RadNet, Inc. (RDNT) Earnings Call Transcript & Summary

Thanks for joining us for our next presentation and with that as always, I very much appreciate Mark taking the time to join us here in Boca as always. I think most of you know Mark. Mark Stolper, Chief Finance Officer of RadNet.

Mark and I were just talking. Real quickly, to level set, maybe just a summary review of the kind of results and initiatives and so forth and I think one of the things for me, given my incompetence with technology when you talk about AI and your opportunity in AI, I would love if you could just embed in that conversation, the whole AI, your rollout, what opportunity you see and how it becomes a bigger part of your platform longer term?

Sure. So just to give a background. Our core business is performing nicely. We've had a good year and the year seems to be ending with a nice level of strength. Our same-center performance on a volumes basis in the first quarter was up over 7%; over -- the second quarter, over 6%; third quarter, 4.2%. So we're growing nicely on an organic basis. We are -- our centers are very busy. We -- because of capacity issues and constraints, we launched almost 2 years ago, a big move or push towards de novo centers. So today we have -- as I speak now, we have about 13 de novo centers in various stages of construction and development, that should come online throughout 2024 and into 2025, which should provide us a significant growth in those years and beyond. We continue to drive the initiative towards hospital joint ventures. We announced a pretty substantial expansion of an existing joint venture with the Cedars-Sinai medical system in Los Angeles. We had 2 previous joint ventures with them, one of which existed in the San Fernando Valley of Los Angeles. The other, a small joint venture with Cedars in Santa Monica in the west side of Los Angeles. We established a third joint venture with Cedars in the third quarter in the East Los Angeles area and then substantially increased the size of the Santa Monica joint venture, where Cedars contributed its 5 outpatient imaging centers into that joint venture, RadNet contributed our 2 large flagship centers in the Beverly Hills area and that now puts RadNet and Cedars squarely as partners in the vast majority of the L.A. area. So -- or L.A. County. So JVs now represent about 36% of all of our imaging centers. 130 of our 366 centers are now held within partnerships with some of the large health systems. And that seems to be growing in demand as the health systems are looking for strategies to recapture revenue that they're losing to lower-cost ambulatory outpatient businesses, they're looking for a strategy for long-term growth and sustainability, and we offer that opportunity for them. So the core business is growing nicely. We've got a pipeline of acquisitions, small tuck-in acquisitions in the markets in which we currently operate, which we can do very efficiently and integrate effectively and easily. And then as you asked at the last part of your question, our AI vision, which is now its own operating segment as of January 1 of this year, is growing nicely. We grew in the third quarter, 220% over the prior year's third quarter. Sequentially, we grew over 20% from the second quarter of this year. The big contributor of that growth is the EBCD program that we launched in November of last year, which stands for Enhanced Breast Cancer Detection service, where we are offering to women when they come in for their annual screening mammography exam, the opportunity to have artificial intelligence do the first interpretation of their exam. And in a number of medical journeys, who have -- medical journals, excuse me, who have published studies on our AI. It has been shown to detect and diagnose cancer up to 2 years sooner than the human eye can. And so we're now having about a 35% adoption rate on the East Coast where we are now fully rolled out with this mammography product. And we started about 3 weeks ago, the rollout on the West Coast in Southern California and by the end of the first quarter of next year, our EBCD mammography AI program will be fully rolled out in our entire network. And we believe that exiting next year given the trajectory of growth and our projections that we could reach break-even point in our AI segment, which will be pretty spectacular given the nascent nature of this technology and the fact that we're not certain there are any other people, at least in radiology, who've monetized artificial intelligence in any meaningful way. So I'll pause there and give you another opportunity to ask questions.

At a risk of asking you a new question but 35% adoption rate, meaning what exactly does that metric refer to?

Sure. So today, given the run rate of RadNet, we perform about 2 million mammography exams on an annual basis. About 80% of that or around 1.6 million of those mammography exams are screening mammography exams, which is where our AI is applicable. So -- today, our business is about 55% weighted on the East Coast and the other 45% on the West Coast. We have it fully rolled out in all of our mammography locations on the East Coast. And we're seeing about 35% of the women who come in for their annual screening mammography exams electing to be in the program. And what that means is they are paying $40 out of pocket.

[indiscernible] of pocket.

Yes. For us to use AI in the interpretation of their exam. And what we're seeing is tremendously positive data coming out of this program. First, we've detected over 500 cancers that would have otherwise gone undetected at least at this stage in the diagnostic workup. And second, we're seeing an over 20% decrease in the callback rates of these women. So when women come in for their screening mammography, often the radiologist will identify something that looks suspicious in the exam, but they're not certain that it's cancer. So we'll call back the patient for further diagnostic workup, whether it's a -- what we call a diagnostic mammography exam or an ultrasound, a breast MRI or even a biopsy. And many of those callbacks are false positives. They're just doctors trying to be conservative with respect to their diagnoses. And what we're seeing with this AI is that the doctor is more certain and more definitive in his initial read or her initial read utilizing the AI as a diagnostic tool.

I mean, ultimately, you're saying that it's -- current rollout, 55% East Coast, so to speak.

Correct. All of our East Coast operations, which represent 55% of the overall...

And the [indiscernible] the West Coast will be...

Yes. So we started the roll out on the West Coast about 3 weeks ago in Southern California. We expect to complete Southern California towards the end of this year. And then we will be rolling out Northern California and our Arizona marketplace and expect to have it in all of our West Coast operations by the end of the first quarter of next year.

Okay. You touched on -- I know Cedars, the JV opportunity. What -- I guess what is -- we're hearing that from multiple providers that -- [indiscernible] talking about it [indiscernible] and so on and so forth. What -- asking a similar question. Has something changed in the environment where the JV model to the core acute care setting, has that become more attractive in some ways for you? Like you're -- 4, 5 years ago, we weren't having this conversation.

Sure. And I think it's not a function of us, meaning RadNet, trying to drive the business this way or even the hospitals. It's just really a function of what's happening in the dynamic landscape of health care. And this is not unique to radiology in that more and more outpatient or more and more services that can be performed on an outpatient basis in lower-cost sites of care are moving in that direction. The payors over the last 5 years have gotten more and more aggressive in attacking cost, meaning they're trying to move more and more patient volumes into lower-cost sites of care. And you're seeing it in radiology, of course, you're seeing it in the growth of outpatient surgery centers, you're seeing it in the growth of urgent care centers taking emergency room visits away, home health, and I can go on and on. And so the health systems are recognizing that they're on the losing side of this trend, and they are looking for ways to recapture the revenue that they're losing to the ambulatory outpatient providers, and we offer them a construct where they can participate in the growth of the outpatient business and recapture some of those revenue. And it's good for them, one, in terms of recapturing the revenue. It also gives them a way to participate in community health, where they can expand their operations into catchment zones that then can feed their acute care services, which is really where they make their money. And for us, it's been a wonderful structure because these health systems have relationships with community-based physicians. In many cases, over time, they have actually acquired community-based physicians, both on the primary care side and the specialists' side, and they use their influence to try to drive those referrals into now our jointly-owned facilities. So they've been instrumental in increasing or providing incremental patient volume to our centers. The second thing that they do for RadNet is that to the extent that we need it, they provide additional leverage with the commercial insurance companies to make sure that we receive fair and equitable outpatient pricing for the long term.

Okay. To -- I guess another part of your strategy, you're talking about the de novos, the rollout of de novos. Again, it's something that we hadn't heard a handful of years ago. As -- and you've talked about somewhat of a backlog in the market, if you will. Can you kind of pair those -- kind of that process of those thoughts together? And what you're seeing deploying dollars into de novos right now? What is that really a reflection of? Is it the backlog? Is it just there hasn't been investment in recent years? Is it coming out of COVID, you're seeing more activity? Is it the shift to inpatient, outpatient? I mean what is it that's kind of driving that strategy today?

Sure. I think it's a little of all of those things. If you remember when we first met maybe 20 years ago, there was an overcapacity in outpatient diagnostic imaging where you didn't have to be a real efficient operator given the pricing that existed 20 years ago to be able to make a business out of outpatient diagnostic imaging. And over the last couple of decades, essentially, the industry has grown into that overcapacity to a point where today operators are busy. And when we look at our centers across virtually all of our markets, we have backlogs, we have capacity constraints. And typically in the past, we've liked to acquire facilities who weren't at capacity and consolidate our centers and create that additional capacity, which we could then fill with increasing demand over time. What's happened now is that there -- depending upon the market you're talking about, there's been fewer suitable tuck-in acquisitions of companies that weren't busy or didn't have that excess capacity and so we need that. We have backlogs in many of our markets. It sounds like a good problem to have. And in some ways, it's the best problem to have, on the -- however, it's also a problem because to the extent that someone -- a patient needs to get in for a serious illness or serious injury for an advanced imaging modality and they're told that we can't get you in for a week, that business is going to go elsewhere. So you have to be able to work through the capacity issues. And so we have -- about 1.5 years ago, we launched more of a directive in creating that capacity through de novo facilities. So as we stand today, we have 13 de novo facilities that are in various stages of construction and development that will come online throughout 2024 and into 2025. We spent very liberally on CapEx in both -- in 2020 to -- and here in 2023. Some of that will bleed into 2024. But we're anticipating a significant acceleration of growth when these come -- when these centers come online. The average center should do in the $3 million to $5 million of revenue and bring down about 20% EBITDA margins pre corporate, about 15% after corporate. So we're looking for significant growth from this initiative.

Okay. Mark, does -- could you walk us through -- on the JV side, not to talk to a specific transaction, but how the economics would look on a typical JV transaction? Like are we -- are you taking a 51% stake typically for now? Or is it kind of -- who funds it? And if you could just kind of walk us through kind of a...

Sure. There really is no typical JV strategy for us. If you've seen one of our JVs -- you've seen one of our JVs.

Yes.

It really depends upon the needs and requirements of the hospital partner, and each looks a little different. So let me give you some examples of what they may look like. Sometimes the hospital owns facilities that they contribute to a joint venture, and so we'll contribute facilities, they'll contribute assets. We do relative valuation, a fair market value of the assets that each has contributed based upon the individual assets they're contributing, based upon the performance of those particular centers. And that creates an equity split. And then if one partner wants to buy up or sell down some cash can change hands. There might be a requirement from the hospital or from our standpoint about who wants to consolidate this from an accounting perspective. Today, the accounting rules, it used to be if you owned over 50%, you'd consolidate. If you owned under 50%, you don't consolidate. Today, the accounting rules are more complicated, where it's really about how much operating control you have and not about economic control. So there are certain situations where we don't want to lose the revenue by contributing our centers or the hospital needs to consolidate for whatever its reasons are. So we're generally more -- we're generally flexible about the structure. In a typical situation, RadNet will own at the low end, 40% of the equity. At the high end, I think we have a couple of joint ventures where we own 75% of the equity. But I would say most of them are generally between 40% and 60% ownership. In 100% of the instances of these joint ventures, RadNet is the operating partner. So it's RadNet's burden and responsibility to provide the day-to-day operations, all of the support functions, such as accounting, IT, human resources, you name it. And for that, we -- RadNet receives a management fee off the top as a percentage of the revenue. And then when there's distributable cash, we make distributions that are proportional to our equity. So it really goes all over the map.

[indiscernible] consolidated?

In -- we have about 25 joint ventures. About half of them are consolidated from an accounting perspective, about half of them are not. What I will say, and I think this is important to note is that -- in all cases, we charge outpatient rates, meaning we use the provider numbers of RadNet's physician groups, our physician groups or our affiliated physician groups, and we bill under their provider numbers, and we bill under outpatient rates. We are not using hospital inpatient rates. We're not trying to arbitrage that. Although maybe we could make a lot more money in the short run by doing that, we think that's a long-term losing proposition because ultimately, the payors, the commercial payers who are doing everything they can to move the business out of the hospitals would view a relationship where we're charging hospital pricing as no different than inpatient radiology.

Acquisitions, you've talked about some -- I mean kind of highlighted the stressed assets potentially in the market there and other x opportunities. What are your thoughts on it? Smaller scale or larger scale, flexibility, you have liquidity, obviously. What are your thoughts on that...

Sure. With respect to acquisitions, I mean, we continually have a pipeline of opportunities that we work through. The vast majority of them are small tuck-in transactions that are in our markets. And this is something that we've been doing for a couple of decades. I think the last real scale acquisition we did was in November of 2006 so 17 years ago, where we acquired Radiologix, which became ultimately the platform that RadNet is today. And since then, most of the acquisitions we've done have been small one-offs or small groups that are in market from which we can grow. And we've been able to buy some of these small tuck-in transactions, let's call it, 4 to 6x EBITDA. Really, we're in a different financial position today than we have been in much of our history in terms of low leverage and high liquidity. We've got -- we ended last quarter with $338 million of cash on our balance sheet. Our net debt-to-EBITDA ratio today is 2.2x. So we certainly have a lot of liquidity and low leverage that gives us availability to look at more substantial transactions that can be more transformational to the business. We're open to evaluating those types of transactions. We'd like to do something on a grander scale. We think that, that could represent great opportunities for all our stakeholders, both our equity and our debt. But if we were to do something on a grander scale, we'd have to be -- have tremendous level of confidence, not only in terms of the -- the quality of the assets that we would purchase but the quality of the EBITDA, the earnings that we had purchased. If there are synergies, which would be I think instrumental to any large-scale acquisition for us, we'd have to be very confident of those synergies. We'd have to pre identify them before announcing a transaction, and we'd have to be confident that we can achieve those synergies within a 12- to 24-month period. So I think we would potentially lever the balance sheet further than where it is today at 2.2x. But given the cost of capital, we want to be judicious.

And there are larger-scale acquisitions out there, so to speak?

Yes. There -- I put them into two camps. They're some private equity backed come -- consolidators in our industry, they don't overlap with us in our existing markets, but they've had reasonable success in consolidating other markets that would have -- that would be attractive to us. And then there are potentially some assets out there that are more stressed with respect to their capital structures and might need to sell assets to raise capital or to help with their financial position going forward, and we would look at bits and pieces of some of those other companies.

Right. Okay. Great. Questions in the audience?

Yes. On the AI -- you [indiscernible] about the AI, it's obviously super interesting. You mentioned there's like 500 breast cancer cases detected that would have been missed. So just -- I guess, just the human doctor read, you said they -- the women sign up for the AI to read the first read, I believe you said -- are you saying then the human doctor is like the second read and then you compare the results and there's like 500 false negatives there, like with the human doctor? Is that how it works?

Sure -- yes, similar to that. As the FDA has gone and approved or cleared some of these algorithms for clinical use, they have, to this point, taken the position that these are tools to be used by a radiologist, they're diagnostic tools as opposed to approving them for autonomous reading. So they want a licensed physician, in this case, a radiologist, to sign off on every exam, whether he or she has used a diagnostic tool or not. So the way our EBCD program works is that the algorithm reads the exam before the radiologist even opens up the scan. So when the radiologist walks in, in the morning and has a work list of mammography exams to work through, and most of our mammographers are specialists, meaning that's all they do. They -- the exams are triaged into 3 groups: a green, yellow and red group. Green meaning the algorithm saw -- was pretty certain that there was no pathology that had to -- that would concern the radiologist. A yellow group where it saw something that the radiologist needs to pay attention to. And then the red group where the algorithm is pretty certain that there's some serious pathology there. And so the radiologist can use this as a triage to make him or her more productive by managing his or her day along the lines of these 3 camps. When the radiologist opens the exam, whether regardless of which group it's been characterized in, if there's any pathology, the algorithm has already circled the areas of concern. So the radiologists can hone in on that area and determine whether he or she agrees or disagrees with the algorithm around the pathology. In our model with the EBCD, if the algorithm and the initial radiologist disagree on the diagnosis, it gets sent to a super radiologist, super specialist who's essentially the arbitrator of that difference. And in our model, we've had over 500 situations where the initial radiologists did not think that there was cancer, the algorithm did, the second radiologist thought there was cancer, the patient was called back for further diagnostic work up, whether it's a biopsy, breast MRI or another type of diagnostic exam, and it was proven that there was cancer. So we're finding cancers that otherwise would have gone undiagnosed at this stage, and it's extremely powerful.

That's super helpful. And so just to follow up on that, like I guess, the -- when we put it into like false positive and false negative percentages, like sensitivity, specificity, those type of language. Is that -- the 500 like [indiscernible] so far, what is the -- what are the stats like for the humans -- human docs?

Yes. So...

And also related to that, what is the AI false positive? What's the AI ability in terms of its false positive rate? In terms of like where the doctor is right so there was no cancer and the AI was wrong, saying it was cancer. Do you have a data like that?

Yes. I don't have the stats off the top of my head. As -- because I don't know about the whole subset of all the exams that were read. But if it's your wife or your daughter or your mother who was one of those 500 people, and now we've caught cancer at stage 0 or stage 1, and the treatment and the long-term prognosis is altered because of that, it's incredibly powerful. What I can tell you, and I do know the stats is that the callback rate has been reduced by over 20%. So when a woman is called back because there's something suspicious and ultimately, it turns out that she's fine, the woman goes through the inconvenience of having to go back to the imaging center, might have to go through a biopsy or an invasive procedure to determine whether or not she has had cancer. And then the cost of -- to the health care delivery system for doing essentially unnecessary exams is tremendous. So this is the type of data that we're accumulating for the payors to argue for reimbursement, which is where this is headed. Ultimately, we think that these AI tools are going to be reimbursed by Medicare and by the commercial payers because they are effective.

[indiscernible].

Sure. So we have an algorithm that's already FDA approved in prostate MRI imaging. So we are looking to do a similar type of self-pay test that will be focused on prostate cancer. And then we have an algorithm that we're already monetizing about -- on lung cancer which we're monetizing, not in the United States because it's yet to be FDA cleared here, but in Europe, especially in the U.K., where we are -- where the NHS is rolling out a 4-country -- a program called the targeted Lung Health Check program where they are scanning all of the current and past smokers in the NHS system for lung cancer, and they mandated the use of AI, along with the radiologist read. And our Aidence lung cancer software has over an 80% market share in that rollout. So this is the future of our industry, and we think it's going to be transformational not only for our business but for the industry.

We should probably wrap it up [indiscernible]. Thank you, Mark. Thank you for the time as always.

My pleasure. Thanks for having us. Thank you.