Identifying Periods of Cyclical Stress in University Students Using Wearables In-the-Wild

TL;DR

This study investigates the feasibility of detecting cyclical stress at the population level among university students using in-the-wild wearable ring data. It employs seasonal detrending, baseline normalization, and daily-maximum waking HR within REML-based mixed-effects models to identify stress signals tied to academic periods. The findings show population-wide stress elevations during exams, New Year, and job-hunting periods, demonstrating that unobtrusive wearables can reveal group-level mental-state dynamics while preserving individual privacy. The work has practical implications for university wellbeing policies and highlights avenues for integrating contextual data and self-reports in future research.

Abstract

University students encounter various forms of stress during their academic journey, including cyclical stress associated with final exams. Supporting their well-being means helping them manage their stress levels. In this study, we used a wearable health-tracking ring on a cohort of 103 Japanese university students for up to 28 months in the wild. The study aimed to investigate whether group-wide biomarkers of stress can be identified in a sample having similar daily schedules and whether these occurrences can be pinpointed to specific periods of the academic year. We found population-wide increased stress markers during exams, New Year's, and job hunting season, a Japanese job market peculiarity. Our results highlight the available potential of unobtrusive, in-situ detection of the current mental state of university student populations using off-the-shelf wearables from noisy data, with significant implications for the well-being of the users. Our approach and method of analysis allows for monitoring the student body's stress level without singling out individuals and therefore represents a privacy-preserving method. This way, new and sudden stress increases can be recognized, which can help identify the stressor and inform the design and introduction of counter measures.

Figures (6)

• Figure 1: Daily Oura ring data availability (indicative of usage) of participants.
• Figure 2: User data distribution per calendar week. Every separated color band corresponds to a single user and the width of the band shows what percentage of the total data available for that calendar week is from that user. No single user or small group of users dominates the data of any given calendar week.
• Figure 3: Left: HR recorded during sleep for one participant (blue). The seasonal fluctuation (in unison with daily sunlight hours) is clearly visible. We fit a sinusoidal model (orange) in order to detrend the data. Right: Sleep HR after detrending.
• Figure 4: Exemplary illustration of estimation of baseline. A skewed normal distribution is fitted (blue line) to the distribution of a participant's HRV values (scattered dots) from the whole observed period. The argument maximum of the fit, the mode $m^0$, is considered the baseline of that measure (here: HRV) for that participant.
• Figure 5: Mixed effects linear model coefficients (unit: MAD) plotted against the respective calendar week for normalized sleep HR, sleep HRV, waking HR and waking HR max. Every graph represents a single fitted model for a single variable denoted at the top of the graph. The background coloring indicates the period according to section \ref{['sec:perioddef']}: pink - pre-exam, red - exam, green - break, gold - golden week, orange lines - grade result release. The grey box gives the intercept for the reference value -- the median of all values during the semester -- as well as the model fit ($R^2$). Circle dots denote a p-value $<0.1$, square dots a p-value $<0.05$.