Predicting recurrence of depression using lifelog data: an explanatory feasibility study with a panel VAR approach

Abstract

Background: Although depression has a high rate of recurrence, no prior studies have established a method that could identify the warning signs of its recurrence. Methods: We collected digital data consisting of individual activity records such as location or mobility information (lifelog data) from 89 patients who were on maintenance therapy for depression for a year, using a smartphone application and a wearable device. We assessed depression and its recurrence using both the Kessler Psychological Distress Scale (K6) and the Patient Health Questionnaire-9. Results: A panel vector autoregressive analysis indicated that long sleep time was a important risk factor for the recurrence of depression. Long sleep predicted the recurrence of depression after 3 weeks. Conclusions: The panel vector autoregressive approach can identify the warning signs of depression recurrence; however, the convenient sampling of the present cohort may limit the scope towards drawing a generalised conclusion.