Depression is among the most widespread mental health disorders worldwide, affecting an estimated 1 in 20 people. It is characterized by persistent sadness, hopelessness, disrupted sleep patterns, changes in appetite and a loss of interest in everyday activities.
While there are now various treatments for depression, including different types of antidepressant medications and psychotherapeutic approaches, not all depressed individuals have access to these resources or benefit from them. Reliably detecting the first signs of depression could be highly advantageous, as it could ultimately allow mental health services to intervene early, before symptoms worsen and the disorder becomes debilitating.
The analysis of data collected by smartphones, smartwatches and other wearable devices could potentially help to detect some early signs of depression, such as a lower mood, increased stress levels and behavioral changes. While various past studies explored the potential of mobile technologies for the early detection of depressive symptoms, the factors influencing the effectiveness of these tools remain poorly understood.
Researchers at Ghent University recently set out to better understand what contributes to the effectiveness of these technology-based solutions, by reviewing earlier papers that assessed their potential. The team’s review paper, published in Nature Mental Health, pinpoints types of data that are particularly helpful for detecting signs of depression, while also identifying computational models that appear to be the most effective for this specific application.
“Early detection of depressive symptom changes is vital for timely interventions,” Yannick Vander Zwalmen and Matthias Maerevoet wrote in their paper. “Mobile and wearable technologies enable continuous, unobtrusive monitoring of behavioral, psychological and physiological data, offering new possibilities for digital phenotyping and just-in-time prediction of depression. This scoping review synthesized findings from 52 studies to identify commonly used features, evaluate their predictive value and assess methodological approaches.”
Using smartphone data to predict mood changes
Vander Zwalmen, Maerevoet and their colleagues reviewed 52 past research studies that focused on predicting early signs of depression. These studies collected data using smartphones or wearable devices, then analyzed it with computational models to predict early signs of depression.
The data collected ranged from movement or location-related information, sleep patterns, physical activity patterns, communication patterns (i.e., how many calls users made and how many messages they sent or received), heart rate variability (HRV) and self-reported mood ratings. By reviewing the findings of earlier studies, the team tried to identify the data patterns that were most closely linked to early symptoms of depression.
“Features such as time spent at home, sleep variability and reduced mobility were strongly associated with depressive symptoms,” wrote the authors. “Combining physiological, behavioral and self-report data enhanced predictive performance. Personalized models and anomaly detection approaches outperformed generalized ones in predicting individual symptom changes.”
The researchers’ analyses revealed that depression symptoms were typically linked with irregular sleep patterns, a reduction in movement, little physical activity and a self-reported bad mood. In addition, models that were adjusted to consider a user’s unique habits and average biological signals appeared to predict early signs of depression better than general models.
Towards better mental health monitoring tools
Overall, this review study confirmed the potential of data collected by portable and wearable devices for the prediction of early depressive symptoms. In the future, it could guide the development of new mental health apps or other technological tools that detect signs of depression and share useful resources or the contacts of local mental health services with users.
“Mobile and wearable data show strong potential for just-in-time depression prediction,” wrote Vander Zwalmen, Maerevoet and their colleagues. “Future research should emphasize new features, diverse populations and personalized models to improve accuracy and real-world applicability.”