Developing biomarker-informed and socioeconomically contextualised artificial intelligence models to map dementia-related hospital care pathways and quantify the associated NHS costs
Older people with dementia are frequent users of hospital services, yet we lack comprehensive understanding of their care pathways and associated costs. Multiple comorbidities, insufficient community social care and varying socioeconomic circumstances create complex patterns of hospital utilisation. Evidence suggests that those with stronger family support and higher socioeconomic status are less likely to experience unplanned A&E (accident and emergency) visits and emergency admissions, and are more likely to receive regular health reviews and planned care.
Understanding pathways of hospital use among older people with dementia requires integrating socioeconomic factors with biological indicators of disease progression. By linking the English Longitudinal Study of Ageing (ELSA) to Hospital Episode Statistics (HES), via UK Longitudinal Linkage Collaboration, this project will examine how physiological markers of ageing, such as APOE-ε4, DNA methylation, C-reactive protein and fibrinogen, interact with socioeconomic conditions to influence cognitive decline and hospital care utilisation. Employing advanced AI methods including gradient boosting and random forests, we aim to map a typology of dementia-related hospital care pathways and to identify individuals at highest risk of incurring substantial NHS costs. Illustrative care pathway types include stable use of health care, repeated alternation between community and hospital (the “community–hospital shuttle”) or intensive hospital use.
This project is part of the 'Careforce', 'Communities' and 'Frugal Innovation' clusters within King’s Better Health & Care Hub.
Aims
This study aims to map dementia hospital care pathways, generating evidence on care types, transitions and intensity to help policymakers understand carer challenges and improve support. By identifying population segments at risk of cognitive decline and those bearing the greatest costs, the project will inform targeted interventions for communities and vulnerable families.
Leveraging 'Frugal Innovation', the study uses AI models on existing linked datasets (ELSA and HES) to provide cost-effective, scalable insights without new data collection. This adaptable framework supports local councils and policymakers in evidence-based policymaking, maximizing publicly funded resources to enhance continuity of care.
Methods
Employing advanced AI methods including gradient boosting and random forests.
Principal Investigator
Reader in Health Economics and Policy
Investigators
Professor of Ageing and Health
Director, Centre for Healthy Brain Ageing
Professor of Health and Social Care
Professor of Social Care Research
Affiliations
Funding
Funding Body: Better Health & Care Hub
Amount: £19,880
Period: April 2026 - March 2027