Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Student Mental Health Screening via Fitbit Data Collected During the COVID-19 Pandemic

Rebecca Lopez, Avantika Shrestha, ML Tlachac, Kevin Hickey, Xingtong Guo, Shichao Liu, Elke Rundensteiner

TL;DR

This study investigates passive screening of college student mental health using Fitbit data collected during the COVID-19 pandemic. It combines five Fitbit modalities with self-reported measures for depression, anxiety, and perceived stress to evaluate classifiers across multiple time-aggregation levels. Key findings show that sleep and heart-rate signals provide strong screening power, with F1 scores reaching up to 0.83 for anxiety and 0.81 for stress, and F1 around 0.78 for depression in several configurations; AdaBoost and Random Forest emerge as particularly effective depending on modality and target. The work demonstrates the feasibility of scalable, unobtrusive campus mental health monitoring and provides a transparent data-processing and modeling pipeline, including an open-source code repository.

Abstract

College students experience many stressors, resulting in high levels of anxiety and depression. Wearable technology provides unobtrusive sensor data that can be used for the early detection of mental illness. However, current research is limited concerning the variety of psychological instruments administered, physiological modalities, and time series parameters. In this research, we collect the Student Mental and Environmental Health (StudentMEH) Fitbit dataset from students at our institution during the pandemic. We provide a comprehensive assessment of the ability of predictive machine learning models to screen for depression, anxiety, and stress using different Fitbit modalities. Our findings indicate potential in physiological modalities such as heart rate and sleep to screen for mental illness with the F1 scores as high as 0.79 for anxiety, the former modality reaching 0.77 for stress screening, and the latter modality achieving 0.78 for depression. This research highlights the potential of wearable devices to support continuous mental health monitoring, the importance of identifying best data aggregation levels and appropriate modalities for screening for different mental ailments.

Student Mental Health Screening via Fitbit Data Collected During the COVID-19 Pandemic

TL;DR

This study investigates passive screening of college student mental health using Fitbit data collected during the COVID-19 pandemic. It combines five Fitbit modalities with self-reported measures for depression, anxiety, and perceived stress to evaluate classifiers across multiple time-aggregation levels. Key findings show that sleep and heart-rate signals provide strong screening power, with F1 scores reaching up to 0.83 for anxiety and 0.81 for stress, and F1 around 0.78 for depression in several configurations; AdaBoost and Random Forest emerge as particularly effective depending on modality and target. The work demonstrates the feasibility of scalable, unobtrusive campus mental health monitoring and provides a transparent data-processing and modeling pipeline, including an open-source code repository.

Abstract

College students experience many stressors, resulting in high levels of anxiety and depression. Wearable technology provides unobtrusive sensor data that can be used for the early detection of mental illness. However, current research is limited concerning the variety of psychological instruments administered, physiological modalities, and time series parameters. In this research, we collect the Student Mental and Environmental Health (StudentMEH) Fitbit dataset from students at our institution during the pandemic. We provide a comprehensive assessment of the ability of predictive machine learning models to screen for depression, anxiety, and stress using different Fitbit modalities. Our findings indicate potential in physiological modalities such as heart rate and sleep to screen for mental illness with the F1 scores as high as 0.79 for anxiety, the former modality reaching 0.77 for stress screening, and the latter modality achieving 0.78 for depression. This research highlights the potential of wearable devices to support continuous mental health monitoring, the importance of identifying best data aggregation levels and appropriate modalities for screening for different mental ailments.
Paper Structure (23 sections, 10 equations, 6 figures, 6 tables)

This paper contains 23 sections, 10 equations, 6 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Burden and Prevalence of Mental Illnesses
  3. Barriers to Diagnosis and Digital Screening
  4. Our Approach and Contributions
  5. Related Literature
  6. The Student Fitbit Dataset
  7. Self-Reported Measures
  8. Wearable Sensor Data
  9. Methodology
  10. Data Cleaning and Imputation
  11. Aggregation Design Choices
  12. Weekly Segmentation and Feature Extraction
  13. Machine Learning Models
  14. Feature Selection and Hyperparameter Tuning
  15. Evaluation Methodology and Metrics
  16. ...and 8 more sections

Figures (6)

  • Figure 1: Screening for Mental Illnesses using Fitbit Data Modalities.
  • Figure 2: Distributions of CES-D, STAI, and PSS scores. Dashed line indicates the binary classification threshold.
  • Figure 3: Data Processing Pipeline: (a) Aggregation of fine-grained irregular time-series data into a unified coarser-grained granularity of 1 or more hours, with imputation performed during aggregation and experimentation with duration units (1, 4, 6, 8, 12, 24 hour aggregation choices); (b) Partitioning into week-long time series augmented by illness labels. (c) Transformation of each weekly time series into a record of derived features.
  • Figure : (A) Performance Across Modalities
  • Figure : (A) Performance Across Modalities
  • ...and 1 more figures