Abstract
BACKGROUND: More than 1 in 8 people potentially live with a mental health disorder, yet fewer than half receive treatment. Poor mental health awareness may contribute to this treatment gap, and digital health technologies, like wearables and their associated phone- and web-based applications, have the potential to reduce the mental health awareness gap due to their ease of adoption, objective feedback, and high rate of engagement. OBJECTIVE: This study aimed to better understand the relationships between mental health and objective wearable-derived metrics. METHODS: We examined the longitudinal results of monthly mental health surveys (Patient Health Questionnaire-2, Generalized Anxiety Disorder 2-item, and Perceived Stress Scale) delivered over 13 months to 181,574 individuals wearing a device (WHOOP, Inc.) that measures sleep, cardiorespiratory parameters, and physical activity (up to 307,860 survey responses and 7,942,176 days of total wear time). Generalized linear mixed models, cross-lag analyses, and intrapersonal scaling were used to assess interpersonal and intrapersonal associations between wearable-derived metrics and mental health outcomes. Age, gender, BMI, and time of year were used as covariates in the models. RESULTS: Interpersonal associations between wearable-derived metrics and mental health outcomes indicate that individuals with better sleep characteristics (ie, longer sleep durations and more consistent wake and sleep times), higher heart rate variabilities (HRV), lower resting heart rates (RHR), and higher levels of physical activity report lower levels of depression, anxiety, and stress. Intrapersonal associations between wearable-derived metrics and mental health outcomes displayed similar results as the between-person analyses, with higher HRVs, lower RHRs, and more physical activity generally coinciding with improved mental health outcomes. However, intrapersonal wearable-derived sleep metric associations diverged from the interpersonal association findings when specifically looking at sleep duration and depression, whereby increased sleep durations within an individual were associated with higher levels of depression. In interpersonal analyses, the largest association observed was between the Perceived Stress Scale scores and RHR, with a standardized coefficient of 0.09 (P<.001); in intrapersonal analyses, the largest association observed was between the Patient Health Questionnaire-2 scores and summated heart rate zones-a proxy for physical activity-with a standardized coefficient of -0.04 (P<.001). Cross-lagged models demonstrated that higher levels of reported stress preceded higher RHRs, respiratory rates, and sleep duration variabilities, as well as lower HRVs. CONCLUSIONS: Overall, this investigation reveals that numerous physiological variables measured by wearables are associated with mental state in free-living environments. These findings underscore the potential of wearable-derived physiological and behavioral monitoring to serve as objective complements to traditional subjective assessments in mental health research and care. However, given the complex nature of mental health disorders, further research is needed to determine how these metrics can be effectively integrated into clinical practice.