Astrana Health, Inc. ($ASTH)

Great. Welcome back to the Barclays Global Healthcare Conference. My name is Andrew Mok. I'm the facilities and managed care analyst here at Barclays, and I'm pleased to welcome Astrana Health to the stage. And joining me here is Brandon Sim, CEO. Brandon, welcome.

Thank you so much, Andrew. Thanks to Barclays for having us here.

Absolutely. Brandon, you just finished up Q4 results guided for 2026. Why don't you give us a current state of affairs of the business and run through kind of like the highlights of what you see for the year ahead?

Yes, absolutely. So for those of you who are not familiar, Astrana Health is a risk-bearing provider group. We operate a unique single payer model in which we take reimbursement percentage of premium from our payer partners, our large nationals, as well as regional payer partners, and we operate a unified multiline of business and multi-payer downstream provider network that's well coordinated, delivers better care at a lower cost. I know the industry has obviously been very volatile of late, and we're very proud to continue to deliver on the consistency and stability of our single payer diversified model. Over the last 30 years, we've been profitable. We've continued to grow the business. And since I joined the company around 7 years ago, we've sixfolded revenue, close to eight-folded free cash flow and continue to grow at a very, very rapid clip in the 20% to 30% CAGR over a very period of time. We just reported results, as Andrew mentioned last week or a couple of weeks ago, where we delivered record revenue, free cash flow and adjusted EBITDA, continuing to grow at a -- I think we grew 40%, 50% year-over-year and have very strong guidance planned for this year as well with another 20% to 30% growth on both the top and the bottom lines. here are obviously a lot of puts and takes to that given the regulatory volatility in the environment and the macro environment that we're in. We believe that our unique care model, the unique technology platform and the diversity of the multi-payer multiline of business model has allowed us to weather some of this volatility better than the market has. Going forward into this year, we expect tailwinds from strong Medicare Advantage rate updates resulting in -- resulting from the 2026 Medicare Advantage rate notice from last year. We expect some incremental headwinds due to Medicaid and some incremental headwinds from the exchange populations, but well kind of made up for by the increases in the Medicare Advantage rate. We also expect continuing operating leverage from our G&A platform and our technology platform, which we built completely internally in-house proprietary and from '24 to '25, we improved G&A as a percentage of revenue by 75 basis points, 110 on an adjusted basis, and we continue to look for opportunities to scale our platform efficiently. As a reference point, we ran G&A at 6.8% of revenue last year, the full year of 2025, despite also taking on a lot of payer-related responsibilities such as claims processing, prior ops, network building, credentialing, contracting and even the customer service lines, even patient service lines, we are responsible for. And we're running that at kind of mid- to slightly high single-digit percentages of revenue versus payers who badly run at 10-plus percent of G&A as a percentage of revenue. So we really focus on efficiency and investing and taking those dollars and investing in our patients to prevent long-term, more acute issues, which are a greater burden to the health care ecosystem and the health care dollar. Overall, we can continue to believe that although there are headwinds that we are relatively more insulated from those headwinds than our peers, we continue to deliver extremely strong growth, both on the top and the bottom lines, and we're well positioned to expand market share across cycles.

Great. Brandon, over the last few years, Astrana has consistently held medical cost trends in that mid-single-digit range even as many value-based care peers have struggled with the volatility. And 2025, trends finished slightly better than your original 4.5% expectation. So to start, can you walk us through what you think has enabled that level of consistency in recent years?

Yes, absolutely. I mean, the biggest part of it is really the business model, which is our ability to engage patients longitudinally over time, the higher LTV of the patient in our system versus in any given insurance company system or any of our peer systems and the ability for that to translate into better outcomes both for the patient and from a financial perspective for us and for the payer partners that we serve. What I mean by that is that we actually engage the patient across all lines of business regardless of which payer they're with. So obviously, Medicare Advantage patients select a different plan every year and seniors have become quite efficient shoppers. -- of late in terms of understanding which benefit designs are better for their particular situation or shopping around for the best supplemental rebate, whatever the case might be. Regardless of the plan that a senior chooses in any given year, as long as they remain with their PCP or with the PCP that's in our network, they are still Astrana kind of risk-bearing members. And so we're able to not only have a more complete data map of their history because we are seeing their claims in their chart history and their risk factors throughout time across different plans, but we're also able to manage and build a relationship with the patient, engage them more effectively, have them be more involved in their own health and more effectively invest in them proactively, knowing that the LTV patient and our system is going to be longer and we'll have a longer period of time to so to speak, reap the award of the proactive preventive care. And so from an economic standpoint, we have more -- we have a greater return on investment in investing proactively in the patient's health than structurally any insurance might have -- company might have because they have no certainty that is that the patient will remain their risk to bear at all in the near future. Whereas for us, given our multi-payer, multiline of business approach, we have a much longer LTV of the patient in our system and therefore, better ROI if we would invest proactively in their health. So there are a lot of reasons why, from a stability standpoint, sure, there have been headwinds, and we have been hit by some of those headwinds, but the volatility has been smoothed out because of that longer-term relationship and the fundamental business model that we operate.

Great. Moving on to membership. In your 2026 guidance, you're assuming about mid-single-digit growth in Medicare. Roughly a 10% decline in Medicaid and a 20% to 30% decline in ACA exchange lives? So first, how much visibility do you have into those membership forecast today? And how are trends tracking relative to those assumptions so far this year?

Right. So it really varies by line of business, Medicare Advantage, we are -- we already see some of the new enrollment from annual loan from the AEP period. So we are pretty certain good visibility into that M&A growth. Reasonable growth, not extraordinary, but obviously, the industry has had a pullback in terms of benefits this year, and we are downstream of some of those benefit design choices. On Medicaid, we've -- we have some unique dynamics facing our Medicaid book because a lot of our Medicaid members are in California -- sorry, in the Medicare program. And so in Medicaid, especially in California, we've seen declines in Medicaid enrollment even through last year, I think 7% or 8% last year. We're expecting 10% to 13% this year because of some of the undocumented immigrant status members that were enrolled into Medicaid over the past years. Our payer partners are actually seeing mid-teen impact from the UIS status members. We tend to have lower indexed UIS membership because those members typically do not have a long-term PCP relationship. So they are less likely to be attributed to our kind of provider groups. That being said, we still expect conservatively 10% to 13% decline in Medicaid, as well as actually built into our guidance are something that there will be a slight mix shift as well because of the acuity of the departing populations. And then finally, in exchange, while we are actually seeing flat to slightly up, exchange membership early in the year, we do anticipate, as is consistent with the industry commentary, I think that there will be 20% to 30% declines once some of the auto reenrollments or auto reenrolled members drop off in April.

Right. And what gives you the confidence to maintain that mid-single-digit sort of trend forecast throughout all this disruption?

Yes. I mean think the first part of it is really our actuarial team has been very strong in predicting where trend is going to be. I mean last year, even in a time of high single-digit trend for the industry, we predict 4.5%, and we came in under 4.5% on trend. We actually predict -- we actually built into our models a higher trend this year than last year despite trends typically decelerating in our populations for two reasons. One, because of the adverse selection affected exchange and Medicaid, they wanted to be thought about what that looks like. And then two, because of the recently acquired population in terms of our prospect Health deal that came in operating at a slightly higher trend than our legacy business. So those 2 factors combined led us to be a little more conservative in terms of trend taking kind of a mid like 5-ish percent trend versus the 4.5% and the low 4s that we actually saw last year, again, out of conservatism to ensure that we have room to meet our expectations for the year.

Great. Move on to the Medicare Advantage rate notice. There's been a lot of discussion on the rebasing of the risk model, which places greater coefficient weight on the skin substitutes at the expense of more chronic conditions that seems like a really obvious flow on methodology. Is that argument resonating with the administration? And how likely do you think it is that CMS will revisit that calibration?

Yes. So I think you're referring to the 3.4-ish percent of the risk model renormalization, the new regressed weights on the post-COVID claims. I think administration has a lot of really smart people in it now. They really recruited a lot of very talented folks from the private sector, from entrepreneurs from the private sector to help bolster their ranks. So actually a lot of respect for the administration and kind of the talent level they brought in. It is very different from well, the government just has a reputation for not having that level of talent perhaps, but that's very different in today's CMS. So really crops to them there. Now I'm not just saying that I truly mean it. And the reason I say that is because they really have, I think, a pride, which they should have in kind of following the appropriate math mathematics essentially, the right math and the right actuarial analysis to come up with the effective growth rate and the risk model normalization that they end up coming up with. And I think the math is correct. In fact, if you think about what happened in fee-for-service Medicare post-COVID, if there is a lot -- there are a lot of -- if there is skin substitute or skin graft fraud, then the way that, that would manifest in the data is that there will be a very high claims cost for patients that have diagnosis of skin conditions because of the fraud. It was very high because of the fraud. And so the question really is, if you run a regression on a data set in which you have very high claims cost, correlated to people with skin conditions, what coefficient comes out of that model for someone with the skin condition. Well, obviously, having a skin condition is explanatory for a lot of claims costs. And so therefore, the coefficient is very high. And the HCC or the chronic condition risk factor that when you add them all up, you get to the risk adjustment factor ultimately, that is essentially the coefficient for the skin condition. So it almost is very natural coming out of the math, so to speak, that the coefficient for skin conditions would be high simply because of the fraud, almost by definition. Now I think the fix to that, there have been a couple of fixes that industry has talked about. A lot of folks have lobbied for maybe a phase-in of that over a 3-year period, kind of like B28 and this kind of pseudo B29 renormalization. I would argue that, that fundamentally is just the wrong thing to incentivize altogether, as you said in your question. And so we've been in front of DC or NDC in front of CMS helping them see that you can still run the math, but if you put garbage in, you get garbage out. If you put the right things in, you get the right results out. So if you put in the right types of claims in, we strip out the kind of extenuating factors, things that relate to fraud, then you might get more reasonable set of patients that truly incentivize a lot of the -- well, frankly, a lot of the things that the administration itself espouses, things like removing fraud waste and beast, things like investing in patients' health proactively and incredibly to preserve long-term health and lower the cost of health care in the United States. So I think there is receptivity. We've been in front of them. It seems like something less -- it seems more likely anyway than the removal of the disallowed diagnosis, which to me seems more philosophical in the sense that they just fundamentally don't believe that, that should be allowed. And I think there's some room, hopefully, in the effective growth rate as well. I think there are some assumptions in the growth rate that don't seem congress with the data that we're seeing. So again, I think they're really smart folks, and I think they'll figure it out and hopefully, that manifests in the final rate.

I think the industry is aligned on some of those comments as well. despite the roughly flat industry-wide rate update, you've indicated that as Astrana's MA rates are likely to be approximately 2.5% to 3%, so much better than the industry. Can you help us unpack why your experience diverges from the broader industry?

Yes. I mean, historically, we've been very conservative on risk adjustment, as you know. Our RAS score today is 1.02, just above the industry average despite having a very high percentage of duals relative to industry average eligible that is. So we've typically been very conservative. We -- our RAS score has actually gone up 28 because of that conservatism. We haven't been impacted by a lot of the headwinds that our peers have faced because of aggressive risk coding practices. Also because of that, we're unaffected or less affected by some of the changes in the advanced notice. So the 1.53% of the disallowed diagnosis, for example, we don't do audio-only calls for risk adjustment. We don't do only charge cases for risk adjustment. Our model is an encounter-based model that is supported by the single payer model that we have where we're paying our claims and we're actually reporting our quality and reporting our risk adjustment ourselves. And so because of that, the 9 basis point increase, net increase that CMS reported, the 153 to 1.53%, that gets us back up to around 1.6%. You add back kind of even the current coefficients that we -- that CMS has published, even if there's no change with the skin substitutes. If you take those coefficients and run them on our HCC profile in our population, you end up with a still a headwind because of the pushdown of the chronic condition coefficients, but less of a headwind than the industry is facing at 3.4%, closer to around 2.5% to 3%. And so if you add all this together and add even a bit of the risk adjustment, 2.5% that CMS kind of expects for the industry, you get to that 2.5% to 3% effective rate increase for Astrana. And I'll note that, that 2.5% to 3% is only slightly margin dilutive for us given that our MA book at low 4s% last year. So even a small increase in the advanced final would make MA actually margin accretive for us in '27. So we continue to believe that our ability to control trend allows us to differentiate ourselves even at a time of regulatory volatility.

Right. And there's still room for payers to change benefit or network strategy to protect those margins. Where does Astrana fit into those conversations?

Yes. That varies from plan to plan. Typically, we're not as involved in the benefit design process. That's more of an actuarial thing that's upstream of us. There are certain plans where we've designed plans together products together with the plan and what's called a provider-specific plan. And so we have that with a few different large plans now where we've designed the benefits, we've designed what the bid will look like, and we work together with the plan to figure out actuarially how we're going to help them grow and design differentiated product for patients while managing cost effectively. So it really depends from plan to plan. But typically, in an environment where plans are cutting back or looking to preserve margin, that would actually pass through to our percentage of premium contracts as well.

Right. Understood. Let's move on to the prospects deal. Shortly after you announced that deal, you laid out a $350 million EBITDA target in 2027 that you framed as conservative at the time. Even though the prospect integration appears to be tracking in line, if not ahead, the regulatory and operating environment has changed such that you're deemphasizing, but not necessarily ruling out that target. So what needs to go right on cost, rates, membership or execution for that target to come back into view?

Right, right. So prospects, I remind folks, we announced back in 2024, November of 2024. So that was prior to actually unfortunate timing, but prior to the election date prior to 03b -- or 0B3 prior to the '27 notice and prior to a lot of kind of seismic shifts in the regulatory environment and even macro shifts in terms of oil prices and so on and so forth. So it's a very different environment before $1 trillion were being planned to be taken out of the health care ecosystem. That being said, the fact that $350 is still within view does reflect, in my view, the conservatism of the $350 million guide to start with. Now that wouldn't be my modal or median outcome necessarily today, if I had to guess. The factors that would swing it kind of more to where $350 million would be the midpoint would really be around rate and acuity mismatch. So this year, I think in our bridge from '25 to '26, which I talked about on our earnings call, we are baking in a $25 million to $30 million headwind related to Medicaid related disenrollments related to adverse selection from disenrolled members. Even the reversal, for example, of that $30 million headwind would already take us, I think, from Street consensus of $315 million or whatever it is. I don't know the exact number, but I don't know where you have us, but low 300s, let's say, closer to $34 close to 50. Not to mention any kind of incremental headwind or tailwind rather that we get from Medicare Advantage rate in QD kind of mismatch the positive way, which we saw this year as a $30 million tailwind. Next year, for example, we talked about some of the math getting us to almost a margin-neutral environment and even 150, 200 basis points of improvement could get us to a margin accretive territory in which that would add another $15 million, $20 million. Just to run through a quick math. I mean, we're at $4 billion, $3.95 billion at the midpoint for '26. We've grown historically at a 30% CAGR. But even if you assume 10% into next year, you're talking about well above $4 billion of revenue in '27. And 60% of the revenue is Medicare. And so if you -- and out of that 60%, 2/3 of Medicare Advantage. So at least for this year's numbers, if you take the $4 billion, 60%, you're at $2.4 billion, take 2/3 of that, you're at $1.8 billion, something like that, $1.6 billion. So we take that number and you Think about how much each percentage point of rate acuity is worth, right? 1 percentage point of $1.6 billion is worth $16 million. And so each kind of 100 basis points of improvement, we're going to get $16 million headwind -- or tailwind, sorry, year-over-year. So if you kind of even just roll back the $30 million and even say there's a 1 percentage point tailwind on MA, you're already at $46 million, and you can easily bridge from kind of where Street has us to well above the $350 million number. And so those are some of the ways that we think about kind of how the shifting landscape has affected what that number is. Regardless of what that number ends up being, depending on where regulatory shakes out, we still continue to believe that our industry-leading ability to maintain trend in the mid-single digits versus high single digits, our resistance and resilience in terms of our conservative risk adjustment posture and our ability to operate in an industry-leading kind of G&A load will continue to allow us to grow market share even at the kind of 6% EBITDA margins that we're running at today.

Great. On the 4Q call, you spent some time discussing the AI tools that are available. Can you help us distinguish which capabilities are furthest along and already deeply embedded in day-to-day operations versus those that are still in the earlier or pilot stages?

Yes, absolutely. And we're very excited about our AI tooling and our overall just the entire technology platform as a whole. When I started the company 7 years ago, I was the first software engineer at the company. Some people here may know that. And my background is actually in computer science and AI. I published papers in AI and machine learning. I did graduate studies in that field. And so part of my enthusiasm for Astrana in the first place and why I wanted to say was because of the massive opportunity we saw to make the health care system more efficient. Some of you may know that I used to be a quantum trader. I used to spend all my time submitting jobs, Google Cloud, building large models, doing feature engineering and trying to predict and get like 2 basis points of alpha on E-minis. And it was a very fun and competitive experience, but also ultimately not a rewarding one for me personally because 2 basis points on E-minis is just not that exciting even if you put $1 billion through it. What we found in health care was you could -- I could build something kind of over a week and not be measuring percentage point increases in basis points, but in like tens of percent, like I could do something and we'd be improving the efficiency of the process by like 25%, not 2 basis points. And so that level of progress, the leaps and bounds that we're able to make in health care was very exciting to me. And so where we're furthest along, we've been developing this for 7 years, even before ChatGPT or LLMs or even a thing are primarily in our payer tools business. So 90% of our claims are automatically adjudicated without human intervention. Over 70% of our prior auths are automatically approved. Nothing denied, of course, via AI. And even for those who require human intervention, we're giving tools to our nurses and our medical directors to allow them to make decisions much more quickly. For example, we now have tools that kind of aggregate all the data, both structured and unstructured data about a patient and our teams can actually ask questions to that patient record. So they can type in, hey, does this patient have CKD? And the AI will scan through OCR, the documents necessary, scan through all the documents, figure out which things are supportive of that diagnosis or even find the page where the 1,000-page PDF that exists and then give all the citations and the nurse or the medical director and click in, automatically scan to the right page in the PDF and apply their own human judgment as to whether the AI was correct or not. Again, nothing is automatically denied using AI. So on the per tool side, a lot of really great improvements. And increasingly, on the care management side, we built our own care management platform, which risk stratifies our patient populations and also handles a lot of the low-value tasks that allow patients -- or that allow our teams to operate at the top of their license. So for example, making calls to patients automatically in a post discharge or reaching out to patients for their annual wellness visits, things of that nature are all now conducted by AI versus having a nurse have to make those calls despite them being able to do much more with their knowledge.

Right. Well, with that, we're out of time. So Brandon, thank you so much for joining us here today. Please enjoy the rest of the conference.

Thank you, everyone. Thanks, Andrew.