Researchers win Wellcome funding for early intervention data tool
Informatics
King’s researchers are one of six successful teams selected for the Wellcome Mental Health Data Prize UK 2026. The award will support the team to develop a data tool that uses Large Language Models (LLMs) to transform fragmented medication information in mental health records into research-ready longitudinal medication histories.
The tool will combine structured prescribing data captured in free-text clinical notes and will be used by researchers, NHS analysts, and policy teams to analyse prescribing trajectories, medication switching, and multiple use of different medications to give a clearer picture of medication use over time.
The project will be carried out by a multidisciplinary team of researchers, led by Dr Yamiko Msosa, based within the NIHR Biomedical Research Centre (BRC): Maudsley Informatics Theme and use secure NHS-approved infrastructure.
The tool will be developed through iterative input from clinicians, researchers, and people with lived experience through targeted Patient and Public Involvement and Engagement, ensuring outputs are ethical, meaningful, and interpretable. By improving the accessibility and analytic value of existing mental health data, the tool will strengthen evidence to inform early intervention research, policy, and service design, with potential to scale across datasets and settings.
Dr Yamiko MsosaResearch Fellow in Health Informatics at King’s IoPPN and lead applicant of the award
Mental health datasets contain rich information about medication exposure, changes over time, and associated outcomes for people with anxiety, depression and psychosis. However, much of this information is embedded in fragmented Electronic Health Records (EHRs) in the form of free text clinical documentation, making it difficult to use systematically for research, service evaluation, and early intervention analysis. While there is growing policy and research interest in improving early intervention and prevention in mental health, especially for psychosis, current data infrastructures limit the ability to analyse longitudinal prescribing trajectories, switching, and multiple medication use at scale.
Recent advances in natural language processing and LLMs provide an opportunity to transform how existing mental health data are structured and analysed. To date, most applications of LLMs in healthcare have focused on point‑of‑care or decision‑support use cases. There remains a clear gap in tools that responsibly apply these methods to improve access to, and interpretation of, complex mental health data for research and service‑level insight.
The tool will improve access to and interpretation of medication data enabling new analyses of treatment pathways and prescribing trajectories relevant to early intervention in psychosis, depression, and anxiety. Researchers will be able to investigate medication changes, complex regimens, and patterns of multiple medication use over time that may be associated with risk, relapse, or unmet need at population level. This project addresses a major barrier in mental health research by making complex medication data more accessible and interpretable, supporting new insights for early intervention in psychosis, depression, and anxiety.
