WEARABLE BIOMETRICS AS PREDICTORS OF MENTAL HEALTH DECLINE: A LONGITUDINAL STUDY

Abstract

The mental health disorders have turned out to be the major cause of disability globally as millions of people are affected by depression, anxiety and stress related disorders in diverse groups of people. Timely intervention of mental health decline is pivotal in terms of future disease burden, early intervention and quality of life. Most recent developments in wearable biometric devices such as smartwatches, fitness bands, and physiological monitors have made it possible to monitor physiological measures continuously and in real-time (heart rate variability (HRV), sleep, physical activity, galvanic skin response (GSR), and electrodermal activity (EDA)) on the wearable. These biomarkers have been associated with stress, mood variations, and cognitive functioning, and thus they have aroused the desire to use them as predictors of mental deterioration. This longitudinal research focuses on the hypothesis of whether wearable biometric data can be trusted to predict the onset or worsening of mental health symptoms in the future. Multi-sensor wearable devices were used to follow a cohort of 600 adult participants during 18 months. Continuous data streams on biometrics data were gathered, and regularized psychological surveys (e.g., PHQ-9, GAD-7, PSS) and clinical interview were conducted every quarter. Complex machine learning algorithms such as random forests, support features, and recurrent neural networks were trained on biometric and self-report data to identify early trends related to the worsening of mental health.