of my career has been focused on risk adjustment coding and value-based care. I was involved in one of the early CMS RADVs. I was lucky enough [to be in the industry] in 2017, when a small number of plans—I think at that time there were seven plans—were audited. And so I was lucky to have had the opportunity to learn what happened in the early days and see how it's evolved. I’ve been with Charta since 2024, and I'm working with some of the brightest minds to use the latest technologies to solve some of health care's toughest problems. So thanks for having me here today to share some insights.
CMS RADV: The changing regulatory landscape
Rachel: We are very lucky to have you here, Tam. So this three-part webinar series is designed to share the practical lessons learned, industry best practices, and the operational strategies that orgs can adopt to improve RADV outcomes. We'll cover a couple things. In the first phase of the RADV audit, we're going to cover chart retrieval. We'll first talk about the new regulatory landscape, then we'll go to how to best juggle overlapping audits and navigate challenges. Last but not least, we'll share some retrieval strategies to set you up for success for this year and beyond. If you have questions at any point during today's webinar, use the comment feature on the side window. We'll try to get to your questions as we go. Otherwise, we'll follow up individually on anything we cannot get to today.
So let's start here, Tam, with a quick voiceover on the new regulatory landscape. Most folks dialing in here today have been through a RADV before, but as we all know, it is 10x or 100x the chaos this year. Tam, what's been the most challenging change and why?
Tam: It's changed quite significantly. I think to really understand it, we need to zoom out a little bit and look at the space itself. When we think back to Payment Year 2020, lots has changed: Membership has grown. More people are in MA now than ever. Right? There are more MA lives now than fee-for-service lives. So as more lives are in this payment model, naturally, CMS has an eye towards accuracy because the majority of Medicare dollars, or more than half of Medicare dollars, are now spent in risk adjustment payments.
Historically, not all plans got a RADV. Right? The probability was very low. Prior to this most recent announcement of every single payer getting audited, it was only about 60 plans being targeted. Most organizations thought, "Well, when it happens, it happens, but likely it may not happen to us." So many people and many organizations, especially the smaller ones, may not have planned for it. We all know that's just not the case anymore. With 550 active MA contracts today, every single one will get audited—which is a massive increase in audit scale, right? Think what that means for not only the payers, but the providers who have these records: There's so much provider abrasion nowadays.
So that landscape definitely has changed. [And] the model itself has changed. We know that CMS is getting rid of non-specific diagnoses; something like major depressive disorder, unspecified, no longer risk adjusts. So we know there are much more implications with more diagnosis specificity, and how much harder it is to validate a particular HCC if the documentation or the specificity isn't there.
Another change is the two records per HCC rule. In an example of three HCCs that need to be validated, you can submit up to six best records. But which records should you submit? If there was one record that validated all three, should you submit six best records or only three? We'll have a [later] webinar to shed some light on some of the best practices there, on what you should be doing and how to select the best records. But the landscape has changed in terms of the methodology and how hard it is to retrieve all of these records.
Chart retrieval is the hardest part of RADV. Best record selection is the most critical, but retrieval is the hardest part, and that's what we'll be talking about today: retrieval and what can be done.
The other major change, I believe, is how RADV gets expanded with technology that is available today. CMS now has access to technology that can do many more audits.
Let me bring up what the RADV timeline looks like today. This is sort of overwhelming for some of the payers out there, right? Five RADVs in a twelve-month period is tremendous. What many organizations are having problems managing is when that third RADV comes. The first two, if you're a smaller payer, are not that challenging. But when that third RADV comes into place and you're right at that August timeframe where you have to submit the best records for the March RADV (payment year 2020), and then you're getting hit with the 2024 RADV while you're already underway with 2021... it gets much more complex as continual and overlapping RADVs occur.
I just want to show you the next slide here—for some payers, hundreds of audits are going on. So I have great appreciation for all those out there undergoing these RADVs. Those of you attending today can feel the pain, and I really appreciate it.
Rachel: That is a lot of contract audits to cover in just a short period of time. We got a question early on here, Tam, which actually transitions us to the next topic here. How can retrieval teams keep everything organized and prevent unnecessary back-and-forth to get all the charts completed as fast as possible, knowing that they're going to have hundreds of RADVs coming down the pipe very soon?
Retrieval strategies and minimizing provider abrasion
Tam: Everyone needs to have a retrieval strategy in place when there are overlapping RADVs. The number one thing and the hardest thing is how you minimize provider abrasion, because doctors are pretty burnt out. You could see, with hundreds of RADVs, continual knocking on their door to make sure you're getting records. Doctors are burnt out with some of these retrieval efforts because it's not only retrieving for RADV, it's retrieving for all the other risk adjustment programs everyone has going on right now.
Another thing you want to look at for successful retrieval is looking at your processes and the data you have in-house. You want to make sure you're improving those data and processes. The third is how you streamline some of your team's operations and your vendor operations. What are some of those things that you can do? Number four is really to maximize every retrieval that is occurring today and make sure you understand the value in every chart that you have.
Rachel: You mentioned minimizing provider abrasion pretty early on, and that's actually not a phrase you typically hear as a starting point for a lot of payers here. But given your background on both payer and provider sides, this makes a lot of sense to me. Can you unpack that a little bit for the audience? What does minimizing provider abrasion actually mean in practice, and why does it matter so much for retrieval?
Tam: To minimize provider abrasion, I really think it starts again with the data you have in-house. If you have bad data and you're knocking on someone's office door, calling them, or emailing them—whichever way you are contacting that doctor's office—you need to make sure you clean up your data. Let me bring up an example on this slide.
What I've seen, especially in the charts that we receive here at Charta, is where one patient [record] is retrieved three or four times from the same office. In this example, you can see that whoever is pulling the data for you didn't group three of the providers into one office or TIN. This member has three claims with three different doctors who all belong to the same medical group. Under that medical group, it's a staff model—it's one record, yet someone is calling that office three times. What makes it even worse is the downstream impact of this: now you have that record three times. The coder is going to read it possibly three times, and if you have three different coders, you might get different answers. Then what do you do? Whose coding results are you going to use? Avoiding provider abrasion starts with cleaning up that data before you even outreach to that office.
Second, try to understand: do I even have to reach out to this doctor's office? Do I have the records in-house today? If you know your organization has been building a record repository, make sure that record repository gets queried first. Second, leverage a vendor who has APIs into these doctors' offices that can retrieve records automatically. I know some payers out there have started their own method of connecting automatically to these offices. So before calling any offices for these records, see if you have them, or see if they can be obtained electronically before asking the doctor's staff, office manager, or back-office personnel to pull that record. That will really help reduce provider abrasion, because we all know doctors are just inundated with these requests nowadays.
Rachel: Got it. So clean data and smarter provider workflows have to go hand in hand to make sure your RADV programs are set up for success from the get-go. Now tell us, what about when the records actually start coming in? What's mission-critical at this step?
Tam: Mission-critical is opening up the record, which I don't think necessarily happens often enough. I think what happens is the record is retrieved—whether electronically, by sending someone out, or getting it through a fax—and someone just checks a box saying, "Oh, the record is received." But I believe there needs to be a triage process. There should be someone saying, "I expect five dates of service in here." If you have a record with only three dates of service, you want to make sure you note it as an incomplete record so you can go back out and retrieve it again. That triage process before it goes to coding becomes very important, because you don't want to find out later on that the record or date of service hasn't been retrieved yet. I've seen that happen, especially—hopefully those involved in the payment year 2019 RADV learned a lesson there—and a triage process would improve your ability to validate things at a faster rate, versus waiting until the end where you may not have enough time to pursue those missing dates of service.
Rachel: Oh, man. We've all been there. You check the box on "got the records," open the file, and then see that there is nothing usable. It happens to the best of us. But let's fast forward through that pain. Let's assume we've all watched this webinar, we're following Tam's best practices, and we've done the work right, so every record retrieved is complete and we're in that state of nirvana. Moving on to the next phase then: all of that effort we just put in, does it have to evaporate after the string of audits? Or can we build something more here so that all the effort we're putting in to build the perfect retrieval program doesn't get wasted?
Maximizing chart value and managing vendors
Tam: Yeah, and I think many more people are thinking about this right now. Everyone's in the middle of a retro season. Hundreds of thousands or millions of records are getting retrieved in certain organizations. So what can you do now with those records?
Let's take a look at all the different efforts underway when retrieval happens. There could be retro records being retrieved, HEDIS records being retrieved, or records for prospective chart review. Maximizing the value of every chart retrieved is really about understanding what is inside that record. What is inside that record that can be stored? With today's advances, you can have AI look into that record. If you expected five dates of service and only got three, you know that upfront. But if you want to know this for the future, you index that record. That record has five dates of service from five providers, and when you get hit with a RADV, you realize you already have every single one of those dates of service. Then you don't have to go after that chart again, which minimizes provider abrasion. Minimizing provider abrasion is critical because everyone is very protective of their provider network and avoiding doctor burnout. Technology today can read the content—not only dates of service and providers, but whether it validates a particular HCC or what HCC is hidden in the record. You can store that type of metadata with today's technology. You can see which HEDIS measures could be a positive hit. So there's so much that can be done with the records you get today.
For those retrieving records for retro season, ask your vendors to use technology to understand what is inside that record so you can store it in your clinical data repositories along with the chart. That's how technology can be leveraged to maximize retrieval efforts.
Rachel: Love that. Jot it down, folks: every chart you're already pulling is a potential data asset, and one that you can query later for future RADVs, retrospective coding, or prospective programs. I really love that framing. So let's go through some questions from the audience. I think you're going to like this one, Tam. Someone asked a loaded question: What are some strategies for managing retrieval vendors more efficiently?
Tam: This is a tough one because many of you have existing SLAs in place with your retrieval vendors already, and today's SLAs probably will not fulfill what you really need for review purposes. You need to work with your vendors and say, "Hey, I need you to implement chart triage." What I mean by chart triage is having someone check off whether that chart has every single date of service being requested. A lot of times, when you give the retrieval vendor the list to pursue, you may not even be giving them expected dates of service for that chart. Data processing for RADV is different from traditional retro retrieval. You shouldn't just use your traditional retrospective chart retrieval workflow. For RADV retrieval, define what a complete record looks like and put all of that into the SLAs and vendor contracts.
Rachel: Put it in the contracts. That all sounds great. So let's get into a few more of the audience questions here. This is a good one. Tam, what do you do when a vendor comes back to you and says, "Can't get the records"?
Tam: That happens, and it's one of the most frustrating things because you wonder if they're really doing their job or if they truly can't get the record. When they call, are they calling offices at the same time every day?
Part of this goes back to defining what it means when they say they "can't get the record," and putting those expectations into SLAs. But sometimes you want to question: if they did everything, what else could be wrong? Often, the data itself is wrong.
Doctors might be burnt out and hesitant to give you the record right away, so you need to ask specific questions: Why can't we get the record? Is it a lack of available staff? Can't they find the record because payment year 2020 or 2019 dates of service were archived? Really get to the reason why they can't get the record, and then establish different escalation paths. Most doctors are in risk arrangements and want to help you validate, while others are burnt out, or no longer have the record due to EMR changes, member terminations from five years ago, or patient passing. Understand and try to help the doctor through that by deploying resources if needed. When it's an outright refusal, you have your legal team and contractual language to fall back on, but you don't necessarily want to go that route immediately. Really understand the "why" first before taking the legal route.
Rachel: Awesome. Thank you for sharing that. Now this is a great build-versus-buy question. Tam, how do you determine if you need to vend out chart retrieval or the entire RADV itself?
Tam: As it relates to chart retrieval, do you have a team who knows how to do chart retrieval? It takes a lot of time. It's not just sending a request list out to the doctor's office: It involves calling them, asking for the best transmission method, and following up with different call scripts. You might need to call 10 times or more before someone picks up the phone to confirm if the record exists. If you don't have the infrastructure, call scripts, or personnel, it makes sense to vend it out based on volume. You might think 35 patients is easy, but if you only have one team member who is already busy with other duties, you may want to ask a vendor to help retrieve those 200 records, especially since vendors may be gatekeepers for some of those records. Whatever you can do to minimize provider abrasion, consider who the gatekeepers are and vend it out on the retrieval side if necessary.
On the actual best record selection and coding side, if you have a 35-member sample and one experienced coder, you should be able to handle it internally. But if your team lacks experience with CMS's latest guidelines, or if it's the very first RADV your organization has faced, I advise bringing in a consultant. If you have a team that can do it and has done it before, definitely keep it internal. Some vendors will do the work, but you'll still need to provide heavy oversight, which can be as time-consuming as doing it yourself. Feel free to ping me on the side with your specific scenario, and I'm happy to walk you through considerations.
Future outlook on CMS RADV
Rachel: Awesome. Very cool. And now for our final question. Looking a little bit forward, Tam: In your point of view, what's going to happen after CMS is done catching up [with Payment Year RADVs]? What is gonna happen next?
Tam: Let me bring up the RADV timeline overlap slide. You can see that the very last submission will be due in August 2027. Looking at this timeline, we might think we're going to be done with this cycle catching the industry up for payment year 2024. But right as we think we're done, in August 2027, the industry will likely be hit with payment year 2025 RADVs because CMS no longer wants to be behind.
I wouldn't be surprised if CMS then starts doing condition-specific RADVs—like OIG acute conditions that warrant an inpatient admission where there are no inpatient claims for a stroke that occurred. Why not expand data validations from a specific HCC? CMS will get there, and the technology is available for them to do it. Everyone needs to be much better about HCC validations, adds, and deletes. If you're proactive about your adds and deletes—specifically deletes—extrapolation can be mitigated. I encourage the industry to do much more data validation because more is coming.
Rachel: More, more, more. Just never ends.
Tam: Exactly.
Rachel: Well, that's all the time we have for today. Thank you all so much for spending the last thirty minutes here with us. We'd love to have you join for the next session, which is airing July 23, covering how to review records efficiently and consistently.
If you have a question you'd like us to cover during that session, drop it into the comments now, or you can send me a note directly at rachel@charterhealth.com.
Thank you all for staying on, and see you on July 23.
