Astrana Health, Inc. ($ASTH)

My name is Craig Jones, I'm of the health care analysts here at Bank of America. And today, I have the pleasure of hosting Brandon Sim, CEO of Astrana Health. So thanks for being here. Do you want to start with an intro or you just want to go straight to Q&A. Yes, go for it.

Yes. Well, thanks so much for having us here. It's beautiful to be in my hometown, I guess. So it's nice. Thank you. It's always a great conference every year. And I'm Brandon Sim. I'm the CEO and President here at Astrana Health. We're a value-based care company, trying to realign the system to build infrastructure that allows for better patient outcomes and ultimately, hopefully, at a lower cost. We've been growing very rapidly over the last 6-, 7-year period. Last year, it grew over 50% year-over-year and continue to grow in a profitable and free cash flowing manner. Some puts and takes, but broadly, we just reported our Q1 results and things came in pretty well ahead of expectations. We grew free cash flow, I think, 3x year-over-year. Revenue grew well over 40%, 50%. So we're pretty pleased with where things are going and look forward to a strong rest of the year here.

Great. All right. So why don't we start with some of your inorganic growth about a year ago or so, closed the acquisition of Prospect. So after the first year, you've guided the high end of the $12 million to $15 million in synergies. So why don't you talk us -- just walk us through how the first year has gone versus your initial expectations and maybe any positive or negative surprises along the way?

Yes, of course. So it feels like we've been doing the Prospect thing for like multiple years. Craig, we've been certainly chatting about it for over a year. But reality, we actually closed the deal only 9 months ago. It hasn't -- we haven't gotten into the year mark yet. So it's been three quarters of reporting so far. Prospect has really been on track to ahead of schedule. We're really pleased with how it's turned out. Of course, with any transaction, there are of that size, [ $707 ] million transaction, there are going to be puts and takes. There are going to be some challenges to work through, and even more so because we acquired it out of a bankruptcy situation. But broadly, we've been able to very quickly integrate from an operational standpoint, get their teams over 1,500 employees that we took on integrated with our teams under a singular operating structure, a single reporting structure and more importantly, using the exact same care plans, care pathways and kind of clinical protocols. And then also kind of supporting all of that is the integration of the technology. So having the prospect teams utilize kind of our unified data layer, our application layer, our agents that we built on top of the data and integrating those into a singular data platform so that you can see members across both. There isn't kind of a silo of these are Prospect members and these are Astrana members. And then implementing the exact same care planning tools that we built, the care navigation tools that we built the care management tools that we've built for the legacy Astrana population and the Prospect membership, which is substantial, 600,000 members. I think that's gone well. And because of all that, we've been able to, like you said, track towards the high end of the synergy range, $12 million to $15 million. We've been able to extract some of the revenue synergies already as well by combining contracts with the same payers that we might have held separately, but now are under one unified contract. We've been able to improve or get their trend, Prospect's trend to look a little closer to ours. For example, when we bought them, we were running legacy core Astrana around 4.5% trend. Prospect was probably around 200 basis points higher than that. We underwrote a 50 basis point improvement year-over-year on Prospect trend, and it's -- even in Q1, that's coming in even better than that 50 basis point improvement. So we're starting to see some of the impact on the care model side, yielding some early dividends. Of course, the full impact is probably going to take 18, 24 months on the MLR side.

Yes, absolutely. So I don't follow up on the 50 basis points there of medical cost trend. I think that's pretty powerful. Can you -- any details in the secret sauce of why adding Prospect to the Astrana platform really enables that you better manage trend better or however this is working?

Yes. I mean there's -- I think I said this before, I mean there's really no magic to what we do. Ultimately, it's a services business although we're taking care of patients. So the -- our doctors are being empowered to take care of patients. I think part of the secret is to not allow any patients to fall through the cracks. And if there's something anomalous or suspicious happening to be on top of the ball and kind of being able to see that early on and address it without it spiling out of control and causing an issue in MLR. So what the software really does for us, and I think you came over very kindly yesterday to our clinics and got a bit of a demo of some of the software platform is that, one, has everything in a longitudinal data record. And two, has -- you can almost think of it as a harness on top to ensure that no patients are falling through the cracks. So historically, if someone forgot to call a patient upon discharge, then they forgot and that patient doesn't get a call and they maybe have a readmission that's unnecessary and cost the system $20,000 that didn't need to happen. That can still happen today in a human setting. But if someone forgets, the software is there to remind the patient or increasingly genically just call the patient automatically so that no one is actually falling through the cracks. That's just one example from a transition of care standpoint. There are tons of other examples in terms of gaps in care, preventive care, their examples around ensuring that patients who need it have their blood pressure controlled or their hemoglobin A1c controlled or even something as simple as making sure that a patient actually gets an appointment for a follow-up with a specialist instead of them trying to call during their lunch break, not being able to find someone picks up their phone and then putting off that visit for 1, 2, 6, 12 weeks and then ultimately having chest pain ending up in the hospital, for example. So it's a variety of all the little things, kind of maybe each of them doesn't amount to much a basis point here or there, but you kind of automate and ensure that all of those basis points are being picked up, ultimately, you get 50 basis points maybe of improvement, right? And over time, that really compounds, especially as you drive more and more members through that care model.

Some tasty secret sauce there. So maybe you mentioned the revenue synergies a minute ago, you've already started to unlock those, which it's great to hear. And my guessing that was probably a bigger angle when you did make this acquisition. It wasn't necessarily the $12 million to $15 million we're going after. It was this revenue synergy. So can you talk us through sort of what you've already been able to unlock, how big the opportunity is there and maybe some kind of time frame, if you want to throw any numbers out there?

Yes, sure. And I think you're hitting on an important point. It's not necessarily just the G&A synergies, which are nice, certainly to have. But I think more importantly, the opportunity to inflect cost trend, which I guess is, is a cost synergy, but kind of is more core to the mission. And then as you mentioned, to improve the revenue opportunities for the combined business as well. On the revenue side, I don't want people to think we're just taking kind of the higher of the rates and then asking for that higher rate across both books. I wish we could do that, but it's really not as simple as that because the payers on the other side, obviously, are not incentivized or inclined to allow us to do that for no reason. It really is about a partnership with the payers that we have had for decades, some of these payers and working with them to figure out a way that our now broader network and our broader care management and quality management capabilities can better -- can lead to a win-win for both the payer and us. I know that's kind of vague talk, but I'll give a concrete example. I mean there are many payers who are still in their margin recovery path. Part of that margin recovery means that they have to pay their capitated networks a lower amount. We would, of course, want a higher amount if we're going to capture revenue synergies. So some of the ways that we can have that work is, well, if there is a group out there that is even higher cost than we are, that either we or legacy Prospect are, could we, with our combined network and our capabilities, combined geographic presence, care management ability, so on and so forth, now take on some of the membership that, that payer had assigned to a more expensive group, thereby helping the payer lower their cost, but also getting us part of what we want. There are other examples of this kind of collaborative thinking. But ultimately, it's in a choppy environment for MCOs, better this quarter, but in a choppy environment broadly, it's trying to work really closely and figure out how we can both get what we need, so to speak.

Yes. Yes, choppy for sure, but I'm just saying getting better. So that's a good segue into let's pivot to Medicare Advantage or Medicare. cautious optimism, right, out of some of the bigger national Medicare Advantage players around trend specifically. So maybe what are you seeing in trend in Medicare Advantage? What do you start with? What are your assumptions going into the year versus last year? And then how has that played out year-to-date?

Yes, sure. So last year, we had a pretty successful year controlling medical cost trend. We're around 4.5%. We actually ran just under 4.5% for the year, year-over-year. This year, we actually underwrote to 5.2% medical cost trend across the business, so actually a slight increase. And part of -- and all of that really related to kind of the weighted average of the 4.5% from the legacy Astrana and then the 6.5% from the legacy Prospect underwriting a bit of improvement on the 50 basis points for legacy Prospect. So far this year, we've come in it's early in the year, so I don't want to be too excited, but we've come in well relative to our trend assumptions. Of the 5.2%, breaking that down line of business-wise, we expected Medicare Advantage to be slightly better than that. We expected commercial and Medicaid to be slightly worse than that. And relative to those expectations across all lines of business, we've outperformed in terms of trend. We do have a lot of California membership. We've got Texas membership. We've got Nevada membership. So of course, weather and flu were not necessarily a large impact for our businesses broadly. And then kind of across the board, we had lower admits. We didn't see any nominal spikes in any kind of particular area. So we feel very comfortable coming out of Q1 around trend.

All right. Awesome. So maybe final rate notice came in better. They delayed the implementation of the new risk adjustment data, but they did remove unlinked charter views. I think you said previously minimal impact from charter views and you think even maybe a lower-than-expected impact from some of the other changes due to lower RAF. So you want to just help us explain why you are more insulated here? And then if CMS does go ahead and implement that -- the new data for next year or for '28, I guess, what would you expect that impact to be?

Yes, for sure. So where we ended up, as you alluded to, the 2.5 net kind of average impact. There's another approximately 1.5, I think, for the disallowed diagnoses, the unlinked charge cases and the audio-only calls. So those are not parts of our model. We don't regularly make audio calls and code based on that. We don't do unlinked drug cases. So we believe that to be a very, very minimal or basically 0 impact to the business. So the real effective kind of rate for us on our average rate book is probably closer to 4%, the 2.5% plus 1.5%. And that's not including the kind of anticipated RAF improvements that CMS thinks kind of all organizations will get better at RAF year-over-year by 2%, 2.5%. So we feel pretty comfortable that our rate is going to be 4% to 6%, call it, and depending on our ability to inflect RAF next year. And trend, as I mentioned, is around 5.2% and coming down because of the Prospect improvement. So we feel pretty comfortable that this 27 is a margin neutral to a margin accretive year. I think CMS took some of the feedback around the coefficients. Now to be fair, they also did say that they didn't think [ skin subs ] were a big part of it, so they put out to those [indiscernible] anyway, which is fine. But we took the coefficients that they put out in the initial rate notice and applied them to our population's HCCs, chronic condition prevalence. And we tried to recalculate kind of what would be the net impact to Astrana had they gone live as they were projected to go live. And we found that it was around a 1.5% impact to Astrana headwind. So 1.5% versus a 2.5% for average for the industry as originally constructed. So to your question, if they were to go live next year, without any changes, assuming that they didn't like reregress on kind of additional years of data, that would be an additional 1.5% headwind for us. I just still think it's obviously bad, but better than industry impact.

And net advantage against the industry, I guess.

And then kind of a quick point, [indiscernible] I think you really want to brought this up maybe, but the further away we get from some of the skin sub -- high skin sub fraudulent utilizer years, which were like '24, '25, et cetera, the more those claims kind of move out of the sliding claims window upon which the regression will take place, right? So hypothetically, I mean, to take an extreme, if we went to like 2035 rates and they reregress then, they would use 2032, '33, '34 data. And obviously, none of those years would have the skin sub data in it. And so really, the further away we get from '27 the less skin sub impact we'll have, hopefully.

Yes. Okay. That does make sense. So maybe let's say they potentially can reintroduce those -- that dynamic. Is there anything else like I think [indiscernible] out link chart reviews potentially or maybe health risk to come up. If they were to go after that in 2028, any idea on what the impact would be to you and just -- and maybe to you and just Medicare Advantage in general?

Yes. I mean maybe I can lump all of those into a category. And the category would be broadly ways to make your patient look more risky to the system, like higher cost to the system than they actually are, regardless of what the method is, right? Like it's kind of a whack-a-mole chase of like the folks will try to take advantage of one thing and then you'll get rid of it and then you'll take advantage of something else. So maybe broadly across the entire category. And the way that [indiscernible] and others at CMS and have thought about doing that is which they're piloting in the lead model, they see a lead model is to just say, I can't always keep guessing what the next form of up-coding will be. It could be on linked today, it could be audio-only tomorrow. It could be linked chart chases the next day. It could be HRAs the following day or it could be anything. So instead, I'm not going to let you guys do coding at all. I'm going to tell you what I think the risk should be using AI or whatever statistical modeling they want to do. From all the information that I have, I CMS have about the patient. And I will just pay you based on what I think the risk is. And that's it. No one gets to code anything. You guys all get your toys taken away. Only I get to decide what I pay you. So -- and they're piloting that in lead, right, that they are working on this AI and risk model and lead. And so ultimately, I think the convergence of all the things that you're talking about, whether it's HRA or like chart chases or whatever, converge to the idea that CMS ultimately will just pay you what they deserve what they think you deserve, frankly. And I think in that environment, we are pretty okay. I think we feel very comfortable with that, because I think we're actually getting paid less than what we deserve, and we're spending dollars. We're having to spend dollars in investment doing kind of more mundane and boring things that don't really help the patient like coding more accurately than spending those dollars actually benefiting the patient. So if CMS were to say, you know what, no one's coding anymore. We're just going to pay you what we think you deserve. We might actually get paid more than we're getting paid today and not have to spend all the dollars and waste the doctor's time doing all the coding. So I would welcome it, frankly, and I'm excited to see how it looks like in lead.

Yes. So yes, that would be very cool. Definitely be interesting to watch the AI risk adjustment model in lead. So it does sound like they're keen to do something, right, CMS and potentially as early as '28. So -- but maybe this AI model, maybe that's a little too early to go right into '28. So is there something we think like a middle ground? Like if they want to -- what could they do in the meantime to semi blow up the risk adjustment model, then maybe we'll see a technical notice this fall? Or how would you design this maybe some halfway there type of risk adjustment model that we could see next year or 2?

Yes. I mean even in lead, to your point, it's not phasing in the first year of lead or the second year. I think it's like the third year of lead. And even then it's a 3-year phase in kind of like v28 is like 1/3 at a time. So -- and that's a CMI program, not even Medicare Advantage or MSSP. So probably after that, if it's successful, then they'll start thinking about phasing it into MA. So I agree with you, 2035, again, like you just. Yes, exactly. We're a long ways away from that. And probably the Medicare Trust one is going to be out of money by then. So it will be too little too late. But -- so I agree with you. I think the -- there are some intermediary measures. I think they're doing a good job, honestly. I think you saw Chris [ Komps ] paper that said that the v28 already decreased the overpayments by not all, but quite a lot of the 10% impact from MedPac down to maybe 2% or 3%. I forget the number. You take out these unlinked charge cases, you take these diagnoses, maybe you get that down even closer to maybe 1% or 2%, I don't know. I haven't run the numbers and you keep kind of hammering around the edges and you get maybe to approximately equal. I think one thing, this is not be lobbying at all, but naive thought is like one thing is instead of regressing claims on Medicare fee-for-service data to give you coefficients for Medicare Advantage, we all know that Medicare Advantage behavior for providers and for patients is very different than an original Medicare. So maybe one thought is maybe we just regress claims data against Medicare Advantage claims data instead of Medicare fee-for-service data because part of the Medicare fee -- part of the skin sub issue was that the skin sub fraud was happening a lot more in fee-for-service. Why not just -- or maybe all of it yet, because the incentives were not aligned -- I mean, are aligned that way in fee-for-service for fraud and not aligned that way in Medicare Advantage. So why not just use the same universe to regress, right? That seems reasonable, but...

Yes. No, totally. That definitely makes a lot of sense. So why don't we switch over to Medicaid. Let's start again with enrollment and trend. What do you assume for '26? I think you said Medicaid enrollment maybe a little worse than you initially expected, but the margin is actually better. Does that imply maybe acuity not as bad as you thought? Maybe any way you can dimension those 2 dynamics for us would be great.

Yes, sure. So we originally expected, I think, 0.75% to 1% disenrollment a month, call it, like 10-ish to 12% for the year. And we'd assumed a 150 basis point headwind in terms of rate acuity mismatch. So we thought the trend would rise faster than revenue would rise. And we had originally sized the combination of those effects in the 20, call it, mid-20s million $25 million headwind to EBITDA. What we're seeing in Q1 so far is that disenrollment has been on the high end of that assumption range, so closer to the 1% a month, so closer to 12% a year, whereas acuity has been -- or kind of adverse selection from the disenrolling members has been less impactful than we thought. And those have broadly canceled out for the most part because we have more members dropping, but the existing members, but the members dropping were less healthy than we thought they would be. And we ran a couple of different slices. We took the members with no claims, for example, and looked at that as a fraction of the total membership before and after this enrollments didn't find a statically different -- statistically meaningful delta. We took various sensitivity thresholds to not just 0 claims because what if you see the doctor once and then you don't use the system for the rest of the year. So we took members with less than 10% MLR. We took members less than 20% MLR. We took members less than 30% MLR and checked the prevalence of those members pre and post disenrollments and still didn't find a statistically significant delta in any of those kind of thresholds. So the histogram, I suppose, of like MLR distribution looked pretty similar before and after distribution. So we feel pretty comfortable that that's the case that there is not a huge amount of adverse selection in our model. Part of it, we think, could be due to our -- the way that we receive attribution, which is maybe different from a plan. We receive attribution -- members are attributed to us as risk-bearing members if the member selects an Astrana PCP as their primary care physician or if they see a plurality of of encounters with an Astrana PCP. Sometimes you can get out of assigned members, too, but these other mechanisms also play a role. So it's pretty hard for us to be a plurality of visits if the member literally has no visits, right? And it's also pretty hard for -- it's atypical for a member to choose -- actively log on, choose a PCP, but then like not use the health care system at all. Typically, those members are not even bothering to choose anyone. And maybe that stays with the plan.

Yes, that's a good point. But -- so in terms of the acuity shift, so it's good to hear you're potentially not as high as you thought. We've heard that from other larger national Medicaid players as well. And it seems like everyone is kind of calling for 2026 as the trough for Medicaid margins. And I know you still have that disparity between rate and trend. Do you think -- is this the bottom? Is this the bottom for the Medicaid margins?

I hope so, but I think maybe we -- last for -- maybe I'm pessimistic, but I think maybe we're -- it's this year, maybe it's a little bit of the next year. And then I feel a lot more comfortable saying '28 is kind of going to be a better year. Look, '27 is being a better year, and we get good rate increases in Medicaid in '27, but that's also great. It just hasn't necessarily been -- I'm not going to be on it.

Yes. Fair enough. So maybe thinking back -- maybe ask you a different way, like thinking back when we have seen this big acuity shift from the disenrollment, how long -- when was sort of the biggest shift? And how long do you think it will take California and others to incorporate that into the rate, right? How much longer is it going to get until we see that big acuity shift finally incorporated?

Yes. I think that's, that's why there's disagreement -- like I may not think it happens until '28, frankly, and maybe some things -- some folks think that that's going to happen next year. But again, we get a percentage of the premium that our partners have. So we're all on the same side here. I would much rather prefer that California or other states, Nevada, Texas, Georgia, where we have Medicaid presences have an appropriate rate update to match the true acuity and the cost of the members that we're paying for.

Okay. Yes. And so it sounds like we don't know when this is going to happen, but so we get stable trend and the rates come back, I mean, how profitable could Medicaid be for you? Like could there be some very profitable years after maybe some not ones?

I think so. I mean all these lines of business kind of operate on cycle, right? We saw the doom and gloom around Medicare Advantage a few years ago and now folks feel good about Medicare Advantage again, maybe or better than they did before. And then that was bookended on the other side, way before during COVID of everyone loving Medicare Advantage and wanting to expose themselves to Medicare Advantage as much as possible. And Medicaid, we had many years of Medicaid being quite profitable as well. So I think there's certainly an opportunity once states figure out the budget and we figure out the impact of OBVA and that we get back to higher margins in Medicaid. I don't mean 15%. I mean 5% to 10%, but I think that's possible.

All right. So maybe then 2027, we've got potentially work requirements, 6 months coming in. Have you heard any -- we're still waiting from some ruling of the CMS in June. Anything you're hearing from California? I would imagine they're not going to start right away, but anything you're hearing on when we might see work requirements come in for you?

I assume not. We're not hearing -- it's unclear. To us, Yes, it's unclear at this moment. I mean you would think you would think maybe not just given the blue going in the state, but you never know these days.

All right. Interesting. So maybe we've got a couple of minutes left here. We definitely want to hit on AI. We had a great type demo from yesterday. You want to give some use cases you've got implemented, maybe ROI? How do you see AI incorporated into Astrana now and next 2, 3 years?

Yes, sure. I think before you even get into AI, you need to substrate upon which AI has to act, right? And so you need the data, you need the agent needs to be able to see across -- need to see across your enterprise and actually act on that data and then perform actions in that ecosystem. It almost be like you work at a bank. So you guys have a [ pennies ] wall, right, between certain parts of your bank. And so some things that happen on one side of the wall, you may not be privy to you and vice versa. And so it's somewhat of that. If you have a part of the bank that's doing something on the other side of the wall, you're not going to know about it. No matter how smart you are, even if you're in AI, you wouldn't know about it because...

That's why don't know about it.

Definitely don't know about it, because it's the other side of the wall. And I think a lot of organizations operate that way when they don't really have to -- I mean, you guys have to, but a lot of organizations operate that way when they don't have to. They have all their information in different silos with different artificially constructed walls. And so even if they were to implement AI or hire a fancy vendor or a consulting firm or whatever to implement AI for them, it's fundamentally going to be stuck inside of its own four walls and not be able to see the side of stuff on the other side of the company. So what I mean by the substrate is that we have built every time we integrate something, every time we JV with someone, every time we add a new provider group to our platform, we integrate them consistently without fail into the unified kind of data layer and then the conceptual concepts, the semantic layer on top of that, the strong oncology so that the AI has an understanding of the concepts of what value-based care, managed care and health care are and so that it can operate on that unified data layer so that it can see across the enterprise. We don't have that to start with. No amount of AI is going to do anything because it's just going to be stuck in its little box, right? Now -- of course, once you have that, that's not enough, you've got to build the application layers on top that are patient, provider and care management team facing. And then you've got to build the agents on top that can autonomously act to hopefully lower G&A or to increase the amount of care or engagement that you're offering to your patient base. So yesterday, you were at some of our clinics, you saw a bit of that. You saw some of our provider-facing tools, single pane of glass because we act as a single payer and we act in a delegated model. We can show not just the quality, the risk adjustment, the pop health, the risk stratification, which a lot of pop health platforms can do, but we can actually combine that with the ability to submit prior auths directly from the same platform. We can combine that with the ability to see your prior auths and submit your claims and see your claims and see your explanation of payments all on the same platform because we serve as both the provider and the payer entity. So to put that into practice here today, a provider typically who uses, I won't name a name, like a pop health vendor, for example. Sure, they can see what the next best action could be, maybe. They might be able to see what their gaps in care. I need to fill this [ AWV ], I need to get this mammogram done for December. Ultimately, when it comes time to submit the prior auth to the payer, United or Aetna, whoever it is, I've got to all tab, log into the United website, log into the Aetna website, copy and paste my member information in, copy and paste my diagnosis in, press submit, then wait 3 days or however long for the prior auth to come back. Then tell my patient, hey, now you can go call such and such doctor, you can go schedule something with the radiology to get your mammogram done, then they've got to wait another 4 weeks to get an appointment and then so on and so forth. And because we have all that under one roof, because we're the provider and the payer, we can really make all that seamless. So you saw yesterday for example, the doctor can just click a button, an agent will actually fill in the prior auth details based on the information that's integrated in the EHR, submitted to us. Our back-end payer agents are auto approving 70% of prior auth. 70% of the time that is coming back to the doctor's office within a few seconds saying, yes, you're approved. Now you can go see the specialists. And then you can put another button you saw where they can make calls to all the in-network prioritized specialists of a certain type, figure out which of them has the next available appointment or the following available appointment, present those to the patient, they can choose one, book it and then they can just kind of do their visit right away. Or you see, I think we tested the transition of care accended that I was discharged from a hospital and could follow up with me. We have over 500 patients far more than that discharge from the hospital on a daily basis across our 1.55 million members. Now we can reach out to call every single one of them, make sure that they're not at risk for readmission, make sure that they've gotten the medicine, the meds that they've been prescribed, make sure that they have a follow-up appointment with their PCP. And all of that filters into the back-end care management platform, which you saw where if it doesn't happen, then it's being escalated to a human where someone is calling or even we have a care home team that can go and visit them in person and make sure that they're okay. So those are some of the examples where we're obviously inflecting G&A downward, 70 basis point improvement year-over-year in Q1, but also over time, that's going to inflect medical cost downward, hopefully, too.

Yes. Future is exciting. I think we're out of time, but thank you.

Thanks so much. Appreciate it.