Machine learning will sift through millions of anonymised GP and hospital records from England and Wales to find hidden patterns in the health of people with intellectual disabilities. Around two-thirds of people with intellectual disabilities have two or more long-term conditions, yet their physical symptoms are often mistaken for behavioural problems or written off as part of their disability. This misattribution drives severe health inequalities. No existing model can predict how these conditions interact or what care a given person actually needs, making coordinated care between health and social services nearly impossible. If successful, the project will produce a visual toolkit that lets clinicians, carers, and people with intellectual disabilities see—in plain, transparent form—which clusters of conditions tend to occur together and how they evolve over time. The team will also combine relevant clinical guidelines into a single practical framework. The ultimate goal is actionable scenarios for holistic care coordination, shaped directly by interviews and workshops with the people who would use them. National bodies including NHS England, Public Health Wales, and the Royal College of Psychiatrists are already linked to the project, giving the findings a clear route into policy and practice.
View original technical description
Research question: What actionable insights can be developed from machine learning aided analysis of "multiple long term conditions in people with Intellectual Disability" to support a model of effective coordination of care? Background: Approximately 1% of people identified in general practice have ID and of these, around two-thirds have two or more long-term conditions. Physical ill-health symptoms in this population are often mistakenly attributed to either a mental health/behavioural problem or as being inherent to the person s ID which contributes to health inequalities. A principal manner in which support could be improved involves care coordination, which is defined as joined-up care between health and social services. However, the lack of ability to understand/predict the complex interactions between MLTCs and care needs of individuals makes it challenging to provide effective coordination of person-centred and holistic care. Aim: To apply machine learning approaches to identify clusters and trajectories of MLTCs in people with ID and to utilise them to develop actionable insights and practical usage scenarios for effective care coordination to improve the health and wellbeing of people with ID. Objectives: Our research is organised into six objectives: To mine and clean data from two independent data sources and platforms: the Clinical Practice Research Datalink (CPRD) for England and the Secure Anonymised Information Linkage (SAIL) Databank for Wales. To identify the clusters of MLTCs that exist in people with ID using appropriate machine learning approaches. To analyse the trajectories of the most dominant/important clusters and their interactions with risk factors and outcomes. To develop an approach to combine multiple clinical guidelines relevant to the dominant/important clusters. To design user-friendly visualisations that present the outputs of AI analytics in a transparent, meaningful and trusted way to professionals, people with ID and carers. To develop actionable insights and practical usage scenarios of AI with/for multiple stakeholders for the effective coordination of holistic care. Methods: WP1: Data mining/cleaning/bias mitigation. WP2: statistical analysis, hierarchical and non-hierarchical computational clustering methods, feature selection algorithms, deep learning classification methods, deep recurrent neural networks, computer-interpretable transition-based medical recommendation model, and AI argumentation methodology. WP3: user requirements analysis, prototyping and front-end design. WP4: reflective semi-structured interviews and stakeholder workshops (participatory design and World Café approach) Timeline for delivery: Month6: protocol/ethical approval, data mining/cleaning. Month12: MLTCs clustering and visualisation. Month18: risk factors analysis, temporal clustering and visualisation. Month24: advanced visual toolkit and system analysis of care coordination. Month30: actionable insights and practical usage scenarios. Anticipated impact and dissemination: Influence shaping policy and guidelines at the national level with support from National Learning Disability Professional Senate and our additional links to Royal College of Psychiatrists, Health Education England, Public Health Wales, NHS England and NHS Wales. Disseminate the outputs with support from our collaborators including Learning Disability Professional Senate and Learning Disability England. Further develop the proof of concept of practical usage scenarios, with the support of appropriate networks like AHSN, through piloting and adoption across services in the UK and beyond.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know