We are developing a programme of work around the acute care of people living with dementia. Existing strands include: expertise in delirium epidemiology, interface geriatrics services (step-up/step-down pathways between acute and community care), established acute frailty improvement platforms (with ICS, pan-London, national and international reach). Expanding these in the context of a capacity-building post-doctoral training programme will enable transformative dementia care in the acute setting.
Delirium and dementia care in acute hospitals have consistently been highlighted as under-researched areas (e.g. James Lind Alliance, Cochrane Dementia and Cognitive Improvement Group). Our PPI work with the Alzheimer’s Society specifically prioritised research into cognitive recovery after delirium. Though we adopt a whole-system approach for health services research across ARC-NT, we will build on our partnership with Care City in the London Borough of Barking and Dagenham - an area facing particular urban health inequality challenges. All of our work is underpinned by established stakeholder engagement, and we have a history of innovative and inclusive PPI.
We will address cognitive and functional recovery after delirium, one of the commonest complications of dementia. Delirium is distressing for patients and their families, and associated with multiple poor outcomes, including increased length of stay, higher rates of institutionalisation, and cognitive and functional decline. Despite this great need, there has been little research to date on how best to maximise delirium recovery. Our aim is to develop an evidence-based integrated inpatient / outpatient delirium recovery programme. This could have high impact for people living with dementia.
This project involves the use of multiple longitudinal population datasets of acute illness in older people with cognitive and other adverse outcome measures:
- First, investigators will work on the DELPHIC study, a prospective longitudinal cohort with baseline and long-term follow up data, as well as intervening inpatient admission episode delirium ascertainment, including patients over age of 70 in the London borough of Camden;
- Then, they will move to the DECIDE study, a comparable study nested within CFAS based in Newcastle, which will be employed as an external validation dataset. Here, investigators will utilise unsupervised non-linear machine techniques to cluster data, such as t-stochastic neighbour embedding (TSNE) as well as autoencoders, then comparing the clinical efficacy of data driven clusters to high-dimensional prediction models of outcomes.
Finally, these techniques will be employed across to larger, broader but perhaps less detailed datasets, assessing the generalisability of each model depending on available data.