The RESILIENT Dataset: Multimodal Monitoring of Ageing-Related Comorbidities and Cognitive Decline

The growing ageing population and prevalence of comorbidities pose significant healthcare challenges, from increasing hospitalisations to dementia risk. Current healthcare systems primarily treat single conditions, overlooking the complex interplay of chronic diseases. Advances in wearable technology and remote healthcare monitoring technologies offer opportunities to enhance management of comorbidities and early intervention to improve healthcare outcomes. This study presents the RESILIENT dataset, a collection of physiological, sleep, and mental health assessment data conducted as part of an ageing-related comorbidities and dementia study. The RESILIENT study has developed a digital platform to integrate data from wearable devices and in-home monitoring technologies to track physiological, sleep, and cognitive patterns. The validation analysis using the Resilient data highlights correlations between cognitive function, mental health, physical activity, and sleep, aligning with existing literature. By leveraging this dataset, researchers can develop predictive models for early detection and personalised interventions aimed at reducing unplanned hospital admissions and improving health outcomes. 

• The Resilient digital platform, an open-source repository that provides software for collecting, storing, and analyzing in-home monitoring data is available at: https://github.com/tmi-lab/resilient.
• The open-source software for aggregating and analysing the dataset, including summary statistics, stratified analyses by gender and age group, and data visualizations, can be found at: https://github.com/tmi-lab/Resilient-Dataset.

The RESILIENT dataset is organised into four main components: 1) A CSV file containing demographic information and baseline assessments related to mental health (PHQ-9, GAD-7, GDS-12) and cognitive functioning (ACE-III) for all participants. For ACE-III, both baseline and 6-month follow-up scores are included; 2) a metadata CSV files describing variables present in the demographic and devices data; 3) a CSV summary file providing per-participant data coverage statistics, including the number of recorded days, average records per day, and the earliest and latest timestamps; and 4) individual participant folders containing raw time-series data, including sleep states and physiological features captured by sleep mats, as well as step counts and heart rate data recorded by smart watches. More specifically, there are four tables included in each participant folder: ScanWatch Steps, ScanWatch HeartRate, Sleep States, and Sleep Physiology. Each folder is named after the participant's unique identifier (UID), allowing cross-referencing between the device data and the demographic information.

Ethics statement:

The RESILIENT study has been reviewed and approved by the London-Surrey Borders Research Ethics Committee and the Health Research Authority and is registered on the Integrated Research Application System (IRAS) under reference number 321104. This publicly available dataset includes remote healthcare monitoring data and baseline mental health and cognitive assessments conducted throughout the monitoring period, providing a comprehensive resource for analysing health trends and detecting early signs of cognitive and physiological decline.

Dataset anonymisation:

A two-stage de-identification process was applied to the data. In the first stage, the data was pseudo-anonymised to develop analytical methods for the study. In the second stage, data was fully anonymised by removing all personally identifying information and any identifiable attributes. Participants are randomly assigned a Universally Unique Identifier (UID) to enhance security during de-identification. This ensures demographics and raw monitoring data from sleep mats and scan watches cannot be traced back to individuals while preserving the data’s utility for analysis.

*Note (13/07/2025) : ACE-III scores taken at 6 months are subject to a second quality control. If any changes are identified, an updated version will be issued.