Using wearable device data can help in the early detection of depression

The method is based on “digital phenotyping,” which involves continuous monitoring of behavioral and physiological markers. The study confirmed that the development of depression correlates with a set of specific changes in the data, including an increase in time spent at home, a sharp instability in sleep cycles, and a general decrease in physical mobility. In addition to physical activity, significant predictors were identified, including changes in communication patterns (frequency of calls and messages), heart rate variability, and subjective mood assessments provided by users.

An analysis of computational methods has shown that personalized models and anomaly detection algorithms are significantly superior to generalized systems. The effectiveness of the forecast directly depends on whether the algorithm takes into account the unique habits and biological norms of a particular individual. This allows the system to identify deviations from the usual rhythm of life as potential signs of a depressive episode.

The practical application of the research results is aimed at creating “just-in-time prediction” systems. In the future, mobile applications will be able to automatically notify users of risks and offer contacts to relevant services before the disorder takes a destructive form. The authors point out the need for further testing of these models on more diverse population samples to improve the accuracy of automated diagnostics.