Real-Time Comparisons of Adolescents' Self-Reported Stress and Physiological Stress Using Smartwatches and Ecological Momentary Assessments

Background and Purpose: Social work researchers and clinicians are searching for innovative tools that can help address the mental health crisis among adolescents. Just-in-time adaptive interventions (JITAIs) are flexible, personalized interventions designed to provide the right type and amount of support at the right time. JITAIs have begun to be implemented through wearables (e.g., smartwatches) linked to mobile devices. They are programmed using a sequence of decision rules that dictate how real-time treatment will be implemented based on each person’s internal physiology and external environment. While these interventions have the potential to be an important piece of adolescents’ care, it is imperative to understand whether the digital biomarkers they utilize accurately reflect adolescents’ emotional experiences. The purpose of this study was to examine the extent to which adolescents’ biological stress levels, measured through physiological readings on Garmin smartwatches, align with ecological momentary assessments (EMAs) of their self-reported stress levels.

Methods: Participants ages 13-18 were recruited through an outpatient clinic in New York City. Each participant wore a Garmin Vívosmart 4 smartwatch which tracked heart rate variability, a measure of stress. The watch continuously recorded physiological stress values from 0-100 for the duration of the one-month study period. Participants also completed brief EMAs that asked about self-reported stress on a 0-10 scale. To compare Garmin stress readings to participants’ self-reported stress in real-time, we identified the time when an EMA survey was opened, then compared the participant’s self-reported stress on the EMA to the average of all Garmin stress readings within sixty seconds after the moment the survey was opened. Multilevel modeling with Bayesian estimation was used to account for the nested structure of repeated measures within individuals.

Results: Average physiological stress values as measured by the Garmin watch varied widely both within and between participants (n = 33). While some participants had high variation on self-reported stress via the EMA, others had little to no variation, with 18% of participants self-reporting stress levels within a 2-point range or less. For the average adolescent, there was no significant association between self-reported stress levels and physiological stress levels as measured through the Garmin watch (b = -0.02, 95% CI = -0.13, 0.10). Across the group, adolescents varied from one another in their average self-reported stress levels (sd = 1.79, 95% CI 1.22, 2.38), and the correlation between physiological stress and self-reported stress was low across the scale range, regardless of whether a participant’s reported stress level was low or high (sd = 0.08, heterogeneity interval: -0.18, 0.14).

Conclusions and Implications: In this study, smartwatch measurements of stress were not an accurate reflection of adolescents’ self-reported psychological experiences. While there’s much enthusiasm around using smartwatches for real-time mental health interventions, social workers should be cautious about using physiological stress data as a proxy for real-time emotional distress. It is critical to resist scaling these tools quickly without young people’s input; instead, we must engage with adolescents to better understand whether they would find these technologies helpful in managing stress throughout their daily lives.