data.bythebay.io: Sai Moturu, Ginger.io: a Data-Driven Mental Health Care Provider
Recording: data.bythebay.io: Sai Moturu, Ginger.io: a Data-Driven Mental Health Care Provider
title of the talk is a datadriven mental health care provider. So, Ginger.io is a mental health care provider that really uses data to integrate um into care. So today I'll talk a little bit about what we're doing um in terms of the care delivery and then get into how data is helping us deliver better care. So there's a lot of recent headlines around mental health. But this is not a new problem. This has existed for a while but we tend to not focus on it or think about it as much. But we should care because uh if you think about the area where we are the top mental the top spend for health care is maternity and then next is mental health. So this is a problem that actually exists all around us
We may not realize it but it's happening everywhere and it persists year after year. So we really need to find a solution to this problem of mental health. And it's not just around about the identification or the lack of it. There's a lot of other challenges. Wait times to get care for mental health are huge. The average weight time across the US is about 45 uh 45 days to get mental health care. In rural areas, it can take up to several months. And beyond that, there's the stigma around mental health
No one wants to really think that this is they have a problem when it comes to mental health. Also, the resources today are too narrowly focused. Most of the resources that exist are focused on mental illness, the chronic condition for depression, anxiety. That is something that people are living with and they're trying to keep in check. So, you have a couple of solutions there. There's therapy where you can go try to identify your sol uh problems, try to understand how you can cope with that. And then there's psychiatry where you're trying to get medications uh to help with your problem. But there's a whole range of other problems that most of us go through at some point in our lives
Stress, anxiety, loneliness, frustration. Let's say you lost a loved one, right? That's not easy to recover from for many folks. And there is no support in this area today. And a lot of us don't really identify this as an issue. We say, "Oh, I didn't get enough sleep yesterday. Oh, I'm just frustrated at work." We brush this and say that there's nothing wrong with us. And as we talked about on the mental illness side, there's the inconvenience, there's the wait times, there's the stigma, and there's no support on the emotional challenges. So at Ginger.io, IO what we're doing is we're trying to build a new mental health solution right so in our world data scientists and clinicians are working together and they are behind all the care that is delivered to user and now think about it this is not something that happens uh in the traditional healthare system or you know most places where care is delivered even if you have tools to gather data and generate insights that is separated from the person who's delivering the care
The benefit that we have by having the data scientists and clinicians under the same roof is that we can build this integrated solution that learns from what the clinicians know that learns from what the data scientists know and puts that together in a way that care the the care providers can actually use. And beyond that, we're trying to solve some of these challenges that I I've talked about, trying to provide care wherever it's needed. So, our care is delivered through smartphone. So, Wednesday night at 8:00 p.m., you can talk to someone and get some help. And so, the idea is that this is a very convenient and private way. So, we might we'll probably be able to get over the stigma barrier because this is from your phone. This is in your pocket. It's available all the time
No one gets to look at it. So we hope that we can actually reach a much wider population and our solution also has some of the same layers as other care providers. The clinic down the street, we have therapists, we have psychiatrists, but we have a couple of additional layers that helps us uh add onto that solution. So we have personal coaches. So if you you know download the app and install it today, you get a personal coach. This is your person. you can talk to this person anytime and they will help you navigate through your journey and we have a lot of self-care tools within the app that also helps build the resilience uh within the users and so our goal is really to help users begin on their unique path to wellness and there's a couple of words here that are important path is really important because this is a journey especially for the chronic condition this is something that people are living through and everyone goes through this journey very differently. So it's unique and so our goal is to actually leverage this data to provide care that helps them on that journey
And beyond getting them better, we really want to keep them better. If you think about the traditional health care system, you typically go in when you're sick. With mental health, you go in when you're really sick. And this was something that I never realized, but really sick for depression means that you can be hospitalized for several days or weeks, right? So we never want people to get there once they're on our system. So the idea is that we'll try to keep them at that level. So what does the patient need? We talked about easy access on your phone anytime, right level of care. You need a coach for, you know, you have a simple problem, you're stressed about work, you can talk to a coach. you don't need to go to a therapist who's very costly, right? You have more severe condition, you go to therapist
You need medications, you can go to a psychiatrist. And the two pieces after that are really where data plays a huge role, we're trying to do proactive outreach, which means that even when you are not reaching out for help, we are reaching out to you to see if we can support you. Right? So with depression and anxiety and these mental health conditions, people are not necessarily always comfortable with reaching out and they may not be able to communicate. But we have the data to understand how someone's doing. And that's what we used to leverage to say, coach, reach out to this person because they don't seem to be sleeping well this week. And then we have personalized programs that adapt to you. The the thing here is everyone requires different level of support, different types of care and we're trying to use data to understand how we can build that and even for the same individual the type of care that they need at different points is different. And so data is really driving a lot of this
So how does this work? download the app, you get connected to a personal coach and you you've gotten started on your journey and the coach will direct you to the right level of care whether it's tools within the app, whether it's the coaching itself, whether it's therapy or psychiatry. So, let's get into the data portion, right? So, we talked about delivering care that adapts to you and data is really the key piece here that allows us to do that. So we're trying to provide the right care at the right time whether you're not whether or not you're asking for it. To give you a little context of what we mean by data, we have the digital side of our uh solution which is the app on your smartphone. We have the human layer with coaches, therapists, psychiatrists and we gather data from both of these people. Uh on the digital side, we gather passive behavioral data from the smartphone, communication movement, we gather self-reports from the patient on how they think they're doing. We have information from all the interactions that they have with their coaches and therapists. Now, we put all of this together and build solutions that allow our care providers to make decisions
So before we go into what decisions need to be made, you need to understand what a care provider needs to know. So a care provider wants to know the patient's state. Is this person depressed? Has he been depressed for a while or is this new? What has been that person's journey over the last few months? Do we know what their symptoms are? This is sleep an issue? Is lethargy an issue? And there's various symptoms associated with the different mental health conditions. And when we're delivering care, you really want to understand the outcomes. Are we actually able to get these people better? And so all of this information is what the care provider needs to know. And we are using data to target each of those elements. And so we continuously monitor all our patients and the state that they're in. So we know how they've been doing in the past, we know how they're doing now, and we know or we try to predict how they will do moving on
And that's where some of the proactive stuff comes in. So data is really driving the care choice. So here's some examples of the data and how we're use leveraging that. We have movement information. Based on that, we can try to understand if the person is isolating this themselves, right? This is something that's very common with depression. We can understand if someone's lethargic based on their actions. We can try to predict sleep. Are you know are they having disturbed sleep? Is there insomnia? Communication
Again, people tend to isolate themselves and change in patterns gives us a lot of good information. So these are some of the things around the symptoms that we're doing. And then we also build predictive models based on all of this data to understand the overall state and forecast how they may do. And all of this analytics is really driving care decisions. So these are some examples of where data is driving those care decisions. Personalized care plans, right? We have self-care tools. We have what the coach is doing, what the therapist is doing. We put those all into a care plan for the user that is personalized based on all the data that we're gathering
And that plan adapts as we uncover new information. So let's say as a user you came in, you're having, you know, trouble with sleep and we're helping you out with that. A few days later, we realize you have a substance abuse problem. That takes priority. So your plan adapts based on the information. Coaches can talk to users, right? So it seems like now they can talk to the user so they probably know everything. But that's not really true because users, especially people with mental health issues, don't always open up as easily. So there's been a lot of examples where the data that the coaches have available have allowed them to ask the right questions to get them to open up and then provide them with the care that they need
Escalation of care is another one. Right? If we know that this person has been on this plan, they haven't been doing as well as we'd hoped, then we need to figure out what else needs to happen, right? So, if you come in and you, you know, you're just having trouble with sleep, you're not necessarily depressed, coach can help. If you move to a state where there's more severe depression, we want to escalate you to a therapist because that's the kind of help you need. Let's say that has been working to some extent, but it's not really getting you to where you need to be. Then we'll escalate to your psychiatrist to see if meds can help. And proactive outreach is where we're trying to predict what the patient feel is like today without having any reports from them, any conversations with the coach, right? Or how they may do in the future. So if we can identify people in the early stages of these episodes, they'll never get to that severe state where they have to go in and get hospitalized, right? So that's really uh the core goal of proactive outreach. You want to keep people healthy, right? And coaches can do this and understand and intervene as required
So data is really supercharging the care that we deliver. Data is allowing us to build a unique solution that does not exist anywhere else. And what we really want to do is deliver outcomes much better than anyone else. We want to get people much healthier than they can be with any of the other solutions out there. And we have strong data advantage. We screened over half a million people. We gathered over a billion data points. And we have over 1.3 million self-reports, which means we have labels for all our data
and we're using that to build our predictive models and we've gathered several years of evidence. Uh the precursor to Ginger IO was the work that our co-founder Anmol and I did when we were at the MIT Media Lab. So we have a few academic publications from back then that really got us started on this path. But over the last few years, we've worked with over 40 medical institutions across the country and we worked with eight of the top 10 academic medical centers. So what we've been doing is really gathering this data, understanding the value, building the evidence, but also in working with all these care providers, we were able to understand how care is delivered, where the gaps are, where we can actually leverage our data to deliver better care. And that's what our solution is really built on. So without data, the care that we can provide is like an ill-fitting suit, right? It does the basics. You know, the suit still covers your body, but it's not really doing the job in the best possible manner
In our solution, we really integrated data to get to this kind of state where, you know, it's a perfectly fitting suit. So, data really helps us personalize, adapt, and provide the kind of care that people uh really need to get to be their best selves. And that's the end of my talk. Uh I'll open it up for questions now. Yes. Sorry if I missed this. How are you measuring success? So success is measured by the outcomes. Are people getting better or give me an example of an outcome
So for example, depression. If you think about depression and how traditional care is delivered around depression, there's various protocols around this. If you take one of the common ones, there's a self-report called PHK9. It goes over the symptoms of depression that are called out in the diagnostic manuals and that's what people use to say this is where the patient started and over time this is what happened. So if you go into a clinic where let's say there's the impact protocol which is uh a famous one for depression they're trying to get you to 50% of your score in 12 weeks right that's the goal that they're setting for stem cells and we're doing similar things we also using the PHQ9 to judge how well we're doing so what's an example of a data element that predicts outcome so we are using so some of the things that I've showed here. So, we have a bunch of different information that we're gathering, passive behavioral data from the phones. We're trying to really use that to either predict symptoms or predict overall state. So like a typical machine learning models we have uh you know we have a bunch of features that go in and we have outcome labels from all the self reports that we've gathered over time that I was mentioning here
So we have data where we know what the patient state is and we have all these variables and we built models based on that information and use that to predict for all new users. and what's your control? So in this case, it's not about a control because we've gathered all this data and built these models, right? You we may use uh you know cross validation and things like that when we're building our models to make sure that it's not overfitting and things like that. But in some of the studies that we've done with our partners, we've had control groups and we've been able to show that our uh our solution is actually better. There's a few publications out there right now and there's more uh more in the works. Well, thank you for a great talk. Um and if anyone else has any questions, uh hopefully Sure you'll be around uh at lunch and uh Yeah. Yeah. [Music] [Applause]