Guest: David Kelly, Chief of Ambulatory Operations & Vice President of Revenue Cycle, Mary Rutan Health Host: Praveen Chandran
Introduction and David's Return to the Show
Praveen Chandran: Hi everyone, welcome to another episode of Season 2 of the RC Executive Lounge podcast series. I'm your host Praveen, and I'm super excited that you're all joining us for this episode. I'm very excited to welcome back David Kelly, Chief of Ambulatory Operations and Vice President of Revenue Cycle at Mary Rutan Health in Bellefontaine, Ohio.
David brings over 19 years of healthcare experience, including more than 14 years in revenue cycle and acute care, with deep expertise in patient access, managed care, physician relations, and platforms like Epic and Meditech. His career has spanned community hospitals like Mary Rutan, large health systems including Bon Secours Mercy Health and Premier Health Partners, and a consulting stint at Deloitte. That combination gives him a perspective that bridges the practical realities of small community hospitals with the operational rigor of large national health systems.
He holds a Master's in Healthcare Administration from Xavier University and is a Fellow of HFMA. What I find particularly compelling is David's commitment to the independent mission of a community hospital. He works to ensure Mary Rutan remains a vital resource for the people it serves, while advancing modern operational strategies and financial sustainability in a complex environment. At Mary Rutan, he has led end-to-end revenue cycle transformation work focused on denial management, billing and collections overhaul, pre-service process improvement, Patient Access consolidation, and patient experience. He also created a centralized charge integrity team at Mary Rutan, an organizational model that has been recognized in industry publications for how it sources and centralizes chargemaster and revenue integrity workflow.
Today we want to dig into a topic David has been particularly focused on: claims denial management. He has written publicly about how denial root causes start far upstream of the claim itself, in places like documentation gaps, diagnosis-procedure conflicts, coding from preliminary diagnoses, documentation and coding misalignment, and limited denial analytics. As he has put it, the challenge isn't just the denial. It's everything happening upstream that makes the denial inevitable.
We will talk today about how David is thinking about denial prevention at Mary Rutan, what's actually working to bring denial rates down in a community hospital environment, how he is balancing his expanded ambulatory operations role with revenue cycle leadership, and where he sees the biggest opportunities and risks for community hospitals over the next few years. Let's welcome David Kelly.
David, you have already been on the show in Season 1 and now you're joining us in Season 2 to talk about a very important topic that's really plaguing the entire industry: claims denial management. I'm personally super excited that you are joining us here. David, welcome.
David Kelly: Thank you. Thank you for having me back. Hopefully I'll be able to go from the high-level stuff we talked about last time to some more targeted elements around denials today in a way that'll be useful for your audience. I appreciate the opportunity, as ever. I'm always humbled that anybody would care what I have to say, so thank you.
Praveen Chandran: That's very nice of you to say. But given your experience, I'm sure our audience will appreciate all the deep insights you're about to bring into this episode. We've had conversations, so I'm really looking forward to this. With that, let's dive in.
Denials as a Multi-Year Program: Never Events, Prioritization, and the Shifting Field of Play
Praveen Chandran: Most organizations have claims denial management as a multi-year ongoing initiative. How does it work in your organization, whether current or previous? And in general, how do you determine the priorities for such a program?
David Kelly: Across my experience in both my current organization and prior organizations, you're right that denials management is a multi-year effort. Not only because getting your hands around it can take years, but because the field of play is always changing underneath you. Payers change rules, and you may be entering or exiting payer relationships as an organization. You might contract with a new plan with new rules, or you might exit a relationship with a plan. So not only are you running a strategic initiative to improve denials, but just maintaining your current position is itself a multi-year effort.
How do I prioritize it? I really want to sit down with my team at whatever organization and decide what is avoidable and what is not. In clinical settings they talk about never events, things that should just never happen. What types of denials do we want to call never events? That is where you start.
One of the things I have been a stickler about for a long time is timely filing. Timely filing is an own goal. If you don't get the claim out the door, you did that to yourself. Yes, there are some fringe circumstances where you can incur timely filing denials. Maybe the patient came in as a John Doe, unresponsive in the ER, and you couldn't get their insurance information, and by the time you figured it out 93 days later you're outside of the 90-day window. But you need to exclude that from your reporting if possible so you can really look at avoidable timely filing.
Some organizations might look at provider enrollment or credentialing as a never event or fully avoidable. That depends on strategies that are often outside of revenue cycle's control, because there are organizations that say we need the doctor right now so that patients can get an appointment and we don't lose market share. Depending on your organization, you might classify credentialing denials as a never event, but you need to be in the room for strategy to understand whether that is a realistic commitment to make.
Other categories where you could try to get to never but may not be fully achievable are around aggressive front-end processes for eligibility accuracy. If you cast too wide a net on what you define as eligibility accuracy for a never event, you will probably never get there. But if there are very specific code combinations that you want to target, you might be able to achieve that in a narrower space. Those are the ways I have chosen to prioritize and work with my team, though the definitions can vary from organization to organization based on your strategy.
Top Denial Drivers and the Rise of Records Requests
Praveen Chandran: You touched on this partially in the previous answer. When you have looked at data across multiple organizations in denial management, what are some of the top reasons typically driving denials, and what patterns have you observed?
David Kelly: The patterns have changed over my career. What I have certainly seen in more recent years is that prior authorization is always a burden and always present, but we have actually had good luck getting prior auth out of our top five denials by being assertive on the front end with clearing patients for service. Much of the time, prior auth denials are overturned, both at the point of the auth not being given and then on the clinical appeal after a claim denial, so you can usually fight that.
What I have seen a significant rise in over the last few years is records requests. The plan approves on the front end, and then we get to the back end and either they have already paid the claim or they are kicking the payment can down the road. Either way, here they come with a request for a chart, and we see a large volume of these coming across. When I talk with my peers at conferences, everybody is trying to find a good way to handle it. There are good automation opportunities out there, but some of the automation path is not clean because plans are not requesting records in a consistent way in the remittance that would make automation straightforward. I don't know whether that is intentional on their part or just a limitation of their claims system, but it feels like an arms race.
Realistic Denial Rate Targets, Payer Mix Variability, and the Case for Internal Benchmarking
Praveen Chandran: Based on the data you have seen, what do you think is a realistic target for denial rates for any organization, especially given agentic AI, automation, and operational changes that can be made? I have seen organizations where the denial rate is very low because they have invested heavily, and others where for various reasons they could not invest and they are high, in the range of 25 to 26 percent. What should organizations aspire to?
David Kelly: The answer is a bit nuanced, so let me give some context before coming back to it. I want to advocate that healthcare has done ourselves a disservice by using the same words to mean different things. When we say denials in this conversation, we are talking about claims denials. But broadly we use denials to mean claims denials, denials of service from prior auth, denials of a stay, underpayments that some people call partial denials, and takeback or recoupment situations that some folks call post-service denials. We need to develop new vocabulary or advocate through our industry organizations to show that it is not just claims denials that are the problem. It is the compounded impact of all these different types of denials that creates the administrative burden we all go home and try to explain to our families.
Back to your question: denial rates are going to vary wildly by payer mix and by state. Ohio's Hospital Association has done a great job putting together a payer scorecard over the last few years, and they have tracked the gross denial rate rise from around 12 to 13 percent several years ago to an average of 17 percent across all payers in the state of Ohio today, over about an eight-to-nine-year period. That is 4 percent of claims, which aggregated across the state of Ohio is probably millions of claims.
That is why it is hard for me to give a specific number. It is going to vary from state to state and even within branches of a multi-site provider, because you might have more Medicare Advantage in one location versus commercial or Medicaid. My advice is that if your state has an organization with payer scorecards, benchmark against that. If not, there are nationwide groups, but those are highly variable, and I would honestly recommend internal benchmarking. Can you keep the high-water mark where it is on the wall? If you are at 12 percent, can you keep it at 12 percent next year, and the year after that? Because honestly, if you are holding steady, you are actually winning the race.
Competing Operational and Innovation Priorities: The Paper Cuts Team Model
Praveen Chandran: What are some of the competing priorities that teams face when dealing with this program? There is always a tug-of-war between operational requirements and innovation requirements. Your technology team wants to take innovation bets that are medium to long-term, while operations wants immediate impact. As a leader you have to play ball with both sides while still delivering on the competing priorities. How do you deal with this situation when defining the program every year?
David Kelly: I start with broad concepts and then narrow in. So much of modern management is scarcity of resources. This is not an avoidable problem. It will come up no matter whether you have a 50-person team or a 5,000-person team. A lot of it is routine and repeated prioritization and reprioritization of your strategic priorities, whether you operate in an agile framework or a more traditional organization.
The other thing we have tried to do, with what I would call moderate success, is I borrowed a concept from Jeff Bezos. I heard him on the Lex Fridman podcast, and he talked about how they built a team at Amazon to deal with what he called paper cuts, the thousand tiny operational things that never get attention because the big things always get the priorities. One-click purchasing apparently came from that paper cuts team. So even at our size, we carved out a role designed to focus on those small operational frictions inside revenue cycle, to optimize and streamline over time. I say moderate success because that person keeps getting pulled into big-picture work, but the concept is right. If organizations can segment out even one person to go after those small things that are constantly in the way of your team and remove friction, I think that is a positive investment.
Measuring Success in a Multi-Year Denials Program
Praveen Chandran: For a program like this across multiple years, what do success targets look like in any given year? The nature of the question is not to come up with a quantitative value but more about when would a leader like you say you are very close to achieving the vision for this program, given that there can be diminishing returns beyond a certain point?
David Kelly: The good news from a job security standpoint for rev cycle leaders is that shifting field of play I mentioned. To some extent, success looks like treading water. Can you keep your same position and not lose territory? That is one key element.
Another element is going back to those never events. Have you eliminated them? If you have gotten timely filing or credentialing or some narrow combination of codes out of your top three or five denial causes, you can probably consider yourself successful. We have had some good success with that at Mary Rutan across my two stints here. That said, it is a never-ending battle, and every month or quarter there is a new hot topic we are managing.
The other element is whether you have folks coming to you not just with here is a denial problem, but with here is a potential prevention. If they are doing that, even if the preventions are not always successful, it means you have ingrained a prevention mindset into your team. So many people correctly say that denials are all about prevention. You will never prevent all of them, but if you can prevent two, three, or four percent, that could be thousands or hundreds of thousands of claims depending on the size of the provider.
Praveen Chandran: Great set of insights, David. With that, let's take a short break. When we come back, we're going to talk to David about a recent claims denial management initiative that his team led. Please stay tuned.
The Clearing House Migration: Initiative, Lessons Learned, and Tracking Success
Praveen Chandran: Welcome back. David, could you walk our audience through a recent claims denial management initiative that you and your team led? Let's start with what challenges you were solving for.
David Kelly: The most recent significant effort at my organization has been a clearing house migration, which we executed in the first part of 2026. Some people might ask how that is a denial management effort. It is a denial management effort in the sense that the question is: are you implementing it properly so that you do not see major spikes in your denial rates? I am very proud of my team for this. We had a fairly aggressive timeline to go live, so we built in a longer stabilization period than the typical 90 days, ours was closer to 150. But we did not see major spikes in gross denial rates. I am very much looking forward to our new partner being able to help me get better data on write-downs and net denial rates. The prior partner was not helping us achieve that, and getting better visibility into net denial rates was one of the main reasons we opted to change clearing houses in the first place.
I can't claim much of the credit. I picked the vendor, that's about it. My team implemented successfully with very stable metrics throughout.
Praveen Chandran: It sounds like the implementation was done internally rather than bringing in an external vendor to run it. How did you make the decision of internal implementation versus bringing outside help?
David Kelly: There was not much of a debate about whether to get outside implementation help. We have a fairly seasoned team, a number of strong performers and a couple of rock stars, and I am genuinely blessed by that. These folks already went through a clearing house migration about six years ago, and most of those people are still around. They have been through it before. The vendor we chose made commitments I trusted them to live up to in terms of implementation support, and our IT team is similarly strong on interfaces. We have a lot of veteran players across the organization, so we trusted what the vendor had said, trusted the expertise of the team, and budgetarily I probably just couldn't have afforded outside help anyway. It was somewhat risky, but the vendor lived up to their commitments and the team delivered on their potential.
I wish I could tell you it was a big framework with 27 Excel spreadsheets where we debated everything in a pros-and-cons analysis. It was mostly: can we do it, on what timeline, let's do it. There was more financial analysis around the vendor change itself, but in terms of whether to bring outside help for implementation, it was pretty much as simple as that.
Praveen Chandran: Despite having an experienced team on both sides, a migration of this scale is obviously risky. Were there any pitfalls, surprises, or lessons learned that you would like to share?
David Kelly: The implementation from kickoff to go-live was about 100 days, which is relatively quick for a clearing house given the operational risk involved. I have heard horror stories of organizations not being able to get claims out for months at a time. We were proud of that timeline.
In terms of pitfalls, the number one lesson is training specificity. The vendor had their way of doing training and we had our expectations of how it should be done, and those were not perfectly aligned. We had some snags. I went back to the sales team, told them we needed better training, more sessions, and more remedial options, and to their credit they jumped right on it and got their trainers lined up to meet the needs of my team. The practical issue was that people understood the product was okay but didn't necessarily know how to work within it.
Training gets left out of implementation plans a lot. My word of advice is to make sure your project plan has adequate training and that it doesn't just say "user training" as a line item. Specify whether these are live sessions or web sessions. If there are multiple shifts and multiple sites, how are you dealing with that? Laying all that out with enough specificity is really where the rubber meets the road, because that is what determines whether people can actually use the software.
Praveen Chandran: You mentioned going through financial analysis for this change. How does a leader like yourself track success for a migration program like this, given both the direct financial impact and the impact on claims management metrics?
David Kelly: It is relatively common sense. Is it on time? Is it on budget? Did the scope stay as planned? If you get halfway through and have an argument with your vendor about scope, you didn't scope it right. So avoiding that debate is itself a binary success indicator.
And then whatever your normal operational metrics are: if you were measuring clean claim rate before and you're measuring clean claim rate after, is it in the ballpark, certainly post-stabilization? If you can get back to your pre-migration performance level inside of the stabilization period, whether that's 30, 60, 90, 120 days or more as ours was, I would call that very strong success.
Our stabilization was on the upper edge of what I would call tolerable, but I was okay with it because the go-live was relatively quick. When you add it all up, maybe seven or eight months total, and that to me is quite reasonable.
Praveen Chandran: Thank you so much, David, for talking through that migration initiative in detail. With that, we are going to take a very short break. When we come back, we are going to discuss the role David sees agentic AI and other up-and-coming technologies playing in reducing claim denials. Please stay tuned.
Agentic AI and the Vision of Zero-Touch RCM
Praveen Chandran: Welcome back, David. The last time we had a conversation, you made a very interesting comment about agentic AI being the new shiny object in the world of RCM. Not only myself, but a lot of our audience connected deeply with that message because they all hear about agentic AI everywhere but there is limited clarity around what can actually be done with this technology in an area like claims denial management. Can you talk us through how a leader like yourself sees the role agentic AI plays in claims management over the next 12 months versus the next three to five years?
David Kelly: I have a mindset at war with itself when it comes to agentic AI, because I see the potential of it to dramatically and positively overhaul the revenue cycle from both the provider and payer perspectives. Think about how many interactions between us and plans could happen autonomously, more rapidly, at more scale, and hopefully with more accuracy for both parties. That could be extraordinary. We are not there yet, and the challenge is what does it take to get there, how much wasted time, motion, effort, and money will be spent on the way, and what are some of the pitfalls on that pathway.
One of them is general public backlash against agentic bots. Between the time we last talked and today when we are recording this, I have seen several videos where people are catching out the AIs in call centers, making social media posts about exposing them. These may be influencers who don't fully speak for the general populace, but the message is clear: when people say they want to speak to a human, don't tell them they're speaking to a human and then send them to an AI. There are potential pitfalls around people's willingness to engage with AI in patient-facing scenarios. But in scenarios where we engage not with the public, but either internally or between us and plans, I am very optimistic. The road map itself is just not clear to me yet.
Praveen Chandran: Last time we spoke, you brought up the term zero-touch RCM with respect to agentic AI, which really caught the attention of our audience. You were one of the first executives to talk about zero-touch RCM in this context. Do you think there are areas where agentic AI can truly shift the game for claims denial management? Not incremental improvements, but a genuine two to five to ten times shift in claims management performance and financial impact?
David Kelly: Yes, I do think so. Claims denial management as we have scoped it for this conversation is one of the larger payoff areas in rev cycle for agentic AI. Go back to that Ohio Hospital Association gross denial rate of 17 percent. If an organization is submitting a million claims a year, you are talking about 170,000 claims that have to be touched, assuming you appeal all of them or resubmit or file corrected claims. And I am starting to see stats emerge now that not all denials are even getting touched.
So first: can an agentic AI allow you to touch everything? Historically, I don't think most places are touching every denial, and it is hard to walk away from revenue. If an agent can help you get from touching large denials to touching all denials, that alone statistically is going to yield more appropriate payment. As long as the cost of that agent is not egregious, there will be a financial return.
Beyond that, can the agent work more quickly? Then you have time value of money. Can it get to resolution in fewer passes? A lot of denials need two or three passes to actually get paid. But can an agent figure out, based on a pattern, that step one will lead to step two, and step two to step three, and then jump directly to the resolution? Fewer passes is more of a time-value-of-money play, but it could also be a get-paid-at-all scenario, because if it takes too long to jump through all those hoops, you might hit a timely filing issue along the way.
And then the agents could stay on top of payer rule changes more quickly. If you can build a good method to upload payer policies to a model that feeds an agent, the agent can know those policies far better than a human can. I read something a couple of years ago that tracked something like 3,000 policy changes a year across a large client base. Who could stay on top of that? But if you are loading those updates into a model that feeds an agent, it probably would be able to.
Adjacent Technologies and the Emerging Role of Robotics
Praveen Chandran: One last question for this section. Outside of agentic AI, are there other technologies that you are watching closely that could have an impact on claims denial management?
David Kelly: This might feel like a wild tangent because it is more in my ambulatory space than strictly rev cycle, but robotics. I think robotics could be dramatic in terms of helping us deal with labor shortages in the healthcare delivery system and broadly in society. We have various demographic and workforce challenges facing us, and the robotics potential is growing every year. There are hospitals starting to have robotic delivery of medications or meals, and I think that is probably just the beginning. Who knows, it could come into the rev cycle space through patient access or financial engagement at some point, but I see it primarily in the care delivery space for now. And there's no arguing with demographics.
Praveen Chandran: Great set of insights on emerging technologies. Thank you so much, David. With that again, we're going to take a short break. We'll be back.
Key Metrics: Yield as the A1C of Revenue Cycle
Praveen Chandran: Welcome back, David. Every morning when you're in the office, is there a specific claims-related metric that you obsess over, something that if it goes down there are serious repercussions?
David Kelly: The number one metric I am routinely looking at, though unfortunately it doesn't get reported daily, is yield. It is the old accountancy principle that cash is king. By yield I mean net patient services revenue as a percentage of gross revenue. If that number is down, it could be denials, it could be underpayments, it could be claims locked up and not getting out the door, it could be unbilled revenue. I think of it the way a diabetic patient thinks about A1C: for the health of my rev cycle, I'm looking at yield. That is the number one indicator, and then I spin off from there to look at more specific metrics with my rev cycle team.
Biggest Barriers to Step-Function Improvement and Leadership Advice
Praveen Chandran: The concept of claims management has existed for a very long time, and denial management as a multi-year initiative exists in many organizations. What do you think are the biggest challenges faced by organizational leaders in being able to bring step-function improvements in denial metrics?
David Kelly: If I had to pick the top three biggest challenges, the first, ironically, is the new shiny. Rev cycle often gets pulled from one new shiny to the next. A few years ago it was all the rage at conferences to talk about blockchain. Fending off new shinyism can come from legitimate emerging technologies or from a physician who heard about something at a professional society meeting. Chasing those diversions takes focus away from the sustained discipline that denials prevention requires.
Second, going back to something I mentioned earlier: scarcity. There are only so many people and so many resources that any given organization has, and there is always more to do in rev cycle. I don't think most organizations are anywhere near the point of diminishing returns. Most have a lot of ways to go before they get even close.
Third is personnel and workforce challenges, specifically churn. A lot of organizations have turnover in roles, whether that is staff turnover from attrition or the more concerning pattern of leadership turnover. Maybe they have great staff, but leadership is turning over. Rev cycle folks are aspirational, we want to move around, or we are in a very metrics-driven role and if we are not hitting the metrics we get let go. When you add those three together, you have versions of distraction, shortfalls in resourcing, and lack of institutional knowledge, and that combination makes it hard to make progress.
Praveen Chandran: If you were to give one piece of advice to another CFO or RCM leader tackling similar denial management challenges, what would it be?
David Kelly: You have to bring denials to the operators. And I don't mean the rev cycle operators. I mean the department and the clinical operators of the organization. Because first, it is their book of business, they need to understand it. Second, a large number of denials start in their shop. And third, in doing so they bring non-rev cycle and non-finance perspectives to the efforts you are undertaking and can help you think about things differently and solve problems in a way that might have taken you three years to figure out on your own, because clinicians think differently than financial people do, and that's a good thing.
Praveen Chandran: Has there been a recent book, podcast, or framework that significantly influenced your leadership thinking, especially around claims denial management?
David Kelly: Nothing I have read or listened to recently hit me specifically on claims denial management, and I wish I could say that. I am an avid consumer of podcasts and try to listen to all different kinds of thought leadership, life coaching, history, all of it, because I feel it all helps build context for tackling different problems. I just got Scott Becker's book about building great businesses and I am interested to read it. Scott has a great perspective on healthcare and has brought a lot of people and connections together over the years, much the same way you all are trying to build thought leadership in the rev cycle space.
Stepping away from the specific question, the broader advice I would give is just the value of community. Being able to have a network that you can reach out to about a problem, whether by text, LinkedIn, or a podcast, is so valuable. There is a consultant cliche that if you have seen one hospital, you have seen one hospital, and it is sadly true in the sense that you would hope there would be more similarities, but sometimes it is just so different. Getting somebody else's perspective can make something click for you that you have been working on for months.
Praveen Chandran: David, today's conversation has been nothing short of amazing. I'm sure our audience will get a lot of insights from such a critical topic like claims denial management. Thank you so much for candidly sharing all these wonderful insights. Our audience would be looking forward to hopefully having you on the show again. Thank you so much for joining us today.
David Kelly: Thank you very much. I appreciate the opportunity. I just hope somebody gets something out of listening to it. I want to share and help us all grow and improve.
