JMIR Publications

ABSTRACT

Background:

Knowing a student’s depressive symptoms can facilitate significantly more precise diagnosis and treatment. But very few studies focused on depressive symptom predictions through unobtrusive systems and these studies are limited by smaller sample sizes, lower performance, and the requirement for higher resources. Besides, it is unexplored whether there exist statistically significant rhythms based on different app usage behavioral markers (e.g., app usage sessions) which could be potential in finding subtle differences to predict with higher accuracy like the models based on rhythms of physiological data.

Objective:

The main objective of this study is to explore whether there exist statistically significant rhythms in resource-insensitive app usage behavioral markers and predict depressive symptoms through these markers-based rhythmic features. Another objective of this study is to understand whether there is a potential link between rhythmic features and depressive symptoms.

Methods:

Through a countrywide study, we collected 2952 students’ raw app usage data and responses to the 9 depressive symptoms of the Patient Health Questionnaire (PHQ-9). The students’ app usage behavioral data were retrieved through our developed tool which was previously used in our pilot studies in Bangladesh on different research problems. To explore whether there is a rhythm based on app usage data, we will do a zero-amplitude test. Additionally, we will develop a cosinor model for each participant to extract the rhythmic parameters (e.g., acrophase). Besides, to get a comprehensive picture of the rhythms, we will explore the non-parametric rhythmic features (e.g., interdaily stability) as well. Apart from these, we will do a regression analysis to understand the association of rhythmic features with depressive symptoms. Finally, we will develop a personalized multi-task learning (MTL) framework to predict symptoms through rhythmic features.

Results:

After excluding participants to satisfy different requirements (e.g., having app usage data of at least 2 days to explore rhythmicity), 2902 students’ are kept for analysis where there are 24.48 million app usage events data and 7 days’ app usage data of 98.17% (n=2849) students. The participants were from all divisions of Bangladesh, both public and private universities, 52 different departments, and 19 different universities of Bangladesh. We are analyzing the data and will be publishing the findings in a peer-reviewed venue.

Conclusions:

Having an in-depth understanding of the app usage rhythms and their connection with depressive symptoms through a countrywide study can significantly help healthcare professionals and researchers to better understand depressed students and may create possibilities for using app usage-based rhythms for intervention. Additionally, the MTL framework based on app usage rhythmic features may more accurately predict the symptoms due to the rhythms' capability to find subtle differences.