JMIR Publications

ABSTRACT

Background:

Digital biomarkers are increasingly used to diagnose depression. Most existing research has focused on individual digital biomarkers (e.g., HRV or sleep data), whereas there are few integrative assessments or comprehensive explorations of the multidimensional aspects of depression.

Objective:

We systematically reviewed the role of digital biomarkers in depression assessment and management. Digital biomarkers, including physiological and behavioral indicators collected through wearable devices, smartphones, and sensors, offer objective, continuous, and real-time data; minimize recall bias; and provide insights into the multifaceted nature of depression.

Methods:

A comprehensive search of multiple international and Korean databases, including the Cochrane Library, OVID-MEDLINE, PsycINFO, CINAHL, IEEE Xplore, Web of Science, KISS, RISS, KMbase, and KoreaMed, was conducted without language restrictions. The search terms included variations of “depression,” “MDD,” “wearable,” “smartwatch,” “biomarker,” “sleep,” “speech,” “behavioral parameter,” “electroencephalogram,” and “electrocardiogram.” We identified 121 studies meeting the inclusion criteria, encompassing diverse biomarker categories, such as sleep, physical activity, cardiac parameters, speech, global positioning system data, and circadian rhythms.

Results:

The key findings revealed significant associations between depression and parameters. Meta-analyses showed that individuals with depression had significantly lower sleep efficiency (–3.10%, 95% confidence interval [CI]: –5.81 to –0.40, p = 0.02), longer sleep onset latency (+4.72 min, 95% CI: 2.46 to 7.04, p < 0.01), longer time in bed (+31.43 min, 95% CI: 24.19 to 38.68, p < 0.001), and more wake after sleep onset (+0.41 min, 95% CI: 0.01 to 0.80, p = 0.04). Among cardiac parameters, the nighttime mean heart rate was significantly higher in depressed individuals (+2.38 bpm, 95% CI: 0.16 to 4.60, p = 0.04). Moderate-to-vigorous physical activity was also significantly lower (–8.99 min, 95% CI: –10.84 to –7.13, p < 0.001). In contrast, other markers such as light physical activity and certain cardiac measures showed inconsistent results. Personalization emerged as a crucial factor, with specific biomarkers displaying stronger predictive power based on age, time, and individual variability. Furthermore, personalized multi-biomarker models may provide a more reliable means of tracking depression if they also feature temporal and longitudinal monitoring. Despite promising advancements, challenges, such as heterogeneity in methodologies and limited longitudinal research, suggest the need for standardized approaches and further exploration of qualitative metrics.

Conclusions:

This review highlights the potential of integrating multimodal digital biomarkers to enhance the precision of depression assessment and monitoring. By synthesizing evidence across diverse biomarker categories, this study identified several clinically meaningful indicators, providing a valuable foundation for future personalized and data-driven depression diagnostics. Clinical Trial: The protocol for this review was prospectively registered in the PROSPERO systematic review database (registry number: CRD42024518136).