We are the Association for Child and Adolescent Mental Health, or ACAMH for short. And this is ACAMH Learn. Welcome to Mind the Kids podcast from the Association for Child and Adolescent Mental Health known as ACAMH. Today, the key question is, can we project child's mental health risk as early as birth? And what would that mean for prevention and early intervention? I'm Mark Tebbs. I'm your host for today. Spent my whole career working in mental health, from frontline service delivery to director of mental health commissioning. I'm currently a chief executive of a charity. And I'm a trustee of a mental Health organisation, and also a career coach. I'm really pleased to be hosting these podcasts. It gives me the opportunity to speak to some of the leading clinicians and academics in the forefront of children and young people's mental health. So if you're a clinician, if you're an educator, a parent, or just curious about mental health, I hope these discussions leave you with some fresh insights and practical tools to help. Don't forget to subscribe where you get your podcasts from. And also on your YouTube channel, just search A-C-A-M-H. And you'll find us across all socials. Today, I'm joined by Emma Butler of the Royal College of Surgeons in Ireland, who is an occupational therapist focusing on prevention and early intervention. In this episode, we'll be discussing her research, recently published in ACAMH's JCPP Advances journal on predictive modelling at birth and then identifying risk at age seven. We'll also look at the role of psychosocial factors and fine motor skills and how we might use this to support families more proactively. So let's get started. Welcome, Emma. Really great to be speaking to you. Thank you, Mark. Nice to be here. Great stuff. Really looking forward to this podcast. Do you want to just introduce yourself, maybe say a little bit about your research interests, and maybe introduce the title of the paper and any people that you worked with on the research? Yeah, so my name is Emma Butler. Clinically, I work as an occupational therapist, and I have worked in neonates and paediatrics for over 20 years. And I suppose that journey has led me into prevention and early life because I wonder why our health system is reactive rather than proactive in terms of supporting children's mental health. And there's just, I suppose, context to why children might be struggling in terms of, can we change the environment? Is the school classroom working for them? How are things at home? rather than seeing the problem being viewed as something that's inherently within the child. So for this paper, I wanted to see, well, can we predict childhood mental health as early at birth with a view to, if we could, that we'd be able to intervene proactively and preventively with children and their families from the very beginning. And so this paper was in collaboration with my three wonderful supervisors as part of my PhD. So Professor Mary Clarke is based in the Royal College of Surgeons in Ireland. Michelle Spiritos is based in school of occupational therapy in Trinity College Dublin, and Doctor Linda O'Keeffe is developed in the School of Public Health in University College Cork. So I suppose another interesting or novel thing about us as a team is that we were a multi-institutional and multidisciplinary clinical background in terms of working together on this paper. How did that multidisciplinary work and shape how the research evolved? Yeah, I guess it really made us all learn, one, from and with each other, but two, to really look and consider things holistically. Because as I'm sure a lot of your listeners are aware, sometimes in research you can get so funnelled and so niche into your very specific topic, which is obviously really important for bringing on that specific area but might be completely then ignoring all of the other things or factors that are contributing to whatever it is that we're studying. And I suppose myself, being trained as an occupational therapist, we're taught to literally think about the person, the environment, and the things that they do. So we are a jack of all trades and master of none in terms of that we try to see everything and look at how everything fits together in terms of people's life and their well-being. And so it was really lovely then to have the support and expertise of experts in their specific field to help me think about child mental health from a holistic-wide stance. Amazing. So we're recording this in March '26. There's been loads in the media this week about children's and young people's mental health, and particularly about the impact of social media as one of those environmental factors that's changed quite a lot over the last decade. So it'd be really useful just to set some kind of context in terms of, from a research perspective, what do we know about the scale and trajectory of the increase in mental health needs? And you mentioned about prevention. So why is it so important that we look at that rise within the lens of prevention? Sure. So I guess, we know from research that mental health is on the increase, and it is a global problem. So the global burden of diseases have showed, I suppose, what they would call alarming changes in children's mental health from 1990 to 2019. And a recent systematic review and meta-analysis in Europe found that there was a pooled prevalence rate of 15.5% of mental disorders among 5 to 18-year-olds. And similarly, in England, we can see the same trend. So in 2017, it was about 11% of 8 to 25-year-olds reported poor mental health. This rose to 1 in 6 in 2020 to 1 in 5 in 2023. So I suppose it's important for us as well to know that this isn't just a COVID or a post-COVID problem, that this increase has been coming from the '90s. And obviously, COVID certainly didn't help. And I suppose why I believe prevention is so important is that poor mental health in childhood has detrimental effects not only in childhood, but for later life outcomes. And from research, we know that poor mental health and childhood has more of an impact on our later health and well being than a physical health problem in childhood. And this is partly probably because it impacts things like educational attainment, school dropout, the type of job you might get, whether you experience incarceration or suicide or homelessness. So there's, again, that broad view of the scale and scope of what poor mental health impacts in your life. So preventing poor mental health would not just benefit the child in childhood, but also potentially their entire life trajectory and, therefore, society as a whole. Yeah, OK. So that focus around prevention, I guess, the ambition is to be able to identify mental health risks before it emerges. So why has that been so difficult? And what was the gaps in evidence base that maybe your study was trying to investigate? So I suppose, it's important to think about what mental illness prevention is. So it's not just that, oh, we'll stop that from ever happening, but that there's also benefit in delaying when symptoms might emerge and/or the time or the level of symptoms that someone might experience. So it's not a binary, we'll prevent mental health or not. It's more even reducing the impact or the point in life at which you start to experience mental health symptoms. And it's widely accepted that adult mental health has its roots in early life. A statistic around this is that 75% of all mental health disorders that persist into adulthood have emerged before 25 years old, but that then allows us to use 0 to 25 as a real area of prevention. But unfortunately, a lot of mental health research focuses on adolescents and young adults and completely ignores the mental health needs of children less than 10. And yet, we know that children less than 10 are experiencing symptoms. they may not necessarily have diagnoses yet. But a systematic review found that symptoms in childhood, rather than a specific diagnosis, was more strongly associated with adult disorder. And despite of all of the research and the funding that goes into mental health, there is still a lack of complete understanding of the specific causes of poor mental health, despite considerable efforts trying to find their genetic and neural basis. And I suppose another complexity of the field is, a lot of [? CASB's ?] work would highlight this, that people move in and across different symptoms and different diagnoses. So there's not one clear pathway prevent anxiety or prevent depression, because a lot of children and young people who experience anxiety at one point, a couple of years later, might be experiencing depression. So there's huge heterogeneity there. And very worryingly, a large systematic review and meta-analysis found that among children with known mental conditions and high income countries, only 44.2% received any services. So if that little amount of children with known diagnosis they receive in services, one can only imagine the level of unmet need that's potentially out there in terms of symptoms. So then from the gap in research, I suppose, is that there is a lot of prediction models that have been developed in the area of mental health. Most have been on adult samples and not children and young people. Of the ones that have been developed to be used for children and young people, a systematic review by senior et Al. Found that most of those models weren't sufficiently powered or calibrated, tested. So they weren't, I suppose, designed in a way that they're ready for us to use, and none were designed to be used in the perinatal period, which is where I wanted to focus on because it's the earliest point in life. OK, so it's important that we look at this in a nuanced way, that it's not a kind of binary prevention. There's a nuance in terms of frequency, severity, types of symptoms. And that the research has been more focused around the adolescent period. And you've focused on this kind of perinatal risk. Want to tell us a little bit more around what it is that you like studied? Or what were the predictors that you were looking for at that perinatal period that you hypothesise would impact on later mental health? Yeah. So I suppose, obviously mental health problems are highly heritable. So that, we know that's going to be a predictor in and of itself as your genetic risk. But there's also a lot of evidence that a range of prenatal stressors is associated with increased symptoms even after we've accounted for that kind of familial risk. So some of those prenatal factors that have previously been identified are things like maternal physical and mental health in pregnancy, pregnancy complications, smoking and alcohol use so that kind of health behaviours in pregnancy, intimate partner violence, and environmental exposures, and then your sociodemographic indicators, like maternal education and household income. But there's also a lot of literature looking at that early postnatal period as well. So risks in that period are, again, socioeconomic disadvantage, the parents' mental health both previously and in that time period, and then any, I suppose, disharmony or relationship difficulties between the parents and/or the parent and the child. But it's really important for us to think about that pregnancy or the perinatal period can't be separated out from the previous life experience that a woman had before she was pregnant. So it's important for us to think about this from a life course perspective and thinking about all the strains and stresses that a person might have been experienced before they became a parent. But any of these associations that have previously been found aren't deterministic. So just because you have that risk factor does not mean that you will go on to develop mental health difficulties. And so the challenge remains in trying to identify which children are most at risk because all children who have any one of those risk factors will not have poor mental health later in life. But we do know that the more risks you have, the chance of you having mental health difficulties increases proportionately. So I suppose, again, why it lends itself to prediction research rather than causal research is that causal would be trying to separate out all those individual effects and work out how much risk is accrued to each specific one of those risks. Whereas, with prediction research, we want to look at the capacity of how we use all of that risk information to predict later mental health. Before we go into that kind of predictive modelling bit, could you just say a little bit about the sample. I think you used the aspect kind of like data sets. Yeah. So it's based at the University of Bristol. So it used to be called Children of the 90s because they were first recruited in the '90s. But now it's known as the Avon Longitudinal Study of Parents and Children or shortened to ALSPAC. Because as the decades have gone on, it's more than just children of the '90s because they actually now have second-generation children in the study. But it was a world-leading pregnancy cohort study. So between April 1991 and December 1992, more than 14,000 pregnant women were recruited into the study. And it collected detailed data, both objective and subjective, from the pre and perinatal period. It's also a large representative sample. And it's prospective in design, all of which made it really suitable for using for this research question. Brilliant. Thank you. So the study's around variable predictive modelling, so looking at risks in the perinatal period and then the outcomes at age seven. So for listeners who may be less familiar with predictive modelling, how should we think about what these models are designed to do in the context of child and adolescent mental health? Sure. So clinical prediction models estimate an individual's level of risk of a particular outcome with a view to, I suppose, guiding clinical decision making. But it focuses on probabilistic risk rather than absolute certainty of disorder. So again, that's the difference between prediction and cause. Cause would say you absolutely are definitely going to experience this thing. Prediction is just looking at your probabilistic risk. But we can use the risk score then to support the clinical reasoning about whether we need to provide intervention or supports. And I suppose, depending on the level of risk, what type of support that might be. OK, righteo. That's super clear. What were the predictors that you were looking at at that perinatal period? Yep. So I suppose main things were things like maternal health, pregnancy experiences, and health behaviours, and sociodemographic factors. So as we had touched on earlier, that group of things were chosen because prior research has repeatedly demonstrated that, individually, they have been associated with mental health in varying degrees. And in a previous paper of mine, I had shown a dose response relationship between the number of pregnancy complications and the level of sociodemographic risk and the child's mental health in childhood using the growing up in Ireland. So that's the Irish birth cohort. I suppose theoretically, all of those predictors are impacting the developing brain. And we know how important the first 1,001 days are of life for the developing brain in general, for every single outcome. And then pragmatically, in a prior paper using the French birth cohort, which is called ELFE cohort, I had used those predictors to do a prediction model in that sample or that data set. So then for this paper, I had actually intended to initially externally validate the model that I had developed in the French data set. But two of the predictors were systematically missing from Outback, which means they didn't collect them on anybody. So instead of throwing the baby out with the bathwater, I use the predictors that were available to, I suppose, mimic to a degree the one in the French data set. But it definitely is not true external validation because it wasn't identical. And this allowed me then to use the Lasso method, which is basically where the computer or the software picks the predictors to be in the final model because that avoids any sort of P hacking, basically from the researcher in terms of finding the best group or cluster of predictors to go together in the model. So which of the factors emerge as the most important? And how does that align with the kind of existing evidence on early risk? Sure. So I suppose, again, because we're using prediction modelling, you can't really infer things about the magnitude of each predictor's coefficient like you would, for example, if you were coming at this from a causal perspective. But in general, the psychosocial predictors or people's circumstances emerged as the most consistent or pertinent predictors. And I suppose there's growing evidence that poor mental health is partly socially determined. So for example, a systematic review and meta-analysis of 1.4 million adults provided evidence for the association between socioeconomic status and mental disorders. And a further systematic review and meta-analysis showed that while current socioeconomic status has a stronger association with adult mental health than childhood socioeconomic status, it is really important to consider the impact of early life and childhood socioeconomic status and, indeed, socioeconomic status of the prior generation to account for the lingering effects, I suppose, of transgenerational or early disadvantage. And a meta-analysis of population-based studies in the States examine the association between socioeconomic status and child mental health, and they had nearly 27,000 participants in this meta-analysis, ranging from 3 to 19-year-olds. And they found small-to-moderate associations for things like low family income, low Hollingshead Index-- can ever say that word-- low subjective socioeconomic status, parental education, poverty status, and receipt of public assistance. So I suppose in the arena of socioeconomic status, how we define it or what we call socioeconomic status to be, there's different metrics used in different studies. But again, I wanted to consider the wholeness or the holistic of all those different situations or indicators that a child might be exposed or experiencing to because hazards and opportunities are not equally distributed amongst society. So rather than me trying to pick out one specific thing, like, for example, low maternal education-- because we know if you have low maternal education, you're probably more likely that your mum might have a lower-paid job, and, therefore, your household has less income. So all of those things together matter. It isn't the effect of just one of those. Sure. Is it possible to bring that to life with a lived experience example? Yeah, so I suppose for example, a lot of times, there might be studies or things that would say there's an association between, say, having a young mum and the child having poor mental health. But it's not having a young mum that's not causal. It's likely that a young mum, for example, might be single, so she might not have another parent or caregiver to help her with-- like, consist of co-parenting. She might be really worried about how she's going to put food on the table and clothes on her children's back to go to school. They might have less nutrition. There might be less books in the home, not because any of this is a person's choice. But it's the circumstances that life has found them being in or led them to. And so it's really important that we provide extra supports to those who need more and to have an equitable sense of a health service rather than an equal sense. Yeah, absolutely. So turning to the findings a little bit more, so I think the conclusion in the paper was that the model predicts mental health at age seven with moderate accuracy. So could you tell us a little bit more about how we interpret that finding. Yeah. So I suppose, the level of accuracy that is acceptable in any model will depend on what decisions you're going to make from the model. So for example, if it was a diagnostic model, you would want to be sure that it's really, really highly accurate because that will, therefore-- let's say, if it's a model for predicting cancer, it'll say, for example, that you need to start chemotherapy drugs because you have a really-- it's highly predictive of having cancer. So I suppose personally, for me as a parent, I feel if the intended use of this model was to provide pharmacological intervention, for example, to my child, I would want this model to have super high accuracy for me to sign up to that. If, however, you are saying to me-- and again, this is speaking as a parent of teenagers and adolescents-- your daughter has a slightly higher risk, and this model is moderately accurate, and what we're offering you is to come and attend a parenting course, I would be less concerned that the accuracy was only moderate because I wouldn't view any negative or unintended consequences of me attending the parenting course, other than I might have used time. But if you are providing it to me for free, the parenting information can only be of benefit to me being a better parent and to supporting my child. And so even if the accuracy was slightly off, and I didn't actually need it, and my child didn't have a higher risk, it wasn't going to cause harm by doing it. So the level of accuracy-- there's no agreed threshold of what an accurate number needs to be to say that any model is good or bad. It's more related to what are you going to do with the information from the model. Yeah. Is that clear? [CHUCKLES] Yeah, no, that makes sense. And as you were speaking, I was just wondering whether that acceptability would be universally felt, whether some parents might think that a predictive model might label their child, or whether there was some sense of self-fulfilling prophecies, I don't know, a little bit more of a philosophical question. For sure. Well, I suppose like it comes down to firstly being really transparent that-- so sorry, at the very first, this model isn't ready to be used in practise yet. But we can probably discuss that or explore that a bit more later. But I suppose from a theoretical standpoint of using models in the area of child and adolescent mental health, it's really important, one, to emphasise that the overall risk is still low, as in, in the entire population, it's not like 50% of people are going to experience poor mental health. So it's lower than that. And also emphasising that any prediction model is probabilistic and not deterministic. So whilst you're saying to the parent or the caregiver, yes, your child's risk score is slightly higher, you're being really transparent that is not definitive. And things happen between birth and seven. So even if my prediction from my model at birth was accurate, something could happen between birth and seven positively or negatively that changes my risk. So it's not definitive, and that's what we really need to be clear about. I also strongly feel that the parent or the caregiver should be an equal participant in deciding whether they want to know the risk score. So some parents, health users in any area of health will say, I actually don't want to know my future risk for anything. And that needs to be parental or patient choice. But I do like to think of it as, I suppose, a prediction model can give you a life jacket jackass to support you with thriving in whatever life circumstances are, rather than giving you a life buoy when you're really struggling. Yeah, that makes sense. I guess it's another tool, potential tool in the armoury that people could choose to engage with, rather than something that's communistic or compulsory or any of that nature. And I guess, related to that, I was also interested that you talked about in the paper about the predictive tool being particularly good at ruling children out. So yeah, could you say a little bit more about that? So I suppose, again, if we think about a bell curve in a population or anything like that, most of the financial and health care use and personal burden of mental health problems in the population doesn't come from that small group of people with the highest level of risk, but actually comes from the people in the middle, just because there's more of them who have a slightly higher level of risk. So being able to consider every child's risk in the population is important because of the level of unmet need out there, as in, we know even the needs being met of the people who haven't identified need is low, let alone the level of children who have an unidentified need. But it would be really important that if we're using this model at population level to say you don't need something, that we need to be confident that you don't need something. And then of the children that are predicted to have higher risk, albeit emphasising still a low risk, that we're not as certain that that positive risk is positive, but that we can work with those children and their families to get to know what kind of supports and interventions they might need. So I suppose having a tiered level of approach in how you might use the model to inform delivery of services. So it can help with allocation of scarce resources to where potentially there's the most need. Yeah. So if you have 60,000 children born a year in Ireland, just as an example, that you would, for example, be confident, if you were to check everyone's risk, that these 55,000 children really have little to no risk and they can just have the universal level of support. What actually, of these 5,000 children with slightly increased risk, they might need a selected or an indicated level of support. OK, brilliant. And then there was a bit of the paper about fine motor skills and the predictive power of those. So yeah, it'd be great if you could unpack that a little bit for us. So you can, again, see my bias as an occupational therapist coming in here, why I wanted to look at motor skills. So I suppose, both anecdotally and from a research perspective, we know that motor skills and poor mental health are correlated or related to each other. But most of the studies that inform that are cross-sectional in design. So we can't infer anything about the temporal order or whether one predicts the other. And whilst prediction research is definitely not causal-- we're going to count how many times I say that today-- the predictors that you use in a model can be causal or non-causal in nature. But they need to be clinically or biologically plausible as to why you'd put them in there. So for example, I'm not just going to put the colour red in my prediction model, because why would I say red is related to poor mental health? So motor skills are plausibly related to mental health via direct routes. So for example, things like cerebellar, basal ganglia, frontal cortex brain function differences. Or we could argue an indirect route. So for example, if I have poor fine motor difficulties, that might impact my handwriting in school. That might impact my academic attainment, which impacts my confidence and can lead to poor mental health. So as I said, prior research that has associated motor skills and mental health as being correlated with each other is cross-sectional. So what was exciting about this finding is that because ALSPAC is prospective in design, and we were looking at motor skills at 18 months and mental health at 7, we were able to see that the fine motor skills did predict later mental health. And a recent study also found that fine motor skills are not only phenotypically linked to later neurodevelopment and psychopathology, but are also genetically associated with these domains. But future research would need to explore whether this is a causal relationship or not. Because if it is causal, then intervening on a fine motor skills in infancy would potentially also benefit later mental health in addition to just impacting the baby or the child's fine motor skills. And a lot of early life developmental research tends to focus on gross motor skills, and fine motor skills aren't looked at. So it will be good for more research to start to focus a bit more on fine motor skills. Just thinking a little bit more about the implications, I guess, and thinking maybe about policies and next stages, so what you-- and I think we've mentioned before about the idea of a tiered model and patient resources. So what might it look like in practise for us to introduce a kind of predictive modelling? Yeah, so I guess the dream-- [CHUCKLING] --would be something like, for example, from a universal level, that every parent of a newborn is given access to an online webinar, which they can access at no cost at a time suitable to them to understand the concept of infant mental health so that every parent has that information, regardless of what they know or don't or who they are or anything like that. Then at a collective level, so if we have a group of children who the prediction model predicts that they have a slightly higher risk than everybody else, so they might be provided with, for example, group-based interventions but in their local community, but in a face-to-face group-based intervention with other parents and babies of similar ages. Or for example, certainly in the UK, home visiting is much more well developed than it is here in Ireland. There's only very few people who get home visiting here. So it might expand, for example, the groups of people that get offered home visiting. So maybe those with slightly higher risk get offered a place on a home visiting programme or a group-based intervention. And then the children or the group of children with the highest predicted risk, they'd get an indicated level. So what that would mean is, they'd get that universal access to the webinar. They'd also get invited to home visiting or group-based things, but they might also get one-to-one direct infant mental health services. OK, righteo. So yeah, so that's a very clear dream that was very-- [CHUCKLING] So what would be the next steps in research that would need to happen for that to materialise or for that to be kind of like-- A reality? --reality, yeah. [CHUCKLING] So I suppose from this model perspective, it would firstly need to be externally validated in a different sample. So what that means is that a completely independent researcher or research group would take the values of these predictors and plug them into the formula with the predictors in their data set to see whether it gets the same level of reliability, stability, accuracy, all those kind of things. It's basically just to make sure that this isn't just noise, that this relationship or formula happened to work in this data set, but actually it wouldn't do the same thing in a different data set. Another next step would be that, again, someone independent might update this model with another prognostic factor. So they might take all of these predictors but add another predictor to it to see if that increases the accuracy even further or to a higher level than what I got in this data set. Then even if all of that happened and was wonderful, we'd also then need to look at the impact on clinical practise. So that would be, I suppose, more higher-level research of looking at the benefits and risks of using prediction models to direct the provision of early intervention in child mental health as opposed to standard care. So considering things like outcomes, patient satisfaction, time, use of resources, cost implications, staff satisfaction, all those broader health system usage of prediction models. And it would also be really helpful, I think, to do some patient and participant involvement research to investigate those risk thresholds. So at the moment, I picked the level of risk that defined whether you were low or high risk. But it would be really helpful to work with families of children who experiencing poor mental health, maybe adolescents with mental health, and/or different stakeholders to see, what level of risk do we think would be needed to say that a parenting group would be offered or whatever, those different kind of relationships? And obviously always, there's the economic analysis to ascertain the cost implications of balancing false positives and false negatives from using a prediction model. Yeah, so lots of work. Lots of work. [CHUCKLING] [AUDIO OUT] Yeah, so I suppose, I very much intentionally only wanted to put in predictors that a parent could answer or work out their score themselves if they wanted to. Obviously, we'd get much higher accuracy if we included certain, maybe, biological or clinical markers, but that's already then putting a barrier into getting this information or getting used because you then have to go on a wait list till you get whatever assessment or whatever scan or whatever bloods to pop into the model to get your actual risk score. So I feel really strongly for this to be something that can be usable. It needs to be these very easy, low-cost, non-invasive predictors that people can report or know or understand themselves so that they can use it. Or they can ask their public health nurse or their GP. If you think of the number of health contacts that a baby has from prenatal until they're two, there's no other point in the health system in your life where you will come into contact with that many health professionals who can potentially help to identify what supports you might need. So I think it's really important that the model would need to be something that's cross-discipline, cross service provider, cross everything, and that the parent is considered like an equal participant or an equal user in the model. And I suppose, from the ethical perspective, it is the bit that you touched on earlier of that self-fulfilling prophecy. So like, if I'm a parent, and I'm told my child has a high risk of later poor mental health, do I then start to see those attributes or characteristics or worry that those things are present in my child? But I suppose that's where we have to be really, really emphasising that it's probabilistic, and it's not deterministic, and that the brain is so plastic in the first seven years of life. And from the smallest things that we do with our child in terms of disconnecting ourselves from our own social media use, that will be an interesting area of research to look at so that we can be present for them. And we're all guilty of this as parents. It's so hard. You have emails coming in from work. You have 70,000 WhatsApps about the sports groups and the playdates and the PTA, and it is just incessant. But there's so much work looking at the cost of that on the relationship between the parent and the child in terms of you not really listening to them or being distracted or basic things, like, you're not looking at them when you answer them, and how that impacts on the relationship. So I suppose it's, again, a way to view of giving support rather than judgement, and so that we can share the information that we have as researchers or clinicians of the things that we know matter to support infant mental health. And emphasising that good enough parenting is good enough. But what does good enough parenting look like? We're not trying to be perfect parents or perfect humans-- they don't exist-- but trying to do a good enough job. So we're coming to the end of the podcast. I just wondered whether there's-- is there a take-home message? I suppose the take-home message is that children from 0 to 5, 0 to 7, 0 to 10 can and do experiencing poor mental health. Because I suppose it struck me over the years how it's felt that it's not possible for a small child to have poor mental health. And that's something that we need to worry about in adolescence and young adults, and actually just to think about their social and emotional well-being from birth and how, if we can get that right, it's such a critically important foundation for the rest of their life. Brilliant. Thank you. Great summary at the end. It's been lovely speaking to you, Emma. And you, Mark. [INAUDIBLE] Thank you for your time. I think kind of one of the big things for me is about how family circumstances and socioeconomic challenges can predict mental health risk. And I love that focus on prevention and really looking at how we can spot chances early. And obviously, that's not a guarantee about how things will turn out. But it gives the chance to really target support when it can most make a difference. It's also a strong call to think holistically from birth, recognising that early health experiences shape lifelong trajectories, and that equitable services could make a real difference without labelling or discrimination. Once again, don't forget to subscribe to the podcast. Sign up and subscribe to the YouTube channel. Search A-C-A-M-H, and you'll find us all across social media. Join us next week because we'll be speaking to Elaine Clark, whose work focuses on the experience of autistic individuals and the factors that support success. [MUSIC PLAYING]