Lampe EW, Collins AC, Lee A, et al. Interindividual differences in digital phenotypes of major depressive disorder: A passive sensing study using smartphone and wearable sensor data. Behav Res Ther. 2026;199:104986. doi:10.1016/j.brat.2026.104986
This study aimed to illuminate digital phenotypes for major depressive disorder (MDD) symptoms using passive data collected from smartphones and wearable sensors. Five previously established digital biomarkers of depression were used as the foundation of the data collection plan and analysis: 1) sleep disturbance (e.g., shorter sleep duration, more time spent awake during the night), 2) low activity level (measured via accelerometry), 3) affective dysregulation (measured via heart rate variability (HRV)), 4) increased screen time, and 5) social withdrawal (i.e., less time spent in microphone-detected conversations, fewer calls/texts sent). Participants (n=297) used the MLife smartphone application, which continuously collected passive data over a 90-day period. Latent profile analysis (LPA) was used to identify the different depression phenotypes. The analysis identified two to three different phenotypes of depression. One large group showed moderate and relatively stable levels across all digital indicators. The other groups were characterized by shorter sleep duration, more time spent awake during the night, lower heart rate variability, and reduced social engagement. Greater inconsistency in sleep patterns was also common in these groups, suggesting poorer sleep overall. These findings suggest that people with depression may experience different symptom patterns that can be detected using passive sensing data alone. Sleep, heart rate variability, and social engagement appeared to be the most important factors distinguishing these groups and were closely related to difficulties in social and work functioning. The study provides new evidence that passive sensing can identify meaningful differences among people who are already depressed. However, the lack of a control group makes it difficult to consider how these patterns differ from an individual’s non-depressed state. Future research should build on this work and examine additional indicators to further refine the digital phenotypes of depression.