Can Technology Help Detect Emotion Dysregulation in Young People?

Writer’s note: This article has been co-written by Aeron Kim and Asilay Seker

In mental health services, clinicians write thousands of notes every day. These records capture the details of people’s lives: how they feel, what they struggle with and how they respond to treatment. Hidden within these words is an enormous amount of knowledge about mental health but most of it has never been analysed in a systematic way.

The reason is simple. Clinical notes are written in natural language, using words rather than numbers. They are full of nuance and context which makes them invaluable for clinicians but difficult for researchers to analyse on a large scale.

Over the past decade, Natural Language Processing (NLP), a branch of machine learning and artificial intelligence, has begun to change this. NLP enables computers to process and interpret human language, making it possible to study large collections of text and extract meaningful information from them. In mental health science, this means we can now learn from the real-world experiences recorded in electronic health records (EHRs).

Image Source: Unsplash+

What NLP Brings to Mental Health Science

Most people have encountered NLP in everyday life through predictive text, translation tools or voice assistants. In mental health science, its power lies in its ability to interpret clinical language.

Clinicians often use different words to describe similar concepts. For example, “mood swings,” “labile affect,” and “difficulties managing emotions” might all refer to the same experience.

NLP algorithms can identify these variations, group them together and transform them into structured data that researchers can work with. This allows patterns to be detected across thousands of patients, helping answer questions that were previously impossible to study using traditional methods. Rather than replacing clinical expertise, NLP supports clinicians and researchers in making sense of what is already recorded in clinical notes, revealing large-scale trends and relationships that would otherwise remain hidden.

The Clinical Record Interactive Search (CRIS) Platform

King’s College London and the South London and Maudsley NHS Foundation Trust (SLaM), has set up a unique infrastructure to process free-text clinical record for research and quality improvement purposes, which is called the Clinical Record Interactive Search (CRIS) platform.

CRIS provides researchers with secure, anonymised access to millions of mental health records from SLaM services across South London, compatible with numerous NLP algorithms to structure clinical notes into analysable data. Using CRIS, researchers can study real-world clinical data without identifying individuals. Within this system, a growing library of NLP applications has been developed to extract information from text, capturing everything from clinical symptoms and medications to complex emotional phenomena such as mood instability, suicidal thoughts or self-harm behaviours.

Image Source: Unsplash+

Exploring Mood Instability in Neurodevelopmental Disorders

Mood instability, which refers to rapid and/or frequent changes in mood, is common in psychiatric conditions but often lacks standardised screening. Much like the related construct of emotion dysregulation, it is likely to predict adverse outcomes in clinical paediatric populations. However, research investigating this issue is scarce, possibly due to data collection challenges. Traditional studies often rely on small samples or self-reported data, which can miss important patterns in how emotional instability presents in clinical practice.

While most NLP algorithms are traditionally used in adult mental health data, recent research has shown their applicability in child and adolescent mental health. A recent body of research led by the CAMHS Digital Lab at King’s College London explored mood instability in children and adolescents utilising Child and Adolescent Mental Health Services (CAMHS). By applying NLP to large-scale clinical data, the study examined the presence of mood instability in mental health records and how it relates to clinical outcomes such as cannabis use.

How the Study Was Conducted

The study analysed anonymised electronic health records (EHRs) from children and young people aged 11 to 18 who received care through SLaM’s CAMHS between 2008 and 2022. Using a previously validated NLP algorithm, the team searched clinical notes for language related to mood changes, including terms such as “mood swings,” “unstable mood,” and “rapid cycling mood.” These mentions were identified within three months of a depression or ADHD diagnosis. This approach allowed the researchers to estimate how common mood instability is across diagnostic groups and to explore its associations with cannabis use.

Key Findings

• NLP-identified mood instability is associated with increased odds of cannabis use in both depression and ADHD groups.

• There was 25% higher likelihood of cannabis use due to mood instability in adolescents with ADHD, compared to those with depression.

• The prevalence of NLP-identified mood instability aligns with existing literature and estimates of emotion dysregulation in the ADHD population, supporting the utility of this method for CAMHS patients. (Seker et al., 2025)

  
  

Why These Findings Matter

Together, these findings highlight mood instability as a clinically significant but potentially under-recognised feature in young people utilising CAMHS. They also demonstrate the potential of NLP analysis as a scalable identification method, enhanced by tailored algorithms to capture the wider spectrum of emotion dysregulation within clinical records.

Mood instability can have a major impact on young people’s lives affecting relationships, school engagement, and overall wellbeing, and is often associated with increased use of services and medication. The results from this study suggest that clinicians should consider mood instability, or the broader spectrum of emotion dysregulation, as an important aspect of assessment and treatment planning for children and young people with commonly diagnosed conditions such as depression and ADHD.

Early recognition could help improve outcomes and reduce distress for young people and their families. This work further demonstrates the value of NLP in unlocking insights from complex clinical data. Analysing the records of more than thirteen thousand young people manually would have been unfeasible, yet NLP made it possible to carry out this analysis efficiently and objectively while maintaining the strict data governance and anonymity standards of the CRIS platform.

Beyond This Study: NLP in Medical Science

NLP is being increasingly used across medical science to turn text into data that can drive discovery and improve care. In physical health research, it can identify early signs of difficulties in clinical notes or radiology reports. In mental health research, it can detect constructs as complex as mood instability or self-harm, and track emotional changes over time, or monitor how people respond to treatment.

These approaches share a common purpose: to learn from existing data in ways that respect privacy and ethics while deepening our understanding of people’s experiences and care. As NLP models continue to advance, they are beginning to better distinguish contextual factors such as temporality and experiencer, offering a richer picture of mental and physical wellbeing than structured data alone can provide.

Image Source: Unsplash+

Looking Ahead

The CAMHS Digital Lab continues to build on this work by developing new NLP tools to study emotional dysregulation, self-harm and crisis risk in young people. One of these projects, supported by the Psychiatry Research Trust, focuses on improving the early identification of emotion dysregulation in CAMHS by leveraging the NLP methods and the online routine outcome measurement platform, myHealthE. These are innovative steps forward to personalise the CAMHS experience for young people and families.

Language is central to mental health. It is how clinicians describe symptoms and how young people express their experiences. We now have the tools that can analyse these words at scale without losing their meaning. By combining clinical expertise with advances in NLP, we can uncover patterns that were previously invisible and use them to guide earlier, more responsive and more compassionate care. This work is part of a growing effort to ensure that the stories recorded in clinical notes do not remain silent but instead help shape a future where every young person can be understood and supported at the right time.