With the ubiquity of smartphones and wearable accelerometer units, movement data has become easily procurable. The easy availability of such data makes it convenient to detect changes in physical activity patterns in real time. Such changes in physical activity levels are one of the most recognizable features of depression. In this work we conduct recurrence quantification analysis to explore how recurrences of patterns in physical activity data differs between depressed and healthy individuals, collected as part of the MOOVD project. We find significant differences (p<.05) in the mean and entropy of the diagonal line distributions and in the LAM to DET ratio. These seem to suggest that the mean duration of recurrent physical activity patterns and the diversity associated with these periods are less in depressed individuals as compared to a non-depressed group. We further explore whether the changes in these quantifiers precede a transition towards a depressive episode. For this we use a sliding window approach to calculate recurrence quantifiers from actigraphy data leading up to a depressive transition. The data was collected from individuals who were tapering their medication, as part of the TRANS-ID project.
