Voice in Healthcare- Dr. Yared Alemu, CEO, TQIntelligence- Project Voice X

This is the transcript for "Voice in Healthcare," presented by Dr. Yared Alemu, CEO at TQIntelligence, presented on day one of Project Voice X. The transcript below has been modified by the Deepgram team for readability as a blog post, but the original Deepgram ASR-generated transcript was 94% accurate. Features like diarization, custom vocabulary (keyword boosting), redaction, punctuation, profanity filtering and numeral formatting are all available through Deepgram's API. If you want to see if Deepgram is right for your use case, contact us.

[Yared Alemu:] Hi. Good afternoon, everybody. It's a a bit unfair to go after Henry. That's... I should have been the warm-up part, and then and then he should go. So I'm Yared Alemu. This is not this is not by design. So so if you guys are impressed about the the wording underneath, just... I'm not sure what happened. So I'm a psychologist by training, and we are a start-up at Georgia Tech at the Advanced Technology Development Center, and our work is partly as... is funded by the grant from National Science Foundation. That's phase one and phase two.

So we're hoping that by the time we finish, we will have some of the tax money, that tune of about one point five million dollars. So... but we're very... you know, we're taking care of your money, so no worries. We're not we're not wasting your... it's just a... we're also partly funded by Google and other... our active collaborators. So one of the things that have a pretty dramatic impact on on all human beings, right, all included, is things that happen very early in life.

And so a while ago, the CDC and Kaiser had a bunch of data, and the... so they've done this long, continual studies, right, and and and looked at other things that happen early in life. We... it could be trauma, but it is a more expanded version of trauma because it includes emotional, physical, sexual abuse, includes neglect, parental, substance abuse, and so so a whole range of them. Do they have any impact, the person that's transitioning into or not transitioning to becoming an adult?

The answer is not only us, but it's devastating. Yeah. So more especially that if you score... you guys can go to our website. If you score four or above, right, the consequences include a twenty years reduction of your life expectancy in addition to multiple mental health as well as other chronic diseases. Right? And there is a reason why. And the reason is trauma is not just an emotional experience. Trauma is primarily a physiological experience. There is a net changes in a number of your organs that directly negatively impacts not just who you are, but also who you becoming. It's it's very difficult for kids who've been through trauma to transition into becoming a a productive member of society. Right? But because we can't see it... right? So so perhaps some sort of an injury that has impacted me from working, if it is visible, people could empathize with that. But if I have an injury, right, the injury is invisible, it it is on multiple of my organs, right, it is... you know, for a group of kids, that becomes... they're externalizing the suffering. Right? So they share that with you. Right? And and and they share that freely. Right? So if you're in a classroom, I'm sure you recognize who the... who they are. They have contact with, you know, the legal system, substance abuse. And then there's another group of kids. They're internalizing it. Right? So that becomes depression, anxiety, self-harm behavior, suicidality.

And so so so this is something that, in terms of the cost... right? California decided to account for what is the cost to our system as for... when I say system, the state of California, the cost of ace. Right? The cost is about more than a hundred billion dollars in two thousand thirteen. Hundred billion dollars. Right? So that cost is health care. Right? That cost is incarceration. That cost is emergency room visits. That cost was child protective services. So not addressing this, right, mean, the... I... you know, I can't... I grew up poor. Right? I I don't know... I would rather not been born where I was born. Right? This is not a choice. There's nothing outside of being poor. I mean, people sometimes romanticize and write songs and write books about, you know you know, that made me a better person. No. It's not... you know?

I could have been a a much better person if I did not grew up poor. Right? So what are those things that come with being poor is is that's kinda, passive, traumatic experiences, sometimes, active trauma... traumatic experiences. They will add up and doing a number on a person. I am a byproduct of excellent mental health services. Right? Got lucky from the get-go.

But there is a group of people, right, there's a group of kids that that is not the case. So there's about forty five million kids that are currently having their health insurance through Medicaid. Right? Forty five million kids. So to qualify for that, right, you have to make, you know, as a family of two, twelve, thirteen thousand dollars or less. So you know what it comes with that kind of income. Right? Not much good happens.

Among those, there's five million of them, right, receiving mental health services currently, five million, at the tune of two hundred fifty billion dollars a year. That is one of the most expensive services for any person in this country. When I say expensive, right, the money it takes to educate a child per year, right, the amount Blue Cross Blue Shield or these private companies spent per patient per year, if you take all those things, this is the most expensive and the least effective.

There is absolutely no data right now that, despite this level of investment, there are improvement in outcomes. It is absolutely ridiculous that you pour into that kind of money into a system and you're not asking for data whether the chart is actually... when I say benefit... kids have three jobs. I have two kids. Right? They try not to do those three job, but have three jobs. One is school. Right? Everybody doesn't to be... doesn't have to be an A/B student, but they're the best of their abilities. Right? Two is the relationship in the community, school. Right? And then three is at home, their relationship with their... other members with the family. Three. Right? So when we say that improving outcomes, we're not inventing something special. We're saying they can be able to do these three things. Right? Or or or they will do better on these three things six months from now if I spent, right, five, six thousand dollars on treatment. Right? So so so the the problem what we have and the why that is, right, is we don't really have we don't really have an objective way of measuring and measuring quickly the severity. Right?

We don't really have an effective way to be able to track outcomes. Right? When I say track outcomes, I'm not talking about this pre-imposed, some fake stuff that these providers do. I'm talking about systematically tracking whether someone is getting better or not. Right? And then and then the other one is changing the model of treatment from episodic. Right? I'm gonna see you today. Until I see you next time. God help you. Right? As if like somehow in between, right, nothing happens. So we don't really have a good way of monitoring when they're not... when not not seen by treatment providers. Right?

Those three things primarily account for why we have significant disparities in mental health treatment. Not just for kids from low-income communities, but people who receiving mental services at all. So what we're doing is... what we've been doing is taking about a forty five to ninety second voice sample and and being able to detect, right, being able to identify the severity, and being able to track whether someone is getting better or not. And so we have pilot sites in three different places currently. Hundred percent of the kids that are now pilot sites are kids from low-income communities. We just started a pilot site in North Carolina with the most severe patient population. These are kids that are... just came out of the hospital, right, trying to keep them from going back to the hospital because that's what happens when you when you don't have good outpatient services.

People show up in emergency rooms asking for medication refills. Right? People show up in patient hospitalizations because there's nobody else to be able to monitor their suicide behaviors. Right? They're seeing four, five, six times a week by provider. And so we're collecting data. When we collect the voice sample, this is kinda what makes us different than most other voice companies. For every voice sample we collect, we know the diagnosis. We know what their medication is on. We know for how long they've been in treatment, and then we collect our own data, right, including collecting the ace, the PHQ nine, and another one, the SFSS. Right?

Whenever you work with kids, you have to ask the parents how the hell they're doing. Right? If you don't ask the parents... so so so the... if you look at the research, there's about a thirty two percent agreement between a mentalist provider and a parent on the perception of how bad the problem is. Just thirty two percent. Right? It's very difficult to make progress working with kids without closing that gap, right, closing the gap between the perception between the parent and the provider. So that's why we intentionally select an instruments that allow us to engage parents. Right? Just because you're poor doesn't mean that that you're not gonna be able to engage them in treatment.

There is no reason, right, for... even if they don't read. Right? Because we have... we... we're in some rural areas in Georgia. Right? The level of deprivation and poverty in those areas are ridiculous. Very limited connectivity. Right? The idea that telemental health services using Zoom is gonna reach there is very unlikely. Right? 'Cause Zoom is not really intended to be able to, you know, be used in those settings that require this... you know, a whole set of these connectivity issues.

So so we have to... so so all that is intended all that is intended is to be able to build our model, right, and and build it from the ground up, because there's no clinical voice samples for just young people in general. Right? So we have adult voice samples, you know, be... but in terms of having voice samples for this this specific population, it does not exist. So what we do, we collect them through... having all these, we individually label the voice samples. And then we have outside anonymous people that that level the voice samples. These are psychologists. Right?

And then we have somebody else that does what they call inter-rater reliability, which means our... between these three or four people that are leveling this voice sample, what the what's the level of agreement? Right? If it is one, that means perfect. Right? And then if it's not any... anywhere about seven... in a point seven or less, we put that voice on on a side. If it's it's point seven or above, but voice sample is included in terms of developing your model. Right? So so it's a... it's data intensive. It's resource intensive. Right? And and it's not quite friendly for for investors to be engaged in in a process that is... so you deal with a high risk. Right?

So the the area of speech emotion recognition, it's... as as as an emerging science, it's a high risk. Working with and developing solutions for poor people is considered a high risk. Right? But the issue is the... if you really wanna make an impact and go where the money is, right, you go to those areas. Right? Mean... so there's... nobody pays more for mental health than Medicaid, by far combined. Right? So there is... you know, we're not as greedy as any other start-up, but we also see an opportunity to be able to impact an area that attracts very limited innovation. So so if you're thinking about this area, please jump in because, you know, it's not... there's no... we don't see anything getting better for the foreseeable future. Right?

We don't see a highly trained group of psychotherapist, right, decided to enter this area. Right? And, by and large, part of the reason there are disparity is because the people who come in in those areas to provide services are new therapists that just graduating. Right? They don't even know what they're going. You don't let anyone who doesn't have his license, right, provide any services for you or for you child. Don't. I mean, you know, you don't go to someone who just graduated from from medical school, right, to get a prescription for medication. They have to go to residency.

I mean, there's a whole part of things that are necessary before they're capable of making a decision on your behalf. Mental health should not be any different. Right? People providing services in the publicly funded system, by and large, are unlicensed. Right? They're in our head because they get a lot of hours. Right? So two or three years, they get the license, and they're out the door. So so so so technology, in this case, is is... that we wanna... this is kinda been telling and augmentation that we we cannot solve the problem of people coming in that are unlicensed. But what we can do is we can come up with solutions, in this case, algorithms that allows them to be able to identify, identify it early, right, and then be able to come up with a treatment plan and then follow up to see whether that treatment... all that is is is based on data. Right? So I think I have one more minute. I'm done. Thank you very much, everybody.

If you have any feedback about this post, or anything else around Deepgram, we'd love to hear from you. Please let us know in our GitHub discussions .