Dr Anna Moore’s experience working as a child psychiatrist within the Child and Adolescent Mental Health Services (CAMHS) has convinced her that something has to change. With mental health problems affecting more than one in five children and young people in England, it’s a problem reaching near-epidemic proportions. Referrals have nearly doubled since the COVID-19 pandemic and resources have not kept pace with the rising demand. Young people are often waiting for many months before they are seen by a clinician and there are just not enough professionals to respond to the escalating need, with around 30% of posts in CAMHS currently unfilled, according to Dr Moore.
“As a child psychiatrist, I am working with patients all the time” she says, “but I’m incredibly frustrated with the system I’m having to work in. The waiting lists are often many months to years long, and by the time I see patients, what they are facing is often so severe that it’s much harder to intervene helpfully. We are working in a system that just doesn’t have the capacity or the funding available to help people in the way that the evidence for best practice says we should.”
As a clinician, Dr Moore admits she often feels powerless to make a difference. But her role as an Assistant Professor in Child Psychiatry and Medical Informatics in the Department of Psychiatry at the University of Cambridge has offered her – and her patients – a ray of hope.
“AI and machine learning give me the potential to be able to help change the system,” she says. “We are currently designing and building the first digitally supported preventative pathway for children’s mental health and we’re doing that by using AI and digital tools to think about how we can streamline the pathways to access.”
Easing the pressure with AI-driven solutions
Over the past six months, Dr Moore and her team have been consulting with paediatricians, mental health workers, GPs, social workers, commissioners, policymakers and a diverse group of patients and their carers to co-design a digitally supported clinical pathway, including a range of AI tools that she hopes will dramatically improve the lives of mental health professionals in the UK, along with the patients and carers they support.
Dr Moore’s passion is palpable when she talks about harnessing the potential of AI to help relieve pressure on the current system, particularly when it comes to assessing patients, which is currently a huge bottleneck. The goal of the AI-driven pathway is threefold, she says: first, to identify where support is needed for children in the community; second, to determine whether the child’s need requires CAMHS intervention or can be addressed within the community; and third, to provide community professionals with the tools to signpost children to appropriate support, helping to prevent mental health issues from escalating.
For those cases that are manageable within the community, AI-prescribed solutions might include digital tools, social support networks and groups, referrals to school counsellors or GPs counsellors and other resources, “almost like a personalised Google for children’s mental health”, Dr Moore explains. Crucially, this would ensure that lower-level mental health needs were identified early and met effectively without needing CAMHS referrals.
“These tools are not actually there to help child psychiatrists diagnose, because they don’t need that – they’ve been trained for over ten years to do that,” she says. “We want to put the knowledge of the psychiatrist into the hands of social workers and others, so they have access to information about whether there is a mental health need, and whether a CAMHS referral is needed, and if not, then how that young person and family can be best supported in the community.”
“We are increasingly understanding that a lot of care can be delivered according to needs, rather than diagnosis,” she adds. “AI can help us develop personalised signposting for young people and their families, allowing them to access help much more quickly. We’re going to try and make tools available in the community that can help professionals and potentially even families to identify need as early as possible. That way, those who need a CAMHS referral will not face such a long waiting list and will be able to access that specialised support more quickly.”
Harnessing the power of data
Underpinning Dr Moore’s research is a huge pool of routinely collected data about children and adolescents that her team has gathered to create CADRE (Child and Adolescent Data Resource), which links electronic health records. In the future it will also include genetic information and other data types. The long-term goal is to create a database that links the health, education and social care data for all children across the region, based at the new Cambridge Children’s Hospital.
Dr Moore and her team have brought together 12 years of anonymised data from around 300,000 children across Cambridge. They are using cutting-edge AI machine learning technology and large language models to help predict children’s mental health outcomes, which will be a key feature of the AI tool she is working on.
“The power of data is transformative,” says Dr Moore. “If I look at all the examples of how there’s been really significant transformations and improvement in quality of care, they’re data-driven – based on how we understand the data and how we use that data to make informed decisions.”
The CADRE concept has now been adopted by the Mental Health Mission, a large national programme to improve mental health research. The plan is to roll the programme out to cities including Birmingham, Liverpool and Manchester to make sure the data set used for training the AI models is as diverse as possible. The team is currently applying for permission to make CADRE data available for ethically approved research projects – once this is in place, the data can be used by other academic teams. This will provide a valuable resource that will speed up health transformation for children across the country.
“One of the big challenges with these digital solutions is that we absolutely must include representative data so that we’re not building tools that exacerbate inequality,” says Dr Moore. “By having this kind of national reach, involving data for around 3 million children, we’ll be able to have access to rare outcomes so we can predict these more accurately – and, with a more diverse cohort, we can build fair tools that can help everybody.”
So far the response from professionals and members of the public towards the project has been remarkably positive, in spite of the highly sensitive nature of the data needed to train the AI model.
“We did some work with the public when we were setting this up because it was really important that we were transparent about how we were planning to use the routinely collected data,” says Dr Moore. “We needed to make sure it was acceptable to the public. The overarching message we heard was: ‘why aren’t you doing this already?’”
“And we heard from older people who said: ‘you know, if I’d had my mental health problems identified early, then my whole life would have been very different and the opportunities available to me could have been very different’,” she adds. “It was actually really moving.”
“Not only did people say it was a good thing because of the benefit it can deliver,” she says, “but they also wanted to improve the quality of the data. So that’s something we’d like to do in future.”
None of this work can happen in isolation, however; collaboration is essential to the project’s success, according to Dr Moore.
“One of the barriers to implementing these tools is that they’re often made in a lab somewhere by someone who thought they were a good idea, but they’re not co-designed,” she says. “In our team, we’ve got engineers, computer scientists, mathematicians, machine-learning experts, data scientists, ethicists and psychologists. We want to really understand what the cutting-edge opportunities are so that we can build a system for the future, rather than a system for last year.”
The ai@cam initiative – which is bringing together experts from across the University of Cambridge to develop responsible AI-driven tools to tackle some of the world’s most pressing challenges – has been crucial to this collaborative effort.
“A platform like ai@cam, gives us the opportunity to bring together this multi-disciplinary team,” she adds. “If we didn’t have that, it wouldn’t be possible to do this project. ai@cam has been incredibly helpful in supporting us achieve what we have so far, and is a critical partner for us as we move into the next steps.”
Further information
From stigma to solutions: Read more about ai@cam’s ‘Responsible AI for better lifelong brain and mental health’ initiative.
