What does electrodermal sensing reveal? Insights from the Pixel Watch & Fitbit Sense 2

Major sporting events are opportunities for in-situ stress measurement. The 57th (LVII) Super Bowl featured the Kansas City Chiefs and the Philadelphia Eagles on February 12th 2023. This exciting back-and-forth game provided another opportunity to observe how a sports team winning or losing can impact our arousal levels — all within the timespan of the same game! In this match-up the teams were tied towards the end of fourth quarter before a game winning field goal in the final seconds that led to victory for Kansas City and the moment of highest sympathetic activation in our cohort. The average SCL for users whose home states was Kansas showed high variations and particularly prominent peaks during that final field goal, the final whistle and during the trophy award. In contrast, data from viewers who consented to this research in Pennsylvania indicated a different, less excited, physiological EDA profile towards the end of the game.

More mundane examples show that the wrist-worn EDA sensor captures higher arousal during meetings at work and when putting two toddlers to bed at the end of the day (shown below).

Circadian patterns in EDA

We also retrospectively analyzed anonymized data consisting of 10 million hours of continuous EDA data from almost 16,000 people, to date the largest EDA dataset developed. Diurnal and circadian patterns play an important role in helping to understand patterns of daily human behavior, and we wanted to understand if EDA can be used to observe these patterns.

For SCL we observed a nadir near 10pm and a zenith near 6am (both local time). This means that sympathetic activity was greatest in the early morning and lowest in the late evening. This result is consistent with previous research based on blood pressure measurements. We also observed a secondary minimum at 10am and a second spike at 6pm, meaning that the pattern is double peaked, unlike other physiological signals that typically have a single cycle. Interestingly, these spike-nadir patterns deviated significantly between weekdays (shown below in blue) and weekends (below in yellow), a difference that applied to several other physiological signals. These variations are not just explained by greater physical activity (e.g., more walking) but are driven by other psychosocial factors likely associated with work and non-work days. It is possible that greater levels of social interaction on weekends drives higher sympathetic activity.

Applications in mental health research

In our second paper, “Evidence of Differences in Diurnal Electrodermal, Temperature, and Heart Rate Patterns by Mental Health Status in Free-Living Data”, we analyze data for a more specific context, mental health and wellbeing. EDA is a measure of sympathetic arousal that has been linked to depression in laboratory experiments. However, outside of controlled settings, the inability to measure EDA passively over time and in the real world has limited insights about EDA as an indicator of mental health status. The availability of an EDA sensor on a consumer device enabled us to conduct research (with informed consent) into mental-health applications of the device.

We used data collected with informed consent as part of a prospective, nonrandomized study to investigate patterns and relationships between digital device use patterns. This data included sensor data from phones and wearables reflecting both behavioral and physiological processes, and self-reported measures of mental health and well-being. The study protocol was four weeks long with passive sensing from a wearable (Fitbit Sense 2) for the complete four-week period. We examined the association between passively collected, diurnal variations in EDA and symptoms of depression, anxiety and perceived stress in this sample.

We found that subjects who exhibited elevated depressive and anxiety symptoms had higher SCL, skin temperature, and heart rate, despite not engaging in greater physical activity, compared to those that were not depressed or anxious. These differences were statistically significant and consistent throughout most of the day for SCL and heart rate. The elevation was most prominent in the early morning/night time hours. In contrast, subjects who only exhibited elevated stress, based on the Perceived Stress Scale (PSS), only exhibited higher skin temperature. Most strikingly, differences in EDA between those with high versus low symptoms were most prominent during the early morning, suggesting there could be a connection to early morning awakenings and sleep disorders that are linked to depressive orders. We did not observe amplitude or phase differences in the diurnal patterns. Our work suggests that electrodermal sensors may be practical and useful in measuring the physiological correlates of mental health symptoms in-situ and that recent consumer smartwatches might be a useful tool for doing so.

Conclusions

The ability to regularly measure EDA on the wrist opens up exciting new avenues for research and personal understanding and could pave the way for personalized mental health support and improved well-being for individuals worldwide. By leveraging wearable technology and large-scale data analysis, we can uncover patterns in our physiological responses to stress, excitement, and even our daily routines. This knowledge can empower us to make more informed decisions about managing our well-being by recognizing the subtle ways our bodies react to the world around us.