Sleep Wearables

Table of Contents

  1. Measuring Sleep Health with Wearables
  2. Challenges and Opportunities to Measure Sleep Health
  3. Validation of sleep measurements from wearables

Measuring Sleep Health with Wearables

Wearable devices for sleep monitoring hold particular significance for older adults and those with Alzheimer’s Disease and Related Dementias (AD/ADRD). These devices are increasingly important given their ability to continuously monitor sleep patterns with minimal intrusion, a crucial feature in managing and understanding the health dynamics of older populations. Wearables such as smartwatches and fitness bands come equipped with sensors designed to track various sleep stages, interruptions, and overall sleep quality. Despite their potential, using these technologies in elderly care requires careful consideration of several factors, including device accuracy, user-friendliness, potential biases, and the burden of continuous wear, especially in populations that might struggle with technology use or have varying physical limitations.

The adoption of wearables for sleep monitoring in older adults and those with AD/ADRD presents unique challenges and opportunities. While these devices are convenient and packed with advanced sensors, their effectiveness can be limited by factors such as the physiological and behavioral differences inherent in aging populations. Issues such as skin elasticity, peripheral perfusion, and movement disorders can affect sensor accuracy. Moreover, the cognitive impairments typical of AD/ADRD may complicate the operation of these devices and influence compliance. Despite these challenges, the continuous data collected from such wearables are invaluable in detecting early signs of sleep disturbances, which are often precursors or exacerbating factors for cognitive decline, mood disorders, and decreased quality of life. This section will explore these issues in depth, offering insights into the reliability and practicality of wearable sleep trackers, helping researchers understand their potential and limitations before implementation in studies.

Challenges and Opportunities to Measure Sleep Health

Clinical Validation Core Co-lead Dr. Rebecca Spencer gave a broad introduction to sleep, its importance for physical, mental, and cognitive health, and how various factors such as aging and lifestyle can impact sleep in the above MassAITC webinar. She explained the different stages of sleep, including slow wave sleep, REM sleep, and stage 2 non-REM sleep, and their associated functions. For example, slow wave sleep is crucial for memory consolidation, emotion processing, and clearing neurotoxic waste from the brain through the glymphatic system. REM sleep is linked to creativity, decision-making, and emotion regulation, while stage 2 sleep is associated with motor learning and plasticity.

Dr. Spencer then delved into sleep measurement techniques, focusing on the gold standard polysomnography (PSG) and its challenges, such as discomfort for participants and time-consuming data analysis. She also discussed the evolution of actigraphy, from basic wrist-worn devices to more advanced commercial devices that incorporate additional physiological measures like heart rate, respiration, and temperature to improve sleep stage estimation. However, she highlighted the limitations of these devices, including inconsistencies across different models, lack of validation in diverse populations, and the influence of factors like age, BMI, and alcohol use on their accuracy.

Specific challenges in measuring sleep in older adults and individuals with Alzheimer’s disease were also discussed. These include forgetfulness, agitation, and physiological changes that can affect the accuracy of sleep-tracking devices. Dr. Spencer emphasized the importance of validating devices for specific populations and considering factors such as battery life, comfort, and ease of use when selecting a device for a particular study.

Finally, Dr. Spencer discussed opportunities for improving sleep measurement and intervention. She suggested focusing on developing devices tailored to the abilities and physiology of older adults, validating devices for specific populations and study designs, and exploring non-pharmacological interventions to enhance specific sleep stages. For example, targeted memory reactivation techniques using olfactory or auditory cues during slow wave sleep or stage 2 sleep have shown promise in enhancing memory consolidation and motor learning. Overall, while there are challenges in accurately measuring and intervening in sleep, particularly in older adults and those with neurodegenerative diseases, there are also exciting opportunities for future research and development in this field.

Validation of sleep measurements from wearables

Accurate sleep monitoring is crucial for older adults, particularly those with Alzheimer’s disease (AD), Alzheimer’s disease-related dementias (ADRD), or other forms of dementia. Sleep disturbances, such as insomnia, excessive daytime sleepiness, and irregular sleep patterns, are common in these populations and can exacerbate cognitive decline, behavioral symptoms, and overall quality of life. Monitoring sleep metrics, such as sleep efficiency, sleep latency, and rapid eye movement (REM) latency, can help identify sleep disturbances early, facilitate timely interventions, and potentially improve disease management and care planning for older adults with cognitive impairments.

Kainec et al, Sensors 2024 [1] A study from Dr. Rebecca Spencer compared the accuracy of five consumer sleep-tracking devices (Fitbit Inspire HR, Fitbit Versa 2, Garmin Vivosmart 4, Oura Ring Gen 2, and Withings Sleep Tracking Mat) and research-grade actigraphy (Actiwatch Spectrum Plus) against polysomnography (PSG) in measuring sleep. The study recruited 53 healthy young adults aged 18-30 years.

The key results are summarized below:

  • Total sleep time: All devices except the Garmin Vivosmart estimated total sleep time comparably to research-grade actigraphy. Agreement with PSG was good for Fitbit Inspire, Fitbit Versa, Withings Sleep Mat, and Oura Ring; moderate for Actiwatch; and poor for Garmin Vivosmart.
  • Wake after sleep onset: All devices overestimated short wake times and underestimated long wake times. Agreement with PSG was poor for all devices.
  • Light sleep: Fitbit Inspire and Versa had low absolute bias. Withings Mat and Garmin Vivosmart overestimated short light sleep and underestimated long light sleep. Oura Ring underestimated light sleep. Agreement with PSG was poor for all devices.
  • Deep sleep: Bias was low for Withings Mat and Garmin Vivosmart. Other devices overestimated short deep sleep and underestimated long deep sleep. Agreement with PSG was poor for all devices.
  • REM sleep: Bias was low for all devices. Agreement with PSG was moderate for Fitbit Inspire and Versa, and poor for others.

Overall the results suggest that proportional bias patterns are prevalent in consumer sleep-tracking technologies and could impact their overall accuracy. While some devices performed well for certain measures like total sleep time, further improvements are needed in the reliability and accuracy of sleep staging by consumer devices.

Lee, JMIR 2023 A study by Lee et al. [2] compared the accuracy of 11 consumer sleep trackers (CSTs) in estimating sleep metrics against polysomnography. Among the wearables, the Galaxy Watch 5 demonstrated the best performance in estimating sleep efficiency with a minimal bias of -0.4%, while the Apple Watch 8 showed the lowest bias of 0.81 minutes for sleep latency estimation. For REM latency, the airable SleepRoutine exhibited the best performance with a bias of 1.85 minutes. However, the study also revealed distinct trends in sleep measure estimation based on the type of device. Wearables showed high proportional bias in sleep efficiency, indicating a tendency to overestimate sleep, particularly in individuals with low sleep efficiency. On the other hand, nearables exhibited high proportional bias in sleep latency, suggesting an overestimation of sleep onset latency, especially in those with prolonged sleep latency.

The implications of these findings suggest that while CSTs have the potential to monitor sleep in older adults and those with AD/ADRD/dementia, their accuracy varies significantly across devices and sleep metrics. Wearables like the Galaxy Watch 5 and Apple Watch 8 may be more suitable for estimating sleep efficiency and sleep latency, respectively, while airables like SleepRoutine could be a better choice for REM latency estimation. However, the biases observed in wearables and nearables should be considered when interpreting sleep data, especially in populations with pre-existing sleep disturbances. Further research is needed to validate these findings in older adults and those with cognitive impairments, as well as to develop more accurate and reliable sleep monitoring technologies tailored to the specific needs of these populations.

Miller, Sensors 2022 A study by Miller et al [3] examined the validity of six commonly used wearable devices (Apple Watch S6, Garmin Forerunner 245 Music, Polar Vantage V, Oura Ring Generation 2, WHOOP 3.0, and Somfit) for assessing sleep compared to polysomnography (PSG). The study found that all six devices demonstrated moderate to high sensitivity (>90%) in detecting sleep epochs, but Polar, Oura Gen 2, WHOOP 3.0, and Somfit outperformed Apple Watch and Garmin in detecting wake epochs (specificity). For multi-state sleep classification (e.g., light, deep, and REM sleep), the devices ranged from 50% to 65% agreement with PSG. Oura Gen 2, WHOOP 3.0, and Somfit had the highest relative agreement for multi-state sleep. The findings suggest that these devices can be used as an alternative to PSG for estimating total sleep time in older adults and those with AD/ADRD/dementia. However, further improvements are needed for accurate sleep stage classification, and the better-performing devices (Oura Gen 2, WHOOP 3.0, and Somfit) may be used to monitor sustained, meaningful changes in sleep architecture.

References

[1] Kainec, Kyle A., Jamie Caccavaro, Morgan Barnes, Chloe Hoff, Annika Berlin, and Rebecca MC Spencer. “Evaluating Accuracy in Five Commercial Sleep-Tracking Devices Compared to Research-Grade Actigraphy and Polysomnography.” Sensors 24, no. 2 (2024): 635.

[2] T Lee, Y Cho, KS Cha, J Jung, J Cho, H Kim, D Kim, J Hong, D Lee, M Keum, CA Kushida, Accuracy of 11 wearable, nearable, and airable consumer sleep trackers: Prospective multicenter validation study, JMIR mHealth and uHealth, 2023

[3] Miller, Dean J., Charli Sargent, and Gregory D. Roach. “A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults.” Sensors 22, no. 16 (2022): 6317.