Pioneering Data-Driven Growth in Behavioral Health at Holmusk

Daniel Levine (00:00.398)

The Life Sciences DNA podcast is sponsored by Agilisium Labs, a collaborative space where Agilisium works with its clients ranging from early stage biotechs to pharmaceutical giants to co -develop and incubate POCs, products and solutions that improve patient outcomes and accelerate the development of therapies to the market. To learn how Agilisium Labs can use the power of its generative AI for life sciences analytics,

To help you turn your visionary ideas into realities, visit them at labs.agilisium .com.

You're tuned to Life Sciences DNA with Dr. Amar Drawid.

Daniel Levine (01:08.942)

Amar, we've got Eze Abosi on the show today. For listeners not familiar with Eze, who is he? Eze Abosi is the Chief Growth Officer of Holmusk. He's got nearly two decades of experience leading commercialization of data, analytics and consulting solutions, supporting the life sciences industry. He joined Holmusk after serving as the Vice President of New Products and Partnership with Optum Life Sciences and has held commercial leadership roles with Decision Resources Group, which is now part of Clarivate

as well as IQVIA via and OptimizeRx. Rx. And for people not familiar with Holmusk, what is it? Holmusk is harnessing real -world data to transform both research and care for behavioral health. It's using data analytics and AI to bring precision medicine to the treatment of mental health by unlocking much needed evidence to guide treatment decisions, drug development, and improve outcomes for patients.

And what are you hoping to hear from Eze today? So I would like to understand the company's platform, how it works, and what type of insights they're able to glean. And also like to understand how companies are leveraging this data and these insights and how receptive they are to then bringing precision medicine to the area of behavioral health, which we know the doctors have relied quite a bit on trial and error type of approach of care. So how we can bring evidence generation

to the treatment of mental health. And a reminder to our audience, if you want to keep up on the latest Life Sciences DNA podcast episodes, hit the subscribe button. If you enjoy the content, hit like and let us know your thoughts in the comments. If you want to listen to this while you're on the move, an audio only version is available on major podcast platforms. And with that, let's welcome Eze to the show.

Daniel Levine (03:07.054)

Eze, thanks for joining us. We're going to talk today about mental health, Holmusk, -Musk, and its neuro-blue platform technology. But before we get into this, let's start with the problem that you're trying to address. And so when we talk about mental health, can you tell us about what are the diseases and conditions that we're talking about? Yes. And so, Amar, once again, thank you so much for the opportunity to join the DNA podcast. Very humbled, very exciting to be part of this dialogue.

In terms of your question about the disease categories related to mental health, we are focused on addressing problems and bettering outcomes in disease categories like depression or anxiety or schizophrenia, PTSD or broadly addiction or specifically, for example, opioid use associated with addiction disorders. And so that very unique category focused on

the brain and the mental aspects of how the brain impacts the patient and ultimately outcomes. That is the problem that we are focused on today. So when we talk about like these mental health, so there's a psychiatric, there's neurological diseases. Can you talk a little bit about like, so there's Alzheimer's, there's Parkinson's and all of that. So how do you think about all of those problems? Are they all together or do you separate those out or so?

Yeah, no, that's a great question. Oh, wow, that's also a loaded question. I would say that generally in terms of how the industry, specifically the pharmaceutical industry, thinks about drug developments, there is a major just caveat in terms of the blood -brain barrier. And the implication of that is the additional complexities associated with developing a drug that ultimately does to some degree

alter, if you will, the mind or the state of the brain. And so within that broader kind of central nervous system approach to drug developments, there is the broader category that is neuroscience. Within neuroscience, we can think about it in two ways, neurology versus psychiatry. With neurology, I think there's much more of a biological, if you will, kind of implication to the disease states.

Daniel Levine (05:24.334)

And so when you think about, for example, Parkinson's, multiple sclerosis, ALS, or Alzheimer's disease, the diagnostic workup often includes imaging technologies. In other words, there are potential variations or imbalances within the brain matter itself that may help us diagnose the patients. With psychiatric disorders, the picture is not as clear.

There may be potential imbalances in the chemical kind of structure or makeup of a given patient, but it's a very heterogeneous population in terms of depression versus anxiety in particular. But what's also very complicating is that there is a significant amount of comorbidities. And so a substantial proportion of patients with major depressive disorder

may also have and will also have, for example, comorbid generalized anxiety disorder. And so very similar to the complexities of developing a drug, although you're focused on one disease category, you have to look at the patient holistically, whether that's in terms of neurology or neuroscience broadly, whether that's in terms of the psychiatric conditions that are implicated in a given patient or patient cohort.

Likewise, there's a mental aspect to everything. And so thinking about the implications of depression and an overall population that has, for example, late stage cancer is also a very unique value proposition and a very unique problem to solve within our industry, which is biopharma. And so those generally statements, those statements really do kind of speak to the nuances that are relevant to its psychiatry. We can look at it

in terms of the specific disease, like depression, we can look at the comorbid aspects of the disease. And when you say comorbid, it's the other diseases at the same time, right? Yes. So both for audience, right? Exactly. Exactly. And so, for instance, an individual with an addiction disorder may also have comorbid depression. And so those complexities are going to some degree kind of just be implicated in one disease versus the other.

Daniel Levine (07:45.55)

So that makes drug development, which is really kind of bettering patient outcomes for the mental health patients even that much more complex. And so we are, it's a global epidemic. And so it's a problem very worthwhile to try and solve, if you will. Yeah, yeah. And we keep hearing, right, like that the mental illness problems, those are increasing right now. But can you tell us about like how big is that need and what is the general experience for people seeking treatment?

Absolutely. And so in the in the Western world, especially as our reliance on technology increases, we're seeing mental health outcomes to some degree, being negatively affected. That being said, I think the pandemic environment was also a very unique experience and a very unique set of circumstances for all individuals, whether you're an adolescent that was, for example, forced to some degree to go through their schooling in a virtual environments.

There are many kind of mental health implications there. There's the adult population that may have been significantly impacted in terms of their housing stability, their job stability, and the subsequent kind of implications on their mental health and wellbeing. Likewise, the elderly population. And so for example, as the person ages and becomes more predisposed to, for example, having a neurological disorder like dementia,

there may be some psychiatric aspects to that unique individual that become more salient as they age. And so mental health truly does permeate through really every individual in all aspects of our society, which again makes it a very notable undertaking to try to solve in some way, shape or form. So we have seen precision medicine emerge in areas like cancer.

but we haven't really seen that emerge in a significant way in the area of behavioral health, right? Even though things are worsening there. Why is that? That is a great question. And so I think from a quantitative perspective with cancer treatments and outcomes, there have been significant steps forward in

Daniel Levine (10:10.254)

biomarkers and diagnostic tests that help the providers, the caregivers, and overall just the stakeholders better understand the most appropriate ways forward in managing their disease. It's very much evidence -based in terms of many cancers, many rare diseases, many orphan disorders, and likewise even like some autoimmune conditions. With mental health,

the biopsy options are much more limited. And given that limitation historically, I think it's fair to say that we have not seen precision, dare I say it, psychiatry evolve in the same manner that we have seen in other therapeutic areas and most notably cancer. But there is a growing interest

in not only understanding kind of the biological as well as genomic and certainly kind of epigenetic factors associated with mental health outcomes. And so although we are not there just yet, I think there's quite a bit of promise in how we can deliver the promise or deliver the value of personalized medicine within psychiatry. Based on just some

conversations as well as observations that I have had, laboratory testing and understanding, for example, biomarkers associated with a particular patient is becoming incredibly relevant towards psychiatric treatments and also psychiatric drug developments. And we're seeing this not only in academia, not only in pharmaceutical companies, but also in the provider entities that are innovating

within the mental health space by not only providing robust psychotherapy services, but also understanding the genetic makeup of a given patient, as well as their laboratory measures prior to, during, and of course, longitudinally following therapeutic initiation. And so we're getting there, but to your point, Amar, we're not there just yet. Okay. Now we talked about this more like the biomarker development area, but what about

Daniel Levine (12:36.621)

like accessibility and actionable health care. Like, so having the data around that. So can you talk a little bit about like, is mental health, like the solutions for that are accessible at this point? And what kind of data is being collected at this point? Yeah, that's another great point, as well as question. From my perspective, kind of the broader topic of social determinants or SDOH, social determinants of health

is relevant across all healthcare and across all therapeutic categories. I would argue within mental health, it's critically important because the individuals with severe or acute mental health disorders, I think that it's a fair intuitive leap to theorize that outcomes for that population are

are more extreme in a negative way because they're likely to not have, for example, a stable source of employment or a stable household or stable just place to live. And so because of the instability, for example, in their income status, they may not be commercially insured. Instead, they may be leaning on state level Medicaid

in the US category at least, to be the primary kind of insurance provider to help them manage their disease. And so you combine their Medicaid status, which to some degree has limitations versus a commercially insured patient population. You combine their insurance status, their lower socioeconomic income bracket, and potentially their location in the rural part of the US,

their lack of availability to access technology, or even just Wi -Fi, and really just the kind of the issue snowballs. And from what we have seen in various evidence -based publications, the outlook is not so positive. And so in short, Amar, I believe that the complexities of mental health are exacerbated in the most extreme ways when you think about

Daniel Levine (15:05.421)

the circumstance of the mental health patients that's let's say suicidal versus patients in other therapeutic categories or disease categories. Okay. Now, can you talk about like, so you have been collecting a lot of real world data, right? So can you talk a little bit about how you're collecting that? Let's start with that first.

Yeah, absolutely. So just generally, just to maybe just give you a little bit of a baseline. And so Holmusk -Musk was started by a gentleman named Noel Roy, who recognized that there was a global crisis in terms of mental health, and this was prior to the pandemic. And so how has Holmusk -Musk tried to solve that problem? Really using two key value adding services. First and foremost is data, real world data in particular

that uniquely has the ability to understand the psychometric assessments that are associated with mental health treatments from the provider perspective. And what do you mean by psychometric? Ultimately, what makes you mental health or psychiatry unique from other therapeutic categories is the questionnaires that are dictated by the provider to the patients. It can be

very straightforward and very simple, such as how happy are you today? That's a great question that has incredibly valuable implications, depending on how the patient answers in one moment in time. So everything from the psychometric assessments in terms of the data that we're normalizing and curating to the platform. And so the platform may be relevant to the patient.

The platform may be relevant to the provider. The platform may be something utilized by the payer. And so the platform could be something that is ultimately kind of serving all stakeholders in the value chain. But Holmusk is trying to basically solve an aspect of the US specific mental health crises by leveraging data and platform. And so in terms of the data, the primary interest

Daniel Levine (17:29.037)

for our team at Holmusk is electronic health records data, or simply put, clinical data. And so our approach is to partner with hospital systems that are medium to large scale who have a robust behavioral health or psychiatric practice integrated within their system. And we really do hone in on the patients,

interacting with providers in that system who have any type of mental health diagnoses. And so we will, in a HIPAA compliant way, of course, normalize and curate the clinical records or the EHR data for those patient populations that are being managed by the provider entity that partners with Holmusk -Musk, which is pretty standard. Nothing there is unique. And so our differentiator

is being able to work with the unstructured provider notes from that same hospital partner. And uniquely, we are extracting the psychometric assessments, again, those unique questions, fielded by the provider to the patient, or sometimes often completed directly, let's say in survey format, by the patient. We're extracting that information from the unstructured provider notes.

Why that's so important is because after you normalize and curate that information and make it available for evidence generation, you're essentially empowering the industry, pharma, academia, government, anybody associated with or interested in better health outcomes for psychiatry or for mental health patients overall. We're empowering them to perform a more granular type of analyses

because there's essentially more data for them to interrogate and to generate evidence than there ever has been kind of historically in the industry. In other words, Amar, historically, researchers with an interest in depression have worked with claims data, tried and true incredible resource that's available at scale. But you're missing the clinical detail.

Daniel Levine (19:52.653)

And so that clinical detail is often available in the EHR. But Holmusk -Musk in particular, through our neuroblu capability and our neuroblu suite, is able to further enrich the clinical data by applying unique techniques like NLP or natural language processing, or you're simply extracting information that is usually not available for evidence generation.

So we can ultimately have a deeper phenotype of the patient of interest. So let's talk a bit. I want to double down a bit on this because our audience may not know about claims versus EHR, electronic health records, right? And you're an expert in that. So can you talk about, so these are the medical claims, right? Like the claims that are made to get paid, or so, right? But the kind of data that's in the claims

versus the kind of data you can get from the electronic health records or EHR or you can call electronic medical records or EMRs, right? So can you talk a little bit about that?

Because it'll be important for people to understand, when we're talking about real world data, what kind of data people can get from EHRs versus from claims data. Yeah, absolutely. And so the claims data comes from insurance companies. From my perspective, in terms of

understanding the nuance of a patient journey, one of the best sources to do that is the insurance company or the payer that is ultimately reimbursing for services associated with set patients. And so in other words, Amar, if you want to really understand the healthcare resource utilization of a patient population,

HIPAA -compliantly work with a payer, a group of payers, or an aggregator of claims data representing U .S. payers to understand the financial transactions that a patient population undertakes to manage their disease. To your point, the medical expense is a really, really great resource. One of the best examples of a medical expense or medical claim,

Daniel Levine (22:14.829)

is something that confirms the patient was hospitalized. And so being able to tease out or to understand whether that hospitalization was based on an ER visit, was based on an inpatient service, was based on a surgery, and literally the listing goes on, that's incredibly valuable. And because ultimately, there's a financial component to the claim, we can not only understand the events at one point in time,

but we can also understand the cost. And so that's why the medical claim, especially the hospital claim is so valuable. Above and beyond that, there are other different types of claims, such as the claim that transpires when the consumer goes to a retail or an online pharmacy and fills their prescription. First and foremost, are they filling the prescription? If they're not filling the prescription, why?

Is it because their prior authorization was denied? Is it because the side effect profile of the drug is just not manageable? Is it because they simply can't afford it? And so understanding the reasons why consumers or patients are not being adherent, that is a really unique value out of the claim. Kind of switching gears to the office setting. Infusions are very relevant to neuroscience because many of the different therapies that are managing

neuropsychology or CNS patients are infused. And so understanding the relevance of what's being infused, the associated cost, who is infusing the medication, and those different details are also just incredibly valuable insights from the claims. And so switch gears a bit and let's talk about clinical data or the EMR. And so generally, the EMR is incredibly powerful because it includes...

the medical history, of course, of the patients. And so while the claim will tell you that the patient had a certain healthcare service delivered, the EMR record allows you to get that much more granular in the clinical or the medical aspects of their journey. That value is, from my perspective, compounded when you think about the insights, not only in the EHR record

Daniel Levine (24:40.045)

that are structured such as the medication by NDC, the diagnosis at the ICD -10 code level, and the listing goes on. But once you start to work with, let's say, the free text that's directly entered by the physician, things get really interesting. A few examples to consider. The claim will confirm that the patient had a lab test. The result of said test is in the EHR record

but typically not in the structured elements. It's typically in the unstructured elements of the EHR. And so going back to this original kind of theme in question about precision psychiatry, we know that precision medicine is typically reliant on some type of diagnostic to guide the provider and guide the patients, caregivers, and all stakeholders on what is the most appropriate course of treatments. And so great to see

the laboratory order on the client, even better if you can understand what the results of that laboratory or diagnostic test is and how that impacted the patient's journey moving forward. And so kind of that illustrative use case is relevant across all therapeutic areas. But I want to just pause there just to make sure that what I'm saying is consistent and does make sense before I delve into maybe the nuance specifically in mental and behavioral.

So yeah, thank you, Eze, for the explanation. So it's pretty interesting about thinking about the structured and the unstructured data. So structured is where you have the data in the tables, but then unstructured is someone has typed the notes. And so that unstructured data, it's the content, but you need to process and analyze both the structured and the unstructured data to really understand what is going on with the patient, right? That is absolutely correct.

And so the kind of the unmet need that we're describing here today, there have been some very formidable companies that have tried to solve this need across various therapeutic categories. Some of my favorites and kudos to the industry are, for example, Flatiron. Flatiron, when it comes to solid tumors, did some really good work in making personalized medicine

Daniel Levine (26:59.981)

a reality from the real world data and evidence generation perspective. Likewise, ConcertAI AI and many other companies that are focused just in solid tumors, if not oncology. Another really interesting play is Koda, specifically within hematology, generally across therapeutic categories. There are some legacy players like Optum Life Sciences, my former colleagues there have done a great job of supporting drug development across

a variety of therapeutic categories, likewise, Komodo. But within mental health specifically, there are really only a few players. And Holmusk is uniquely positioned as the only entity today that is solely focused on mental health. Therefore, there's not a broader kind of consideration or interest, at least in this point, in neuroscience, we're really just deep in psychiatry and nothing but psychiatry. And so...

whether the topic is addiction disorders or anxiety or the comorbid depression patients. If it's relevant to mental health, Holmusk -Musk is trying to uniquely extract the associated psychometric assessments. That way your team in governments, in academia, in pharmaceutical developments is well equipped to develop a proper drug for a very unique subpopulation. And that is kind of helping us inch towards,

that again, that promise of eliminating the evidence gap and realizing the benefits and the value of precision psychiatry. Okay, so tell me a bit more about this, right? So when you're collecting this real world data, a lot of the patient data, so like how many patients' data do you have? And then what is kind of like the coverage? Like for which of these disorders do you have a lot of data? And then, like how are you using it? Like what kind of...

insights are you getting from this? Yeah, absolutely. And so, Holmusk has evolved quite a bit in the last 12 to 18 months. 18 months ago, the number of enriched patient lives, the patients not only have kind of the standard structured elements of clinical data, but also have these unique psychometric assessments available for evidence generation. That number was about 540 ,000. Good number.

Daniel Levine (29:20.109)

Great for trending nationally, can't go deep. That number has magnified by about 8x at this point. And so today we're 4 .1 million lives that are enriched. We're basically acquiring additional partnerships at scale in this calendar year. And we think there'll be three more hospital partners that'll interest up towards 10 million enriched lives.

So 10 million lives, 10 million patients who have some kind of psychiatric or like mental disorder, you have the data for all of those? Yes, yes. In other words, for 10 million patients, we have at scale applied NLP and various enrichment techniques to deliver high quality EMR data or clinical data, as well as these unique psychometric assessments. And that is truly a differentiator because again, the standard for

the pharmaceutical workflow is to simply work with claims data only when thinking about the patient journey, healthcare utilization, or ultimately the cost of managing mental health disorders. And so the shift that Paul Holmusk is leading to leveraging EMR for the mental health population, that's new and that's very exciting. But the NLP enrichment is extremely unique.

And for the first time in my 20 year career, Amar, I'm having conversations with colleagues internally about how we commercialize data points that the industry has never worked with before. And so it truly is a nascent opportunity because the data that we're extracting has never been available at scale in the US healthcare marketplace. And that is what excites me. Okay. Now you talked about NLP enrichment. Can you please tell us more about that? Yeah. So NLP, it's AI.

There it is. We finally said it. On the DNA podcast, we finally reference AI. And so AI can mean quite a bit of things. In this particular case, we are applying an AI, specifically machine learning technique, called natural language processing, to understand the way that providers are articulating observations, symptoms, severity, or stressors.

Daniel Levine (31:44.237)

of mental health patients within their Epic systems, for example. But it's not just Epic. We work with all the EMRs. And so it could be Cerner, it could be Athena, it could be Veridigm. Doesn't matter. But in terms of NLP, I think I'm not a data scientist, but I certainly appreciate the data science. And so in terms of NLP, I think about it as a spectrum. To some degree, the most kind of rudimentary form of NLP is a simple keyword search.

And so just going back to my kind of emphasis on being able to understand not only the standard kind of elements of a good clinical data set, but also the psychometric assessments that make behavioral health unique. Being able to search for whether this patient had a suicide attempt or their level of anxiety at one point in time. Literally, control F and just search away.

to some degree, that's NLP. And so that's the, I wouldn't necessarily call that NLP, but just for argument's sake, just that assume that very simplified example was one example of NLP. On the other side of the coin is a true machine learning algorithm trained on terabytes of data that is at scale understanding concepts,

under accounting for potential misspellings, abbreviations, or just variations in the way that human beings speak the English language or non English language. That is the, the, the, at least today, kind of the alpha state of NLP. And so within Holmusk -Musk, our enrichment techniques are really transpiring in three different ways at this point in our journey.

First and foremost is leveraging our enrichment capabilities, which is more of like the rudimentary version of NLP, to extract semi -structured data. And so there are some assessments or questionnaires given by the provider to the patients that are absolutely structured. But the EMR system the provider is working with is not formatted today to ingest that information in the structured elements.

Daniel Levine (34:06.989)

And so although it's technically unstructured, when we pull it out of the notes, for instance, it's essentially structured data. I wish it was always that simple, but it's just not. Kind of the next iteration are going to be different questionnaires. And there's one in particular called the mental status examination or the MSE that we work with quite a bit that are quick prompt

questions from the psychiatrist to the patients that really delve into symptomology. And so although these questionnaires are basically standardized in clinical practice, from our perspective, this information has not been used to generate evidence. And so when a formal disease -specific psychometric assessment is not available,

What we are doing at Holmusk is leveraging the mental status examination that's available at scale to understand how severity is trending or how symptoms are evolving, even when the formal test is not there. So that's another kind of version, if you will, of our NLP enrichment techniques. Then there is the true NLP machine learning exercises that involve clinical annotation, medical directors, physicians, and of course, data scientists kind of working

in sync to do something really special. And we're building those at this point at Holmusk to examine two disease categories. They're the two most important from a pharmaceutical pipeline perspective, and those are depression and schizophrenia. And so within depression, the problem we're trying to examine and solve is comorbid anhedonia.

And so think about the hedonist and then the opposite of that profile. The individual that rather than completely seeking to be happy all the time, the individual that is not happy at all is unable to feel happiness. And so there is a formal ICD -10 code for anhedonia, but it's underutilized. And so we actually published this with our full methodology. But what we found was

Daniel Levine (36:27.789)

above and beyond the ICD -10 code that's utilized for or underutilized for anhedonia. Essentially, once you apply and build a model, train the model, deploy it appropriately, you will confirm that anhedonia in terms of a comorbidity associated with an MDD or major depressive disorder patients, it's much more prevalent than what we would think. And those findings were so stark that we published not only on the key insight, but also on the methodology.

That way there's transparency on what we at Holmusk are doing. Take that same theme of ultimately finding comorbidities, symptoms, and additional insights in disease severity and apply it to psychotic disorders like schizophrenia or bipolar depression. And so being able to not only understand positive symptoms, things like hallucinations, but also

really kind of delving into the cognitive and negative symptoms such as social withdrawal, being able to deeply phenotype a schizophrenia patient or a bipolar disorder patients by further examining the unstructured provider notes using NLP. That is some of the groundbreaking work that we are undertaking at Holmusk today, and it's not going to slow down. And in transparency, Amar.

Like this innovation of really enriching the NLP and other techniques enriching RWD. That is why I joined Holmusk. It's because of the promise of delivering the same value companies like Flatiron, Coda, ConcertAI and the like have done within their respective therapeutic areas. Yeah, I mean, that's amazing, right? And these are underserved diseases. So you're I mean, there's just a lot to be discovered here based on the rich amount of data

that you have access to. So tell us about, with the data of so many million patients, how do you deal with the patient privacy or making sure that this data is anonymized or so? Yeah, interesting. And so that's a great question. Thank you for proactively kind of raising it. I was going to touch on this at some point in our conversation. But mental health data is incredibly sensitive.

Daniel Levine (38:51.885)

PHI and PII, of course, as we can both appreciate, is a very sensitive topic and you must not only address the issue, but you must move forward in the appropriate way. Mental health data is even more complicated, is even more complex. For instance, that there are, I'm not going to specifically cite any names, but there are companies

that are well known for their real world data capabilities and they proactively kind of shy away from focusing their efforts on mental health because of the sensitivities of their data partners. And so I think it says quite a bit about Holmusk, our vision and the value that we deliver to providers since they're enabling us access to their mental health data at scale.

And so first and foremost, the data is de -identified by the provider before it enters our environment. Likewise, the... When you say de -identified, any information about the patient is taken out, right? It's removed. Exactly. And it's not only done appropriately by the provider entry or the hospital system, but there are other parties

that we work with that provide expert determination to ensure that the data is truly de -identified given the context of the setting and the patients that it represents. So basically then based on the way it is done is that if someone looks at the patient data, there is no way they can identify which patient, what would be the name of that patient. There's no way for anyone to go back to the patient.

Correct? That's the goal. Absolutely. Yeah. Absolutely. Which actually is a great lead into my next point, which is the use cases that we serve. And so basically, in short, we are supporting research. Research with the full intent of bettering patient outcomes in the mental and behavioral health arena. And so there are certain entities that may aggregate data,

Daniel Levine (41:16.493)

aggregate data for, call it commercial purposes, to, for example, sell more units of a certain technology device, whatever it might be. Holmusk -Musk is more so, is absolutely focused on supporting research that ultimately helps better understand as well as develop options to manage mental health.

And so we not only appropriately

Daniel Levine (41:49.773)

blind if not encrypt the data, but we have the technology infrastructure and security protocols in place to ensure that it's protected. And we only enable use to our platform for the appropriate R & D purposes. And so again, pharmaceutical discovery R & D teams are the key user of our insights, academia or government entities that are thinking about, for example, policy to better mental health at scale.

Okay, and so who are your customers? So these are the way you talk to us. So who are your customers? How do you work and how do they work with your data? Right? So if you can explain that a bit. Yeah, that's a great question. So again, it's more so those broader themes of academia, government as well as life sciences. I think what you're asking, Amar, is more so kind of just visibility on the title or the type of individual or user. No, no, no, it's more about like what like these entities who are using that, right? Like what is the kind of

How do they use the data? What is it that they get from it? Absolutely. As you said, it's for R & D, right? So is that they're looking for specific drugs that they can then use? Can you talk a bit about that? Yeah. And so if you're an epidemiologist, a health outcomes researcher, or somebody interested in communicating value and outcomes in mental or behavioral health to a provider base or physician base,

You are you are a potential user of our services in terms of why you're using your services to provide a very deep phenotype of a certain mental health population our chief operating officer and president McKinsey guy's name is Tony Trimonti, the way that he frames it and I think it's very eloquent is that we conduct a digital biopsy of the mental health patients again going back to evidence

and precision medicine, or in this case, precision psychiatry, our insights at Holmusk are to some degree equivalent to the biopsy results that a cancer patient may review with their provider when it's time to really kind of think about the next best option in managing their disease. And so, really deeply profiling a patient population from an epidemiology perspective.

Daniel Levine (44:12.301)

Absolutely a use case for how we support the industry. Likewise, outcomes research. And so what are the economic implications of managing a disease like depression or schizophrenia using this kind of pathway or journey versus a different pathway or journey? For example, psychotherapy versus psychotherapy plus pharmacotherapy or a new pharmaceutical intervention

that is a novel approach to the better side effect profile. The examples go on, but really kind of thinking about the explicit implications from a monetary perspective, as well as in terms of healthcare utilization and the cost, especially when you think about the cost across different insurance types. Like the medicated population, for example, versus the...

the Medicaid population, for example, versus the commercially insured population, Medicare, so on and so forth. Okay. Now, like all this work that you're doing, right, So how do you ultimately see this impacting the way mental health is diagnosed and treated? Absolutely. And so there, again, our ability to deliver value or to deliver on our value proposition,

hinges on the buy -in of provider networks or hospital systems with robust behavioral health practices. And so why would they ever work with us? And so there is a grant component, so how we partner with our hospital systems, but it's not just that. It's the access or enabling access to our platform. So the post -docs or the researchers within an academic

institution, like an academic medical center, can perform high quality research using Neuroblu. For the practitioners or the executives that are managing the behavioral health unit itself, we offer a unique tool that helps predict which patients under your management are most likely to go into crises. That way you can allocate resources accordingly

Daniel Levine (46:35.533)

to really focus your attention on the patients that need it most. And we have generated evidence through our partners, government partners actually, government entities, year over year over year, I think we're about five years into it at this point, that verifies we're having a very significant impact on healthcare utilization rates, producing ER visits,

and ultimately bettering patient outcomes. And so, yeah, that's kind of how we're thinking about it, and that's the approach. That's great, yeah. And finally, how do you see the world of mental health five years down the road? Five years down the road, I think we will start to really think about in the same way that the term, let's say, lung cancer has evolved.

I think that the term depression or the like is going to evolve where we think about not just the top level disease as a category, we still have to really think about the key subpopulations that are relevant to different journeys downstream. And so breast cancer treatments 20, 25 years ago was very consistent across the board.

Breast cancer treatment today is incredibly, or lung cancer or pick your cancer, is very dynamic based on, for example, the genomic markers or biomarkers implicated in your particular tumor, i .e. precision medicine. And so I think moving forward, we'll start to see, for example, drug development focus on not just major depressive disorder patients,

but major depressive disorder patients with a high risk of suicidality versus the major depression disorder patients with comorbid anhedonia. And in fact, Johnson & Johnson, kudos to J &J, has really pushed this topic forward with the approval of esketamine, which is indicated specifically for the MDD patients with a high risk of suicidality.

Daniel Levine (48:58.701)

And so it's real and I think that we're going to propel the industry forward with additional interest, investment and positive outcomes for patients. Ultimately making the vision of precision psychiatry real and eliminating the best of our ability, the evidence gap within psychiatry. Eze Abouzi, Chief Growth Officer for Holmusk. Eze, thank you very much for your insights into mental health today.

Thank you, DNA podcast. Glad to join the conversation and thank you again for having me. It's a lot of fun.

Well, Amar, what did you think? Yeah, it was pretty interesting to understand about the mental health, how big the issues are, and getting more like an in -depth understanding of the different types of psychiatric disorders and the way that the real world data gets collected for those, right? So psychometric measurements that as they talked about, but also the fact that a lot of the data that they collect is the notes, is the notes that the...

physicians are writing and so collecting all of that, putting that together. So that was pretty interesting. But then also I see like a lot more challenges with that, right? Because you have a lot of this what we call unstructured data, right? Text data. So there are definitely much more challenges in terms of analyzing that data and getting the insights from there. Mental health has lagged behind other disease areas when it comes to precision medicine. To what extent do you think AI has a chance to change that and improve care?

I would say we have to see in terms of... so AI definitely can help with the analysis of these notes. As they talked about the NLP, the natural language processing, and how that's helping in gaining insights of this. So I I think the AI definitely can help with that, and especially now with generative AI, where getting insights from texts and content has become much easier than it was before.

Daniel Levine (51:05.901)

So definitely, I think we're going in the right direction. We still have to see about how AI can really come up with the insights that are going to change the treatment, right? So some of the interesting points that he made about like, so looking at the subtypes, right? What are the subtypes of depression? Like, we really don't know. We just think, okay, depression is depression. And clearly a very, very interesting example about lung cancer and breast cancer where they were...

thought of as, or at least from a treatment point of view, they were like homogeneous diseases about a generation ago. They're not anymore. And we're able to go much more into personalization and precision medicine in those. So AI can bring that, but I think getting insights is one thing. I also want to see more about how we can understand these subpopulations better. So I would say both from the generative AI point of view and also from the machine learning point of view, I think there's a lot to be done

to really go deeper into this. I mean, one thing I really like what they're doing at Holmusk is creating this data set of millions of patients. And the big thing for AI or data analytics is that you've got to have good, rich data to be able to come up with new answers, right? So I think this definitely is a step in the right direction

having this many people's data available so that you can actually try to see what the signals are and then try to understand what are the segments of the patients in this. Well, how much of a competitive advantage do you think that provides Holmusk? I think it provides a lot of competitive advantage. I mean, when he talked about, like, there are not many other companies who are really focused on mental health, right? And, I mean, we all know that mental health is a huge issue. And then also getting data for that is...

complicated. But then at the same time, these people are focused in that, putting all this data together, processing all this data together, making sure of course everything is HIPAA compliant or so. But having this kind of data set, they have a huge competitive advantage in the mental health space because I mean, any company, drug development company who would want to develop...

Daniel Levine (53:25.933)

medicines in this they they they have to talk to Holmusk because Holmusk has the real world data of the patients so to understand you know more about these patients and I would say also in terms of like the The the clinical care right? I mean they have so much data that you're gonna I mean they are able to, they will be able to help in terms of how how the physician should even treat the patients. I don't like every should treat the patient so you can get a lot of

a lot of insights from this volume of data. Like a lot of things, AI, the results are going to be very dependent on having the right data. Yeah. Do you think AI can help solve some of the data challenges by helping identify what the right data is? I believe so. I would say with generative AI or so, you can do that. I mean, one thing that he talked about, which is,

a lot of information about, well, okay, well, did the patient try suicide or so? That's in this text, right? I mean, there is usually when you're looking at, when a physician is talking to a patient, they don't ask, okay, well, how many suicide attempts have you made in your life? And like having a box there, okay, two or three. Doesn't happen that way, right? I mean, all of that is information that gets captured in the conversations. And so then making sense of that,

and then coming up with some of these parameters, I think generally AI especially will be very helpful in that. And then also really understanding in terms of the right treatment or even the right populations, I think we're going to have machine learning go through that and identify that. Because a lot of times we don't even know what the right parameters are for mental health, right? Because there is no biopsy, right? So,

you don't know exactly what are the variables that we need to actually measure and then they will be able to segment patients. We don't know that. So I think with AI, if we program it in the right way, we should be able to then try out a lot of different parameters to see what are the ones that are more relevant, what are the ones not. And then based on that, we segment the patients and then we develop drugs for specific areas and then we can use biomarkers or so.

Daniel Levine (55:47.629)

for those patient populations to develop the drugs. And part of the issue that you're speaking to is the complexity of these conditions. It would seem to me that would make it well suited for the application of AI. Is that a fair statement? Is there a relationship between the complexity of problem and the benefits AI can bring to bear? I would say with AI now,

Yes, you can solve more complex problems, but it also again goes back to do we have the right data? Do we have the good data? And so if you have good right data and a lot of data, then AI is able to solve the complex problems. So I would say here, right, I don't see this being like everything being solved in mental health in the next couple of years, right? I think it's gonna take time. We're going the right direction with AI is going to help speed up

the research in mental health. But I would say it's going to be gradual. We will keep getting, we'll achieve some milestones as time goes on. But I think it's going to take time. Mental health is very tough to solve. AI will help with it, but it's going to happen gradually. That would be my prediction. Well, it was really compelling to see what they're doing and it's a great case of how AI is being used. Amar, until next time.

Thank you, Danny.

Daniel Levine (57:20.621)

Thanks again to our sponsor, Agilisium Labs. Life Sciences DNA is a bi -monthly podcast produced by the Levine Media Group with production support from Fullview Media. Be sure to follow us on your preferred podcast platform. Music for this podcast is provided courtesy of the Jonah Levine Collective. We'd love to hear from you. Pop us a note at danny at levinemediagroup .com.

For Life Sciences DNA and Dr. Amar Drawid, I'm Daniel Levine. Thanks for joining us.