DECODE: mapping the challenges and requirements for Data-driven, machinE learning aided stratification and management of multiple long-term COnditions in adults with intellectual DisabilitiEs

Intellectual disability (ID) is a life-long condition affecting diagnosed in 0.5-1.0% of the UK general practice population.

Intellectual disability affects two aspects of an individual's functioning and these areas are; intellectual functioning (such as learning, problem-solving, judgement) and adaptive functioning (activities of daily life such as communication and independent living).

People with ID have a high prevalence of premature mortality, a large contribution to of which is from is the increased prevalence of multiple long-term conditions.

Two-thirds of people with ID have two or more additional long-term conditions, higher still among older individuals and people with severe ID.

This has led to policy recommendations to improve care for these individuals, including the provision of a named care coordinator for people with ID who also have multiple long-term conditions. However, this has not been implemented due to the complexity of developing a model of support.

There is very little information on the pattern of multiple long- term conditions and modifiable risk factors in this population.

Machine learning and data analytics have the potential to help us understand the co-occurrence, relationship and interaction of multiple-long-term conditions over time by leveraging insights from multiple sources of data.

Information analysed using this approach could go beyond diagnoses to include data about medication, other interventions, laboratory results as well as information on human phenotypes, genotypes and environmental factors. Information processed at this level could lead to the development of a data-driven approach to targeted interventions.

However, such a research would need a thorough preparation focusing on methodology, data availability and completeness, data governance, data interpretation/interaction and explainability.

The aim of the proposed development grant is to map key technology, human factor challenges, requirements and priorities of using machine learning to uncover the relationship between multiple clusters of long-term conditions and factors that may influence their course.

Methods include a review of the literature, focus groups and PPI consultations appropriate to meet the communication needs of people with ID and their carers.

A specification documenting the key technological, methodological and ethical requirements for preparing a future AIM Research Collaboration application will include: A list of key research questions that machine learning should address in order to understand the relationship between multiple long-term conditions in people with ID that is produced in collaboration with key stakeholders (experts, people with ID and their carers).

A map showing available machine learning methods and their appropriateness to explore the relationship between multiple clusters of long-term conditions and factors (including lifestyle) that may influence their course.

A record of available datasets – their relevance, completeness and access requirements – that should be considered when using machine learning to examine the relationship of multiple long term conditions in people with ID.

A checklist of ethical and human factors considerations that should be taken into account when designing data-driven interventions for the stratification and management of multiple long-term conditions in people with ID.

A network of research collaborators and project partners who will collaboratively support the submission of the final project concept.