Mahir, A., Luong, N., Baryshnikov, I., Martikkala, A., Isometsa, E. and Aledavood, T.

Multi-Modal Sleep Measurement and Alignment Analysis in Outpatients with Major Depressive Episode

In review

  • Sleep plays a critical role in mental health, and this study explores how sleep can be tracked in real-world settings among individuals experiencing major depressive episodes (MDE). Using actigraphy, smartphone data, bed sensors, and ecological momentary assessment (EMA), the study evaluates how these different tools compare and where discrepancies arise.
    Sleep onset, offset, and total sleep time (TST) were measured over a two-week period in 172 participants, including healthy controls and individuals from three MDE subgroups: borderline personality disorder, major depressive disorder, and bipolar disorder. The consistency between the various measurement tools was assessed through Bland-Altman plots and Pearson correlation, and linear mixed-effects models were used to analyze factors influencing sleep alignment, taking into account demographics, daylight duration, and participant subgroup.
    Results showed that patients exhibited greater variability in sleep patterns than healthy controls. Actigraphy tended to overestimate TST when compared to bed sensors (by 0.48 minutes) and smartphones (by 0.99 minutes), while the smartphone generally underestimated TST relative to the other modalities. Older participants showed better alignment between actigraphy and bed sensors, as well as between smartphone and bed sensor sleep offset. However, TST alignment between smartphones and bed sensors was less accurate in females and in patients with bipolar or borderline personality disorder. Longer daylight duration was associated with improved alignment in TST and sleep offset measurements across all modalities.
    These findings emphasize the presence of measurement biases, the influence of seasonal and demographic factors, and the need to consider these discrepancies when using objective sleep tracking tools. Despite their promise for long-term monitoring, awareness of their limitations is essential for both research and clinical applications.

Ikäheimonen, A., Luong, N., Baryshnikov, I., John, T., Martikkala, A., Isometsä, E. and Aledavood, T.

Variability in Self-reported Depression Symptomology and Associated Mobile-Sensed Behavioral Patterns in Digital Phenotyping: An Observational Study

In review

  • Digital phenotyping studies have linked smartphone-sensed behaviors to depression, but variability between and within individuals limits generalizability. This study analyzed 64 patients with major depressive episodes (MDD, BPD, BD) to examine variability in depression symptoms and behavioral markers. Group-level differences were found in five PHQ-9 items, and behavioral associations with depression were weak at the sample level but highly mixed across individuals. Multilevel modeling showed between-person differences explained 58.5% of depression severity variance, increasing to 63.7% when behavioral, demographic, and contextual factors were included. Significant within-person predictors included nighttime communication use, morning screen use, battery variability, sleep duration, and SMS counts. These findings emphasize the need to account for symptom and behavior variability to improve generalization in digital phenotyping research.

Aledavood, T., Luong, N., Baryshnikov, I., Darst, R., Heikkilä, R., Holmén, J., Ikäheimonen, A., Martikkala, A., Riihimäki, K., Saleva, O., Triana, AM., Isometsä, E.

Multimodal Digital Phenotyping Study in Patients With Major Depressive Episodes and Healthy Controls (Mobile Monitoring of Mood): Observational Longitudinal Study

JMIR Ment Health 2025;12:e63622

  • Mood disorders are common worldwide, and digital traces from wearables and smartphones offer a way to monitor individuals in natural settings. In this study, 188 participants — including patients with major depressive, bipolar, or borderline personality disorders, and healthy controls — were followed for up to one year. We examined group-level differences in smartphone use, communication, sleep, mobility, and physical activity, as well as associations with depression severity. Patients showed lower location variance and entropy, less diverse communication patterns, but more varied smartphone use rhythms compared to controls. Incoming call duration and morning physical activity variability were negatively associated with depression severity, while outgoing call duration was positively associated. Although some behavioral differences were observed, they were modest at the group level, highlighting the need for multimodal approaches and strategies to address high dropout rates in long-term digital phenotyping studies.

Ikäheimonen, A., Luong, N., Baryshnikov, I., Darst, R., Heikkilä, R., Holmen, J., Martikkala, A., Riihimäki, K., Saleva, O., Isometsä, E., Aledavood, T.

Predicting and Monitoring Symptoms in Patients Diagnosed With Depression Using Smartphone Data: Observational Study

J Med Internet Res 2024;26:e56874

  • This study explored the use of smartphone-sensed behavioral data to detect and monitor depression symptoms in a cohort of 164 participants (patients and controls) over up to one year. Using PHQ-9 scores for labeling, statistical analyses and machine learning classified the presence of depression with 82% accuracy and changes in depression state with 75% accuracy. Key predictive features included screen-off events, battery levels, communication patterns, app use, and location data. The findings suggest that smartphone-based behavioral markers can effectively supplement traditional clinical assessments for detecting and monitoring depression.

Next

Behavioral sensing