Abstract

Passive sensing, which collects behavioral, physiological, social, and environmental indices with smart devices in an objective and automatic manner, is recently being explored as a tool for evaluating mental disorders. Among them, studies on depressive disorder are the most commonly performed. Although the expression patterns of depressive disorder may differ according to the symptoms and levels, studies that take this point into account while examining the relationship with sensor data are very limited. The purpose of this study was to identify sensor data that is highly related to depression symptoms, and to determine whether there is a difference in sensor data according to the level of depression. A total of 64 college students and graduate students were classified into three groups (normal, mildly depressed, and severely depressed) according to the level of depression. For a total of 30 days, self-reported data on 9 symptoms of depressive disorder were repeatedly collected 4 times a day. At the same time, a total of 8 sensor data were collected throughout the day using a smartphone and a smartwatch, and a total of 14 features were extracted as 4 values ​​per day. According to the results of this study, different features were found to have a significant relationship with each symptom. Some symptoms were found to be related with all features, some symptoms were related with only partial features, and some symptoms were not related with any features. In addition, it was found that there were significant differences in all features ​​according to the level of depression, and detailed patterns were different for each feature. Implications and limitations of the current study and directions for future research were further discussed.