We are the Association for Child and Adolescent Mental Health, or ACAMH for short. And this is ACAMH Learn. Welcome to Mind the Kids, a podcast series dedicated to exploring the latest advancements in child and adolescent mental health research and practise. As a junior doctor, and as an academic clinical fellow in child psychiatry, I'm really passionate about understanding and addressing diverse mental health challenges faced by young people every day. And today, we have the pleasure of receiving Dr. Alice Wickersham, Zoe Firth, and Professor Johnny Downs from the CAMHS Digital Lab at the IoPPN in King's College London. We will be discussing their groundbreaking work translating digital innovations into CAMHS services and their recently published paper in the Child and Adolescent Mental Health Journal. So, Alice, Johnny, and Zoe, thank you so much for being here with us today. And, Zoe, if you want to start by introducing yourself. Well, thanks, Clara. So my name is Johnny Downs. I'm a professor of child psychiatry and health informatics at the Department of Child & Adolescent Psychiatry, Institute of Psychiatry, Psychology and Neuroscience at King's College London, and part of the King's partnership. And I am very lucky that I lead a group called the CAMHS Digital Lab. And we've got a few members here today. Hi, my name is Zoe Firth. I am a research assistant at the CAMHS Digital Lab and also an assistant psychologist at Croydon CAMHS, which is in South London Maudsley NHS Trust. My research background is in clinical linguistics, so my research interests are a lot to do with how we can use language data to better understand mental health, and how we can use language technologies to make that analysis more efficient. And I also do some work with school mental health and sustainability in digital mental health. Hi, I'm Alice. I'm a research fellow at King's College London in the same department as Johnny. I'm also the academic lead for the CAMHS Digital Lab. And I also hold a part-time role as a lecturer at UCL in the division of psychiatry. So my research interests are primarily in child and adolescent mental health and how it intersects with education and other areas like social care and criminal justice. And so we produce lots of information on our website, which is the camhsdlab.co.uk. And there's contact details or how to get in touch with us or if you're interested in any of the programmes on the website. And we also have a few social media handles, et cetera, which Zoe might point people to. Absolutely. We're camhsdlab everywhere basically. So we're cahmsdlab@kcl.ac.uk.. We're also camhsdlab on Bluesky, on Instagram. We're also on LinkedIn. So yes, please do reach out and get in touch. Amazing. To start, I wanted to ask you. So I know that one of the main problems that CAMHS services and patients face at the moment is the exceeding demand. And in this sense, digital technologies could help in so many different ways and areas, and lots have been done in the field of digital interventions. However, I think we all know there is a real problem in the implementation phase where we have lots of papers published in digital interventions, but when we get to implementing those in the community and in clinical services, real barriers exist. So I know you work in many different work streams, but I was wondering if you could tell us a little bit about your work in the implementation area and in overcoming those barriers. Thanks, Clara. So maybe I'll give a bit of an overview about some of-- when you mentioned those barriers, those implementation barriers, just to contextualise what they might be and your first part of your question was, how we get digital mental health technologies into regular practise? What we struggle with, what clinicians struggle with, I struggle with as an honorary child and adolescent psychiatrist, is that we don't really fully understand what, in terms of digital mental health technologies for children and young people, whether they legitimately work. Are they safe? Are they effective in everyday clinical practise rather than slightly more potentially esoteric trial settings, et cetera? And actually, are they expedient to use in services? Do we know actually how they fit into our care pathways, when's the right time to introduce them, when's the right time to advise people to not use them. We don't really have much clarity over guidance notes, how to purchase them and then how to actually embed them and implement them into everyday practise, and then what the evaluation frameworks should be for making sure that they're working for the right people. And we don't have really much good guidance at the moment about what the governance wrappers should be. I think one of the other dry aspects, which is kind of quite a crucial aspect in terms of the way NHS organisations currently work, is that you're not really rewarded or it's not really counted as a sort of meaningful activity if you recommend a digital therapeutic or if you purchase licences for a digital mental health therapeutic. So that might be some of the computerised CBT work that we're seeing that's being advertised, or some of the online parenting tools that are coming or coming out. And I know that you've done previous podcasts relating to some of those particular tools. Or the more digital single session intervention. At the moment, we're not sure whether it's really counted as important clinical time spent. So in terms of the usual audit pathways, both locally and nationally, trusts can be doing the work, but it's not really essentially put against an activity metric that might be judged as good clinical care. So I think there is issues around trusts being recognised if they do purchase and they go to some risks, maybe to purchase these and disseminate them amongst children and young people, that it's not really being actively counted as meaningful work. I think these are things that everyone's aware of. Not to say that they're being ignored. Everyone's aware, but there are current issues to be resolved, I think, both from a national and local perspective. I think the final bit or big part is this workforce issue. At the moment, we're not really trained. Clara, probably as you're going through your training at the moment, I don't know how much you're finding into your curriculum, you've got an idea about what digital mental health interventions might look like, how they might be adopted into your everyday clinical practise, what suits when and for who. So is also a matter of workforce training and it being part of our armoury of potential treatments that we now reach for, beyond the talking therapies that we're often trained up in and also some of the medications, treatments that we're trained up in. So second strand to your question was how has a CAMHS Digital Lab tried to tackle some of those areas? What we've been trying to do and sometimes, well, sometimes we're learning as we're going as well, is trying to understand what is-- firstly, what we think would be clinically meaningful for clinicians and young people in terms of priority setting. So what are some of the areas that we think are most useful? And as Alice laid out in the paper, we have a digital therapies and assessment group, which is trying to go through, sift through the evidence, sift through some of the glaring literature, sift through some of the lobbying and marketing work that commercial providers provide and saying, well, what of this looks like it's going to meet the criteria of being effective? Going to meet where we feel real population burden is on our services and then where we think we can actually get an adequate measure on how to judge whether it's useful or not? And then start to fold those into doing the hearts and minds work with clinicians to say, this is where we think it would be very useful to put a particular digital therapy. And one of the pieces of work that we're doing is thinking around computerised CBT and sparks, et cetera. Also then thinking about what some of our online guidance provision might be as well. So we do a lot of guidance that we issue out through more manual methods. How do we fold that into everyday practise? And then also be able to describe and make sure that we reassure people around the governance issues, to make sure that data isn't being sent off to various parts of the world where there's unknown provenance. Making sure that we've got the right safety evaluation that goes through our clinical safety officers, and which members of the group have been trained up in. So there's ways to try and put these wrappers around introducing these technologies and making sure that we're evaluating them effectively as they're being bedded into our care pathways, but also providing the trusted governance reassurances that it's been done in the right, cautious way. So we're not putting people at unnecessary risks in terms of privacy or reputational damage to the organisation. Thank you, John. That was a very comprehensive answer. And I think you've laid out the challenges around digital technologies implementation. I know that the CAMHS Digital Mental Health Lab does a lot of work around public engagement. I think about 10 PPI groups a year. And this is also a question that I wanted to ask about acceptability because you talked a lot about getting stakeholder buy in and getting clinician buy in, which is fundamental. Because I know, for example, you guys developed two different apps, myJournE and myHealthE. And I know there are other trials at the moment in the NHS for brief psychosocial intervention for patients on the waitlist. I was just wondering, in the work you're doing around acceptability and co-production of those apps, what you've learned along the way. It's a big and important area, usability and acceptability of these digital mental health technologies. And I think when we all first started, we were all-- my kids use this phrase, which my fairies are being crushed a bit, which essentially means that all of our optimism about how young people and families as well may engage in these tools. How essentially, they get a good dose of whatever the therapeutic intervention may be, guidance or these in the right time, in the right place psychological support or all of these things that we thought we could operationalize from our talking therapies that we could then put into digital tools, and that people then use them routinely. We realise that actually, quite quickly, we were not getting the right volume that one would hope from engagement. So we had lots of studies that said people set out, they signed it up, and then only 5% of people actually regularly used it as the protocol intended. So I think it's a very big issue that we still have to reconcile with. One of the important things I think we do within the group is firstly, own that and realise that it's an uphill battle to get people engaging in these tools. So crucially, it's building in this co-design work really very early on with these assumptions that people probably aren't going to use them. And if they are going to use them, what are the right contexts in terms of what sort of persuasions, how often do you have to invite people to check in with the system? When's the right time to be using these tools and be nudged in terms of engaging with them? And then critically, it's not to get them hooked on those tools to become dependent. And I think this is the other thing that slightly goes a little bit against some of the general commercial aspects of the way people want to use these tools is that they've always had the idea that the market is this almost consumer subscriber model where you've got to be constantly looking to engage. But we know within mental health services, we really want people to use tools when they're useful and then not use them when they actually feel like they're doing OK. So we've been really lucky to bring in some great user experience and user interface designers now to become full-time members of the lab. But can really start to understand right from the beginning, if you were to have a tool, would you want a digital tool to help you? If you were to then use the digital tool, how often? When? What is the general cycle of your everyday? And it's been wonderful learning for me as being part of some of these great co-design sessions, led by our lead designer Stephen Dorsch, and also colleagues with Anna Morris as well, that have been saying to well, let's pare it all back. Tell us about your day. We were doing work with parents saying, tell us about the layout of your house. If you're going to use a digital technology like a wearable device, how often do you wear a watch? Where do you put it down? How often are you going to pick it up again? All these really important things about understanding what the real environment's like for people and then how the technology might fit within it. And we've learned an awful lot through doing-- pairing all the way back and doing this initial discovery work. Although their children are enduring. And moving from end user and user acceptability to stakeholder engagement, so at the moment in the UK, there have been some UK wide initiatives like the Mental Health Mission, really putting emphasis on digital technologies to streamline research. There's also been a lot about digital technologies on the NHS 10-year health plan. And I think we all know there is this big momentum now for digital technologies and the potential they have to facilitate research, which we know sometimes it can be a slow process in the UK. And I wanted to ask you guys, yeah, how do you envision digital technology's potential to streamline research processes? And if there are any initiatives in the lab that you can share with us at the moment in that arena. Yeah, no, so you've touched on something very important, which is where does this all fit in the research ecosystem? And crucially, what is the future look like for tackling a really important issue, which is-- I know that we're struggling within across all research, clinical research, is getting good participation in research studies. So I'll briefly tell you what I think are some of the key domains. And then we'll touch on some of the aspects within the lab that cover it. So firstly, I think one of the things that's really helpful is developing the idea. We know that you want to go out. When you're trying to discover problems to fix, you want to really do some proper user engagement. You want to properly go out to your populations of interest and saying, how are you coping with your particular illness or how did that illness develop? What do you think may have helped along the way to be done earlier? So these advisory exercises, these patient and parent experience groups, these lived experience groups, they're really good. They're often done in quite small, discrete groups. One of the great things I think about now where we're potentially using chatbots and using these other AI methods of engagement is actually to surface ideas for very large groups. So we may be able to get to the point where we've got much better representative takes from people in terms of what they think is important. And I think methods of engagement to do. And what's been found is that if you want to get useful information from people and to feed people that they've done some meaningful engagement, sending out structured forms, sending out even free text forms. You're not surfacing the right levels of information from people in the depth that you'd probably want. And you tend to be favouring people that have enough time to sit down or have the literacy skills to be able to sit down and formulate those responses on the form. So what we're finding, and this is work that's been led by Laurence Telesia in our group alongside groups within the department of informatics, is can you use some of these AI chatbot tools to actually get advice early on about what kind of research areas are really valuable? What problem statements that people are having and actually how you can scale the way that you roll those chatbot tools out to get information from people about what they think the priority areas are? And actually, whether some of the research ideas that some people are having in higher education institutes, et cetera are valid. So that's one thing. That's literally about developing an idea and using these chatbot technologies to develop the idea. One of the crucial things, and one of the pieces of work that we're doing a lot of is getting better phenotypic information from young people and their families that visit our clinics. So this is the ability to be able to run case finding tools over our very large electronic health repositories. And we've got a large one called CRIS. But lots of trusts have other areas where you're able to do the de-identified, so these are anonymized searches, against people that might meet eligibility criteria for a particular study. And that's been very powerful. And one of the ways that we've really boosted that through our group is being able to ask very early on, when people are referred into services, would they be happy to be part of this research register. And would they then be happy to potentially be contacted by researchers if a research study seems pertinent to them. And what we found is that we were able to ask that very early, because our digital tools now are recruiting people. And the myHealthE tool is recruiting people right at the point of accepted referral. So traditionally, young people had to normally wait until they were maybe already engaged fairly comprehensively in the service to potentially access some of the research studies that were going on within that. So the great thing is now is that they're able to be asked early. So we've seen huge increases. We've gone from about 1,500 to about 4,500 in the space of two years of families and young people that are saying they're happy to be contacted. And that's just locally within our own caseload in the Maudsley. So the other joy of that is that because the way we're asking and we're asking in a fairly convenient way, doesn't have to wait to be seen by a clinician, doesn't have to wait for your clinician to find the time within the clinical session to ask, we're getting much better representation of the families that we see. So we're getting much better, higher proportions of those that traditionally didn't have the opportunity to engage in research. So these are people from lower socioeconomic groups, people from minoritized groups that commonly use our services. So we're getting better intake of, I'd say, people to be much more representative of the young people that are actually seen in child and adolescent mental health services. The other thing is this the same methods of engagement. So the tools that we're now using are-- I'd say that we're able to start to surface studies and what's going on within studies much more rapidly. So we talked about recruitment, but also we're now able to publish, in part of our patient experience, what studies are available and whether people would be interested. And we provide surfacing of that type of information quite readily to young people and families that are potentially pertinent to them. And then also telling people what's happening. So I think through a lot of our digital channels, it's basically-- and Zoe does fantastic work being able to tell people what might be going on within our organisation and within the lab about studies that they think where they are, how they're being tracked. That's very helpful in terms of making sure that people are still are engaged in the follow-up, and they fill in the measures and they feel like they're building to a growing evidence. And people are making use of that particular types of information. So that's really sustaining the interest. And one of the things that we found really valuable, and hopefully, Zoe, we have chance-- if I let you have a chance to get a word in, is actually then talking and working with young people and schools about how to find good, acceptable, interesting methods of engaging them in research and getting them to think about some of the research findings that are coming out and getting their views on that. And then finally, your last point, I think, in terms of digital work. So digital is not just about obviously the apps and the technologies about what people are accessing, but it's also around the way that data is curated and the methodologies that allow data to be linked together. And this is where Alice has done tremendous work in this area across the educational sector. But you can now start to have efficient follow-up. So you can have methods of being able to integrate data from other systems that can track how they're doing in terms of their education, how they're doing in terms of their employment, without needing to repeatedly ask them. And we know that the attrition biases, where we tend to lose people that have the most difficulties in these trials and cohort study follow-up, we can now actually get methods of actually capturing what their outcomes are. So yes, that's it in a nutshell. I think a huge part of that is the work your group and Alice has been leading on electronic health care record linkage. And in that sense, I was just reflecting the other day because it's a bit of a paradox we have in CAMHS, isn't it? We have this huge waiting lists. But then I think the STADIA trial, which was published a few months ago, also in GCP, showed about 30% of CAMHS referrals are rejected. And that just led me to think about there's so much more we need to know about the pathway young people go throughout their lives when they are referred to a service. And I know that Alice, you've done-- I know this is old work now, but I know you've published, a few years ago, a really nice paper looking at CRIS and create databases, looking at the pathways, the trajectories of young people with depression. And I just wanted to ask, yeah, if you want to talk more about your work with looking at trajectories of young people in mental health services, but also with the linkage that I know you've been doing. Yeah, sure. Thank you. So I think as we've touched on going back to the very beginning, all of this work stream touches on the real need for public services to understand the mental health needs and risk factors in inequalities affecting their services, to help inform their provision. And in terms of that piece of work that you mentioned, where I looked at pathways through CAMHS for kids with depression, I have to admit, that was actually quite a selfishly motivated paper for me. So I had been doing a PhD on child and adolescent depression, and I don't have any clinical background. And so the inner workings of CAMHS just felt like a complete mystery to me. Now, what I've since learned is that it's a complete mystery to a lot of people, including sometimes the people in CAMHS So I don't think it was just me. But at the time, I just wanted to very descriptively lay out the sorts of people that we see presenting to CAMHS with depression and what a journey through CAMHS often looks like for them. The paper, I think, primarily just reinforced a lot of the things that we already know anecdotally about this group, high levels of anxiety, comorbidity. We see a lot of females and adolescents presenting with this and often treated with antidepressants and talking therapies, just reinforcing stuff that we already know. As you've mentioned, the other elements of this paper was the fact that I did the work in two different regions. So we did it here in South London and Maudsley, but also replicated the work in Cambridgeshire and Peterborough NHS Foundation Trust. And while we found that a lot of the patterns were very similar, there are a lot of differences that came up. And it was quite difficult to unpick to what extent that reflected genuine differences in how those services operate, or to what extent we were just picking up in differences in how aspects of the patient journey are recorded in clinical notes, so that we're then able to appraise it in this sort of work. So it was actually quite a hypothesis generating study in the end, which I think raised quite a few more questions than it answered. But as you've touched on, it's all part of a wider story anyway that's not just limited to CAMHS. Other public services are really tied into the work that CAMHS does, and also need this work to inform how they provide their services. So the education sector, of course, is also affected by this as well as social care. Even the criminal justice system can end up being impacted by the sorts of things that we see earlier on in CAMHS and in schools. So as Johnny's touched on, one of our most exciting resources that we draw on in the lab is this large linkage we have between our local CAMHS records and education records sourced from the Department for Education. And so one of the packages of work that I've led on over the years is showing how school performance and changes in school performance over time correlates with other factors that might be going on in a young person's life. So we've done several pieces of work showing that kids who are diagnosed with depression sometime before the age of 18 tend to show average or sometimes even slightly above average levels of school performance in primary school. But by the time they get to the end of secondary school, they're showing levels of school performance that are very much below the rest of their cohort. So we're seeing quite a significant loss of potential through secondary school in this group. In another parallel but separate piece of work, we've also shown that declining school performance can be associated with subsequent involvement in the criminal justice system during young adulthood. So essentially, what this body of work is trying to show is that school performance is potentially a really sensitive signal for other things that might be going on in a pupil's life, whether it's difficulties with their mental health like depression, or whether it's issues with their family, with their peers and other factors that can then lead to difficulties long-term and have an impact downstream. Did you also find an association for involvement with the criminal justice system at a later age? Or it was just for mental health difficulties? So yes, although these are two separate linkages. That's something I should say. So in our linkage work between mental health and education records, we found that depression was associated with declines in school performance. In a separate linkage between education and crime records, we found that declines in school performance are potentially associated with increased risk for subsequent involvement in the criminal justice system. So yes, on the whole, it seems like changes in school performance could potentially be quite a sensitive signal for other difficulties that pupils might be facing. And I think that raises all sorts of questions about how we can identify that these issues are happening earlier and what we can do to support those pupils. Now, in terms of the kind of support that these pupils might need, that will obviously vary hugely on a person by person basis. So I think the take home message here is that there's something really important in terms of close liaison between child and adolescent mental health services and schools, and the sorts of provision that both are offering to the child and making sure that those efforts are coordinated and appropriate. So in that vein, another thing that we think a lot about in our lab is that liaison between CAMHS and schools. And actually, Zoe has been doing a lot of important work in terms of outreach to schools in the lab. And I wonder, Zoe, whether you might want to say something about that. Yes, thank you, Alice. So some of our work with schools falls under the scope of some of our digital products that we're using actually within schools. So the app myJournE, I can't remember if you were mentioning this earlier, Johnny, but that aims to capture the mental health needs and experiences of students in schools. We're doing research on this, but also a lot of our work in that area is very programmatic. It's going into schools, really trying to give back information to those schools that is useful for them in terms of understanding what are student's experiences of bullying like, maybe self-harm. And some of those issues are really prevalent in the schools. And at every stage, engagement is really key. We gather all this data from the survey, some qualitative data, some quantitative data, and we take that data, analyse it, and produce a report. And then bring the report back to schools, bring it back to the students themselves and also the parents and say, we got some data indicating that anxiety was higher in year six students than year five students. What do you think that might mean? Does this reflect your experiences and what you're seeing in schools? And that's really crucial for us getting to understand both across schools but especially within individual schools, what those needs are. And we're also really keen to engage with CAMHS services, but also schools in new ways. As of autumn 2025, we're launching a new partnership with Elstree Screen Arts Academy. So this is a screen arts academy in Elstree, as you might have been able to tell. It trains 14 to 19-year-olds in screen arts across the board. So that's everything from costuming to writing, to grips, sound design, absolutely everything. So we've partnered with them. And we're launching a new project called Research Explored, which is a series of three short films about different projects at the lab. And we're sort of imminently launching our first one. And this is about a bit of work that Alice led, actually, in working with a longitudinal twin study. There was some data within this twin study, mothers gave short speech samples when the twins were 10 years old. And then we looked at-- Allison and co analysed whether those short speech samples at age 10 actually could help us predict the trajectory of mental health of those twins across adolescents. And yes, there were some significant findings. More negative speech was associated with some of the worst mental health outcomes or some of the more negative mental health outcomes later on. But likewise, warmth, warmer, more positive speech also held associations with more positive health later on. So that's the subject of the first film. And it's really kind of opened our eyes as well, I think, to the different kinds of engagement we can be doing. We're really tapping into these young people's voices, not just as potential service users or service user researchers, which is maybe what we do more often in our capacity when we're working within PPIE, the traditional research engagement, public and patient engagement. But actually getting to hear their voices, creatives, and hear how this research really resonates with their experiences of mental health. So yeah, we're excited to be working more with the academy. Can I just chip in there, Zoe? And so Chris Mitchell, thr executive head of Elstree Screen Arts Academy, and his team have been brilliant about also saying young people really know how to put messages across to young people. They really know actually what types of media. We can learn an enormous amount from actually how they think about what is an engaging message. So what I really like about it is that they're training up and using the tools. And Chris and his team are training the young people in using these tools. The subject matter, it just happens to be mental health and children and young people's mental health and the research that's been generated within that field. But then the outputs of it, we're learning from how to do it better. But then also they're providing stuff that allows us to disseminate the information on our services too. And then coupled with, they're actually able to give us real insights on the domains that we're talking about around the impact of depression, anxiety, concern for others, exam stress, or actually the kind of content of being an adolescent as well in the modern era. Well, I'm overdoing it a little bit, but I think this is just magnificent because we're not only just is the training and the apprenticeship work that they're doing within that plus also focusing on domain feels like, it's fulfilling lots and lots of needs mutually between us. And that's where I'm very proud and very proud of Zoe and the group and Jess and others in actually being able to set up that collaboration alongside the great partnership that we've had with Chris and his team at Elstree Screen Arts. It's really fantastic. I think the most successful co-production and co-design partnerships are exactly like these ones like that. It's a mutual benefit. You guys are benefiting because you're engaging with this really fantastic school that's giving you really good insight into how to better implement digital mental health technologies into their reality. Yeah. So circling back to the initial point of how to really overcome these barriers and how to use digital technologies to streamline research, I think, Alice, you've mentioned, all mentioned actually, the work of linkage you've been doing and using electronic health records linked with other databases. And I think for people who work with linkages, they know that we need to extract all that information somehow. And we use, lots of the time, NLP, natural language processing, to be able to extract information out of linkage. And that can be a tricky process. And that's because in real life, clinical systems can be quite clunky. And I think now working as a full-time clinician again, sometimes I find myself asking, how do I find that again? Where is that again? Depending on the clinical system you use, it can be very, very clunky and information is scattered around. So I think that must be a real challenge in terms of extracting information from electronic health records. And I just wanted to ask you guys if you have any current initiatives in the lab at the moment. Someone, I think it was you, Zoe, mentioned Sophie's work. I know Sophie worked a lot with NLP tools in the self-harm realm. So yeah, if you guys can talk a little bit about that and those real-life research challenges. Yeah, I'm happy to talk about that. So this is obviously referring to the issue that when you're doing statistical analyses using electronic health records, what you ideally want to run your analyses is always neatly formatted information in columns. But obviously with clinical notes, it's just not that straightforward. Important information is obviously buried in a lot of free form long notes written by clinicians. So one way to get around that would be to manually go through all those notes and pick out the information you need. When you consider that each patient has dozens, if not hundreds of notes on their record, that's a hugely impractical task. So NLP automates this process, and it's a trained model to go through the notes for you and detect mentions of the information you need. So our colleague, Dr. Sophie Epstein is doing a lot of work on a self-harm NLP tool to identify mentions of self-harm in clinical notes. Now, as you can imagine, all sorts of pitfalls can come with this sort of tool. And that's because as humans, we use words in very nuanced and context dependent ways, which can be quite difficult for a computer to pick up on. So if we think of this example of self-harm, cutting is a common word that might be used in this context. But of course, the word cutting can also be used in lots of different ways that don't imply self-harm, cutting my hair, cutting something out of my diet. It can even be used in noun form. Cutting is something you take from a newspaper or from a plant. It's difficult to imagine how a plant cutting might come up in clinical notes, but you get the point. Essentially, we're wading through words that might not mean what we think they mean once we read them in context. And so that is the challenge of refining these sorts of NLP tools to actually get to the information that we are really looking to extract. So with that in mind, we are refining our use of NLP tools and starting to think about different ways that we can get at this information using different methods. So for example, one tool that we are increasingly using for finding mentions of information in clinical notes is medcap. So Johnny, you might want to say more about that. Well, no, thanks. And so obviously models and large language models are becoming increasingly more sophisticated about really clearing out and getting the context right, so we can be positive, or we can affirm that a OD is an acronym, refers to maybe overdose rather than once daily or misspelling of odd. I think the thing to point out is that the assumption is that all this has been fixed. It ain't been fixed. There's quite a lot of work that needs to be done to make sure that the content that is being structured through these large language models are actually producing the outputs that we need to produce. So we have quite a large number of teams actually doing quite a lot of clinical validation work. Five or six now full-time clinical academic researchers have within their work plans actually the ability to look, is this a diagnosis? Is this the way that sleep's referred to? Is it accurate in terms of emotional dysregulation? Is that accurate? And one of the great tools that we've taken advantage of that's been produced out of the health informatics group here at King's is a neural network approach of identifying and training up these data extraction, these natural language processing data extraction methodologies to pull out A, something that is contemporaneous, so it's current rather than past. It's been experienced by the patient rather than a relative. And also you can get closer to what we hope to be as ideas of severity and these other contextual factors that are important to code as well. So we can know not just who and when it's been experienced by maybe the level of that experience. So we're developing these pipelines that allow us to be able to train on small select samples of free text, and then you can deploy that at scale across the 54,000 patient records that we have for children and young people in South London. And crucially, and I know this is what we're furthering work on is that methodology, those algorithms work can be then translated and moved across into other settings. So we've done portability work regarding using the same algorithms, again, not data being transferred, but using the same algorithms between us and a Cornell University based in New York records. But also we've done work looking at can they be ported across into Cambridge, led by Rudolph Cardinal's group as well alongside our researchers here. So the idea is that can we get to the point where we can really start to furnish the types of information in clinical epidemiological studies much better. Case finding, can that be done in a much better way so we can recruit right people in, so we've got much better phenotypes about our eligibility criteria who we're bringing in. And then I think finally where we really want to get to is the point where we can start to populate the clinical dashboards. And there's work that's ongoing within our group, but also within the adult services here, led by the Lucy Group, which is David Codling and Kate Codling. They're doing a really, really important work about how you can structure this information so it presents things like as you're going through Clara, big note summaries where you're saying, well, did they have an adverse reaction to that type of medication? Or who was diagnosed when with a particular condition? And sometimes you be confronted with thousands of pages to go through on your electronic health care record to source that information. And so then being able to use these NLP techniques to actually provide clinical dashboards that quickly summarise what a patient journey has been like within the EHR system so it doesn't take two days worth of note review in order to get an accurate summary of what's going on in a young person's life. So that's the plan is to keep going with these NLP validation approaches to the point where we can start to populate clinical dashboards, so we get better knowledge to clinicians to make decisions and essentially, potentially to young people as well and families through their own portals. Yeah, that's really fantastic. And I think also touches to a point that we didn't talk a lot about but it's also using digital technologies to reduce administrative burden. Because if you can do that, then you don't have-- as a clinician that already has lots of time constraints, you don't need to spend many hours reading notes, but maybe you can spend more of those hours seeing actual patients. And yeah. And I think to wrap up, because my last question is going to be about stakeholder engagement, but Zoe, you've talked plenty about this, and I think also Johnny and Alice, you've also talked about this. So I thought about asking a slightly different question, which is to ask to each one of you, what's the current project or development in the digital technologies realm that you're most excited about? If you can share that with us, of course. Absolutely, if I can kick us off. Well, it relates to what you were just talking about, actually, which is possibly why it was top of mind. But just talking about reducing administrative burden for clinicians or just in the process of starting a trial at the lab with an ambient voice technology product that is designed to record, take an audio recording of clinical assessments, and then use the transcription that it generates from that recording into a clinical report. Now, crucially, as part of any ambient voice technology product or at least how we're using them, a clinician is there at the appointment, and they are also reviewing this kind of automatically generated report, but it's just a huge reduction in administrative burden. So the use case that we're using it for is the first stage of neurodiversity assessments at different CAMHS teams throughout South London. These are assessments that take a couple of hours to conduct, and then a further to possibly more hours just for the person to type up their-- the clinician and the assistant psychologist to type up their notes and turn this into a clinical report. It's incredibly important information that's being captured about the developmental history of the service user of the child or young person. But it just takes a really long time to write down, to turn it to our appointment and into that report. So we're working with an organisation called Anathem, which have this ambient voice technology product. And we're very encouraged to see-- there was previously a trial done with some colleagues in Central and Northwest London. They got some really promising results. They tried it across a couple of different use cases, including this neurodiversity assessment. They saw a 45% reduction in that administrative time. Their paper is under review at the moment, so stay tuned. Watch this space. I'll leave you with the cliffhanger. But actually, attended a talk with them recently in September, and they sort of finished by giving some feedback from a clinician about how just absolutely transformative it's been to use this product. They get to focus more in the appointments. There's less of this administrative time, more time to just focus on the complexities that come up with an individual family, a service user, a child or young person, a parent, to focus on their case. So really, I'll leave you with a little message of hope there that we're really excited to run our own trial and hopefully see the same gains in reducing that administrative burden, positive experience with young people and families of course. Other ambient voice technology companies are available as well just to-- yeah, so thanks, Zoe. Yeah, we're all excited about that. If I was to pick a different strand-- be careful to pick your favourite child really. But there are so many great work and again, very excited about everything that we've talked about, the linkage work, the initiative in terms of the outreach work for Elstree, et cetera. But I think I can just about announce it because I think it's just gone out. The press has just gone out. But I'm really excited about work that has been funded by the Maudsley charity that we're looking at, which is around a primary care integrated neighbourhood team programme that's been running for a while in Lambeth called the Well Centre, which is largely focused about providing a service, primary care service specifically for adolescents, for young people. So it's really trained up lots and lots of great people, focused on the types of issues that young people endure, and actually sets up an environment where they don't mind coming to rather than some of the more austere places in primary care which don't feel particularly well suited for young people. So we're looking at basically the impact of that current service and how it's been running and how it really, I think, acts. But there is some early evidence, but we're going to hopefully increase the quality of that evidence at looking at the impact of that service on ventral referral rates to mental health services, to crisis services. So does it actually really mitigate the issues that young people have with their mental health and actually reduce their reliance on specialist services. But we're also really interested in actually how we can support primary care, a better outcome measurement collection, and try and get some information back into their system so they can track young people more efficiently. Because I know that often sometimes things are a little bit of a mystery in between visits when they see them. And also this way of getting some structured measures into primary care about how best to assess symptomatic recovery over the course of treatment and support. I'm really pleased that we've got funding to do that. And I'm really looking forward to next couple of years and getting those programmes embedded, which take advantage of our digital technologies we use, but also some of the link data between primary and secondary care services that we have locally. And I think for me, the thing that's most exciting, perhaps unsurprisingly, given my research area, is the growth of data linkages that are happening nationally. And I think the reason for me is just that, historically, Scandinavian countries have really led the way in terms of using public service data to properly inform policy and practise. And actually, the UK is now really hot on their heels in that regard. We're doing really quite well, and Administrative Data Research UK received a reinvestment earlier this year in 2025 to continue that really important work. So I think that's really exciting. And within that context, these sorts of linkages present a really good setting for refining quasi experimental trial methods and other techniques that will help us understand what provision can best support our children and young people in a way that positively impacts them, not just in one area like mental health, but also in lots of other associated areas as well, like in social care and education. So I think that's really exciting. l think with one of of the really cool things about all the work we've been doing at CAMHS Digital Lab is that thinking from country wide perspective. So I think we're hugely privileged. I'm based in CPFT, and you guys are in [INAUDIBLE]. And even more because you have the Kings Maudsley partnership, but I think those are trusts that generally CPFT has create. But if we look to other areas of the country, if we look to the north of the country, for example, I know that some places still use paper notes or even use electronic health care records, but don't have a fraction of the research investment that we have. And I think with digital technologies maybe in the future, it would be really cool if we could have this UK wide database of electronic health records. And who knows, maybe in 10 years time, that will be possible. Yeah. Clara, you're not the first person to suggest that, that would be a good idea. You're completely right. And there's various attempts over the years to try and get these very large or work on federated systems and things like that. I think we will get there. Well, I hope within Zoe, Clara analysis lifetime that it will be done, and we'll actually get them to-- it will always be just understood. And it's part of just what is part of the health system policy and programmes are built around these very large resources. But yeah, I think it's a really good space to be in because it's got loads of-- people can see the potential, but there's lots of problems that we need to fix. So research is really needed. As someone who wants to do a PhD in epidemiology, I have to agree with you there, Alice. I think it's really exciting. And I think yes, I think when you go to epidemiology conferences, and you see people from Scandinavia or Sweden presenting their population health research, I always am a bit envious because they have-- Psychiatry Sweden is just-- that they just have-- yeah, you have everything linked. And I think it's really exciting that in the UK we're now catching up. And I didn't know about the administrative data, UK renewed funding. That's really great news. Yeah, and really amazing work. And yeah. And thank you so much Zoe, Johnny, and Alice for being here today, and for sharing all about the wonderful work you do at CAMHS Digital Lab. Thnak you for having us. Thanks, Clara. Thanks for looking after us. [MUSIC PLAYING]