CeriBell, Inc. (CBLL) Earnings Call Transcript & Summary

Good morning, everyone. We'll get started here. Welcome. I'm Robbie Marcus, Med Tech Analyst at JPMorgan. Really happy to kick off the 2025 JPMorgan Healthcare Conference with the first-time presenter, recent IPO, CeriBell. Happy to introduce CEO, Jane Chao. We'll do a presentation followed by some Q&A. Jane?

Good morning, everyone. It's very exciting to give our first JPMorgan presentation as a public company, and thank you all for getting up early to join us. So today, I will still walk through the overall CeriBell story. I will give you the overview of the company and then talk about the unmet need as well as the solution, well focused then switch to our business model and our future pipeline, will wrap up the presentation with our 2025 areas of focus. So with that, our disclaimer. At the highest level, CeriBell invented a novel brain monitoring platform. It has the hardware, which you see here, I dive deeper later, that make EEG signal acquisition, very easy. It also have a software and an algorithm empowered by AI and machine learning. For now, we focus on seizure detection. However, down the road with the combination of the hardware in AI, it could go way beyond seizure. Our current focus is on seizure detection in acute care. That's the ICU and the emergency department. So with that, seizure detection in ICU and ED in the U.S. alone is more than $2 billion market TAM. And our future pipeline is also very exciting. Still in the near term, focus on acute care, but expand beyond seizure detection using the large EEG database we have as well as our AI capacity to indications potentially including delirium, stroke and others. And that would have a significant expansion of our TAM. Our Q3 until end of September last year, the revenue run rate is about $69 million, and we are active in more than 500 hospitals in the U.S. and that translates to year-to-year 46% growth rate, and the company-wide gross margin is already at 86%. So now switching gear. I mentioned we focus on seizure detection. Often, we think about seizure, we think about epilepsy. However, in the ICU is a very different type of seizure, and they are very common. A third of the neurological patients in ICU have seizure. And it's a very high proportion. If you follow here, it could be a patient had a convulsive status epilepticus, likely, it's the first time you come to CeriBell presentations, like the first time you heard the word status epilepticus, you're going to hear this over and over again. It's considered as the neuro emergency. It's defined as seizure lasting for 5 minutes or longer, and I'll show you later lead to high mortality morbidity. So a patient had a convulsive, meaning they are convulsing, 50% of these patients later would have nonconvulsive, basically very little symptom. And also post-stroke, post-cardiac arrest, post brain tumor and even post-brain infection and post sepsis. So it goes from neurological patients to non-neurological patients. Then you have 10% to 30% of your ICU patient have seizure, you need to monitor a lot more. And to make things even more complex, majority of these patients, up to 92%, when they have seizure, they have no obvious symptom. So it could vary from they just get confused or they are in complete coma. So you have to have this device EEG that detect brain wave to diagnose seizure. And there's a lot of parallel between seizure and stroke. We often say time [ is brain ] for stroke. Our physicians often say time [ is brain ] as well for seizure in acute care setting and here's why. If patients have seizure and we delay that treatment, a patient seized for 10 to 20 hours, mortality rate, as you see here, is 33%. A patient seized more than 20 hours, mortality rate can go up to 85%. The gray bar you see here on top of the blue bar is even when patients survive, morbidity increased significantly. That can be permanent secondary brain injury, irreversible memory loss and everything. You might wonder which modern ICU would leave a patient seize for 20 hours. Unfortunately, commonly in the hospitals in the United States, not even rural, even some teaching centers, which I walked you through. The good news, again, parallel to stroke is if you detect seizure early, the treatment is readily available. It's a large dose of benzo. And if you treat the patient within the first 30 minutes of seizure onset, 80% majority of these patients will respond. And in most of cases, seizure resolved for this patient. However, if you just delay that by 1 or 2 hours, the response rate dropped very quickly. First-line medication is the patient's best chance to stop seizing. So again, early detection, early treatment is one of the most important factors for seizure management. And you might ask, why don't we just empirically treat the patient because mortality and morbidity is so high sensitive. The reason is the treatment often lead to suppression of the airway. So that means patient -- physicians often have to intubate the patient and send to the ICU. Even with that, there is a large scale of empirical treatment happening. So to detect and manage seizure only, that means early EEG because it's the only way you can detect seizure in a reliable way, and the guidelines already agreed to it. Neurocritical Care Society, which really is the highest authority for this patient population because they master neuro ICU, more than 10 years ago, they already required to have EEG arrive on the bedside 15 to 16 minutes. And over the past 10 years and especially more recent years, you start to see Level 1 recommendation from American Heart Association, post-cardiac arrest, patients need prompt EEG setup as well as interpretation. And more recent years, you can see Stroke Society, both rapid EEG for stroke management -- post-stroke seizure management, both ischemic stroke as well as hemorrhagic stroke. Before CeriBell, most hospitals cannot be compliant to these guidelines. And if you look at the first [indiscernible] line, not a single hospital globally can be compliant. The reason is that the standard of care for decades has been conventional EEG. Last year, I used to say this year happened to be the 100-year anniversary since EEG was invented and not major improvement has been made there. It was designed for detecting epilepsy in the outpatient clinics where patients can wait for weeks and months. It's not designed for rapid response that's required in the acute care setting where every minute counts for patient's brain. And it has some intrinsic bottlenecks preventing it to be rapid response. Number one, the device is very big. So to set it up is very cumbersome. You have to have EEG technician to set it up. Majority of the hospitals have EEG technician available. Usually, it's a handful of them or a couple of them, Monday to Friday, 9 to 5. So that means 3/4 of the time, majority of the hospitals have no EEG available. And even when EEG tech get to the bedside, it takes them about 30 to 40 minutes just to set up EEG. So it takes a long time. And even after setup, EEG interpretation is highly specialized. Most neurologists do not know how to read EEG, only epileptologists or neurophysiologists, which is a very small group. So large teaching centers usually have 5 to 10 of them. Typical community hospital usually have nobody or 1 or 2 specialized neurologists. So if you combine these bottlenecks together, it's a very -- it doesn't meet the current need. And one more barrier is seizure is dynamic. It's unlike stroke. Patient can seize and stop 5 minutes and get back to seizure again. So it's critical for us to continuously monitor EEG just like EKG. A lot of cardiac events are dynamic. So we need to continuously monitor. And currently, no hospital -- very, very few hospital, if I don't make [indiscernible] statement, can afford to have neurologists to continuously monitor. So in reality, the supply of EEG is always hours and often days away from needing an EEG to getting an EEG, and continuous reviewing of the EEG recording is very rare. So now imagine if you were a -- if you were intensivist or emergency physician, you are left with a few very suboptimal choices. Number one, you can wait. If you worry this patient have seizure, you can order EEG and wait. That's somewhere between 4 hours to 2 days. And you remember the chart, if a patient seized 1 or 2 hours, can increase mortality and morbidity. And therefore, often, they go to the second option, which is empirical treatment. They know it's going to wait for a time, but just in case they would treat this patient and often they lead to unnecessary incubation, send patient to the ICU. And also recent guidelines start to advise against empirical treatment. The third option is barely option. You transfer patients out, which happens all the time. So it's very suboptimal. And then if you zoom out at the macro level, very likely before CeriBell, you already know sepsis and stroke and cardiac arrest. If you look at these patient populations, their mortality rate compared to status epilepticus. And hospitals have dozens of protocols on sepsis, stroke and cardiac arrest. And before CeriBell, there's almost no hospital have protocol for early detection, especially, and that's not compliant with the guidelines. And if you look at the age, that's shocking. This is not often the 80-, 70-year old, many of these patients are very young. So the average age of this patient population is 40-year old. So this is the big picture. This is what we are trying to solve. Instead of waiting for EEG for hours and days, CeriBell system allow setup within 5 minutes. We do not require an EEG technician, nurses, residents, really we can train any bedside providers and caretakers to set up CeriBell EEG in just a few hours -- just a few minutes. It provides clinically equivalent signal quality. Setting up EEG doesn't help patients. It's the result that drive the treatment. And so in a few minutes, our AI algorithm, we call it Clarity, would give the early detection. And also because seizure is dynamic, so it's critical to continuously monitor, our AI doesn't stop and doesn't sleep, doesn't get tired. It will continue to monitor EEG during the entire recording every 10 seconds. So I'll walk you through in more details how we do that. The hardware system is what you see here. You have a recorder that's just similar size to your phone, and the headband is single patient use. It's very intuitive, and this headband and this hardware is what allows nurses after 10 minutes training can set up EEG equivalent to conventional EEG. The data would then stream through hospital WiFi to the Ceribell portal. And that's where neurologists can log in on their phone iPad, the computer in real time to review EEG. And believe it or not, this is often many neurologists for the first time can review their EEG on their phone. So when they're on call, they can just quickly say, hey, what's going on? And these 3 components all largely replace the conventional EEG functionality. And the fourth component is brand new, and that's what we call Clarity, and that's the machine learning algorithm, optimized procedure detection. So let's double-click on Clarity since it's the most new component compared to other EEG systems. Clarity is an algorithm, look at EEG every 10 seconds. And then binary decide is that 10-second EEG seizure or not? And the data is very complex, as you can see -- on the left, you can see it's actually the raw EEG signal. So we said, Clarity would report the relevant information, both on the bedside as well as to the neurologist. You're all staring at this very complicated trace. So let me explain to you. The very complicated up and down wave, that's a raw EEG signal. And that's why it's very complicated to read as you can probably appreciate. And it's also time consuming. So this one page is 15 seconds EEG. For neurologists to read a typical 10, 20-hour EEG is thousands of these pages. So what Clarity does is at the bottom chart, that's only what CeriBell has is the entire recording and then share to the neurologies say, if you look at when patients have seizure, the season burden, which is the percentage of time patients spend during the past 5 minutes, would tell you where patient seize. Without knowing the detail, you will see this is probably where you want to click the first 10 minutes, and that's exactly where neurologists would go. So this would significantly improve both the efficiency of how neurologists will review EEG instead of spending hours, they could be more efficient in triage in terms of where to go. And Clarity also sent the bedside signal, which is critical because on the bedside, that's when the bedside physicians have to make life death decisions on do I treat this patient, do I not treat this patient. So whenever seizure burden passed 90%, which is equivalent to 4.5 minutes of seizure equivalent to status epilepticus, the device would turn red and start beeping. So that's a lot of information. Let me walk you through a real patient case and hope you can see this. So the X-axis you see here is the entire EEG recording. The Y-axis is the seizure burden. Again, it's the percentage of time patients spend in seizure during the past 5 minutes. First thing first, in this recording, it started at 1 a.m. That means this patient without Ceribell in majority of the hospital would never get an EEG at 1 a.m. because the tech is usually not on site. Even when they're on site, they might take longer. And when you set up an EEG, usually you take hours before you get a neurologists to read it for you. In this case, in 5 minutes, the device turned red and start beeping, say, the patient is in status epilepticus, continued seizure. And the little pink needle here, you see it's not photoshopped. It's bedside annotation. The nursing were able to treat the patient within 5 minutes. And if you remember the chart, not all the patients respond to the first-line medication. And without CeriBell, you would not know because likely this patient had comatose. In this case, within about 20, 30 minutes, the device say, hey, patient didn't respond to the first-line medication. And again, within minutes, they were able to escalate the treatment. And you can see in real time at 3:00 a.m., patients start stopping and seizing. However, continuous monitoring is critical because seizure is dynamic. In this case, patient returned to seizure at 4:30 a.m. And again, majority of the cases, this seizure would be missed. And in our case, it's not. And the bedside was able to again elevate the -- escalate the treatment again. This is not an outlier case. This happens all the time. And this patient in majority of the hospital, you don't have to go to rural area, pick any hospital in San Francisco, likely would have seized out the entire night, if not the next day in the morning. And if you remember, a patient seized 10 hours, 33% mortality; more than 20 hours, 85% mortality. So we continue to develop our clinical evidence, and we are -- this is one of the biggest publication we released last year. It's called SAFER trial. Retrospectively, we collected about 1,000 patients from big centers, Yale, Mass General and University of Mexico. About half of the patients are conventional EEG patients and half the patients on CeriBell. And we start to see this very strong signal in ICU length of stay. It's about 4 days. Initially, when the [ PI, ] they are all independent. And so the 4 days ICU, their initial reaction is there must be a selection by us. It's too bigger difference. We cut the data, they cut the data every single way. And in the end, the most reliable way, they carefully matched to cohort of patients carefully selected with similar sickness, diagnosis reason at admission and age and level of seizure submission. And if you compare, it's about 120 patient. CeriBell patients stay in ICU 4.1 days shorter. And then if you look at the one patient from door together EEG, these are top teaching centers, CeriBell get to EEG about 19 hours faster. The biggest surprise in the positive way even for myself is the outcome data. So we looked at the modified ranking score, essentially, it measures the disability when patients get discharged. The conventional EEG arm, 76% of these patients were discharged to be severe disability, which means -- usually means they cannot go home, they have to go to a long-term care nursing facility. With CeriBell, again, they are very similar patients, 58%. So that's 18 percentage point difference, which means out of the same 100 patients, 18 out of 100 patients, CeriBell Group, they were able to go home without severe disability. That's a big thing. Of course, it's association, but we are very -- this has been largely endorsed by our -- especially beside physicians. This is one of the many examples here today. We have published 35 peer-reviewed journals and 75 abstracts. And over and over again, we show physicians changed their treatment decisions, 40% to 50% of the time when they use CeriBell. We show length of say reduction, and we show we largely reduce the patient transfer due to lack of prompt EEG. I want to switch gear a little bit talking about the potential competition. We created a category. And so we continuously monitor the competition. We see our major biggest competition is from status quo, which is the conventional EEG. However, as we're thinking about potential future competition, we have 3 main areas: our IP patents portfolio, our biggest -- one of the biggest EEG database and AI know-how as well as our sticky business model with our customer base. So IP, pretty straightforward. We have 13 patents in the U.S. and another 12 pending. It covers from hardware as well as the algorithm and software development. One of our largest barrier is database because we have the hardware, so we have close to 1 million hours of EEG data already all through CeriBell. And not only our users label it, we have, over the years, invested heavily. We top KOLs epileptologists in labeling these EEGs. And so we have developed a very in-depth know-how, both in signal processing as well as machine learning, AI in this type of signal. This is a very unique time series EEG signal. So a lot of off-shelf machine learning tool from the tech doesn't easily adopt here. So the in-house know-how is critical. And moving forward, we'll talk a little bit later is using this AI know-how to extend beyond seizure potential for delirium and stroke. And our customers also have a very high switch cost. This is not a typical OR tool that you can -- when your rep is not there, your competition can train the physicians, the surgeons and the [indiscernible] lab or technicians switch the gear. When we lauch account, we work closely with our hospital in multiple departments often from ICU to the ED to the step-down units to the rapid response team. And we partner with them, develop what's the right workflow directly anchored back to Clarity performance and Clarity threshold, and we train often hundreds of nurses and [ dozens ] of physicians and sometimes 100 physicians when we launch top teaching centers. So the CeriBell switch cost is very high for the hospital. And moving forward, we are also developing potential integration of the EHR, which would make the workflow even more streamlined and increase the competition. So let me switch gear and talk about our business model. About 1/4 of our revenue is from the software as well as the machine learning subscription. That's the SaaS model. It's reoccurring. It's about $5,000 per month, that's our listed price. The rest of our revenue came from our disposable headband. So we don't really have any capital sales. And these headbands are our single patient use as I mentioned. So with this business model, we have created a very sticky usage pattern. We have a very low attrition rate. We have very consistent reordering. And therefore, as I mentioned earlier, create a strong high barrier of entry. Our current sales force have 2 arms. One arm is our account acquisition team, the territory managers. They focus on getting new accounts to account acquisition. And then the second arm is our clinical account managers. They are more responsible to continue growing accounts once we are in because we might not launch in every single department yet. We might not have trained every single physician. We might not have integrated to the proper patient workflow yet. And jointly, these 2 for us actually launch sites together. And the third unique part of our business model is the reoccurring revenue. As I mentioned earlier, the SaaS part is close to 100% reoccurring. And even though -- even the disposable -- even during the pandemic, when we lose the coverage, the usage has been rather stable and high gross margin, as I mentioned before. We're only reporting to Q4 last year. As you can see, the entire history as we can be reported, we don't have a single quarter that's lower than the previous quarter despite all the rapid change in supply chain and macro environment. It's really a validation of both our business model, pricing model as well as how sticky the usage pattern is. So if you double-click on this revenue driver, there are 2 main drivers. The left is what the TM, territory managers, focus on that's the account acquisition. As you can see, has been growing very steadily and consistently over the past years. On the right side is the same-store growth, and that's the utilization per hospital. The Y-axis you see actually start from 0. So we are not doing any game as you can see in '20 -- end of 2021, about 4 years ago, we decided to also focus on same-store growth, not just drive -- get as many accounts as possible. During the past 3 years, we have roughly more than doubled our usage per account. So we have confidence that we have a playbook on both ends will continue to drive account acquisition as well as the same-store growth. We got the question a lot, what is our TAM? I mentioned in the U.S., seizure detection alone is more than $2 billion. And here's why. We actually did a very bottom-up work closely with physicians and the database. And in the U.S., there are about 3 million patients are at risk of seizure and status epilepticus in ICU and ED. And in terms of number of facilities, we include short-term acute care, critical access and freestanding EDs. And recently, we have gained access to the VA hospitals that further extended our TAM. So we have, in the U.S., about 6,000 hospitals who will benefit from using CeriBell. So our ASP, the estimation is more than USD 2 billion TAM. So in summary, in terms of our future growth in terms of revenue, in our core markets, we'll continue to drive account acquisition as we have been doing during the entire commercialized history, drive utilization as we have been since we started our utilization strategy. One of the barriers still remains the awareness of status epilepticus, both how common it is as well as the severity and some of the recent guidelines that will continue to drive the awareness, the disease state awareness, especially in the emergency department. We are very committed to grow our clinical evidence and continue to improve our system product to further improve the user experience. The more medium-term growth area goes to potentially expand to OUS as well as some adjacent market. The VA is the first time we proved that we did that. And also going beyond seizure. So let's talk about going beyond seizure. We see our future in 3 horizons. We are currently in the first horizon, which is to become standard of care for seizure management in acute care setting. In the medium term, our goal is to make EEG a new vital sign. What does that mean? Brain remain a black box in the best ICUs and emergency department in the United States or any country. Let's say, if you are a cardiac surgeon, you may just finish the major cardiac surgery and your patient didn't wake up within the time window you are expecting, your patient could have a stroke, your patient could have noncompulsive seizure or your patient could have just been overly sedated. Other than sended patient to an urgent CT, which doesn't give you the full -- or MRI, which doesn't give you the full answer or doesn't provide any continuous monitoring. You don't really have a tool. So there are still a lot of guessing games. And EEG is one of the richest signal you can acquire on brain functionality noninvasively. And historically, because the hardware has been so limited and AI is not where it is today. So by combining eliminate the barrier of signal acquisition, building this large EEG database, use the state-of-art AI, we are very excited to potentially really see harness what EEG can do. And so we could potentially see EEG could be a biomarker for detecting area to detect stroke to detect other neurological indications in the ICU. So that just like we monitor patient's heart with EEG if the patient have a chest pain. A patient have altered mental status, any acute cognitive changes in ICU and ED, you put CeriBell will tell you, relax your patient is overly sedated or please rush patient to the CT or CTA, looks like your patient might have a stroke or please initiate your antiseizure medication path. In the third horizon, which is longer term, it goes beyond acute care setting. It could be -- if you search any neurological or psychiatric disorder, you think of depression, OCD, ADHD, dementia and you put EEG and biomarker, you google it. I promise you somewhere some labs have shown EEG can be a biomarker for that disease state. However, again, historically, because the data acquisition has been so limited and this often requires sophisticated AI that has not been possible. So again, that's our third horizon. My last slide of my presentation is, so what are we going to do this year? We naturally will focus on the first and second horizon. For the first horizon, commercially, we'll continue to drive our account acquisition as well as same-store growth utilization. Since we recently acquired ATO from the VA system, we will drive the VA adoption in 2025 as well. On the regulatory front, the main expansion we do here is to extend patient population. Our hardware is approved or cleared for all the entire age group. Clarity, which is our AI selection -- AI detection algorithm only approved for 18 years and above. We already submitted our pediatric Clarity 510(k) and are going to secure a 510(k) clearance for pediatric population. We are also preparing to potentially submit FDA application for neonate Clarity algorithm. And this would allow us to cover every single group starting from NICU. On the second horizon, which is to make EEG a new vital sign, we already received the breakthrough designation on delirium. So we plan to submit FDA application to get clearance approval on delirium. So the submission will happen this year. And on product front, we'll continue to both increase the user experience of our existing product for seizure detection and also will invest in developing potential algorithm to detect stroke as well as the second-gen hardware that would enable and support this new indication expansion. So with that, I won't repeat the last side, but we are a device that have potentially saved many lives, saved hospital a lot of money with a big TAM and even more opportunity to expand the TAM reoccurring, predictable, high-quality revenue and have established reimbursement. We did not get a chance to cover today as well as the very experienced leadership team that took the company so far. Thank you.

All right. So maybe we could start with the last slide. I think it was the first we've heard delirium is going to be submitted in 2025. I think that's the first time line we've gotten. So maybe talk about how delirium helps the selling experience of just the current indication and what another indication adds not only to the clinical necessity but the pitch to the hospital, the experience for the hospital and just any other color you could add?

Yes. So delirium is the most common neurological disorder in acute care setting, especially in the elderly patients. And the unmet need there currently is a nursing behavior test. It's difficult to do. It's unreliable. It's a spot check. And as we potentially launch delirium down the road, the synergy with seizure is huge for couple of reasons. One, it's the same call point. It's the ICU and potential ER. And two, it's the same patient population because the symptom is the same, the patient is altered, cognitively all a sudden, they are confused. So to differentiate delirium and seizure itself has a natural synergy. So we do not have to establish a new sales force for it. And this is one more step for making EEG a new vital sign so that we can really help bedside to understand what's going on with patient's brain, what kind of failure, neurological failure that's happening.

To me, the clinical necessity seems really clear, but it's a new way of treating something that's been around forever. What's the pushback you get when you go into hospitals? And once you go through the process and you show on the product and the technology, what are some of the reasons that hospitals decide not to adopt CeriBell?

And you're talking about seizure or delirium?

Seizure right now.

Seizure right now. Yes. I think the -- for seizure management, when you think about it, the number one, I would say, is resistant to change, as you mentioned. This has been practiced for decades. For your probably context, most emergency departments have never used the EEG before because when you wait a device for 2 hours -- or 4, 5 hours or 2 days, so imagine you're an ED physician, you were never trained in medical school, you need EEG. You always think that's upstairs. And they were battling COVID or they are recovering from COVID. And to tell the physicians and nurses, you have to completely add one more task. I think that's the biggest resistance we have. However, that's why we invest heavily in clinical evidence to show how this can benefit patients as well as the economic benefits. And I think overall, that's the first one. The second, as all the hospitals are under financial pressure. So we have very strong health economics data. However, it's not the simple procedure, say, hey, the procedure is $5,000. It costs you $4,000. It's a more sophisticated DRG discussion, and that takes time.

You recently had an IPO at the end of 2024, raised a pretty substantial amount of cash. What's the priority usage for that? How are you taking this? Where should investors expect it to go? And how does that translate into better business results?