SOS-DR: a social warning system for detecting users at high risk of depression

• Published in: Personal and ubiquitous computing Vol. 26; no. 3; pp. 837 - 848
• Main Authors: Tai, Chih-Hua, Fang, Ying-En, Chang, Yue-Shan
• Format: Journal Article
• Language: English
• Published: London Springer London 01.06.2022; Springer Nature B.V
• Subjects: Bayesian analysis; Computer Science; Decay; Direct reduction; Electronic devices; Epidemiology; Mental depression; Mental disorders; Mental health; Mobile communication systems; Mobile Computing; Original Article; Personal Computing; Risk; Signs and symptoms; User Interfaces and Human Computer Interaction; Warning systems; Depression; Social diary; Warning system; Bayes method
• ISSN: 1617-4909; 1617-4917
• DOI: 10.1007/s00779-017-1092-3

Mental disorders have become a major disease observed in people with a contemporary lifestyle. Similar to many physical diseases, prevention and earlier detection are critical for mental health. This paper proposes a mobile device-based system, referred to as the SOcial warning System for Depression Risk (SOS-DR), to automatically estimate the risk of a user experiencing depression, identify users with a high risk, and provide them with help by recommending useful information (such as the Center for Epidemiological Studies Depression Scale (CES-D) questionnaire) and warning their close friends. For these purposes, SOS-DR provides a friendly interface for users to post on Facebook and derives a risk score of depression for each user by monitoring the frequencies with which symptoms of depression are mentioned in his/her online posts. Once a user’s risk score is detected to be high, SOS-DR recommends useful information and warns the user’s close friends to take additional and timely care of him/her. In the inference of risk scores, SOS-DR adopts a weighted Bayesian method and introduces a time decay factor in the calculations of symptom weights to highlight the impor- tance of recent symptoms. Therefore, the SOS-DR can monitor users’ mental states in daily life and send timely alerts. Evaluations using real online posts have showed that the effectiveness of time decay-weighted Bayesian methods is highly consistent with that of the CES-D. System usability studies and qualitative comparisons with other similar systems have also demonstrated the ease of use and the advantages of SOS-DR.