Identification, recording, and reasonable adjustments for people with a learning disability and autistic people in NHS electronic clinical record systems: studies using routine data, data linkage and machine learning to improve access to, and quality of, NHS care

RESEARCH QUESTIONS

- How well is learning disability and autism recorded in different electronic health records (EHRs)? What factors are associated with these conditions being unrecorded?

- What reasonable adjustments are recorded and for whom, and what is the impact of recording reasonable adjustments for people with a learning disability and autistic people on care processes and patient outcomes?

- Can machine learning (ML) be employed on EHR data to optimise identification of people requiring reasonable adjustments to care, including those with unrecorded learning disability or autism? BACKGROUND

People with a learning disability and autistic people experience significant health inequalities; narrowing these is a priority for the NHS. Providing reasonable adjustments can improve care and lead to better health outcomes. The NHS is implementing a Reasonable Adjustments Flag (RAF), a digital record that alerts staff to a person’s impairments and the need to provide specific adaptations.

However, many people with a learning disability and autistic people do not have their condition recorded in EHRs; this has implications both for research using care datasets and may also lead to people not being flagged. There is little information on how reasonable adjustments are recorded in this population and how this can be improved to support the RAF.

Novel methods of analysing ‘big data’, including ML, can be a powerful adjunct to providing high-quality care but are not fully utilised. AIMS & OBJECTIVES

To examine recording of learning disability and autism in EHRs across the lifespan, to better understand recording of reasonable adjustments, and use ML to identify unrecorded learning disability and autism and suggest individualised reasonable adjustments. METHOD

Leveraging national and regional EHR data from primary care, mental health services, and secondary care through new and established data linkages, we will:

1) Conduct a retrospective study across care settings to examine the recording of learning disability and autism in EHRs and examine individual and service factors associated with unrecorded learning disability and autism

2) Investigate historic recording of reasonable adjustments in primary care, examining disparities in recording and the relation between recording of reasonable adjustments and care processes and outcomes 3) Prospectively monitor changes in recording during the roll-out of the RAF scheme

4) Develop and validate ML models to i) identify people with increased likelihood of having unrecorded learning disability or autism and ii) suggest personalised reasonable adjustments to care

5) Conduct qualitative work to examine feasibility and implementation of predictive ML models for identifying reasonable adjustments ANTICIPATED IMPACT & DISSEMINATION

This project will last 3-years. It will demonstrate the representativeness of different EHRs for NHS commissioning and surveillance purposes and their generalisability when used for research. Novel ML models will automate the identification of people at increased likelihood of needing reasonable adjustments due to having a learning disability or autism and suggest specific adjustments.

Lived experience PPI groups will inform all aspects, including potential implications of ML approaches, help to interpret findings, and co-produce summaries. Multi-pronged dissemination, including specific outputs for policy makers and NHS leaders, will ensure the work has impact.