Accepted for/Published in: JMIR Formative Research
Date Submitted: Jun 16, 2022
Open Peer Review Period: Jun 16, 2022 - Jun 28, 2022
Date Accepted: Nov 22, 2022
(closed for review but you can still tweet)
Passive monitoring of depression and anxiety among workers using digital biomarkers on their physical activity and working conditions: A two-week longitudinal study
ABSTRACT
Background:
Digital data on physical activity is useful for self-monitoring and the prevention of depression and anxiety. Although previous studies have reported machine/deep learning models using physical activity for passive monitoring of depression and anxiety, there are no models for workers using working conditions and feedback connections to deal with a series of pieces of information.
Objective:
The purpose of this study is to evaluate the performance of a deep-learning model for psychological distress optimized for workers.
Methods:
A two-week observational study was conducted for workers in urban areas, Japan. Absent workers at the baseline were excluded. In the daily survey, psychological distress was measured. As features, activity time by intensity was determined using the Google Fit application. In addition, we measured age, gender, occupations, employment status, work shift types, working hours, and whether the response date was a working or non-working day. A deep learning model, using the long short-term memory (LSTM), was developed and validated to predict psychological distress on the next day from the features. The performance of the model was evaluated by the correlation between the predicted and measured values of psychological distress.
Results:
A total of 1,661 days of supervised data were obtained from 236 workers aged 20 to 69. The correlation coefficient between the predicted values by the deep learning model and measured values of psychological distress in test data was 0.619 (R2=0.383). Overall classification accuracy was 79.9%, and the accuracy for the subthreshold-, severe-, and light-level psychological distress were 46.7%, 30.2%, and 94.1%, respectively.
Conclusions:
The developed deep learning model showed the same performance as in previous studies and, in particular, high accuracy for light-level psychological distress. Working conditions and LSTM could be useful in maintaining the model performance for monitoring depression and anxiety through digitally recorded physical activity in workers.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.