Predicting circadian phase across populations: a comparison of mathematical models and wearable devices

Abstract

From smart work scheduling to optimal drug timing, there is enormous potential in translating circadian rhythms research results for precision medicine in the real world. However, the pursuit of such effort requires the ability to accurately estimate circadian phase outside of the laboratory. One approach is to predict circadian phase non-invasively using light and activity measurements and mathematical models of the human circadian clock. Most mathematical models take light as an input and predict the effect of light on the human circadian system. However, consumer-grade wearables that are already owned by millions of individuals record activity instead of light, which prompts an evaluation of the accuracy of predicting circadian phase using motion alone. Here, we evaluate the ability of four different models of the human circadian clock to estimate circadian phase from data acquired by wrist-worn wearable devices. Multiple datasets across populations with varying degrees of circadian disruption were used for generalizability. Though the models we test yield similar predictions, analysis of data from 27 shift workers with high levels of circadian disruption shows that activity, which is recorded in almost every wearable device, is better at predicting circadian phase than measured light levels from wrist-worn devices when processed by mathematical models. In those living under normal living conditions, circadian phase can typically be predicted to within 1 hour, even with data from a widely available commercial device (the Apple Watch). These results show that circadian phase can be predicted using existing data passively collected by millions of individuals with comparable accuracy to much more invasive and expensive methods.

Distinct Circadian Assessments From Wearable Data Reveal Social Distancing Promoted Internal Desynchrony Between Circadian Markers

“…We next examined the alignment between the predicted DLMO phase and the estimated CRHR. We found that 78% of the predicted DLMO timings from the validated methods of Huang et al (5) were within the 80% confidence interval of the CRHR phase. This matches the results shown in Figure 1 and provides additional evidence that separate circadian rhythms in the body are aligned under normal circumstances.…”

“…In these studies, individuals needed to isolate themselves from external Zeitgebers such as light, and researchers had to monitor physiological signals such as the onset of melatonin secretion (DLMO) over intervals ranging from hours to almost 2 days (1)(2)(3)(4). Two recent techniques have been developed which now allow researchers to supplement these typical intensive experiments through mobile assessment of circadian rhythms using wearables (5,6). Wearables typically collect data on wrist movement (actigraphy) and heart rate, each of which can separately be used to estimate various outputs of the circadian clock in the body (5)(6)(7)(8)(9)(10)(11).…”

“…Two recent techniques have been developed which now allow researchers to supplement these typical intensive experiments through mobile assessment of circadian rhythms using wearables (5,6). Wearables typically collect data on wrist movement (actigraphy) and heart rate, each of which can separately be used to estimate various outputs of the circadian clock in the body (5)(6)(7)(8)(9)(10)(11). Using mathematical models and activity measurements collected by the Apple Watch, Huang et al recently showed that DLMO can be predicted to within ∼1 h in non-shift workers (5), which is much more accurate (in terms of mean absolute error between a laboratory DLMO measurement and predicted phase) than cosinor analysis of ambulatory circadian rest activity [the average difference between in laboratory DLMO and the acrophase from cosinor analysis of rest-activity cycle is 4.47 and 4.6 h from Huang et al (5) and Woelders et al (12), respectively].…”

“…All data and reports that are sent to or from the app's servers are transmitted anonymously. Analysis of this rich wearable data set using the methods of Huang et al and Bowman et al could help determine how social distancing and the 2020 COVID-19 lockdowns affected circadian rhythms ( 5 , 6 ).…”

The Promise of Sleep: A Multi-Sensor Approach for Accurate Sleep Stage Detection Using the Oura Ring

“…We have added three sensor-independent features to capture these two components, a cosine function, an exponential decay, and a linear function. The cosine function represents the circadian drive component [26,40]. The second feature represents the decay of homeostatic sleep pressure across the night, most dramatically during the first hours of sleep which are rich in deep NREM sleep.…”