Autonomous coding readiness: What healthcare leaders need to know

0:5:57.906 --> 0:6:13.186
Cofer, Lasha M
Good afternoon, everyone, or good morning, wherever you are in the world. And welcome to our Autonomous Coding Readiness webinar. We're so glad you're here with us today. I am Lasha Cofer. I will be the moderator for today's panel discussion.

0:6:14.466 --> 0:6:32.706
Cofer, Lasha M
Autonomous coding has moved quickly from an emerging technology to an executive level revenue cycle strategy. But success is not determined by accuracy alone. It comes from three things, and they will frame our conversation today.

0:6:33.346 --> 0:6:36.306
Cofer, Lasha M
Discipline qualification of the right work.

0:6:37.266 --> 0:6:58.146
Cofer, Lasha M
Trust built through local validation and governance, and analytics as the control tower. So here's our premise. This is not a coding engine or an isolated automation project. It is a governed, measurable revenue cycle operating model. With that, we have a very full agenda.

0:6:58.226 --> 0:6:59.746
Cofer, Lasha M
So let's go ahead and get started.

0:7:2.466 --> 0:7:20.626
Cofer, Lasha M
It is my honor to introduce our esteemed expert panelists for today's session. Before I do that, I should introduce myself. Again, I'm Lasha Cofer, and I will be moderating the panel discussion for today. I've been in the HIM industry for over 25 years.

0:7:20.786 --> 0:7:39.626
Cofer, Lasha M
spanning operations, coding, consulting, healthcare technology, and I currently serve as a Director of Strategic Product Management at Optum. We're a lead strategy and innovation for our middle revenue cycle technology solutions. When I'm not working at Optum, I serve as an adjunct instructor.

0:7:39.906 --> 0:7:48.946
Cofer, Lasha M
for Ross College. I have a master's in health systems management and I'm credentialed as a registered health information administrator through AHIMA.

0:7:49.986 --> 0:8:9.346
Cofer, Lasha M
Temeka Lewis is the Director of Automation Capability, Middle Revenue Cycle Product, and has over 20 years experience in the healthcare arena as a health information management and revenue cycle professional. She joined Optum as a subject matter expert for our print publications,

0:8:10.226 --> 0:8:29.26
Cofer, Lasha M
then moved to sales within our provider market solutions. In her current role as the Director of Autonomous Capability within MRC Pillar, she leads an agile, multi-layered team of HIM SMEs who guide the artificial intelligence and NLP developers

0:8:29.346 --> 0:8:48.306
Cofer, Lasha M
to optimize performance of AI being used within our solutions. Temeka has an MBA in Health Services Management and is credentialed as a CCS and CDIP through AHIMA. Dan Schofield is a product leader at Optum focused on healthcare analytics,

0:8:48.786 --> 0:9:8.266
Cofer, Lasha M
revenue cycle transformation, and data-driven innovation. He works with health systems to deliver insights that improve performance and support better decision-making. And rounding out our panel discussion for today is Michele Maier. She's the Director of Strategic Product Management at Optum.

0:9:8.666 --> 0:9:28.146
Cofer, Lasha M
where she leads initiatives focused on the Optum Integrity One workflow. She holds a master's degree in health management systems and analytics, and is a faculty member at CCAC, teaching in health informatics programs. With more than 25 years of experience in the healthcare industry,

0:9:28.466 --> 0:9:39.26
Cofer, Lasha M
Michele has worked extensively with hospital organizations to transform the middle revenue cycle and optimize utilization of their solutions.

0:9:40.866 --> 0:10:1.946
Cofer, Lasha M
So with that, we'll go ahead and get started. Here is where we're headed. Our goal is that you leave knowing how organizations are using artificial intelligence or AI to automate routine coding with human oversight and what a scalable compliant path to revenue cycle improvement really looks like.

0:10:2.66 --> 0:10:21.26
Cofer, Lasha M
including the lessons learned in moving from experimentation to enterprise adoption. We will cover 3 themes. First is scaling autonomy through discipline qualification. Not all work should be autonomous. So how do you identify the right work

0:10:21.66 --> 0:10:42.546
Cofer, Lasha M
and know you are ready. Second is building trust through local validation and governance, because trust cannot rest on a single accuracy score. And third, analytics is the control tower, the visibility that lets you manage the whole system, not just individual coding decisions.

0:10:43.346 --> 0:10:45.666
Cofer, Lasha M
So let's start with the very first thing.

0:10:46.786 --> 0:11:6.146
Cofer, Lasha M
Scale autonomy through discipline qualification. So before we begin our panel discussion, we do want to open it up with a poll question. So I'm going to launch the question now. You should see a little window pop up on your screen in just a moment. The question is simply, how is your organization

0:11:6.306 --> 0:11:17.746
Cofer, Lasha M
currently approaching autonomous coding. You'll see about 5 options there. So if you could take just a moment or two to please answer that polling question, we greatly appreciate that.

0:11:21.106 --> 0:11:41.746
Cofer, Lasha M
All right, so while that's going, we'll go ahead and get started with our panel discussion for today. So Dan, I'll start with you. Autonomous coding has moved from a future state concept to an active strategic initiative. What has changed over the last, say, 12 to 18 months?

0:11:42.906 --> 0:12:3.506
Schofield, Daniel A
Perfect. Thank you, Lasha, and thank you for the warm welcome. And also just pleased to be on virtual stage here with my colleagues, Temeka and Michele. And also thanks to all our attendees for coming here this afternoon, at least East Coast. So, I mean, what's changed over the last 12 or 18 months? You know, you kind of have to have been on a rock.

0:11:47.666 --> 0:11:48.66
Cofer, Lasha M
This.

0:12:3.906 --> 0:12:24.866
Schofield, Daniel A
in order to not see all the press and buzz that artificial intelligence is getting. I mean, we're all using it in our day-to-day work. And, you know, as it pertains to autonomous coding, the models and also our here at Optum, our engagement with customers has helped inform progression and innovation

0:12:25.346 --> 0:12:43.826
Schofield, Daniel A
in the autonomous coding practice itself. One more thing I would mention before handing it over to Temeka is we've seen increasing willingness of healthcare and hospital leaders to entertain autonomous coding

0:12:44.226 --> 0:13:0.786
Schofield, Daniel A
as a solution and a contributor to their medical coding staff. And I think that's a big piece of this is to accept autonomous coding and be willing to incorporate that into the process. What do you think, Temeka?

0:13:2.666 --> 0:13:23.506
Lewis, Temeka L
Yeah, I agree. So thank you everyone for being here today. Thank you for that warm welcome and introduction. But one key thing that Dan just stated is we've seen an increase in the willingness of a lot of our acute care health systems to use autonomous coding. So when I think back maybe four years when this was quote un quote,

0:13:23.746 --> 0:13:44.466
Lewis, Temeka L
future state. The question then was, well, can we use AI to autonomously code and encounter without any human intervention? I think it took us maybe about a year and a half to two years to say yes. The answer to that question is yes. But over the last 12 to 18 months, I think because of the evolution of the various

0:13:44.506 --> 0:13:56.946
Lewis, Temeka L
forms of AI. The question has gone from can we to how do we, right? And so again, unless you've been living under a rock, you may not have heard of ChatGPT or Claude or Copilot, any of the videos.

0:13:59.553 --> 0:14:20.273
Lewis, Temeka L
When I think about our flagship product, you know, OIO and our CAC applications, well, we've always had NLP, which is a type of AI that has always read your documentation and suggested code. However, those require human intervention. So now we're thinking through, well, what else exists, right? Can we layer in?

0:14:20.433 --> 0:14:40.433
Lewis, Temeka L
some machine learning models? Can we prompt along certain areas of the pipeline? Can we prompt an LLM? Or has anyone heard the word agentic AI lately? Right, where the agents are given full autonomy and we can allow those agents to come in and evaluate each task and evaluate processes along the workflow.

0:14:40.753 --> 0:14:59.553
Lewis, Temeka L
and determine what exactly needs to happen in order for this case to go autonomously. So over the last 12 to 18 months, I think the conversation has shifted from can we to how do we and what forms of AI do we use for which task in that medical coding space.

0:14:59.913 --> 0:15:7.873
Lewis, Temeka L
So now I'll pass it to Michele to see if she has any additional input to ask with respect to her conversation she's been having over the last year or so.

0:15:9.473 --> 0:15:29.153
Maier, Michele M
Thank you, Temeka. And again, welcome everyone. I completely agree with Dan and Temeka. I think the other piece that I kind of look at from my perspective is really looking at, you know, AI has changed, but what hasn't changed is really looking at those workforce shortages. Do we have the staff, you know, to keep going at the rate that we're growing?

0:15:29.353 --> 0:15:50.273
Maier, Michele M
from an organization perspective when we're looking at hospitals or other organizations. And then also looking at the increasing complexity. Every year we get new regulations, new policies, new procedures that have to be implemented, and how do we do that? And then always those constant making sure that we're maintaining those quality,

0:15:50.593 --> 0:16:10.273
Maier, Michele M
levels that we want to be at. So I think with all of that and the advances in AI, we are definitely seeing some different shifts, especially like we said, you know, that shift from could we to how do we, and really starting to look at what that framework looks like. And I will hand it back to Lasha.

0:16:10.913 --> 0:16:29.473
Cofer, Lasha M
Thank you. Thank you all. That's a great way to get started here. Certainly have seen a shift in the last year or so. So our next question around this theme is when healthcare leaders evaluate autonomous coding today, what problem are they really trying to solve?

0:16:31.433 --> 0:16:32.433
Cofer, Lasha M
Let's start with you, Dan.

0:16:33.73 --> 0:16:53.393
Schofield, Daniel A
Yeah, sure. And Michelle just mentioned the labor stressors, and I'll kind of defer to my 2 colleagues, and especially given their position as professors, really have their finger on the pulse. But, you know, from my perspective, even just anecdotally, I had a couple routine medical appointments last week, and I

0:16:37.633 --> 0:16:37.873
Cofer, Lasha M
Yeah.

0:16:53.753 --> 0:17:13.953
Schofield, Daniel A
showed up and I was met by a screen, no person, no warn, welcome. So there's, and I asked the physician about that and it's hard to find somebody to do the job well. And so I think the health information space is no different. I would posit too that with autonomous coding,

0:17:14.833 --> 0:17:34.353
Schofield, Daniel A
There's a certain level of predictability that comes along with it. You know, humans are subjective and dare I say, we make mistakes. Whereas a human, not that it is a perfect science, can have a more predictable and maybe once calibrated can be

0:17:34.753 --> 0:17:48.833
Schofield, Daniel A
of higher quality. So I think the name of the game here, though, is freeing up humans to do more valuable things as part of the revenue cycle. Temeka, I'm sure you have more to add.

0:17:50.593 --> 0:18:9.473
Lewis, Temeka L
Sure, sure. And so I won't belabor the point. Everyone knows that there are workforce shortages. But when we think through what problems our healthcare leaders can solve with autonomous coding, I'm going to take another Ave. Let's talk through how we develop and upskill our current

0:18:9.553 --> 0:18:29.793
Lewis, Temeka L
coding workforce. So I'll date myself a bit here. When I say, if you carefully design your solution, your autonomous coding product could become your skinny chart coder. Everyone here who has ever worked in HIM, remember the skinny chart coder. It was the person that showed up very early in the morning, grabbed all the skinny charts off the cart,

0:18:30.113 --> 0:18:48.913
Lewis, Temeka L
works through those, right? So this is where we want to efficiently use our AI solutions, right? Again, if we design it carefully enough, right, it can take on some of those rope coding encounters, those lower complexity, higher predictable charts. We can take that out of your traditional coder.

0:18:48.993 --> 0:19:10.513
Lewis, Temeka L
rotation. And this will create capacity for your existing team, right? So what does this mean? This means your experienced coders, right, who may have had some difficulty with some of those complex coding charts in the past, the reason that they probably weren't picking those or prioritizing those, you can now kind of focus. And everyone uses the word, they can now work at the top of their license, but you can focus.

0:19:10.873 --> 0:19:25.633
Lewis, Temeka L
You can educate those who may have lacked the skill set to take on, again, those really complex cases, if you allow your autonomous coding solution to handle a lot of your skinny to intermediate charts. So that's where I'll go with that one. How about you, Michelle?

0:19:26.913 --> 0:19:49.713
Maier, Michele M
So I think I completely agree. It really does allow your organizations to use that talent pool they have, that existing talent pool, in different ways, more strategically, and trying to find those efficiencies where they can, but also value that human expertise. I mean, we're not taking away the expertise that we all have as humans, but it's that idea of

0:19:50.753 --> 0:20:13.393
Maier, Michele M
letting us have the bandwidth to really, one, if we need to, gain that expertise. If you have a newer coder that you're like, okay, we really need help in cardiology, they have that ability now to, that bandwidth, so to speak, to now go after learning those different skill sets that they maybe did not have before, or maybe they're going in a completely different direction.

0:20:14.113 --> 0:20:33.153
Maier, Michele M
I have a few of my students that are like, I love doing the investigation. They're really kind of shifting a little bit more to that revenue integrity side. So I think taking those skills we already have in those, that human workforce and really strategically being able to utilize that while letting the technology kind of do those

0:20:33.873 --> 0:20:36.913
Maier, Michele M
Those skinny charts as Temeka was dating, so...

0:20:38.913 --> 0:20:59.393
Cofer, Lasha M
Excellent. All great points. I'll just add, you know, as an instructor myself, and I mentor a lot of young professionals and, you know, they're struggling to find work. And I just see this as such a huge opportunity for newer coders, right, to kind of get their foot in the door and for organizations to not be so hesitant to give newer coders a chance.

0:21:0.993 --> 0:21:13.713
Cofer, Lasha M
All right, so with that, we'll move on to our next question. So Dan, if a health system asked, are we actually ready for autonomous coding? What evidence would you want to see before answering yes?

0:21:15.313 --> 0:21:37.33
Schofield, Daniel A
Yeah, great question, Lasha, and hopefully not too provocative with my responses. But, you know, I think that the hospital leadership needs to build some consensus. And what I would be looking for is participation from various stakeholders and various levels of leadership.

0:21:38.273 --> 0:21:59.313
Schofield, Daniel A
I think that bringing everybody along for the ride and having everybody bought into the process is a key role that leadership needs to play from the hospital system standpoint. I also think that a health system that's ready for autonomous coding has a good understanding of their current coding practices, their guidelines,

0:21:59.713 --> 0:22:21.673
Schofield, Daniel A
right? Even what processes may be embedded in the existing workflow that might be a little tangential to just the core of coding practice. Lastly, I'd be looking for someone who's willing to experiment and learn from the process. And we're still early days, even though that first question pointed to how far we've come.

0:22:22.113 --> 0:22:40.353
Schofield, Daniel A
there's still a long way to go in terms of, you know, if we think about a future of full autonomy. And so there's a progression here, there's challenges along the way, and a willingness to try and pursue and persevere through those challenges, I think, is a big part of it.

0:22:41.473 --> 0:22:42.593
Schofield, Daniel A
Temeka, what do you think?

0:22:44.193 --> 0:23:3.433
Lewis, Temeka L
Yeah, keyword there, workflow. I think if an organization were to ask me, were they ready? Two things I would want to know. A, have you evaluated your documentation quality? And B, what existing pain points do you have today? What I've learned over these last two years is that

0:23:3.513 --> 0:23:25.713
Lewis, Temeka L
Automation tends to magnify whatever already exists. So if you've got tons of disparate processes, things are being done differently across different sites, different locations, that's going to be magnified when you try to shift from an assistant to an automated coding process. For example, if you ask almost any HIM leader, you know, do you have certain providers

0:23:26.33 --> 0:23:44.953
Lewis, Temeka L
who always document a certain way, certain services, you know, that may be optimal for your humans to recognize, but what's going to happen when we shift to when the AI needs to read that documentation, right? So if you have, you know, a highly standardized, high quality process, that's something that's scalable.

0:23:45.233 --> 0:24:6.593
Lewis, Temeka L
But if you automate things that are highly variable, you may be setting yourself up for some roadblocks down the road. So, and again, like, are they applying the coding guidelines standardly and consistently across all of your sites? What documentation gaps already exist? What denial issues do you already see? We want to make sure that we're tracking and trending.

0:24:6.753 --> 0:24:25.233
Lewis, Temeka L
so that when we do shift to that autonomous coding process, we will understand the actual impact that we are making. So again, any organization that would ask me if they were ready, if they were ready, I would say, do you understand your current documentation quality? And then what existing pain points do you have so that we can talk through those?

0:24:25.473 --> 0:24:30.433
Lewis, Temeka L
earlier on in the process as we begin to implement our solution. What are your thoughts there, Michelle?

0:24:31.553 --> 0:24:49.313
Maier, Michele M
I completely agree, Temeka, but one thing that I also think we have to start with is what is the goal? Like, we can implement and we can do a lot with technology, but not knowing what is that organizational, like, what's the intention? What goals are you trying to solve for? Is it a workflow?

0:24:50.193 --> 0:25:8.593
Maier, Michele M
workforce shortage? Is it that you want to be able to elevate, you know, your current staff? What is that some of those goals as far as, you know, or do you anticipate growing, having massive growth for the next few years, but you don't anticipate being able to hire enough people to kind of take that over?

0:25:9.73 --> 0:25:33.273
Maier, Michele M
understanding what those business goals are and kind of going back to what you were saying, that I always call it fact-based view. What is your current performance, but also what are the variations in that current performance? And it's not just, you know, coder productivity, but more of why is there variation between one coder and another? Do you have an understanding of what that looks like and why is that?

0:25:33.353 --> 0:25:53.393
Maier, Michele M
there. And then at the end of the day, that organizational capacity to manage change, when implementing anything new, technology, AI, anything like that, we are essentially having to manage change because we are changing, you know, we're bringing to light sometimes those workflows that we're not.

0:25:53.713 --> 0:26:15.473
Maier, Michele M
everyone kind of forgot about that we were doing, but we've been doing them. I mean, I remember when I was first out of college going in and doing that workflow consulting and asking questions and it would be like, well, why are you doing that? And it's like, well, we've done it for the last 20 years. Okay. It's also kind of getting used to that, well, is that necessary? Can we change? Is that something that's still needed or not?

0:26:15.873 --> 0:26:24.433
Maier, Michele M
So a lot of those conversations come down to also how well can we manage that change through this process.

0:26:26.673 --> 0:26:46.193
Cofer, Lasha M
Excellent. Thank you all. All right. So our last question for this particular theme, Dan, starting with you again, what baseline data should leaders understand before selecting their first autonomous coding use case? And what can that baseline reveal that intuition alone

0:26:46.273 --> 0:26:47.73
Cofer, Lasha M
May miss.

0:26:47.953 --> 0:27:6.593
Schofield, Daniel A
And I'm glad that, I mean, this will dovetail great with a couple of the things that Michelle just mentioned, productivity, coding behavior, you know, documented workflows and processes, I think are, you know, it doesn't necessarily pertain to data in its truest form, but that's important information.

0:27:6.993 --> 0:27:28.393
Schofield, Daniel A
and we'll identify some of those variations that Michele and Temeka were referring to. Maybe in the scope of more hard data, today, I'm sure many of the participants on today's call are using computer-assisted coding and they're getting code suggestions. I would ask,

0:27:28.713 --> 0:27:49.233
Schofield, Daniel A
Do they know, at least in summary, how their teams are reacting to those code suggestions, right? What's the level of automation that they're experiencing today, and what's leading to how your teams are then reacting to the artificial intelligence that's already baked into your solutions today?

0:27:49.553 --> 0:27:53.153
Schofield, Daniel A
So those are a couple of things that I might think of. Temeka, your thoughts.

0:27:54.313 --> 0:28:13.713
Lewis, Temeka L
Yeah, so baseline data, right? Most of our HIM operations leaders already track some form of KPIs. But I think the question I have is how granular can we get with that data? Dan just mentioned the end user's reaction to those predicted code suggestions, right? So do we have the ability

0:28:14.33 --> 0:28:36.393
Lewis, Temeka L
to slice and dice the data that we have, you know, maybe by specialty, maybe by encounter type. Do we know how well or what the agreement rates are for some of those predicted codes, whether it's, you know, in the assisted pathway or the autonomous solution pathway? Do we understand the complexity? Remember, you know, our goal here is to take away some of the quote un quote skinny charts.

0:28:36.633 --> 0:28:59.713
Lewis, Temeka L
So how much of that is actually complex when this solution was built? And so I would like to know, can you identify where you have high volume, highly standardized workflows with little variation, and what does that data look like, as opposed to us looking at the largest slice of the pie with all of coding productivity and accuracy and turnaround time, right? So I would like to know,

0:29:0.113 --> 0:29:4.993
Lewis, Temeka L
How granular can we get with your existing data models? What do you think, Michelle?

0:29:5.713 --> 0:29:25.273
Maier, Michele M
I completely agree. And I think the other piece that we have to think about also is those coding practices are based off a lot of the times off documentation. So in addition to looking at it from, you know, policies and procedures or coder behavior, sometimes I'm also interested in how consistent is that documentation.

0:29:25.393 --> 0:29:44.353
Maier, Michele M
even on cases that or types of encounters that you're like, it would have to be consistent. Do we actually have that baseline data to know? One of the things the question kind of brings up is, you know, what can data reveal that intuition may miss? And I think that's sometimes where, you know,

0:29:44.393 --> 0:30:3.73
Maier, Michele M
When I've done analysis before, it was trying to answer a question a lot of the times, like, why did this go in this direction or things? But a lot of the times, the patterns we uncover, we knew something was going on. Intuition told us something happened. But until we dove into the data, did we understand really what that underlying pattern was?

0:30:3.553 --> 0:30:22.353
Maier, Michele M
and being able to slice and dice the data in all of those variables is without question needed. So when we talk baseline data, I think it's the, to make it to your point, how granular can we get and how much can we shift and look at it from different directions to be able to truly get data?

0:30:22.433 --> 0:30:39.953
Maier, Michele M
to what is, you know, where do we have those consistencies or where do we not have consistencies that maybe that, you know, you expected that workflow to be very consistent and when you look at it, you're like, oh, that does not have quite the consistency that we thought it was going to and then being able to look to understand why.

0:30:43.313 --> 0:30:44.513
Cofer, Lasha M
Excellent. Thank you all.

0:30:45.473 --> 0:30:50.673
Cofer, Lasha M
All right, so with that, we're going to shift to our next theme for today's discussion.

0:30:52.33 --> 0:31:11.473
Cofer, Lasha M
And that is build trust through local validation and governance. So, Temeka, I'll start with you this time. Our first question here is, trust cannot rest on a single accuracy score. So how should executives interpret autonomous coding performance claims?

0:31:13.233 --> 0:31:32.913
Lewis, Temeka L
Accuracy. That's a big word. I think as an HIM professional, we all have heard the terms, coders must be 95% accurate. But if I were to poll, I don't know, 5 organizations on this call, the way they calculate that accuracy might be different. So as executives are looking at

0:31:32.953 --> 0:31:53.393
Lewis, Temeka L
the accuracy claims that are put forth by several of the vendors, OptumInfo included. I would always challenge everyone to say, what's in your numerator, what's in your denominator, and accurate just at what? Again, every organization has their method of calculating accuracy. You know, is it code over code? Is it case rate accuracy?

0:31:53.793 --> 0:32:15.313
Lewis, Temeka L
At what point was the audit done? You know, is it accurate at this particular point in time? And so I think everyone needs a good understanding of what that accuracy looks like by specialty, you know, by encounter type. I can't say this enough, how much complexity is involved in that numerator and denominator when we determine those accuracy numbers?

0:32:16.33 --> 0:32:36.33
Lewis, Temeka L
So organizations also need to understand how the system performs against their own coding standards. If we're calculating accuracy using one method and you're calculating accuracy using a different method, then there will always be a small amount of misalignment, right? So I challenge everyone to ask the questions.

0:32:36.233 --> 0:32:49.393
Lewis, Temeka L
How is accuracy determined? What are you using in your formula? And then what involves your denominator when you say a specific percentage of accuracy? So what are your thoughts on that one, Michelle?

0:32:50.953 --> 0:33:10.593
Maier, Michele M
Completely agree, Temeka. I think the other thing is looking at those claims and saying, how much does that match up with your workflow and from your organization's perspective? What, you know, does that align? Are you kind of, you know, that old adage, are we comparing apples to oranges when it comes down to it?

0:33:10.833 --> 0:33:31.953
Maier, Michele M
saying you're expecting to be able to get here, but when you're looking at it, it isn't what you were thinking it was going to be. So that understanding of how was that claim, what's the underlying numerator and denominator versus how you guys are looking at data, but also how does that align with the use cases that you feel like would be ideal from an autonomous coding?

0:33:32.273 --> 0:33:35.633
Maier, Michele M
or use of technology perspective. Dan, any thoughts?

0:33:36.753 --> 0:33:59.993
Schofield, Daniel A
I think I just under score what you both said, which, you know, clarifying the universe, understanding the whatever message, marketing message, whatever you're being told about a certain solution is critical. You know, make sure you're comparing apples to apples versus apples to oranges, right? There's all kinds of complexities that

0:34:0.113 --> 0:34:20.33
Schofield, Daniel A
Temeka and Michelle have highlighted here. Just know what you're interested in understanding so that when you are given a piece of information, you're able to then compare and contrast that against what you already know. And the only unique spin I think I would put on what you both said is

0:34:20.433 --> 0:34:41.953
Schofield, Daniel A
Take the solution for a test ride, right? I mean, I know here at OptumInfo, we have all kinds of, you know, behind the scenes workflows that we can perform for customers. You know, treat that as if it were a new coder and put your educational and audit staff on it and find out your 95% accuracy score for itself. I think that's a

0:34:42.73 --> 0:34:50.593
Schofield, Daniel A
an important part and incumbent upon the solution consumer to perform some of those tests as well.

0:34:54.33 --> 0:34:56.993
Cofer, Lasha M
Excellent. Thank you. Very insightful here.

0:34:58.193 --> 0:35:7.873
Cofer, Lasha M
So along that same line, how do organizations build trust in autonomous coding without slowing innovation?

0:35:10.153 --> 0:35:11.393
Cofer, Lasha M
I start with you again, Temeka.

0:35:10.433 --> 0:35:10.993
Lewis, Temeka L
Buildings.

0:35:12.153 --> 0:35:32.113
Lewis, Temeka L
Yeah, so building trust. I like to think of that as something called earned autonomy, right? And so one of the approaches that works well with us here at OptumInfo is we always start with a clearly defined population. We know the target encounters that we feel are good for autonomous coding. And then there's a...

0:35:32.353 --> 0:35:52.753
Lewis, Temeka L
partnership between us and any of the clients who want to use the system. We also validate the results locally. We establish, you know, some performance thresholds internally, and we share those with any of our potential clients. And so we don't expand until evidence supports the expansion of autonomous. And so I think

0:35:53.33 --> 0:36:14.513
Lewis, Temeka L
Organizations also have to have their ongoing auditing processes, you know, the same way that we do here internally, but we don't want to treat it as a one-time validation, you know, an event that happens before a go-live. Definitely, we want transparency into those expectations. We want to make sure we're sharing our findings back and forth and iterating with our clients.

0:36:14.873 --> 0:36:34.353
Lewis, Temeka L
so that we can determine, you know, which cases qualify, should stay in the pipeline, as opposed to when we've hit a roadblock. You know, is this an opportunity to refine or expand, or is this an opportunity to just say, this probably wasn't a good candidate, right? And so again, I think transparency helps to build trust, but nothing slows innovation.

0:36:34.553 --> 0:36:50.513
Lewis, Temeka L
Because again, it should be an earned autonomy. We don't want to just force anything. We want to make sure that we're being transparent, sharing those results, and making sure that everyone is comfortable before we move to an encounter that can actually be sent autonomously. What are your thoughts on that, Michelle?

0:36:52.273 --> 0:37:10.753
Maier, Michele M
I think one of the other big things is bringing everybody in early so that, you know, it, you know, it's not just coding that's going to be impacted by this. Like, Temeka, you were just saying, bring in the auditing team, bring in compliance, bring in those other departments that are going to, you know, see information downstream.

0:37:10.993 --> 0:37:30.953
Maier, Michele M
isn't, you know, in the sense it's bringing everybody in. That also goes back to that concept of change management. The more everyone's included in a change and more everyone understands how the system is going to work and what is it going to do in this, you know, that ability to kind of ensure we're still seeing, you know, innovation.

0:37:31.73 --> 0:37:47.73
Maier, Michele M
but that we can also build that trust as we're going along so that they understand all of those parts and pieces. I think the better everyone's going to, you know, kind of accept and understand that transparent evidence that is being built along the way. Dan, any thoughts?

0:37:48.33 --> 0:38:6.913
Schofield, Daniel A
I love the concept of earned autonomy, you know, and I think continuous validation, testing, validation, just as you would with a human coder, is at least in early days, the name of the game. I totally agree with what you're saying, Michele. You want to bring everybody along for the ride.

0:38:6.993 --> 0:38:27.153
Schofield, Daniel A
You want to make sure everybody is bought in part of the process. There's a role here to play for various people, especially during the implementation and the learning. And as you do learn things, right, back to what Temeka said a couple of questions ago, where the autonomy has a tendency to magnify some of the

0:38:27.433 --> 0:38:52.33
Schofield, Daniel A
the warts or difficulties or challenges that already exist today, right? That's an opportunity for improvement. And you likely just learned something that has always existed. And now the repetitive autonomous process has just repeated that negative outcome. That's an opportunity now to improve your entire organization in that specific area.

0:38:52.993 --> 0:38:53.873
Schofield, Daniel A
Lasha, back to you.

0:38:54.433 --> 0:39:14.273
Cofer, Lasha M
Thank you. All right. So as we wrap up this particular theme with one more question, who should have decision rights over autonomous coding and what controls, governance structures and human oversight should exist before cases move autonomously to billing?

0:39:16.873 --> 0:39:37.73
Lewis, Temeka L
So, great question. And I think Michelle and Dan both just touched on it. Governance, right? I think one of the most important things is to bring everyone along the journey. Your governor's team has to truly be cross-functional. Deciding which encounters and when an encounter can go,

0:39:37.393 --> 0:39:56.513
Lewis, Temeka L
Autonomously is only one part of the process, right? So you need the coding and the HIM leadership at the table because they understand the rules, you know, the documentation requirements, compliance has to be involved. What's involved in those operational workflows? You also need the technical teams and need to make sure that they're aware of.

0:39:56.673 --> 0:40:16.913
Lewis, Temeka L
All of the decisions, so that we ensure that these workflows that we have in mind in ideal state are actually built and deployed properly, and then, likewise, the data and the analytics component, they also have to be a part of the conversation, because they may not know what needs to be measured ahead of time.

0:40:17.273 --> 0:40:36.993
Lewis, Temeka L
But once we build a solution, we have to track and monitor which encounters are moving autonomously. So I think who owns the decision rights is the full governance team, and not just product and HIM, but the technical teams, the data teams. Those decisions have to be made together with full transparency from start to time.

0:40:37.33 --> 0:40:38.993
Lewis, Temeka L
to finish. What are your thoughts on that, Michelle?

0:40:40.353 --> 0:41:1.393
Maier, Michele M
I agree. I think that cross-functional representation, it's not just, we're not just changing coding. It is really going to impact others, especially as we go downstream. And I think the other thing is, is understanding, you know, yes, we're going to have eligibility, we're going to have thresholds, we're going to have audit results, but

0:41:1.753 --> 0:41:26.313
Maier, Michele M
you know, understanding the workflow and kind of also knowing where does that human stay in the loop in this conversation and in these processes, especially when maybe a case doesn't meet the, you know, that ideal path. But what happens then in making sure that we have, you know, processes in place and that we have escalation paths if something isn't working the way that we're expecting it to and it kind of goes back to.

0:41:26.393 --> 0:41:45.713
Maier, Michele M
So again, that overall governance structure, but really making sure from my perspective that it is cross-functional and then we're not just focusing in on one area. And I think that really happens and is much more easily understood and accepted the earlier we kind of start bringing everybody in because then they're kind of...

0:41:45.793 --> 0:41:51.553
Maier, Michele M
bought in on the whole process all the way along. Dan, what are your thoughts?

0:41:52.513 --> 0:42:16.233
Schofield, Daniel A
Yeah, you both said it, multi-dimensional, right? I mean, so you're going to get leadership from OptumInfo, you know, you got your coding teams, finance, analytics, IT, you know, this is a change management process, right? This is something where you know you might need some program or project managers that you want to you want to apply.

0:42:16.313 --> 0:42:36.353
Schofield, Daniel A
to this. So just having that cross-functional, multi-discipline group approach, building consensus, you need leaders. I know we have some CFOs and other senior leaders on this call today. I mean, you got to break the tie at some point, for sure, right? Like there's going to be some trade-offs and there might be some difficult decisions to be made.

0:42:36.833 --> 0:42:56.753
Schofield, Daniel A
But having everybody on board, and I see it more and more, there are health systems out there that are building an AI discipline or an AI vertical within their organizations. Those folks should also have a voice in the process as well. Whatever team members you've

0:42:57.73 --> 0:43:5.33
Schofield, Daniel A
assign that responsibility to, that have that background, should be aware and be able to participate as part of this group.

0:43:6.113 --> 0:43:6.833
Schofield, Daniel A
Over to you, Lasha.

0:43:8.273 --> 0:43:9.313
Cofer, Lasha M
All right, excellent.

0:43:10.393 --> 0:43:30.513
Cofer, Lasha M
All right, so we're going to move to our third theme for this discussion today, which is analytics as the revenue cycle control tower. So the first question here is, why will analytics and reporting become the strategic differentiator for organizations adopting autonomous coding?

0:43:33.273 --> 0:43:34.433
Cofer, Lasha M
Let's start with you, Michelle.

0:43:33.313 --> 0:43:34.273
Maier, Michele M
So, one thing...

0:43:35.433 --> 0:43:55.233
Maier, Michele M
Sorry. One thing I do see is that, again, this isn't just like a one use case, one point kind of solution. It's really starting to take autonomous coding in the technology and say, how can this be managed across the entire, you know, entire platform of all types cases.

0:43:55.393 --> 0:44:13.953
Maier, Michele M
all the case types and everything like that and really expanding. So you have to start looking not at just what are the performance patterns for where it's at now, but what are the exceptions? What interventions make the most sense? And really looking at those downstream and outcomes and saying, where is it working well?

0:44:14.393 --> 0:44:36.593
Maier, Michele M
where could it work well? And then where are some risks or where are there some interventions that may be needed that things may be not working exactly as we expected it to. As we've been kind of preparing for this webinar, the thing that kind of stuck in my brain is that idea of you're taking data and it's those organizations that can kind of take data and turn it into

0:44:36.833 --> 0:44:49.113
Maier, Michele M
information and then really turn it into knowledge. How can we, what should we be doing in the future? Where should we be looking? I think is really going to be that strategic differentiator. Dan, any thoughts?

0:44:50.593 --> 0:45:14.313
Schofield, Daniel A
I have thoughts. You know, there's so many things to measure in the revenue cycle and coding and CDI and audit. I mean, it really is a playground of data. You know, I've had the luxury and the privilege of seeing some of our analytical solutions for our mid-revenue cycle with Integrity One come to market. And there's, I mean,

0:45:14.393 --> 0:45:33.193
Schofield, Daniel A
And it's just a never-ending supply of things that can be measured. Code frequencies, stages of autonomous coding, success rates, impacts of success or lack thereof. I think what I would say, and I'm sure I know Temeka has been on the ground floor with some of this, it's not just to set it and

0:45:33.433 --> 0:45:55.473
Schofield, Daniel A
forget it, right? You want to grow and evolve and learn things and improve. And the only way to do that is monitor through information. One thing I'd say to our audience today is, how well are you using analytics today? And I don't mean that to say, I'm sure you have dashboards and reporting and are measuring certain things. But

0:45:55.513 --> 0:46:18.33
Schofield, Daniel A
but do you have an analytical discipline as part of the revenue cycle? Do you know your information inside and out as it stands today? I think it's an unexplored opportunity, or at least underexplored opportunity for hospital systems to invest additionally in their own views of analytics. And lastly,

0:46:19.153 --> 0:46:42.433
Schofield, Daniel A
All of the AI, right, the TV commercials, the everything, right? All the AI is based on data, right? It's a model on top of data. And so, you know, it is paramount that the information that's being used to drive your automation and autonomous processes is of trusted and high quality. And the only way

0:46:42.993 --> 0:46:50.993
Schofield, Daniel A
To understand that is to become more familiar with it. Temeka, I know you probably have a lot to say on this topic as well.

0:46:52.153 --> 0:47:13.873
Lewis, Temeka L
I don't know if I have a lot to say. In fact, I don't even know if I want to follow, you know, our expert in data and analytics guy, but a couple of things. So when I think analytics, like Michelle touched on one key point, like we all want to know which encounters are going autonomously, but what about the ones that didn't, right? Where are those exceptions occurring?

0:46:57.113 --> 0:46:57.633
Cofer, Lasha M
Yeah.

0:46:59.193 --> 0:46:59.633
Schofield, Daniel A
Yeah.

0:47:14.153 --> 0:47:33.913
Lewis, Temeka L
And that's one thing analytics will give us all insight into, like which specialties perform well. You know, maybe we thought this was a good specialty or a good department to start with. Data is going to boil up to the top. Those areas that were risk may be emerging that we may want to kind of refine a bit. Also, we didn't talk about...

0:47:34.753 --> 0:47:53.793
Lewis, Temeka L
how analytics can help us connect the performance to downstream outcomes, right? It's great to talk about autonomous coding, but we also want to make sure that we're not increasing the amount of edits that our humans or the AI solution is having to work, or we're not increasing those denials.

0:47:54.233 --> 0:48:13.873
Lewis, Temeka L
We're making sure that we're not impacting the revenue and ultimately the cash flow for our organizations. And so organizations that develop really strong analytics capability, as Dan was just mentioning, they will have the ability to continuously refine and qualify use cases for autonomous coding. Back to you, Lasha.

0:48:15.873 --> 0:48:16.673
Cofer, Lasha M
Thank you.

0:48:18.193 --> 0:48:33.713
Cofer, Lasha M
All right, so we're getting close to the end. One more question around this theme. As routine work becomes automated, how will coding, HIM, revenue cycle teams evolve over the next five years?

0:48:37.633 --> 0:48:58.113
Maier, Michele M
Just one that I feel like I have to consider almost every day, especially as an instructor. Lasha, you said it earlier, you know, kind of working with students, but it is one where, you know, routine work. What I did, so like is to make it caught at the skinny charts, this coder should come in for skinny charts. What I thought 25 years ago as a coder

0:48:58.473 --> 0:49:19.473
Maier, Michele M
When I came out, what I started coding, that is what is really going to become autonomous, especially if it isn't already. And really start to work through, okay, so what does that look like? It does allow us to allow our current staff to work at the top of, you know, really become those experts in specific areas.

0:49:19.833 --> 0:49:38.673
Maier, Michele M
and really manage those more complex cases, but really starting to see that they can oversee where we have quality concerns and where we want to be focusing and spending our time. We also can start to dive in, I think, in my mind, to some of those exceptions. Why is an exception an exception and what could be done? So I think a lot of the time.

0:49:39.113 --> 0:50:0.593
Maier, Michele M
those skills that we have in our workforce is just, they're going to shift how they're looking at information. They may not be coding it day-to-day, but they may be looking for why is this not automating the way that we expected it to. Kind of going back to Temeka's point of the data can show us those cases, but sometimes it is still probably going to take some human intervention to understand.

0:50:0.953 --> 0:50:24.113
Maier, Michele M
what are they doing? Also really starting to look at revenue integrity and starting to be able to look in some of those, you know, areas that maybe we haven't had time to focus on just yet. But now with some of those, you know, things automated, some of those standard consistent cases automated, we would have time to kind of dive in and ensure reimbursement and quality measures that we can really

0:50:24.353 --> 0:50:31.313
Maier, Michele M
focus in on everything that we want focused in on from an organizational perspective. Dan, what are your thoughts?

0:50:32.593 --> 0:50:56.833
Schofield, Daniel A
You know, I agree the level of complexity that the humans will need to manage in the process will go up. And that's a good thing, right? And then we can all leverage our skill sets and support these processes, even, you know, it's not a perfect process, right? We've heard that throughout. And so there are opportunities for improvement.

0:50:57.153 --> 0:51:20.953
Schofield, Daniel A
I think a closer partnership as well between the coding profession and the technology in terms of managing the AI process as it grows and becomes more enmeshed, there will be possibilities for configuration and managing the exception queues that fall out of the

0:51:21.193 --> 0:51:44.753
Schofield, Daniel A
the AI process. So, you know, there'll be a, I think, a greater partnership, even more so than there is today between the people and the technology. And I think, you know, even if I'm putting on, you know, I don't know, a crystal ball or maybe a foil hat, I don't know, you know, kind of looking into the future, I could see where people are really,

0:51:44.993 --> 0:52:5.953
Schofield, Daniel A
watching almost like a stock ticker and things are moving incredibly quickly. And it's almost a conveyor belt of activity. And then people are sort of picking the things that fall out, right? The things that need to be managed and put back into line so that it can continue down that conveyor belt. So maybe all three of those things are similar.

0:52:6.993 --> 0:52:9.273
Schofield, Daniel A
I'll hand it over to Temeka.

0:52:11.233 --> 0:52:32.113
Lewis, Temeka L
So great points. I had all of those written down, so I won't repeat anything. But the one point I'd actually like to make here is when people say, how will your coding and HIM in teams evolve over the next five years? If you're thinking that the evolution is happening now, I'm going to challenge you and say that you might be behind the 8 ball.

0:52:32.273 --> 0:52:51.793
Lewis, Temeka L
The evolution started three years ago. I mean, if you think about the credentialing organizations, right? AAPC had an entire 7 or 8 module training on AI and autonomous coding three years ago. AHIMA has the AI toolkit. There's all sorts of articles, all sorts of webinars.

0:52:52.193 --> 0:53:11.313
Lewis, Temeka L
It has been talked about at the AHIMA conventions over the last three years. So the evolution isn't happening now. The evolution started several years ago. And yeah, everything that Dan and Michelle just talked about, right? HIM professionals, you know, they're more so into auditing, exception management, quality oversight.

0:53:11.433 --> 0:53:33.233
Lewis, Temeka L
revenue integrity, you know, many of them are learning the governance process. So perhaps they would be on the front end or the approval end as opposed to just assigning medical codes. And so I think there's going to be a greater need for our HIM professionals who can sit between, right, the coding operations and the technology team, right? That is kind of the team that I manage today.

0:53:33.513 --> 0:53:53.153
Lewis, Temeka L
That is where they thrive, right? They understand the various forms of AI and technologies that we're using. And so when we have conversations with our clients about what's next for your organization, this is where we kind of mesh, right? How to use technology, how to make your team more efficient, because we already know.

0:53:53.393 --> 0:54:6.273
Lewis, Temeka L
The skinny charts are going away. So what do we need your staff, what do we need to understand from your organization in order to make autonomous coding successful? So the evolution is already here. Back to you, Lasha.

0:54:6.913 --> 0:54:24.433
Cofer, Lasha M
Thank you. It sure is. All right. So our last question for today, if autonomous coding reaches its full potential, what will distinguish organizations that lead the industry three to five years from now?

0:54:27.793 --> 0:54:49.33
Maier, Michele M
I think one is that those organizations are already kind of looking and understanding what autonomous coding can do for them. Everyone's starting now. You have to start now in order to be able to kind of lead in the future. I think that's always kind of one of those things, the first people to kind of understand and tackle and really understand all those pieces and parts are going to be very important.

0:54:49.193 --> 0:55:8.233
Maier, Michele M
I think the other thing is, is that those organizations that see AI or autonomous coding as another tool, and they're going to continuously be looking for those additional tools to make things work better and to improve processes and to make things more efficient. At the end of the day, we are all here to make patience

0:55:8.553 --> 0:55:30.553
Maier, Michele M
have the best health care possible. And I think that's kind of it. It's continuously improving and looking to see how things can continuously improve. I think the other thing to think about is, you know, we were talking about the students and Temeka just brought up, you know, how AAPC and AHIMA have had things out there for AI. I can tell you right now in

0:55:30.953 --> 0:55:51.433
Maier, Michele M
my 100 courses, those baby courses, so to speak, that you think about, AI is up front and center. That is one of the first things almost that we tackle at this point in time is starting to introduce them to that. So again, those new professionals that are going to be coming out are going to be driven to understand and

0:55:51.513 --> 0:56:1.513
Maier, Michele M
know how to utilize those skills and that technology. So it'll be interesting to see as we move forward what that looks like for many of our organizations. Dan.

0:56:3.273 --> 0:56:21.473
Schofield, Daniel A
Yeah, I agree. Start now, right? I mean, if you haven't started, you're already behind. I think that's the message. I'd be so interested to see the results of our poll that we offered at the beginning and, you know, where, you know, the attendees are on that continuum.

0:56:22.153 --> 0:56:44.393
Schofield, Daniel A
The longer the wait, then the longer, the further back you're going to be behind. I think investing, those organizations that invest heavily, or at least invest ahead of data and analytics, the understanding so that they are, they have a material understanding of the processes that are driving

0:56:44.673 --> 0:57:5.393
Schofield, Daniel A
the recommendations of the AI solutions will definitely be well ahead. And it'll show too. I mean, that's where it starts to hit the bottom line here is where you've got an understanding of your practices and your data and what's influencing it today. And the autonomous processes come in and

0:57:5.713 --> 0:57:24.633
Schofield, Daniel A
They start accelerating some of those things. You can then partner with the technology. I mean, OptumInfo has some of the biggest technology that we can bring to bear. Your organization and OptumInfo can partner together for a super-powered solution. And then the last one is just...

0:57:25.33 --> 0:57:43.993
Schofield, Daniel A
you know, perseverance and, you know, continual trial and error. You know, those organizations that have a willingness on the revenue cycle to try and continue on through some of the difficulties are going to come out the end.

0:57:44.393 --> 0:57:57.393
Schofield, Daniel A
as opposed to potentially wind up, you know, debating things and get into long decision-making cycles, right? Power through, innovate, and get to the end. Temeka, your last word.

0:57:59.33 --> 0:58:19.833
Lewis, Temeka L
Yeah, those are great because power through and innovate, right? And so the way I look at it is your organizations that are going to be the leaders are the ones that have the ability to continuously learn. None of us who have become credentialed HIM professionals maintain those credentials without CEUs. And so I feel the same way

0:58:19.913 --> 0:58:39.153
Lewis, Temeka L
when it comes to learning about all of the capabilities within revenue cycle and how AI can impact those capabilities. So leading organizations will be continual learners. They're going to study those exceptions. They're going to understand the audit findings and the financial outcomes. And so ultimately, I think

0:58:39.233 --> 0:58:59.793
Lewis, Temeka L
you know, this autonomous coding, while we think of it sometimes as a new coding technology, it's more so an enterprise capability. Like how can we push forward the organization at the enterprise level? And then leaders who have that CEU kind of mindset are going to be the ones that are most successful.

0:59:0.153 --> 0:59:10.113
Lewis, Temeka L
with the transition or the shift or the change from, you know, an assisted coding process to an autonomous coding process. So those are my final thoughts. Lasha.

0:59:11.353 --> 0:59:29.193
Cofer, Lasha M
Excellent. Thank you all very, very much. Temeka, Dan, Michele, such an honor to be here with you today. Thank you for sharing your expertise. Dropped several nuggets over the last hour or so. We want to thank our audience. Thank you for joining us today. We really, really appreciate your time. Have a great day.