William Leonard
Hey, ladies and gentlemen. Welcome back to the Atlanta Startup Podcast. I’m William Leonard, your host, and investor here at Valor Ventures, a leading seed-stage venture capital firm in Atlanta, Georgia. Today, I’m really eager to sit down with Dr. Alemu of TQIntelligence. Dr. Alemu, thanks for joining me today.
Yared Alemu, Ph.D.
Appreciate the privilege.
William Leonard
Awesome. I’ve been doing a lot of research on your background, and I think it’s a really interesting trajectory. I’d love for you to really educate our listeners on how you came to start TQIntelligence and a little bit more into your background.
Yared Alemu, Ph.D.
It’s a good question. I’m a psychologist by training. Most of my work involved children, adolescents, or adults from low-income communities. I run a decent-sized community behavioral health agency, about four thousand patients, almost all of them are children and receiving mental health services through the American system. I got really frustrated, in terms of, there’s no data, right? We don’t know who’s getting better, we don’t know who’s not getting better. We don’t know why they’re not getting better, or we don’t know how long is gonna take for someone to respond to treatment. That led me to TQIntelligence in not just being able to provide data, but also to build a cutting edge technology as it relates to voice analytics and AI to objectively quantify the problem. Mental health is the stepchild of the healthcare system. When you go to your primary care physician for the checkup, they’re just not gonna ask you, like, “How is your cholesterol?” If you say, “I think I’m doing good.” They’re not gonna take that answer, right? They’re gonna do some blood tests. If you have some pain in the back or headaches and all that, maybe they will do a CT scan. And then they will combine what you’re telling them, which is the subjective information and the data from the objective information come up with a diagnosis or treatment plan. We don’t have anything like that in mental health. We listen to you and based on my background perception, I think this is what the problem is. A month from now we may say, “I think I changed my mind. This may be something else.” In the meantime, there’s no way to quantify what’s the baseline severity of this person and then come up with a treatment plan. That’s kind of what led me to TQIntelligence. Poor people get screwed over and over again. The system of care that I’m talking about, it is a $250 billion market for 5 million kids. People always complain about money. This has nothing to do with money. This has to do with transparency, data outcomes, that should be expected, right? Because the private sector has its own quality control. They pay half or less of that, right? If you have Blue Cross, and they pay, like 1/3 of what Medicare is paying, per child per year, right? But they’re very limited, it means someone can find me some sort of a systematic study of thesis outcomes out there anytime. It started with that frustration, on one hand, is just the general maintenance, lack of having instruments to quantify, on the other hand, you got people screwed left and right. When poor kids don’t get quality treatment, they end up showing up at the justice system, showing up in the emergency room, attempting suicide, so the consequences are very severe.
William Leonard
You mentioned an interesting point there. The system of care for underserved underprivileged demographics of people. As you were working at the early stages of your career prior to TQIntelligence being started, was there a particular moment for you where you said, “Okay, enough is enough. I need to go out and build something.” If so, what was that moment for you?
Yared Alemu, Ph.D.
It’s my last job that has probably close to 200 therapists. We’re supervising 4000 or more patients and these insurance companies will come and say this is how much the services that you guys are using. I’m on the other side with no data to balance it with what they’re saying, what data I needed at that point. If you’re complaining about this is how much the services are being utilized, I want to be able to show you the outcome, the quantity of what you have paid for and invested, right? I was in a number of different committees around the state on best practices. I was on three different working groups with what we call the Medicaid Managed Care, which is if you’re in Georgia, your services are managed by an insurance company. For example, Anthem has a medigroup that manages the public sector. In those three working groups, sitting right in the middle of the providers on the side, and appears on the other side, trying to push the conversation into data. And it was not going anywhere, partly because it just does not exist. We’re very subjective, right? I see you today, I come to my computer. I think he’s experiencing some trauma, but I’m not sure what I hear. We write like crazy, right? Most providers don’t use any measurement that has to do with numbers. At the systemic level, that we need to do much better in terms of just other general fundamentals. But that’s more acute when you’re looking at the low-income communities. Because as I said, what if my kid goes to therapy, and then he comes here, and you know, I don’t like this guy, right? I’m the quality control. I have the resources. So I say like, “Okay, let’s find somebody else.” So that’s a privilege, right? I’m the quality control for my family. If you’re going into services, the system that we have right now, a single mother with a kid that is referred to these mental health agencies because he was not doing well in school. One, the family doesn’t choose which agency to go to. They’re just assigned to that once they get there, right? They don’t really have resources, even to say when someone’s not providing good services, or they find out six months to a year that Joe’s not making any progress. When that happens, when the kid comes to services in that system, is usually when things are about to kind of go down. We’re in this kind of crossroads, right? School is trying to kick you out. They’re trying to figure out what to do with you. They send you these alternative schools that are horrible. Quality mental health is a life-and-death situation. It’s not a luxury. Some of the quality services have mental health services, the opportunity to significantly move this individual in the right direction.
William Leonard
I think you hit the nail on the head there. Quality mental health is not a luxury, it’s a necessity. I’d love to kind of pivot here a little bit to the business side of things. You think about TQIntelligence and the kind of the core of the business, right? It’s these voice-based algorithms that you all have, and really pairing that with AI. We’ll get into the AI piece a little bit later in the conversation. But you know, can you talk a little bit about the therapist experience, and then the patient experience as well?
Yared Alemu, Ph.D.
What we also know is that probably 80% or more of the kids who are having these severe mental health issues are driven by trauma. As a trauma, you can start from a generational trauma community that has been experiencing violence and terror over and over again. That gene is passed through generations whether you like it or not. Most of us are walking around on some level of trauma. There’s no way that he can live in a society and not experience some level of trauma. You open the local Channel News, and there are all kinds of craziness that increase my level of anxiety. That stuff shows up as a behavioral problem, it shows up in your suicidal being so sad, it shows up in, you know, having sadness and not having enough energy and cannot concentrate, right. And then you have the trauma in a hidden, so so if you have a kid, that the school is not secure, it’s not safe. The community that you had walking back and forth in school is not safe. You go home, there is scarcity, there’s a lack of resources, right? Every day you have to figure out how you’re going to make it through. A kid coming to services, it’s not just kids from low-income families that experienced trauma, there are kids whose resources come from significant trauma symptoms, because of its emotional, sexual, or physical abuse by the parents. You have to understand from the perspective that we’re going to have to kind of struggle together, right? All the adults that I trusted screwed me, right? The system that’s supposed to protect me, abandoned me. This idea that just because I’m a therapist, are you going to open up? That’s not how it works. You need experienced therapists that can hang on and get connected, and provide treatment. The problem that we have in this system that we’re discussing right now is the therapist in that system of care are the least experienced. They’re just coming out of graduate school, right? They have to get work somewhere so they can collect their clinical hours for three to four years. And then as soon as they’re licensed, most of them are gone to the private sector. You have consistent changes of poorly trained- I don’t want to say poorly trained- when you’re new, you’re new. You have a case that is the most severe, right? And then you have a therapist that’s the least trained. When you talk about the disparity in team and outcomes, it’s not an abstract idea. The disparity is driven by decade after decade, these inconsistencies throughout the system, so it’s not going to change, right? Our effort, is we want to get in the middle of what we call intelligent augmentation. If we provide data, data that matters and that is subjective when the therapist is sitting in front of the kid, and now we’re kind of arming them with what they need to be able to come up with specific treatment plans. It’s not going to change in my lifetime. All these highly trained psychologists and they’re not going to run into the system of care. So this isn’t what you need. You can either complain about it and continue this problem decade after decade. But it’s not just us. There’s not much innovation in the public sector, but there are others coming in. We’re not trying to sell to the therapist, we want to support them with data that matters in real-time, that they can use the data to be able to improve treatment.
William Leonard
I think the framework of bridging the disconnect between the inexperienced therapists, and patients who have really experienced the worst and most severe outcomes is really important. You mentioned that you’re working with youth right now from lower-income communities. Do you see TQIntelligence moving upstream to work with the parents at some point? How do you think about that?
Yared Alemu, Ph.D.
All the kids in our pilot right now, either in low-income communities or it could be in rural Georgia, right? It’s the level of deprivation sometimes even worst. You have to collaborate with the parents. Our workflow or the way that we conduct data, in addition to the voice data that we collect, is that there are specific instruments that we use to ask the parent. “How’s Johnny doing?” Every week, the therapist, as you know in Georgia, a lot of the services are provided in the home or school. Because if you are asked for people to come to your clinic, you’re putting up barriers, they’re not going to come. Georgia’s been progressive in the sense that we’re allowing therapists to see kids in their homes and provide treatment. The therapist as part of the collected data will hand them the instruments that we have. It’s 26 questions, right? We want to make sure that, because the therapist is not there, they’re the ones who sought to deal with this issue on a regular basis. We bring that data, and then what we do, we’ll compare that data, and we’ll compare that data with a therapist, right? The therapist has their own version because there is only a 32% agreement on the perception of the problem between the therapist, the caregiver, and the kid. Everybody has their own misunderstanding, and then hope that everybody’s on the same page. We probe into that. As soon as they submit, we’ll get that data. Here’s the therapist’s data, here’s the caregiver’s data, what are the differences? They don’t have to guess what kind of stuff the parent is experiencing as a result of the child’s behavior. If they’re doing better, they want to know the numbers are moving in the right direction. If it is a high risk, if it’s 65 or above, we’ll consider it to be higher risk, sometimes higher risk for harm to self and harm to others. When you work with poorer people, the parent may not be able to read, or they may be just irritated as hell because they have to deal with this stuff, you have to bring them to the table. You have to make sure you empower them in a way that creates a working relationship. Without that working relationship, it’s very difficult to make progress on the child’s behavior.
William Leonard
As you reflect over the last few years or so, you think about the increase in AI use cases and how it’s becoming more frequent among a plethora of industries, and now, it’s made its way into the mental healthcare space. Why is AI so important in this space now? Do you see it being used in more mental health applications over the next 10 to 20 years or so?
Yared Alemu, Ph.D.
Absolutely. The innovation is exciting, though limited. I was in a meeting with Google and this guy said, “There are about 10,000 people that actually know how to create these algorithms, and there are about 100,000 people that know how to use them.” And so synthesizing is as much as people talk about AI in terms of a use case and hidden that is commercialized and making an impact. There’s a kind of lag, the innovation is not there most of the time, and innovation in healthcare is also limited. Growing, to say the least. And it’s so easy. You can use AI, in this case, to address these decades-old disparities in healthcare, which could be in mental health, right? There’s a potential for misuse. In our case, nobody has the clinical voice samples as an open-source that we can train our model. What we do, we start from scratch by paying attention to bias. For every voice sample that we collect, I label them first. I labeled them, I have the medical record, the mental health records, and I look at that. There is a specific protocol that we have labeled. When I finish labeling them, we send them to other psychologists that will label them. They have nothing on this, they don’t have the name, the age, the gender, no clinical information. All they have are the voice sample that’s anywhere between about 60 to 90 seconds. They’ll hit that and use the same instruments that are used for labeling. That is intended to address what’s called Inter-rater Reliability. As you know, writing the code is not that difficult as really having the proper information to be able to develop this algorithm. We have other psychologists that do this kind of Inter-rater Reliability. What is the percentage of the time that I and the other two people agree that what we’re hitting is what we’re stating here? So between zero and one, zero means complete disagreement, one is perfect agreement. We use the 0.75 or above, and then the other ones that don’t have the disagreement or the agreement are like 0.60 or so, we put that on the side. We use it later on. But for now, training our model requires that level of careful use of data, right? The other biases are unique in the sense that we were collecting these clinical voice samples. Our competitors don’t have access to the level of the system that we have. What they do is they’ll train actors to sound sad or to sound anxious as a way of training the model. The downside of us having those kinds of clinical voice samples, clinical voice samples from probably one of the most severe populations, when it comes to trauma is that we don’t have any nonclinical voice samples. Now we have to go out and collect data from kids with no mental health issues, which is difficult to find, because of COVID, right? Just because they’re not diagnosed, or they live on the other side, doesn’t necessarily mean they’re not experiencing it. We’re struggling to find that data. But we are responsible for making sure that that part of the development of this algorithm, really addresses these issues so we can be able to do these things. Because we have a scientific advisory group by professors from Georgia Tech, we have a person from Google as part of the Google for Startups. He continues for the next year also. He brings in a lot of experience in terms of hobbies. Google is having the same problem that we have. When I talk about voice, we’re not talking about sentimental analysis, we’re not talking about natural language processing, right? The area that we focus on is called speech-emotion recognition. A much more emerging science than natural language processing. This started in 2017 so it was very recent, right? Google’s interest in what we do is partly it’s not a charity, but it is in a way that all this exciting work, and then who’s going to come up with a solution, an algorithm that actually could help someone. That’s kind of our approach when it comes to AI is to make sure that we have the right people on the table, right? You hear a lot about bias these days in terms of AI and all this stuff. Despite having people of color, women, and all that in our team, we still have to be very careful in terms of how we train our model.
William Leonard
As you’re building, it sounds like you’re seeing, and getting a lot of validation from outside third parties about your approach to leveraging AI to help reduce bias and just get a better state of mental health for your patients. I’m really curious about that. As we’re wrapping up the conversation here, you and the TQIntelligence team are really on the front lines in this space and taking a macro look at mental health. Are you seeing any trends in this space that you think are really likely to evolve and insert themselves over this next decade? I say that with a little bit of context, I don’t know if you’ve seen the Olympics this week, but Simone Biles pulled out of an event, and due to mental health reasons. We’ve seen Naomi Osaka as well. Now we’re seeing a lot of companies and corporations providing employees with mental health days or even mandating these mental health days. We’re seeing a lot of activity from a macro level here. I’m wondering, are there any interesting trends that you can point to, or really speak to that you think will emerge over the next 5 to 10 years?
Yared Alemu, Ph.D.
The Simon Biles issue is heartbreaking. Watching her and I was reading about her background, this is kind of been the downside to being a psychologist, there has to be some sort of reason. And trying to kind of explain, it doesn’t mean that we’re anywhere near what the actual explanation is, but you see it commonly. We call these adverse childhood experiences, right? It’s about 10 questions on adverse childhood experiences, the most scientific measurements of trauma. That includes emotional, physical, sexual abuse, and includes parental incarceration, parental substance abuse, neglect. If you have a score of four or above, life is going to be very, very difficult, right? I was kind of listening to her experience being in foster care, and all those things. She’s done a marvelous job to turn that thing around, like to really take the pain, and make something from it. But the problem with that is, at some point, it will catch up with you. And for her, it showed up this week, right? It doesn’t mean that it hasn’t shown up before, it’s just that it becomes more public. She, unfortunately, cannot even process that privately, because everybody has to have several opinions on that. The time to mental health has kind of blocked from a historical perspective, right? It’s kind of started Managed Care, how these insurance companies end up going to approve, and all that. We got to parity, to make benefits for physical houses as much as for mental health, right? If you have insurance, they should not discriminate that they’re not going to pay for mental health, but they’re going to pay for you to go see your doctor. We’ve dealt with that in the last year and a half. We’re at the stage of access, right? What COVID did is remove the legislative barriers, the payment barriers, the technology barriers for virtual therapy. All these companies are merging and raising a significant amount of money. You see Amazon getting into the game, you have Walmart getting into the game, you have Optum getting into the game. Everybody’s rushing through this right, which is great. We’re beginning to work through the access issue. The problem is, if you live in rural Georgia, your connectivity does not support it. If you’re a child with a history of trauma, and you needed to speak to your therapist privately, and if there’s one phone for the whole house? All these logistics that people assumed, just like you and I have this kind of privilege in our office, but I think that should be addressed. What we are moving to, a company like us is quality. You can’t just open access, right? We’re battling the same issue. Just because you’ve seen someone virtually doesn’t mean that you’re benefiting from treatment. We want to make sure that we can be able to plug our algorithm into these tele-mental health providers as they talk to the patients, they can actually see whether the person is getting better or not from prescribing medication. To have a baseline or come back and see you in two to three weeks. We want to be able to make it easy for them to be able to determine. We have to come back to quality. There’s a finite number of finances, right? Mental health is seen as a disease of young people, before it’s ever become a problem for older people, it’s a problem. You cannot really address poverty without really addressing mental health and physical reason. Trauma is a property of physical emotional problems. If you want someone to break the cycle, you have to provide them with quality services. That’s what we are. We think we’re in the right position to be able to kind of on this access to mature. Mature in terms of all the small pockets that are receiving those benefits, and then we’re going to come behind that and say, “If someone pays me this amount of money, you have to demonstrate your value, right?” You can’t just see someone and charge for that. You have to be able to see someone, demonstrate that the service that you provide is beneficial, and then get paid. Even with different amounts of money, not everybody should get paid the same amount of money. A hierarchy that encourages people to really focus on quality services. But the current system doesn’t incentivize that. Generally, $80 per hour? Part of these healthcare companies does not have the data to be able to create this differential pain system to be able to promote quality.
William Leonard
I think you hit the nail on the head there, again. Quality is everything when we think about mental health care and really access to health as well. I think this was a really informative conversation. I think this is a space that is seeing a lot of innovation and a lot of startups trying to solve a multitude of problems. I think your approach to really serving youth and lower-income communities is really interesting, and really appreciate all the work that you and the TQIntelligence team are doing there.
Yared Alemu, Ph.D.
I just encourage more innovation. We’re not a nonprofit, right? This is not charity. We’re as greedy as any other startup. But what you have is, in this case, you have two things that are one, the market is sizable enough and does enough. The opportunity to make an impact is equally available. We don’t have to pick one or the other. We’re talking to investors right now and they’re looking at they don’t have any money, so how are you going to charge kids in low income? They don’t know whether it’s going to show up. But there is really an opportunity right now to be able to, not just in mental health, but in other diseases, diabetes, all these chronic diseases, that these innovations are that we can push, but we really can push the quality issue using these digital house interventions and make an impact and make money at the same time.
William Leonard
Dr. Alemu, I appreciate you joining me today and look forward to continued innovation within this space that you and the TQIntelligence team are building and how can some of our listeners get in contact with you if there’s interest in the product?
Yared Alemu, Ph.D.
TQIntelligence.com, yalemu@tqintelligence.com. We’d be happy to chat and see if we have something that we can be able to collaborate on.
William Leonard
Perfect. Thank you again, Dr. Alemu.
Yared Alemu, Ph.D.
Thank you.
Lisa
We’re thrilled to have you as an Atlanta Startup Podcast listener to help you get the most out of the experience. Let me invite you to three insider opportunities from our host Valor Ventures. First, want to be a guest on this amazing show. Reach out to our booking team at atlantastartuppodcast.com. Click on booking, It’s a no-brainer from there. Are you raising a seed round? Valor definitely wants to hear from you. Share your startup story at valor.vc/pitch. Are you a woman or minority-led startup valor sister program? The Startup Runway Foundation gives away grants to promising startups led by underrepresented founders. The mission of the Startup Runway Foundation is connecting underrepresented founders to their first investors. Startup runway finalists have raised over $40 million. See if you qualify for one of these amazing grants at startuprunway.org. You can also sign up for our next showcase for free there. Let me let you go today with a shout-out to Startup Runway presenting sponsor Cox Enterprises and to our founding partners, American Family Institute, Truist, Georgia Power, Avanta Ventures, and Innovators Legal. These great organizations make Startup Runway possible. Thanks for listening today and see you back next week.
