Video: Scaling Measurement-Based Care with AI: Lessons from Headspace and Brellium | Duration: 1812s | Summary: Scaling Measurement-Based Care with AI: Lessons from Headspace and Brellium | Chapters: Introducing Headspace's Approach (8.4s), Measurement-Based Care Scaling (129.94499s), AI-Powered Quality Tracking (275.025s), Quality and Culture (435.925s), AI Implementation Strategies (658.555s), AI Implementation Strategies (908.52496s), Implementation Challenges Addressed (1097.51s), Future of AI (1236.4401s)
Transcript for "Scaling Measurement-Based Care with AI: Lessons from Headspace and Brellium": Hey. Hey, everyone. Appreciate everyone jumping on our webinar today. Today, we have an exciting one. I am Zach. I am the CEO and one of the cofounders of Brellium. But more importantly, today, we have Kenley from Headspace joining us to speak about she how she and her team are using AI to schedule, to, scale measurement based care within the business. Cammy, it's a pleasure having you on board. Thanks for joining us. You're welcome. Thanks for having me, Zach. It is a pleasure to be here. Awesome. Awesome. Do you mind, giving a a brief intro on yourself, focusing on where you spend your time mostly at Headspace? And then we can jump into things from there. That sounds great. So I am a clinical psychologist by training. And at Headspace, my title is senior director of quality assurance and training. And what that really means is I oversee a team that is responsible for ensuring that the care we provide throughout our scope of services, so coaching, therapy, psychiatry is really high quality. Know who are writing feedback for our providers, and then our providers are well trained and that they have everything they need to do their job as well, and they also have professional development and growth opportunities within our platform. And so I really see those two things as going hand in hand and being very closely tied together. So I really my team is just doing incredible work to make sure that we're delivering the best care that we can here at Headspace. That makes sense. Mhmm. And so I know that that you guys are are known for being some of the best in the business for focusing on just that quality of care. And quality of care is notoriously really hard to get just right, which I think brings us into the measurement based part of the conversation to today. How do you think about scaling quality of care, and how do you think about what actually quality care looks like? Yeah. It's a great question. So, you know, we, as you mentioned, practice measurement based care here at Headspace. And, really, what that is is just a process of using member data throughout the course of treatment to plan and direct care. At its very core, MVC is what allows us to ensure that our members are getting better and that the care we're providing is effective. So it's about using assessment data at the outset of treatment to develop an appropriate treatment plan based on that member, individual member and their needs, and then continuing to collect data throughout the course of that treatment to ensure that progress is being made. And if not, to really recalibrate and change direction of of care or make adjustments to better support that member's needs, which happens from time to time. It's why measurement based care and having that data is really important. And at Headspace, we use the PHQ nine and GAD seven as our foundational assessments, to track member data. So we, you know, track those at an aggregate level, and we also support our providers in using them in their work directly with members to track quality and ensure that their members are improving. We also use some other instruments, like we use a perceived stress scale for, our more subclinical members, and then know, providers also use other types of data in their care. So for example, suds scale or subjective units of distress to assess, you know, a member's own experience around maybe a particular component or symptom that they're presenting with. And it's really important to us that both providers and members are aware and are discussing kinda what data points are gonna be tracked and discussed throughout care so that the the member themselves can also play an active role in monitoring their progress and improvement, in treatment. That makes a ton of sense. And so I imagine that Headspace in the super early days when you were super small, it was a little bit more straightforward to be able to track different indicators between patient sessions. But as you scaled up into a much larger business now, I'd imagine that certain things change when you need to be able to scale how you're tracking quality of care between certain sessions. Can you talk a bit about how you've actually scaled that as the business has grown into what Headspace is today? Yeah. Yeah. It's a great question. So, you know, a lot of tracking quality can be a pretty manual process. And one of the ways we found to really support scale is the use of AI tooling like Brellium. So in the past, for example, Brellium is a great example of of scale, we had humans reviewing clinical note documentation. So clinicians who had expertise were reviewing therapy notes, and there's only so many notes that humans can review. And as the network grows, as the number of sessions expands, it's just impossible to review all of them and to to have, you know, a a complete picture of of what documentation looks like and what care is looking like, with our members. So we partnered with Bralium to be able to move from, like, 2% of note notes being reviewed to a % of notes being reviewed, which is incredibly helpful in allowing us, one, to track trends across our providers. So we, as a team, understand where do we have opportunities for growth, where we maybe need to do some more training or provide some additional support to our network, and to be able to ensure that we are, you know, meeting compliance and quality expectations more holistically than we were able to do before. It's also been really helpful for for providers because we're able to provide them with weekly feedback around how they're doing, areas for improvement. It's super concise and clear. So we've also seen much faster progress in terms of responsiveness to areas for improvement than we were able to do when we used humans. So we're continuing to, like, think about scale in that way. Like, how do we use humans in the areas where we need their expertise, but how do we implement other tools that allow for efficiency, that just isn't possible with with humans in the in the next cell. Yeah. That makes a lot of sense. And and certainly understand the value in improving quality when going from taking a look at 2% of your clinical notes in seeing that to a %. We touched on this a little bit, but the important part is first reviewing and finding out where we're doing well and where we're not doing so well and how we can improve. And the second part is actually putting that into practice and instructing our providers on how to get better over time. How do you think about giving feedback to those providers? You mentioned the the weekly notifications. How do you make sure that that feedback is taken correctly by providers, is clear and actionable, and actually helps your provider base improve quality week over week and month over month? Yeah. So we do that in a couple of ways. One, we work really closely with the Breleum team to, you know, be frequently making adjustments to email templates have been amazing to work with in terms of really tailoring the messaging and even the the amount of content within one of those email notifications to the specific needs of our providers in our system. And so we're frequently looking at that and making adjustments as needed. We also provide other avenues for training and information. So we send a weekly, email to all of our providers that has information that they need, has micro learnings. So we are integrating, you know, themes that we're seeing come up on the quality side into those and really using that to guide, you know, where are we providing additional information or training materials to our provider network. We also offer, you know, clinical consultation opportunities or opportunities to meet one on one when providers have questions or need just further support or depth of information around a particular topic. So we really try and take a broad approach to making sure it's easy to access the information they need and that that quality is staying top of mind. That makes a lot of sense. And I know that when you and I have spoken offline, is that your team has mentioned that Headspace really prioritizes culturally responsive care, which is not always the norm, across, networks, across behavioral health. How does your team ensure that that prioritization continues as your network has scaled? Yeah. Well, so I mentioned the email. So we culture responsive care, we just embed into everything that we do. So it's in it's part of every live training, every continuing education course. We're taking a culturally responsive lens and creating that that material. We also, in the weekly email, start with some knowledge or information or sharing around some culturally responsive topic. It could be, you know, a a history month that's happening and some information. It could be sharing some content within our app, reflective question, but really making sure that that is at the forefront of provider's minds and that everything they were seeing from us has that lens to it. We also have a field within our clinical note where a provider that is about culturally responsive care, and the expectation is that providers are using that field to talk about salient member identities, how they're bringing that into the treatments that they're providing, and how they're they're integrating and taking a a responsive approach to that in their care. So that's one of the fields that we are, QA ing in our quality assurance process and ensuring that providers are using that field appropriately, and then, therefore, continuing to to think about culturally responsive care member identity in the work that they're providing. That makes perfect sense. Shifting gears a little bit and focusing a bit more on the the technology side of the house. Obviously, AI is hot. It's become a bit of a buzzword. And especially when applying AI in health care, we run into the recurring issue of making sure that folks are comfortable with AI's implementation. How did you think through with Breleum and maybe even other AI tools that you're currently leveraging? One, getting your operation staff comfortable with implementation. And then two, also making sure that your providers are on board with new products that deeply affect their day to day work. Yeah. So, you know, I think probably goes without saying that there's a really broad range of perspectives about AI, and some folks are really skeptical or even worried about the impact it may have both from, like, a member security perspective as well as, I think, a common fear is that it will replace the work of, of care providers. And so we are really intentional about messaging early and often about AI, how we're using it, and then our approach at Headspace is to develop and use AI tools in an ethical and rigorous way that are backed by clinical experts and that are you know, protect our member privacy. So anytime we're developing AI, our legal, compliance, security teams are very closely involved in the discovery and implementation process to ensure that anything we're implementing meets our our safety and security standards and privacy expectations. So, you know, you talk about implementation and everything is it's a very cross functional, partnership to be able to implement any of these AI tools. And so, you know, we have deep relationships with cross functional partners, and then we're continuing to communicate to our providers about what we're bringing on board, why we're doing that, and how that's gonna help them or improve efficiency for the providers themselves and are really clear again about our our messaging around where we're using these tools in an ethical and rigorous way to create efficiency but not to replace. That makes a lot of sense. Are there so I know that that there are other folks in the audience today that might work elsewhere in behavioral health or might work in health care in adjacent specialty. And most health care institutions, what we've seen, are exploring or early stage of implementing new AI tools to help improve quality of care and improve patient outcomes. Are there any implementations, maybe apart from Brelean, that your team has found particularly successful, either one, from a quality standpoint or two, from an internal, employee standpoint within your team? Yeah. Yes. There's a couple exam of examples that come to mind. One is we've partnered with, another AI platform called Virtual Sapiens, and they're really, like a communicate use AI to do communication coaching. So we have a group of our health coaches who are transitioning from text based to video coaching sessions, which is a pretty big shift in the way that they're working. And so we've partnered with the Virtual Sapiens to help prepare them in terms of their video presence and how they engage with a member live over video instead of over text. So they get to participate in kind of, like, simulated sessions, get feedback on a variety of different areas that are, you know, core factors in virtual or or video presence with members. We also know that, clinical documentation takes a lot of provider time, and it's probably one of the things providers like least about their jobs, or at least that's true for me. So we AI is an area where we really can use, technology to support reduction in or improved efficiency and just reduction in cognitive load for provider. So we're exploring and and kind of building an AI note taker tool so that providers will be able to spend less of their time and energy on documentation and more time and energy on the actual member in front of them, thinking about treatment planning and how they're gonna provide better care. So there's just a couple examples, that come to mind. I I can't think of other ways I I could see in the future. You know, we have a really robust content library, and so using AI to help better integrate content into the care providers work so that we're not putting the entire burden of knowing our content library, knowing when to make content recommendations on an individual provider, but just reducing some of that load for them by helping surface content that could be supportive of the member they're working with. That makes a lot of sense. Shifting gears a little bit to, kind of the the business and the contracting side of the house is that we've touched on a couple of different new AI tools that Headspace is currently either one looking at or two has implemented. And so it seems that you guys have more experience here than most folks in industry that are really, really, really early stages that you guys have have made the loop made made the jump as one of the the first movers in the industry. When you're talking to other organizations that are looking at implementing new AI tools, are there any, specific criteria or decision making processes that you currently, use at Headspace that you'd recommend other folks think through as they're gathering different stakeholders on different teams to help that new technology, data, legal, and security best practices, etcetera, that you might recommend other folks also think through when they're looking at new AI tools? Yeah. So, you know, we always start the the first step is always think starting with a needs assessment or gap analysis to identify the problem that we're trying to solve or the issue that we're noticing. From there, we then take a step back and bra brainstorm and think really broadly about what are the types of tools or resources that could help us, meet that need and that we could leverage in solving the problem. And, truthfully, AI has opened up just a lot of options, because there are so many tools that that are scalable and efficient in ways that, you know, resources just weren't in the past. So we after we've identified possible tools and started looking at, you know, what vendors are out there or or is this something that we could build in house ourselves, that will help meet that need? We spend a lot of time vetting the options and thinking about, you know, cost and efficiency and kind of the build versus buy question. And, you know, like I mentioned previously, we also, throughout that process, are very closely involving our compliance, security, legal teams, product teams to ensure that whatever options we're looking at meet our our internal expectations around security and compliance, and then also, you know, we understand what the lift would be like from a product, an engineering perspective. And then, again, of course, everyone's least favorite topic, but budget is always a big consideration in those discussions and conversations and really thinking about how are we gonna fund this and how is that gonna create efficiency to to save us money in other ways. And so, ultimately, you know, when we found a solution that meets the gap or the need or the problem we've identified, meets our compliance and security expectations, and fits with budget, we've found the winner. And sometimes it takes a while to get to that place, but I think really being intentional and thorough in the process is important in finding the right solution and and being able to bring them on board. Makes sense. And then after you've gone through now, it looks like a couple of procurement cycles where you've gotten various stakeholders involved, got compliance, legal, security, clinical, all signed off together, and then now walk through implementation. Are there any big learning experiences, basically things that might be different after you've implement some of these tools in practice that you might not have thought of initially that other folks could leverage or any common misconceptions that you might not have immediately realized, before implementation that now are clear, in practice? Yeah. You know, I think initially, we sort of or I sort of thought, like, oh, once we get budget and we get approval and everyone has signed off on it, like, our work is done. And, really, it's that's not true. It's being prepared for the resourcing of the implementation itself that, at least for me, was an oversight and is important. And knowing that, you know, there's there's gonna be a period of time where there's a lot of collaboration between organizations in terms of actually just bringing the system on board and making sure all the security checks are in place and all of that. But then there's also a period of time where you're really calibrating and making sure that it's working effectively and you're QA ing the AI itself and making sure, you know, the things that we're flagging are appropriate or the things that we're not flagging, where we're not overlooking anything. And then the communication piece is is really critical that, you know, you're clear with stakeholders around the time commitment that they might need to allocate to the implementation itself and that everyone is on board with that. And that then we're really clear with whoever the users are gonna be of this particular product or who's gonna be impacted by by bringing it on. And so, you know, we've had very clear communication plans that we work with our internal comms team and, the team standing this up, our our quality assurance team, that we're communicating early and often with providers, and they understand why we're doing this and how it's going to help them because it can be scary. And so I think really, really focusing on that communication piece and and taking the time to do it right is important too. Yeah. I think that that that makes a lot of sense in that. And on our side of the house, that's really what was jumped out to us, really, kind of super high level from working with you and your team is how great the communication has been, obviously, external between two parties, but within your teammates, I e, when you are implementing new technology or new anything. And from an organizational standpoint, just make it clear that this change is happening. Here's when it's going live. It's really hard to get the rest of the organization to truly buy in because they feel like a decision is being made for them and and who who enjoys that, especially in in in a company. But when you explain the why and the logic of how you arrived here and share that process transparently with the team, which sounds like you guys have done quite well, It makes it a lot easier for folks to get on board and get even excited about a new project that's coming through instead of pushing back because they feel like they're more involved in the decision making process, which is just great from an organizational kind of best practice standpoint apart from implementing tool. Absolutely. It's so important to think about stakeholder buy in early and not feel like you're saying people we don't want them to feel surprised or shocked or unclear about why something is happening to them. And I have to give doctor Alexis Hernandez Hans and her team credit because they put so much time and energy into really the whole process of of onboarding Brilliant and other vendors, but especially the communication piece and, making sure this felt like a supportive resource to providers rather than something that, was scary or uncertain or or, you know, would get in the way of their work in some way. Absolutely. Yes. Alexis and and team are absolutely top notch. They're amazing. Oh, yeah. Mhmm. Cool. Let's, let let's just focus a little bit to what's next up and what the future of AI and language models looks like in health care and specifically in behavioral health. So we're we're pretty early stages in implementation technology, where right now we're working on automating processes and taking a few steps towards, helping providers with clinical decision making support at point of care, but we're still we're super, super early stages there. How do you and and your team think about kind of where AI within headspace slash more broadly in behavioral health goes from here, and what excites you the most? Yeah. You know, AI, obviously, is so powerful in its ability to process large amounts of data very quickly and identify trends that we just are not able to do without it. So I'm excited about about the ability to do that and how that could really improve member outcomes in health care. So I mentioned an example a minute ago. One thing that comes to mind is, like, how do we help providers in the moment be making recommendations or treatment planning in a way that is most appropriate for that member? That can be hard to do when you, you know, seen six other people that day and you're tired. But how do we use AI to help surface content recommendations, treatment planning recommendations so that the provider can really focus on what they are uniquely very skilled at doing, which is being with a member, building therapeutic alliance, creating a safe space, and then implementing what other whatever treatment or modality that might be. So I'm really excited about how we can use technology to just relieve provider burden in that way and allow them to, like, really hone their energy and focus in on the member that they're seeing in front of them. I think there's also a lot of opportunities, again, for tooling that supports their work, whether that be patient scheduling or, you know, pairing of of members with appropriate providers, note documentation, all of those kinds of things. I think the more you can release some of those administrative burdens from a provider, the better outcomes are gonna be when they're able to really focus their energy on the member and the treatment themselves. That makes a lot of sense. And I have one more question for you. And then it looks like we have a couple of audience q and a's, and then we will wrap up for today here. But my last question for you, Kenley, is super broad. It's based on your and Headspace's experience with implementing AI to improve quality of care and also broadly in other places within your organization as well as where Headspace is headed from here. Is there a key takeaway that you'd hope that the folks that are listening to you and I chat today walk away from as, you know, hey. Like, I I spent my my 30 to 45 listening to Zach and have Kenley chat. And this is my my big learning, from our time together. Yeah. Well, no pressure on that question. You know, we are, obviously, in a time of just massive innovation and change. And I know, you know, for some, AI can be scary. It can be intimidating. It can be unknown. It can be overwhelming. And I my hope the takeaway for the audience is is that there is also tremendous opportunity to use AI tools in a way that is ethical, that is rigorous, and that truly does enhance mental health care without replacing providers or risking, patient safety or confidentiality. So my hope is that everyone in the audience walks away with that as a glimmer of hope or understanding and then starts thinking about or just sparks a curiosity around thinking through what are ways AI could support you? What are the problems or friction points that you are noticing in your systems? And is there a solution for that out there that is new and innovative and can really help support mental health care system you're working in as a whole? So my hope is that that's really the takeaway is that folks are walking away with more maybe more of an openness or at least a willingness to think through how to use AI in their work and support both scale and and quality care. Yeah. I think that that that makes a lot of sense. I mean, it it's a hard jump to make. Right? It's brand new technology where most folks have at least heard of it, maybe played around with it, And it seems a little bit of a black box mystery with a bunch of potential, but also a bunch of news articles that kind of go go both ways about it. So always really important to make sure that you have a rigorous vetting process. But once you get through that vetting process, really, really, really important potential long term opportunities, once implemented and scaled up in in a similar business. Exactly. Cool. Well, awesome. I think one big question from the audience, and it's coming from Heidi Faust. Question to you, Kenley, is how did using AI to help inform MBC affect your clinical outcomes? That's a great question. And I don't have specific data on that yet simply because we've only gotten to a % scale with Brelium within the last, like, six weeks. So we don't have enough time to fully measure, you know, impact on, implementation and outcome. But what we have seen that I think signals that we that we will get those results is that we've seen much greater adherence to measurement based care practices by our providers. We've seen significant just their documentation, how they're attending to things like, diagnosis. So, you know, one of the things we're tracking is how the symptoms do or don't align with the diagnosis and providing feedback on that. And we all know you know, diagnosis is essential to effective treatment planning, which supports outcomes. So we've seen changes in all those areas, which I do really think will help see a notable difference, once we have an a long enough time horizon to start measuring outcomes more directly. Awesome. Sounds good. Well, I think that that wraps us up for today. So to the audience, really appreciate you spending a few minutes of time with us. Kenley, do you wanna give a quick recap on you, maybe where folks can find you or reach out to you if they have any questions afterwards? Sure. Like I said, I'm Kenley Yurety. I'm, at Headspace. I'm a senior director of quality assurance and training there. You are welcome to email me anytime. My email is just my first dot last name at headspace dot com, and I'm sure we can have Mary send that out, with summary information after as well. And thank you so much, Zach. It was such a joy to be here today, and I just really appreciate you inviting us into this conversation. Likewise. I I appreciate you spending your time with us too. A bunch of great nuggets of information shared over the last few. To you and the rest of the audience, appreciate everyone spending their time here. We will send over a recording of the webinar. You don't have to remember everything. That will go live to all participants. If any questions, feel free to reach out to me too. I'm Zach. I'm one of the founders and CEO here at Brellium, and my email is just my first name, zach@brellium.com. I appreciate everybody joining us today, and, have a great rest of your day.