How ASCs Can Use AI to Turn Data Into Action

Ep. 147 Matthew Obenhaus – How ASCs Can Use AI to Turn Data Into Action

Here’s what to expect on this week’s episode. 🎙️

Matthew Obenhaus is SVP of Design & Engineering at NueHealth, where his work spans product development, artificial intelligence, and direct surgery center operations.

Matt joins us for the next installment of our Trust the Data series to explain how ASCs can move beyond high-level dashboards and use artificial intelligence to turn their data into practical action.

We discuss how AI can help surgery centers analyze physician utilization, block time, projected revenue, case profitability, payer performance, and supply costs. Matt also explains how ASC teams can create repeatable workflows without traditional coding skills, identify potential data-quality issues, and reach useful insights faster.

Just as importantly, Matt shares why accurate data and human validation remain essential before presenting findings to physicians or board members. We also cover protecting PHI, establishing clear standards for AI use, and creating appropriate governance around these tools.

In our news segment, we cover a new forecast projecting significant ASC volume growth over the next decade, preview the return of HST Connect, and share a heartwarming story about a pediatric surgeon helping children feel like superheroes before surgery.

Resources Mentioned:

Matthew Obenhaus and NueHealth
https://www.nuehealth.com/insights/leadership/matt-obenhaus/

Vizient Forecasts Continued Growth in ASC Volume
https://www.ascfocus.org/ascfocus/news/digital-debut/202607-vizient-forcasts-continued-growth-in-asc-volume

HST Connect
https://www.hstpathways.com/connect/

Helping Children Feel Like Superheroes Before Surgery
https://www.businessinsider.com/superhero-costumes-help-kids-before-surgery-2026-8

Brought to you by HST Pathways.

Episode Transcript

[00:00:24] Ryan Cohn: Hey, everyone. Here’s what you can expect on today’s episode. This week, we’re continuing our Trust the Data series with Matthew Obenhaus of New Health. Matt brings a unique perspective to this conversation because his work spans both surgery center operations and technology. He joins us today to explain how ASCs can use artificial intelligence to move beyond high-level dashboards and turn their data into specific practical actions.

We discuss how tools like Claude can help centers analyze physician utilization, block time, projected revenue, case profitability, payor performance, and supply and implant costs.

Matt also explains how ASC teams can create repeatable AI workflows without actually needing to be programmers themselves. And just as importantly, we also talk about the foundation that makes all of this possible. Matt shares why accurate data is essential before presenting findings to physicians or board members, and how AI can help identify problems within a dataset and why human review still remains critical

We also cover the guardrails that centers need to consider, including protecting patient information, establishing clear standards for AI use, and making sure that technology is being used safely and responsibly. After that conversation, I’ll cover a new forecast for surgery center growth.

We’ll preview the return of HST Connect, and we’ll also end with a lighter story about a surgeon who’s helping children feel like superheroes on their way into surgery. I hope everyone enjoys the episode, and here’s what’s going on this week in surgery centers

[00:02:04] Ryan Cohn: thank you so much for being here today, Matt. Can you tell us a little about your background, your role at New Health, and the work that you do with surgery centers?

[00:02:12] Matthew Obenhaus: Yeah. Thanks, Ryan and happy to be here. Thanks for having me on. I started my journey into operations, general management, technology innovation, you name it even as far back as my time straight out of Texas A&M and in the Army, which gave me a really good background in operations, and so I credit that quite a bit.

After that, I was with Raytheon doing supply chain operations and supply chain work for, a little over a year, and then I moved on to Cerner for ten years, various product general management roles. So that was my introduction to the world of healthcare. I’ve been with New Health for five years. I started my journey with New Health in a very focused let’s build out provider performance, partner performance physician network performance across our ASCs.

So I had a real introduction into the ASC setting, how physicians operate how they’re profitable, and building our dashboards and analytics. Some work was already done but it– we really catapulted our focus in more digital experiences and growing our ability to manage and impact key performance indicators, which I’ll call KPIs from here on out.

And currently, that, that grew into a focus just out of expediency. We’re a small-ish company. I started owning and managing facilities directly in more of what we call a group vice president role. Of course, we have our site administrators, but I started overseeing and supervising a number of our facilities and sitting on their boards a-as part of my journey with New Health.

And a couple of years ago, I took on direct responsibilities for our entire technology portfolio including our partners our vendor partners our technology partners such as HST as well as our innovation and design around our homegrown analytics, and here recently, our focus in innovation around AI.

So I essentially wear two hats, Ryan. I still have a footprint, sit on boards and manage facili- a number of facilities, while also managing our technology and our innovation, which is a really nice symbiotic way to make sure that our technologies that we’re that we’re partnering around and that we’re developing work extremely well and I get that real-time feedback of whatever we’re doing downstream and how does it our im-impact our facilities.

So it’s a role I want to continue to play, wearing a h- operational hat and a technology and innovation hat for the foreseeable future. We’ll see how it goes.

[00:04:40] Ryan Cohn: Yeah. Yeah. I mean, I could imagine that’s, it’s a two-prong effort. It’s a lot of work but, like you said, you could come up with the the strategy and tell someone how to do it, but you could tell someone that a million times and it never gets implemented potentially.

So you being there on the board to actually walk them through it and make sure they’re using the systems correctly is so critical.

[00:05:03] Matthew Obenhaus: You’re right. And the accountability of that, of, “Hey, me and my– our team developed this. Let’s see what impact it has,” and it’s we own it, right? And we’re on the board, and we interact with physicians and our board member partners and he-health system joint ventures.

If I’m getting negative feedback or something isn’t quite right in the settings, in the data, in the presentation, then I’m accountable for it. So it is a lot of work, but it’s important, I think, to have that critical accountability and feedback loop in real time.

[00:05:32] Ryan Cohn: Yeah, absolutely. And so you mentioned that you work with centers on, the data aspect and you’re working with them and your technology partners such as HST.

So what led you to sort of combine, the data and all your technology partners and combine it with these AI tools that have recently flooded the market?

[00:05:53] Matthew Obenhaus: Yeah, it’s a good question, Ryan, and I want to give our team and the industry at large credit for… this isn’t– it’s not like I arrived on the scene and all of a sudden I’m doing something extremely all of a sudden out of left field novel and innovative.

I think there’s some innovation involved in what we’re doing in the AI space. However, we have laid a pretty good foundation with our partners down at the facility level And work that we’ve done that’s critical as far as understanding data, cleaning data, mapping data, and making it useful in a presentation format, let’s say a, a Qlik or something like that, a user design that we can do these canned reports.

And we’ve been doing that for several years, and that work frankly predates me. I was able to take work that was already done and grow it and expand it. Now, I will say the world of AI and our particular flavor of choice in this space is Claude. It’s allowed both opportunity to go faster and my own impatience with, “Hey, this is a standard canned report.”

It gives us some really good diagnostic KPIs, but it doesn’t get down to the granular level of actionability. And yes, we have ways of doing that with ma- subject matter experts in supply chain, business office, clinical. Whatever the diagnostic dashboard is telling us, we can dig and we di- and we do dig into figuring out the corrective action.

But the, the world of AI has allowed us and has allowed me to just go so much faster into exploring what’s going on behind the KPI. It’s allowed me to build a lot more ad hoc stuff that is purpose-built for specific facilities, specific physicians, specific payers, vendors, you name it. I’m able to just go so much faster than the traditional product technology roadmap of, we need this new KPI.

We need this new corrective action report. Let’s get it on a roadmap. Let’s do the data mapping. It just, AI allows us to go a lot faster. I will say that w- as we’ve developed a lot of ad hoc AI-generated reports, whether it’s a very deep and detailed physician utilization let’s say anesthesia stipend allocation, labor allocation, these are the sorts of types of modules that AI’s allowed us to crunch through the data and present it in a meaningful and actionable way a lot faster.

I kind of liken it to, our traditional data and analytics development approach and process is more like the, the ants that are on a pheromone trail. They know where the food source is. They keep building on that. AI has just allowed us a lot more of the, the forager ant analogy. Hey, we’re exploring new trails, and we’re developing some really useful things.

But at some point, we also have to standardize and scale those. We just ha- we have more forager ants, but we also are pulling it back into how do we think about scaling this type of report? Hey, we’ve done really good dashboards and analytics using AI for very detailed Position utilization and margin contribution reports in a way that we didn’t have yet in our traditional data and analytics.

Now that we’ve done it and explored that with Claude in this case, now we’re busy thinking about how to scale that and repeat that across all of our facilities, giving highly actionable, highly accurate, and highly competent data to each of our physician owners that tells them exactly where they’re at on their revenue, margin contributions, their efficiency.

So it– to answer your original question, why did I do this? It’s opportunity and impatience and the ability for AI to do it a lot more faster than we had with all the data that w- and analytics that we already had. This just allows us to get faster, a click deeper, and then actionable a lot faster.

And yet I’m still mindful of bringing it back into the full stack of how do we scale and repeat it. So it’s, it is a very nice convergence of where the technology is and what it allows us to do while bringing it back into standards-based scalable reporting.

[00:09:56] Ryan Cohn: Yeah. Yeah. So it sounds like you, you identified that these tools could be– they could be really helpful and you had the data, and you wanted to figure out how can you combine these, these sources together and figure out how to use this data to the best of your abilities.

And yeah, that’s the beautiful thing with AI is, is you really can build so fast and and it’s so customizable to what your needs are. I, I’m curious to hear a little bit more about these, these specific use cases that you’ve built using Claude, and I would imagine you’re using Claude Skills and maybe you’re building some ad hoc pieces of software and reporting dashboards.

What does that really look like in practice? I would imagine our listeners they probably hear about these AI tools all the time and they hear what can be done with it. But it can be a little overwhelming to actually understand what that looks like in, in practice. So- Yeah

could you walk us through a little bit what that actually looks like?

[00:10:53] Matthew Obenhaus: Yeah. I’ll start with a specific use case that I think is relatively profound and obtainable within the audience set here. Which is, look, there’s a wealth and a a wide realm of knowledge coming out of an EMR such as HST, but certainly others, that you, you have all the data at your fingertips that you could ever possibly want to make a lot of really good headway in performance improvement.

Sometimes it’s about organizing and understanding the data. So for me, the outcome was, look, I want to get past these 30,000-foot view KPIs, which are the equivalent of a check engine light. I know there’s a problem. I can see that my supply chain spend as a percent of net revenue and as a per case is much higher than we budgeted and is much higher than a, a facility that looks and feels a lot like this facility with their case mix and so on and so forth.

So I’ve got the diagnostic. We’ve had the diagnostics. Well, how do I get deeper faster on much more of the insights of a per procedure, per physician, per payor, you name it, set of metrics that really allows me to go deep and actionable and understand at least where should I go attack? So in this use case that I’m describing here, the outcome was a, a very robust But aesthetically pleasing and understandable dashboard down to the physician level of how they, how each physician is doing according to their procedures, down to the and looking at their block allocations, their schedule.

How are they doing on a net revenue total, on a margin total, less their supplies and implants, which gets us into a data accuracy conversation that I’m sure we’ll get to. How are they doing on their overall margin and revenue contributions, but also how are they doing on their margin and revenue contributions on a per OR minute blocked and a per OR minute used, which is a little bit of a, a convergence of the efficiency of the time that they use and the dollar contribution.

So that was the holy grail for us, is getting to that level of detailed real-time analysis using the data coming real time on a daily basis out of the EMR. So that is where we started. Now, to get there, right, I’ve given you the outcome and the goal. Yeah. To get there, it says a lot about the data accuracy, which is a different question, a different topic, but let’s assume that the data is highly accurate within a specific facility’s use case.

[00:13:34] Ryan Cohn: Mm-hmm.

[00:13:34] Matthew Obenhaus: What I’ve done to develop… So you can use Claude to say, “Look, here’s the dataset. Here’s what I’m trying to achieve. I want a physician-ready and layperson-ready dashboard, and I want it built in HTML, and I want it built in PowerPoint slides for each physician to view their scorecard.” It’s amazing what just a simple prompt like that to- can do, assum- again, assuming data accuracy, what Claude in this use case can do.

I’m sure ChatGPT and others can do similar things As far as building that sort of analysis. Now to get there, right, I’m really conscious of what my footprint is and my token footprint is, and my usage footprint is for Claude. Yeah. So I do some data cleaning. I do some– And I’ve developed… And Claude has helped me with skills to do this data cleaning and data analysis.

Now, for those of you, I’m not a, I’m not a– Despite my technology title and ownership, I am not a computer scientist. I don’t have developing coding skills. The– This is a great leveler. Something like Claude and skills enablement, you don’t have to be a coder. You’re typing concepts and ideas into a user interface that then in turn, Claude’s doing the heavy lifting of developing that into a repeatable, code.

For those of you that have used like Excel macros, it’s a very similar concept of, hey, I don’t, I have, I just need a rudimentary thought of what I want, and it helps me through the process of coding that into a repeatable skill. So let me get to what those skills look like, Ryan. It’s, “Hey, HST pumps out a ream of information.”

It’s useful for many use cases, but I have a specific thing in mind. So I do a lot of data cleaning and a lot of column collapsing and removing. And the s- I do it the s- I was finding myself doing it over and over again. Like, man, I’m spending 30 minutes just cleaning this thing. I’m gonna hop into Claude and say, “Look, follow my process and enshrine this in a skill.”

So there’s certain columns I don’t want. There’s certain, certain columns like multiple CPT codes that I wanted consolidated and just separated via a comma, ’cause you could still do good data and analytics. I wanted certain specialties to be reconfigured. Hey, we’ve got a lot of ortho-aligned procedures that are actually total joints.

We’ve got a lot of ortho-coded cases that are actually pain management cases. They just happen to be an ortho doc or a spine doc doing them. So there’s some specialty alignment. There’s there’s payor alignment and mapping. Hey, there’s thousands of payors, but you know what? In HST. But you know what?

Each of them map to a Blue Cross Blue Shield, United, Aetna, Cigna, workers’ comp. So it all maps to maybe 10 major categories. Let’s do that. So I’ve got an a, a way for mapping these things, and I want them to be repeatable So a Claude skill allows you to build those things. It works alongside you to build those things, and after it builds it, all you got to do is when you’re in an Excel document, like the, a D5500 report out, out of HST, I’m just ty- I’m just typing /D55cleanup, which is the name of this skill- and it does all these things I’m describing for you.

And then, there’s some other things that I wanted. Look, we’ve always been a hostage of looking at Three to six months back for what our revenue is and like, oh, I’m confident that these things have fi- have been billed out, have been final collected, and we understand that’s the revenue.

Well, why not do more of a projected and real-time revenue based on that facility’s history? Hey, we know this contract, we’re owed this. We also know that this his- that we have five years of history of what our bad debt is. Let’s projec- let’s lean forward and do some projected revenue, and we can see the supplies and implant costs in the data.

Let’s do a projected margin. So we are giving our physicians and our facilities real-time projected revenue and margin. That’s a skill, a repeatable skill that I developed in Claude of, okay, go back and look at this history of this client, look at the contract fee, look at the balance due, and if it’s not closed out, do a, a look back of history of bad debt, and let’s associate that to a projected net revenue and give me a confidence interval.

Hey, we’re ninety-five percent confident that we can get… obtain this financial footprint for this as a projected net revenue on this case that was done yesterday, instead of waiting for it to matriculate fully through, through the cash flow process. So that-that’s just a, a couple of use cases that I’ve developed on…

Again, these aren’t super complex things. They’re repeatable things that you can build skills in. Now, I think we’re gonna continually move up, up chain in the complexity of these things, and companies are already doing really impressive and important things in process stuff, in revenue cycle management, clinical decision support, ambient listening.

But for us at New Health we’ve achieved enough value out of just the mundane, taking the drudgery out of work and getting us to much faster insights and analysis.

[00:18:46] Ryan Cohn: Yeah. Absolutely. I mean, there, there’s so much value in what you were just talking about there. It’s hard to even pick out just one or two things.

But the, the things you were talking about with Claude Skills I, I wanted to comment on because a lot of people, I think they, they have access to these tools like ChatGPT and Claude and, maybe most of us are using, five percent or ten percent of what these tools are really capable of, and it’s r- it’s not any fault to, to us or to anyone.

It’s just these tools have so much, and there’s constantly new features and things that are being rolled out. And Skills is one of those things I’ve been diving into a little bit further recently and the best way I can describe it is if you take a process that you’re doing inside of Claude or ChatGPT, and you’re doing that process all the time, and you have to, re-prompt yourself to, to go through this, lengthy conversation with Claude, that’s something, that’s a workflow that you could turn into a skill basically and tell Claude that I want to take this entire conversation and turn this into something that, that you understand from just a single prompt.

And like Matt said, you can just do a backslash, type your skill, and it’ll know exactly what it needs to do first time. So it’s a really great way to use these tools, and ChatGPT has something similar with creating custom GPTs. You can also create project folders, which is a decent way to do that as well.

So, just to give everyone listening some context of if you only have ChatGPT, there’s other ways you can do this as well potentially. So one thing I wanted to talk about is, you talked about the importance of clean data going into these tools. How important is it to have clean and accurate data when using AI for this type of analysis?

[00:20:34] Matthew Obenhaus: I think it, it depends on the use case, right? One thing I’ll say is, If you feel like you don’t have clean data, and this is where the human to AI interaction works well. We have a relatively new center where we’re still in the journey of making sure that we have accurate information, preference cards, implants captured at every procedure.

AI can actually help with pinpointing those. “Hey, look across this dataset, find inaccuracies and outliers within this dataset, and give me a sense of is there a time issue here? Hey, when we first started, data looked like this, but as we’ve rolled along, it’s improved, and now it’s consistent across these sorts of use cases, these physicians.”

It looks like you’re, for in this example, your total shoulder implants are finally being captured, and they’re accurate every time according to this, preference card and this build set that you should expect. You are good as of December twenty twenty-five, so everything before that is inaccurate, everything after that is accurate.

We’ve used Claude for that sort of use case and analysis for data integrity. But to answer your question pointedly, yes, you’ve got to do that data accuracy work before you get downstream to these board and physician-level conversations, right? Now, there have been occasions where I’m using the journey of data accuracy and understanding to to work with our teams on the ground, our administrators, our materials managers, and supply chain directors for, “Hey, what do you think our data accuracy is?

Here’s what Claude is saying. Let’s work on this.” And we’ve pinpointed issues for preference cards and implant logging per case. It’s been great at that. But it’s also been very good once we feel highly confident that here’s our materials list for these sorts of cases. This is what we shou-we should expect.

And we have ninety to ninety-five percent confidence with everything coming out of our EMR data that we’ve captured that in this month timeframe That feels pretty good to take stuff in front of a board, take it in front of physician owners and members directly for Dr. Jones, generic name. Here is your scorecard and your revenue and margin contributions.

Here’s your outlier cases for high profitability. Here’s your outlier cases for low profitability. Obviously, we, we– that’s a… There’s always nuance to that. Hey, I’m doing this negative or low profitable case because it’s combined with these 10 other profitable cases, and there’s nowhere else to take it.

Okay, that’s a conversation, but it at least allows us a high degree of integrity and accuracy to understand and present these trade-offs and to have the conversation. So to answer your question pointedly, how important is accurate information? It’s essential before we can have that physici- direct physician and board conversation for trade-offs and outliers, and maybe, just maybe sometimes it results in an action plan as far as, “Nope, take that case somewhere else where it’s more profitable.

Payor isn’t paying enough here, but it is paying enough there at the hospital.” Or, “You know what? It looks like we could be more profitable if we got a better rate out of, commercial payor X, and our commercial payor X is using our data across all of our facilities, is under water compared to other facilities.”

Or, “You know what? That particular implant, it looks like we’re paying 30% more than anywhere else that we’re doing that implant. Let’s negotiate.” Yeah. And we’ve used AI to uncover a lot of that stuff a lot more fast and more cleanly with accurate data. So accurate data drives the action plan.

Yeah. And it, it doesn’t get you laughed out or scorned out of a conversation with a physician, because as soon as you present inaccurate data, you may have lost them for months if not longer, as far as trust in the data. So it does become essential for that. Yeah. But again, AI can be used to help drive data accuracy and understanding and confidence in your data as well.

[00:24:34] Ryan Cohn: Yeah. Yeah, that makes perfect sense. I mean, I would imagine, inaccurate data in creates inaccurate results a lot of the times, unless you’re using AI to clean up that data and, improve the data set. But yeah it’s kind of like AI slop, it could turn into that if you’re not careful.

So you definitely need to make sure you’re you have the right tools to be able to capture that data and then to be able to, present it and clean it up.

[00:24:59] Matthew Obenhaus: Yeah. And I liken AI to an extremely bright and capable and hungry and fast data analyst. But it’s only as good as the data and the prompts you feed it.

So all these things are essential. Now- My advice to anyone g- just, “Hey, man I don’t know about my data. I don’t know about my prompting a- ability.” You’ve got to start the journey now to move up the curve. Mm-hmm. But I think you’ll find that i- if you’re conversant in your facility, your organization’s data set, at least nominally, and you have some good in- intuition of what’s going on, don’t be afraid…

Like, getting started now is the most important step, and you’ll find yourself like I was months ago, amazed at what AI is able to do.

What it’s able to pinpoint that the, the mere brain wasn’t able to do. Again, that very speedy analyst. But also, you’re gonna find that it still needs a human checker, a human validator, a human…

And your inherent knowledge in a lot of your roles is gonna naturally figure that out. Yeah. And it’s important for me to be able to blaze trails with AI, but it’s been extremely important for me to have the subject matter experts in supply chain, business office, clinical, some of our administrators.

I’ll give one of our administrators extreme credit for just her ability to interact with me and AI, and she’s asking for things, and I’m able to turn it around. And then she’s able to really quickly spot, “Well, that didn’t turn out quite right,” or, I think this is missing a certain nuance of the way that this payor contract works,” or, “The way that this vendor negotiation works.

You’re missing re-” she’s able to really quickly see what it didn’t capture- … feed back that, feed that back into my process, and refine and improve the products. And that still, the human engagement is still extremely critical at this point- Yeah … working with these AI models.

[00:27:01] Ryan Cohn: Yeah, that, that’s such an important piece.

I think a lot of… There’s a lot of anxiety around AI tools right now. I think a lot of people are worried, about it replacing people. They’re worried about the security aspects of it. And you touched on a really important part there, which the- these tools they’re tools at the end of the day, and they’re meant to improve our workflows and how much we’re able to do in a given day.

And, we, I think to some degree, we should embrace using these tools to make us better at our jobs, really. And-

[00:27:37] Matthew Obenhaus: Absolutely … i,

[00:27:37] Ryan Cohn: I think the people that, that take that sort of mindset and dive into these tools and, treat it like a passion project or like a, a hobby, to some degree you’re the type of people that are going to learn a lot quicker and probably advance faster.

And that’s how I’ve been looking at it, and I, play with these tools on the weekends and at nights and just see what’s possible. And it sounds like that’s sort of the approach you’ve had with it too, Matt.

[00:28:03] Matthew Obenhaus: Yeah. Ryan, let me elaborate on that if I can. Yeah … it’s, I’m a te- I, yeah I don’t have the entire– I don’t think anybody has the crystal ball, and you have all kinds of different techno-optimist, techno-pessimist of where all this is gonna go.

What I have experienced in my personal anecdote, and I was even texting a friend moments ago about she feels almost guilty for, like, sh- her role as building, AI agents and whatnot. She al- she loves it, she enjoys it, but she almost feels guilty about maybe some of its implications.

And my response to her and my thought here is, look, my experience, at least thus far, has been very much, man, this, th- this is giving me and others almost like a new joy in some of the work that we do because we’re able to get to these insights and add value so much faster to our clients and our physician owners and our health system joint venture partners.

It’s taken a lot of the drudgery work that was non-value add.

[00:29:07] Ryan Cohn: Mm-hmm.

[00:29:07] Matthew Obenhaus: Yeah. I pride myself on being a really dang good Excel model developer, but do I really need to be doing that? The models that used to take me 40 to 60 hours to fully develop and flesh out, working with Claude, I can do in two or three.

Yeah. That’s extremely powerful. And what have I found? The work that I used to do in 40 to 60 hours, maybe I can do in five to 10 now. Am I just coasting and not spending 30 hours? Am I just wasting time? No. I’m finding other more useful things to do, and I think that… And it’s just value-added work for our clients.

And I’m finding the same thing down on the ground. The stuff that we’re doing to enable and help our materials managers and business office directors with some of this stuff, h- I have no thought of replacing any of them. If anything, they were already underwater, and this is just allowing us to do a lot more, a lot faster, and it’s allowing us to find those bad pricing arbitrages with, vendors.

It’s allowing us to attack that aging bucket that is just a process error or a documentation error. And it’s just allowing us to attack that stuff faster with the same people that we have. Yeah. And the more we find those revenue streams and cost saving streams, the more I’m gonna be able to say, “Look, we can go hire that next front off- office person.”

We need… We’ve long wanted and needed that extra RN to help us get more efficiency on those flip rooms. I’m not in a space where I’m thinking anything about labor reduction. It’s, if anything, freeing up cash to, to de- redeploy in areas that we’ve long needed.

[00:30:43] Ryan Cohn: Yeah. Yeah, absolutely. I mean, you used to have to be a data scientist and have a degree to really understand, the data that’s coming from some of the tools that ASC leaders have available.

And, now you can use these tools to really dive further and understand it and to give you some takeaways of how you can improve your centers. And I think that’s such an important aspect of it is just learning to use it- Yeah … so, so you’re ahead of the curve instead of behind it, essentially, is what it comes down to.

[00:31:12] Matthew Obenhaus: Yeah. I wanna say I don’t have some special genius. I think this is democratizing intelligence and knowledge and insights and corrective actions- Yeah … which is a beautiful thing for the industry at large, especially as we think about and experience consolidation, competitive pressures, margin pressure with cost increasing, reimbursement rates being anemic.

These are almost like the essential survival tools for us as an industry for creating independence, autonomy for these local centers, for physicians, you name it. So I think it’s a bit of a, a godsend in a lot of ways for us in our long-term survival.

[00:31:51] Ryan Cohn: Yeah. Yeah, I agree. I mean, with reimbursement rates continuing to be more, more challenging as years have gone on.

We need to be a little bit more creative and data-centric, and that’s- Mm-hmm … that’s really why I wanted to have you on. I knew you’d have a wealth of knowledge around this, this topic. And- Thanks … I think a lot of people are g- are gonna find a ton of value from this. Now, one thing I– that’s come to our attention recently is how, some ASC leaders might be a little bit worried to implement AI in their center, due to the data and security, aspect of it.

They wanna make sure that they’re just abiding by the rules and, there, there’s no leakage anywhere or anything like that. What do you– what would you say to those people, and what guardrails, would ASC leaders wanna put in place when using these tools?

[00:32:37] Matthew Obenhaus: Yeah. I think the obvious risk is one of loading up PHI and data breaches.

We feel fairly good about where we’ve arrived from an anthropic perspective and some of the HIPAA compliance and ability to sign a BAA for certain things. For our facilities, and this would be my advice for some of the, the companies that have the breadth and scale of a new health, of, “Hey, think about ways that you’re helping your facility leaders with setting up an AI council.”

And an AI council will help vet use cases, products, will set standards of compliance and behavior as far as these are the things that you can and shouldn’t do with AI. Even with a BAA in place with Anthropic, we’re still layering in certain logging and e- and event tracking tools. We’re considering things like single sign-on and other things for logging into that environment.

And we’re also… The other thing I would think about is, look, do– much like you’ve always been, nothing’s changed as far as how you handle things like HST or EMR generically or source system outputs. You never wanna load up or send via email PHI. Nothing’s changed with this use case. You– it’s a generally a good idea not to load up PHI into these environments.

So we’re helping our facility leaders and our subject matter experts and cross-functional team members that support our facilities with a process that almost just- Dummy-proofs it. We take our data exports from sources like HST. We upload it daily into a share– an accessible SharePoint drive that already has removed the PHI.

There’s not very many people in our organization that need the PHI, and if they are, they’re down at the facility level in those leadership and director roles. They can access HST directly for that. For p- for the vast majority of people, I myself, you don’t need the PHI. You’re looking for broad data and broad trends and we’re able to surface that information to all of our people that are using these tools with a report that strips that stuff out automatically.

So I think for our facility leaders, it’s that may not have the overhead and the coverage of a company of like us or SCA or whoever it is, then you’re just gonna wanna work with your IT team to make sure that your cybersecurity environment is secure. You’re gonna wanna explore things like what a BAA might have with a company like Anthropic and what it doesn’t cover.

And you’re just gonna wanna have a greater degree of caution, trepidation, and management of the data that you’re pulling in, and you’re just gonna wanna make sure that it’s stripped of PHI before you even engage. Those are just some quick off-the-cuff thoughts.

[00:35:22] Ryan Cohn: No, that’s great. That’s amazing. I…

great insights, and I know I definitely learned a lot in this episode and wish we could go for hours. We’ll definitely have to have you come back on another episode. But last question here, Matt. We do this every week with our listeners. What is one thing our listeners can do this week to improve their surgery centers?

[00:35:40] Matthew Obenhaus: Yeah. I, in thinking about this question, it’s- I– There is so many things to go attack. What I would suggest is, and since the theme of today is how to use AI, it’s using AI starting now. Like, plant that seed now, get comfortable conversant with it, see what it can do in your environments safely and securely, of course and start that thread.

You’re gonna fi- you probably have some general sense of your diagnostics. How is my supply spend? How is my SWB? What are my clinical indicators and benchmarks as far as infection readmissions? You probably know those. So start with your general diagnostics and just start pulling on those threads.

Use AI as an enabler to tease out the high-dollar stuff from the low-dollar stuff, the meaningful stuff from the non-meaningful stuff, your block schedules, who’s– which physician is using those well with utilization, and which are contributing well and underutilized. Start pulling on those threads, and pull that thread relentlessly until you get to the final corrective action.

Or you may find that it’s not actionable. Hey, Dr. Smith just happens to be slow. There’s nothing we can do about that. He or she is a brand-new physician. We’re gonna just– we’re gonna live with that until they get faster and they get more reps. You may find an action that isn’t actionable right now. Move on to the next.

So my advice is start those threads. We’re operators. We know the end goal is to drive profitability and clinical optimization and success and efficiency of a center. There’s– Even the best performing centers, our work is never done. There’s always something that can be im-improved or some new thing that pops up.

Pull those threads relentlessly until you’re done with them and move on to the next. That’s my advice. And AI can greatly enable our ability to pull or discard those threads that aren’t meaningful much more quickly than in the past.

[00:37:46] Ryan Cohn: Yeah. Well said. Well, thank you so much for coming on, Matt.

[00:37:55] Ryan Cohn: a new industry forecast points to significant continued growth for surgery centers. Vizient’s twenty twenty-six Impact of Change forecast projects that ASC procedure volume will grow twenty-four percent over the next decade

Orthopedics, gastroenterology, ophthalmology, and general surgery are expected to remain at the foundation of that growth. Cardiovascular procedures are also emerging as one of the industry’s next major opportunities. The forecast points to several forces driving this trend, including advances in minimally invasive procedures, payor demand for lower cost care settings, changing patient expectations, and the continued expansion of procedures that can be performed safely inside of an ASC.

But a twenty-four percent increase in volume won’t automatically translate into stronger financial performance, of course

Growth can create new pressure on staffing and on anesthesia coverage, on scheduling, supplies and implants, and on your overall physical capacity. And centers will also need to determine whether the cases moving into their facilities actually make financial sense based on reimbursement and the total cost of delivering care.

And that actually connects directly back into our conversation with Matt today. National forecasts can help ASC leaders understand where the industry is heading, but each center still needs accurate facility-level data to decide which of these opportunities are right for its market.

ASC leaders should be looking at which specialties are growing locally, where block time is available, which physicians have additional capacity, and which procedures are delivering the most sustainable margins for them

The centers that benefit most from the industry’s growth might not simply be the ones that have the most demand, but they’ll be the ones that are the most prepared to turn that demand into efficient and profitable case volume

Next up, we have HST Connect happening next week, August 11th and 12th at the Joseph Hotel in Nashville

This two-day event marks the return of HST’s customer conference after several years, and it’ll bring ASC administrators, clinical leaders, operational teams, business leaders, industry experts, and most of the HS team together for practical education, hands-on training, product insights, and conversations about the challenges that surgery centers are facing.

The timing connects especially well with today’s Trust the Data discussion. Will Evans, HSC’s vice president of data science and insights, will present findings from the twenty twenty-six State of the ASC Industry Report, and other sessions will explore how ASCs can use their data to understand profitability, improve financial clearance, communicate performance in the boardroom, and strengthen governance and compliance.

Attendees will be able to follow the full patient and case journey through connected surgery center workflows, learn strategies for improving upfront collections

Hear directly from HST customers who have transitioned from paper to electronic charting. Get an inside look at HST’s product roadmap and long-term platform vision. The event will also feature a keynote from Dr. Betsy Grunch, a board-certified neurosurgeon, founder of Southern Neurosurgery, and the creator behind Lady Spine Doc with over four million followers across her social media platforms.

And beyond the educational sessions, HST Connect is really designed to be a fun and memorable experience. Attendees are going to have plenty of opportunities to meet other surgery center leaders, explore Nashville, and build relationships outside of the conference rooms. We’ll close out the event with a conference party featuring food, drinks, networking, and even a live performance from Autumn McIntyre.

And before you start googling and looking into her, yes, she is of the Reba McEntire lineage. It should be a great way to celebrate the return of HST Connect and spend time with ASC leaders from across the country.

I’ll also be on site recording podcast episodes and creating some additional content, and I’m really looking forward to meeting everyone who’s able to attend. And for those of you who couldn’t make it this year, we hope to see you at next year’s event

And finally, we’ll end with a lighter story about a pediatric surgeon who found a creative way to make surgery a little bit less frightening for children. Dr. Leandro Brandão de Marisch is a pediatric ear, nose, and throat surgeon who lets younger patients choose a superhero or princess costume to wear on their way into the operating room. He began this practice about six years ago after seeing how scared children became before surgery. He wanted to help them feel stronger and give them a more positive experience during what can only be described as an overwhelming moment for them

once the children have picked their costume and are all dressed up, they can then choose to run or even fly through the hallway with Dr. Dumaresq carrying them on his shoulders. After the procedure, they get to take the costume home as a reminder of their bravery.

This change has helped make the experience more comfortable for the children, their parents, and even the surgical team. it’s also a good reminder for surgery centers that improving the patient experience doesn’t always require expensive technology or a complicated new program. Sometimes a small, thoughtful change to one part of the patient journey can completely change how patients and their families remember their care

And that’s all for this week. Please make sure to share this podcast with a friend or colleague if you found it helpful. It’s the number one way to show us that you find this kind of content helpful, and it’s also the best way to support the show. Thank you so much for listening to This Week in Surgery Centers.

We’ll see you again next week