Relationship Between Physical Activity Intensity and Physiological Stress Indicators in Wearable Sensor Data: A Correlational and Regression Analysis

This study aimed to investigate the ability of the intensity of physical activity measured with an accelerometer to predict three physiological signals, which are typically considered stress indicators in wrist-worn wearables. The 100 participant-session recordings in a public wearable-sensor dataset of an Empatica E4 (Empatica, Milan, Italy) were screened for artifacts (e.g., implausible physiological values, sensor faults, duplicated data), and means were computed per session for 96 participant sessions that passed quality control for the following signals: accelerometer, heart rate, electrodermal activity (EDA), and skin temperature. Three hypotheses related to the correlation of accelerometer magnitude with HR, EDA, and TEMP were tested using Pearson correlation and simple linear regression. Three hypotheses were tested regarding the correlation of the three measures: HR, EDA, and TEMP with magnitude on the accelerometer using Pearson correlation and simple linear regression. No significant correlation was found between activity intensity and any of the three physiological variables (r = –.005 to –.057, all ps ≥ .58), nor were any of the three regression models significant, accounting for less than 1% of the variance for all of the outcomes. There was a moderate correlation between heart rate and EDA (r = .424, p < .001), which are both linked to the sympathetic branch. These results do not suggest that physical activity does not affect HR, EDA, or TEMP, but rather that the session-level aggregation approach is limited by condition heterogeneity and/or statistical power. It is likely that the signal collected from a wearable sensor is a combination of physical activity and psychological stress, and that these two components can be separated using dynamic, time-resolved, and condition-specific analyses.