Wearable devices show how health conditions change sleep patterns

In a recently published study, NPJ Digital MedicineUsing a large dataset consisting of 5 million nights of sleep monitoring data from wearable devices, the researchers examined changes in an individual's sleep phenotype over time to determine whether changes in sleep patterns and phenotypes could provide information about the duration of acute illnesses such as fever or COVID-19.

study: 5 million nights: Temporal dynamics of human sleep phenotypesImage credit: New Africa/Shutterstock.com

background

Rapid advancements in wearable device technology have made wearable health monitoring devices easily available and affordable. Besides various health parameters, these devices are also widely used to monitor sleep patterns and quality.

However, despite the abundance of sleep monitoring data, translating insights gained from this data into actionable changes has been difficult because inter- and intra-individual combinations of sleep parameters change over time.

The National Institutes of Health recommends that adults should get 7-9 hours of monophasic sleep each night, but sleep research shows that sleep architecture is highly dependent on length and quality, and that these variations are associated with lifestyle and health-related factors.

Studies that have used clustering analysis on large-scale sleep data to quantify variation in sleep characteristics have been effective in characterizing sleep phenotypes but have only used cross-sectional data and have not considered inferences about disease and health that can be derived from longitudinal sleep data.

About the Research

In this study, the researchers used a large dataset of sleep monitoring wearable data from over 33,000 individuals, totaling over 5 million nights of data, to determine changes in sleep phenotypes over time and to understand whether these changes could be informative about health parameters or duration of acute illness.

Sleep periods in the large dataset were treated separately, and through clustering analysis, the researchers obtained a set of sleep phenotypes, including an insomnia-like phenotype consisting of less than 6.5 hours of fragmented sleep per day, and a preferred sleep phenotype of 8 hours of monophasic sleep.

The relevance of these phenotypes and changes in sleep phenotypes to disease and health was tested by examining whether transition probability patterns in a cohort of chronically ill patients significantly differed from those in a healthy cohort.

The researchers also tested whether transition probability patterns differed in the same individuals before and after the disease.

Data were collected via self-reported survey responses and sleep-wake time series from 33,152 participants over a 10-month period. Sleep monitoring data from a wearable smart ring device was also obtained from all participants.

Data were divided into non-overlapping sleep periods of 3–6 consecutive nights and used to determine sleep phenotypes by clustering analysis.

Features typical of each sleep period were used to identify predefined sleep phenotype clusters. Transition patterns and distribution of sleep periods over time for each individual were used to determine changes in sleep phenotype.

Additionally, the study also explored the distribution of sleep phenotype changes among groups of individuals with sleep apnea, influenza, diabetes, fever, and COVID-19.

result

Results reported 13 sleep phenotypes associated with sleep quality and duration, and found evidence of change in an individual's sleep phenotype over time.

Furthermore, sleep phenotype transition patterns showed significant differences over time between groups of individuals with and without chronic diseases and health conditions, and within individuals.

This study found that sleep phenotypes not only reflect individual dynamics, but also that changes in sleep phenotypes provide information about health status.

Furthermore, assessment of the temporal dynamics of sleep patterns revealed that current sleep patterns indicate potential changes in sleep phenotype, for example, shorter duration of deep sleep indicates a transition towards an insomnia-like sleep phenotype.

The temporal dynamics of sleep phenotype transitions were also found to be indicative of an individual's chronic diseases and health-related factors. Dynamic transition models were found to be more informative about an individual's respiratory and cardiometabolic health factors than specific sleep phenotypes.

Conclusion

Overall, the study identified 13 sleep phenotypes associated with sleep duration and quality and found that these phenotypes vary between individuals based on health status and within individuals over time.

Temporal transition patterns of sleep phenotypes were also indicative of chronic disease states, such as respiratory and cardiometabolic diseases.

These findings highlight the importance of long-term sleep analysis and temporal dynamics assessment in drawing actionable inferences from wearable sleep monitoring data.