One Week of Smartwatch Data Matches Clinical Dementia Screening Tools: What This Means for Digital Health
Passive monitoring via wrist accelerometry provides dementia risk assessment equivalent to established clinical tools, enabling population-scale screening without clinical visits.
Passive monitoring via wrist accelerometry provides dementia risk assessment equivalent to established clinical tools, enabling population-scale screening without clinical visits. This groundbreaking finding from a massive UK Biobank study of 47,371 participants demonstrates that we can identify dementia risk using just one week of smartwatch data—no blood tests, brain scans, or lengthy questionnaires required (Chan et al., 2024).
The Research Background
Dementia affects over 55 million people worldwide, with numbers expected to triple by 2050. Early identification of at-risk individuals is crucial for preventive interventions, yet current screening methods face significant barriers: they require clinical visits, blood tests, cognitive assessments, and accurate self-reporting of symptoms and family history. Many of these tools are impractical for population-scale screening due to cost, accessibility, and the burden they place on both patients and healthcare systems.
Researchers from NeuRA and UNSW, led by Lloyd Chan, sought to overcome these barriers by developing a dementia risk assessment using something many people already wear: a smartwatch.
A Groundbreaking Study in Scale and Simplicity
The study analyzed data from 47,371 UK Biobank participants aged 60 and over who wore wrist-worn accelerometers for seven days between 2013 and 2015. These participants were then followed for a median of 7.5 years to track who developed dementia. The scale alone makes this one of the largest studies ever conducted on wearable biomarkers and dementia risk.
What makes this research particularly groundbreaking is its simplicity. The prediction model requires only:
- One week of wrist sensor wear
- Age and sex information
- No dementia-specific information like family history or environmental exposures
- No blood tests or brain scans
- No questionnaires or self-reporting
How the Method Works
Participants wore an Axivity AX3 wrist accelerometer—similar to sensors in modern smartwatches—for one week during their normal daily activities. The device continuously captured movement data, which researchers analyzed using their Watch Walk algorithm (version 1.0) to extract 19 different digital biomarkers.
These biomarkers included:
- Movement patterns: Maximal walking speed, usual walking speed, daily steps, running duration
- Gait characteristics: Step time variability, walk hand positions
- Sleep characteristics: Bedtime, wake time, sleep duration, sleep efficiency
- Activity levels: Time spent in different activity intensities throughout the day
The beauty of this approach lies in capturing real-world behavior. Unlike laboratory tests that measure performance in artificial settings, the wrist sensor recorded how people actually moved and slept in their daily lives—providing a more accurate picture of functional capacity.
The Remarkable Results
During the 7.5-year follow-up period, 387 participants (0.8%) developed dementia. The analysis revealed that movement patterns, particularly maximal walking speed, were powerful predictors of dementia risk.
Key findings include:
- Lower walking speed and increased step time variability were the strongest indicators of future dementia risk
- Maximal walking speed showed the strongest association (hazard ratio of 0.68), meaning each standard deviation increase was associated with a 32% lower risk
- The prediction model achieved a C-statistic of 0.76—matching the performance of established clinical tools like the ANU Alzheimer’s Disease Risk Index (0.76) and the UK Biobank Dementia Risk Score (0.75)
- Early bedtime (before 9 PM) was associated with a 60% increased risk of dementia
Importantly, maximal walking speed proved to be a better predictor than usual walking speed. As the researchers explain, maximal walking speed may reflect individual functional capacities more accurately, while usual walking speed could be influenced by habitual preferences that mask subtle declines.
Why This Method Changes Everything
The study’s authors highlight several game-changing advantages of their approach:
Accessibility Without Compromise: The method achieves clinical-grade accuracy while eliminating nearly every barrier to screening. No need for blood draws, brain scans, or lengthy questionnaires. No reliance on accurate self-reporting, which can be affected by memory issues or social desirability biases.
Real-World Data: By capturing movement in natural environments rather than clinics, the method provides a more accurate assessment of daily functional capacity. People move differently in their homes and communities than they do in medical settings.
Better Compliance: Research shows that wrist sensors are better accepted by older adults and have higher compliance rates compared to waist-worn sensors. This matters for both initial screening and longitudinal monitoring.
Cost-Efficient Scaling: The approach enables population-level screening at a fraction of the cost of traditional methods. Once someone has a smartwatch or fitness tracker, reassessment can happen continuously without additional healthcare visits.
Automation Reduces Burden: The automated analysis eliminates the need for specialized personnel to administer and score assessments, reducing the burden on healthcare systems while maintaining accuracy.
What This Means for Digital Health Platforms
The implications of this research extend far beyond a single study. As the authors note, digital integration platforms “may provide a personalised weekly assessment of risk, enabling its use in clinical settings to evaluate and manage dementia risk and implement early preventive and management measures.”
This is precisely where platforms like Sahha come into play. At Sahha, we specialize in transforming raw biomarker data from smartphones and wearables into actionable health intelligence. Our research-backed intelligence layer doesn’t just collect data—it interprets patterns, identifies trends, and provides meaningful insights that can inform health decisions.
Consider what becomes possible when we combine multiple biomarker streams:
- Gait analysis (as demonstrated in this dementia study)
- Sleep patterns that reveal recovery and stress
- Activity levels indicating physical and mental wellbeing
- Behavioral patterns that signal early health changes
Our platform already provides five validated wellness scores—Sleep, Activity, Mental Wellbeing, Wellbeing, and Readiness—derived from these biomarkers. The dementia study validates what we’ve long believed: passive monitoring through consumer devices can provide clinical-grade health insights when paired with the right intelligence layer.
The Future of Health Prediction Through Wearables
This research represents a pivotal moment in preventive healthcare. We’re witnessing the transformation of everyday consumer devices into powerful health screening tools. The fact that simple movement patterns can match traditional dementia screening tools suggests we’ve only scratched the surface of what’s possible.
Future applications emerging from this type of research include:
- Early detection of Parkinson’s disease through movement pattern changes
- Cardiovascular risk assessment through activity and sleep patterns
- Mental health monitoring through behavioral biomarkers
- Recovery tracking after surgery or illness
- Population health surveillance without clinical infrastructure
As wearable sensors become more sophisticated and machine learning algorithms improve, we’re likely to see even more conditions become detectable through passive monitoring. The key insight from this dementia research is that we don’t need specialized medical devices—consumer wearables, when paired with validated algorithms and an intelligence layer, can provide equivalent screening power.
Building the Intelligence Layer for Healthcare
The gap between raw sensor data and actionable health insights is where platforms like Sahha operate. We provide the intelligence layer that makes sense of the millions of data points generated by wearables and smartphones. Our behavioral archetypes and wellness scores transform overwhelming data streams into understandable, actionable information for developers, researchers, and healthcare providers.
For researchers looking to replicate or extend studies like this dementia research, having easy access to processed biomarkers and behavioral patterns is crucial. Rather than building data collection and processing infrastructure from scratch, research teams can focus on what matters: understanding the relationships between digital biomarkers and health outcomes.
The Path Forward
While this study marks a significant breakthrough, it also highlights the importance of continued research and validation. The UK Biobank participants, while numerous, represent a specific demographic. Expanding this research to diverse populations and combining multiple biomarker streams will only strengthen our ability to predict and prevent disease.
What’s clear is that we’re entering an era where health monitoring becomes continuous, accessible, and integrated into daily life. The ability to identify dementia risk through a week of smartwatch wear demonstrates the transformative potential of digital biomarkers in healthcare.
For Researchers and Developers
If you’re a researcher interested in leveraging biomarkers and behavioral archetypes for your studies, or a developer building health applications that need an intelligence layer, we’d love to help. Sahha provides:
- Processed biomarkers from smartphones and wearables
- Behavioral archetypes that categorize user patterns
- Wellness scores validated through university research
- Easy API integration to accelerate your research or development
The future of healthcare lies in making powerful health insights accessible to everyone—not just those who can afford clinical visits. Studies like this dementia research prove it’s possible. Now it’s time to build it.
Learn more about Sahha’s platform | Explore our API documentation | Contact our research team
References
Chan, L.L.Y., Espinoza Cerda, M.T., Brodie, M.A., Lord, S.R., & Taylor, M.E. (2024). Prediction of incident dementia using digital biomarkers from wearables: The UK Biobank study. International Psychogeriatrics, in press. DOI: 10.1016/j.inpsyc.2024.100031