Decoding Death and Dying in people with Dementia by Digital thanotyping

Project Information

5-D

Grant agreement ID: 101088414

DOI

10.3030/101088414

EC signature date 2 August 2023

Start date 1 January 2024

End date 31 December 2028

Funded under

• European Research Council (ERC)

Total cost

€ 2 000 000,00

EU contribution

€ 2 000 000,00

Coordinated by

UNIVERSITETET I BERGEN
Norway

Periodic Reporting for period 1 - 5-D (Decoding Death and Dying in people with Dementia by Digital thanotyping)

Reporting period: 2024-01-01 to 2026-06-30

Summary of the context and overall objectives of the project

Decoding Death and Dying in people with Dementia by Digital thanotyping (5-D)
How can we recognize that a person with dementia is at the end of life? When we are dying, our physical, mental, and social abilities are gradually declining. No reliable method of predicting perceived dying currently exists although the technology is available (sensors, algorithms).

The aim of Decoding Death and Dying in Dementia by Digital thanotyping (5-D) is to provide methods and tools to diagnose and describe dying to an unprecedented level of accuracy and robustness, within a timespan larger than is possible now, focusing on the case of dying people with dementia as one of the most vulnerable and difficult to study groups. 5-D combines clinical assessment tools with wearable sensing technology to monitor a) pain and distressing symptoms, b) behavioral and psychological symptoms in dementia (BPSD), c) oral changes, and to decode “the point of no return” as the beginning of perceived dying.

To obtain this outcome in nursing home patients with dementia, I will test the main hypothesis: from monitoring the evolution of thanotype components over time and their interdependencies, the prediction of the “point of no return” is possible. The objectives of 5-D are:
O1. Collect data using sensors and validated assessment scales.
O2. Develop estimation methods for BPSD from sensor measurements.
O3. Develop digital tools to capture the expression of pain.
O4. Determine the relationship between breathing and oral symptoms.
O5. Develop models for symptom interdependencies at the end of life and the “point of no return”.
O6. Perform human-in-the-loop validation of developed tools, models, and algorithms.

The ground-breaking interdisciplinary novelty of 5-D endeavors to enhance our understanding of end-of-life underlying pain and symptoms in people with dementia. Advancing our theoretical knowledge to uncover how, when, and why perceived dying can be identified opens the doors for transferable research across several scientific fields.

Work performed from the beginning of the project to the end of the period covered by the report and main results achieved so far

Technical and Scientific Activities and Main Achievements
The 5 D project advances a comprehensive, data driven framework for understanding dying trajectories in people with dementia living in Norwegian nursing homes by integrating multimodal clinical assessments with wearable and contactless sensing technologies. The project systematically collects longitudinal data from nursing home residents across multiple municipalities, enabling large scale characterization of pain, neuropsychiatric symptoms, sleep patterns, activity levels, and oral health changes toward the end of life.
A core scientific activity is the deployment of two complementary sensor systems: the Garmin Venu 3S smartwatch, which captures continuous heartrate variability, respiration, and activity metrics using optical and inertial sensing; and the Somnofy radar system, providing contactless monitoring of respiration, sleep stages, bed movement, and environmental parameters. These technologies generate high resolution physiological and behavioral time series that are synchronized with validated proxy rated clinical symptom scales collected throughout the disease trajectory. This integrated dataset forms the foundation for digital phenotyping and for identifying sensor derived correlates of agitation, pain, functional decline, and sleep disruption.
Achievements to date include the establishment of a large, multimodal dataset, with more than 200 enrolled participants, over 170 completed baseline assessments, and systematic follow up across six and twelve month periods. Sixty end of life events or deaths have been documented, yielding unprecedented opportunities for modeling symptom evolution during the final phase of dementia.
Complementary sub studies enhance the scientific breadth: the DIPH.DEM study tests feasibility and analytic strategies for activity based digital biomarkers; ORAL.DEM integrates oral health diagnostics in 129 participants, saliva and plaque sampling, and dental symptom trajectories; and a qualitative arm evaluates the ethical acceptability of sensor based monitoring from the perspective of family caregivers. These contributions strengthen cross domain interpretation of digital measures and support the development of clinically meaningful multimodal phenotypes.
Across objectives O1–O6, the project has refined methods for detecting agitation, pain, symptom interdependencies, and the transition into the terminal phase ("point of no return"). Early findings indicate substantial interdependence among symptoms and the potential of sensor based signatures to reveal otherwise undetected physiological and behavioral changes.
Overall, the 5 D project delivers a technically innovative, clinically integrated, and ethically robust platform for real time characterization of end of life trajectories in dementia. The resulting digital phenotypes are expected to form the basis for next generation assessment tools and personalized, comfort oriented clinical decision support.

Progress beyond the state of the art and expected potential impact (including the socio-economic impact and the wider societal implications of the project so far)

Results, Potential Impacts, and Key Needs for Further Uptake
The 5 D project has generated a comprehensive multimodal dataset integrating clinical assessments with continuous sensor-based monitoring. The deployment of wearable and contactless sensing systems, Garmin Venu 3S and Somnofy radar, has enabled stable large-scale acquisition of heart rate dynamics, respiratory patterns, sleep architecture, physical activity, movement signatures, and environmental conditions across multiple clinical sites. Combined with validated proxy rated symptom assessments, these datasets support high resolution modelling of pain, neuropsychiatric symptoms, agitation, sleep disturbances, and oral health changes throughout the disease trajectory and into the final phase of life.

The project has already achieved substantial operational progress, establishing a large cohort of more than 200 participants (expected sample size N=340), with structured follow ups and 60 documented end-of-life events, creating an unprecedented scientific foundation for analyzing terminal phase trajectories in dementia. Early analyses reveal consistent interdependencies among symptoms and suggest that multimodal sensor profiles can uncover clinically relevant physiological and behavioral shifts not reliably captured through proxy-based assessments alone, especially at night-time.

Complementary sub studies strengthen the project’s translational and methodological depth. These include proof-of-concept work on activity derived biomarkers (DIPH.DEM) integrated oral health diagnostics with biological sampling (ORAL.DEM) and qualitative research exploring relatives’ perspectives on the ethical acceptability of sensor-based monitoring at the end of life. Together, these contribute to the development of meaningful digital phenotypes and inform future clinical implementation.

Potential impacts include the establishment of next generation tools for automated detection of agitation, pain, sleep disturbance, and functional decline; refined digital biomarkers for identifying the "point of no return" in dying trajectories; and improved symptom recognition to support comfort oriented palliative decision making. These innovations have the potential to shift dementia end-of-life care toward more precise, person centered, and ethically grounded practice. To ensure further uptake and long-term success, several enabling conditions are required:
1. Further research and demonstration: Continued methodological refinement, multi-site validation, and deeper interrogation of sensor-symptom associations using advanced computational modelling.
2. Internationalization: Expansion of recruitment to diverse health system contexts, supported by UiB’s strong interdisciplinary collaboration structures.
3. Access to markets and commercial pathways: Industrial partnerships for device integration, secure data infrastructure, and scalable deployment models.
4. Regulatory and standardization frameworks: Strengthened ethical governance, particularly relevant given the involvement of cognitively impaired individuals, and harmonized standards for digital phenotyping in vulnerable populations.
5. IPR and data governance support: GDPR compliant data handling and institutional support for protecting emerging analytic pipelines and biomarkers.

Overall, the 5 D project provides a scientifically robust and clinically anchored platform with strong potential to transform the detection, interpretation, and management of symptoms in dementia. Following aspects of multimodal digital health research must be considered:
1. Ownership of Multimodal Sensor Data
2. Rights to Algorithms, Models, and Digital Biomarkers
3. Use of Proprietary Sensor Technologies
4. Protection of Software, Pipelines, and Analytical Frameworks
5. Ethical Legal Constraints on IPR in Vulnerable Populations
6. Future Commercialization and Market Access