Predicting depressive and manic episodes in patients with bipolar disorder using statistical process control methods on passive sensing data.

Early detection of emerging affective episodes is crucial in managing bipolar disorders (BD). Passive sensing-passive data collection via smartphone or wearable-offers a promising solution by potentially capturing altered activity, communication, and sleep patterns, indicative of manic and depressive episodes. Recently, statistical process control (SPC) has been introduced to psychopathology as a novel approach to identifying out-of-bounds processes. However, its application to mobile sensing data and to BD remains unexplored. To investigate SPC's potential in detecting emerging affective episodes, we utilized the BipoSense study, which monitored patients with BD. The BipoSense data cover 12 months of continuously collected passive sensing data via smartphone app, daily e-diary data, and biweekly expert interviews, that is, 26 in a row, to assess the psychopathological status. Compliance was excellent. A total of 26 depressive and 20 (hypo)manic emerging episodes in 28 patients were included in the analyses. SPC charts and multilevel analyses revealed heterogeneous results. Passive sensing, despite its potential as a low-burden, continuous measurement tool, did not demonstrate robust detection of affective episodes or preepisode weeks. Self-rated current bipolar mood, assessed via e-diary, outperformed passive sensing parameters in predicting current episodes, whereas predicting preepisode weeks was also limited. Notably, SPC with personalized control limits did not surpass established clinical cutoff scores. Even after systematic optimization of SPC settings, the combination of detected emerging episodes in relation to false alarms was insufficient for clinical use. Future studies warrant mobile sensing parameters closer aligned to psychopathology, thereby increasing validity, sensitivity, and specificity. (PsycInfo Database Record (c) 2025 APA, all rights reserved).