Detecting and monitoring dementia using dynamic digital biomarkers of night-time behaviour and sleep.

Our aim is to produce a digital biomarker, the Dementia Sleep Quality Index (DSQI), for use in identifying neurodegenerative pathology and progression towards dementia. The DSQI is generated from information about sleep quality collected using under-mattress technology that is passive and unobtrusive. it combines night-time behaviour, sleep quality and physiology to provide information about sleep disturbances produced in different stages of dementia. We urgently need improved ways to identify individuals at high risk of developing dementia for targeted interventions to improve brain health, clinical trials, and in due course earlier access to novel therapies. Existing diagnostics including brain imaging and cerebrospinal fluid testing can determine core features of Alzheimer s (amyloid, tau accumulation) and cerebrovascular disease. However, these are not viable means of screening large geographically dispersed populations. Additionally, the presence of pathologies alone is unlikely to be sufficient to identify individuals who should be treated. For example, not everyone with significant brain amyloid will not go on to develop Alzheimer s dementia. Additional functional measures of the behavioural effects of neurodegeneration are necessary to determine an individual s proximity to dementia. Sleep and night-time behaviour are disrupted early by neurodegenerative pathologies and so provide a rational, feasible and acceptable means of non-invasive screening for neurodegenerative risk. At the UK Dementia Research Institute (UK DRI) we have developed a secure digital platform ( Minder ) to integrate digital technologies. We have evaluated and validated an under-mattress sensor that measures bed occupancy, sleep quality (deep sleep) and night-time physiology (heart and respiratory rate). We have analysed data from a large cohort of patients living with dementia, confirming these individuals have altered sleep quality, bed occupancy and high nocturnal awakenings. We have used machine learning to integrate these measures into a single digital biomarker, the Dementia Sleep Quality Index (DSQI) capable of identifying dementia-related abnormalities. Applying this digital biomarker into 3.7 million nights from the general population, we have shown that individuals with abnormal patterns of sleep can be identified. We will test whether the DSQI is: (a) sensitive to the presence of neurodegenerative pathology; and (b) whether it predicts an individual s progression towards dementia, measured by loss of brain tissue and cognitive decline. We will study members of Insight 46, a subsample of the British 1946 birth cohort. Now aged 76, they have had detailed serial measures of cognitive function, MRI, amyloid imaging, fluid biomarkers and tau PET. We will deploy under-mattress technology in 250 participants, 100 of whom have amyloid-positive preclinical Alzheimer s disease. Continuous measures of deep sleep, bed occupancy, awakenings and night-time physiology will be taken for around 2.5 years. This will allow us to validate the PSQI as a way of detecting the core biological underpinnings of dementia symptoms and predict neurodegeneration and cognitive decline. Via patient, public and professional involvement (PPPI) we will explore how this type of digital technology can best be deployed in general practice and memory clinics to inform the co-production of dashboard technology within Minder to communicate results to healthcare professionals.