Online Behavioral Digital Phenotyping for Depression: Scoping Review of Behavioral Features and Exploratory Assessment of Research Domain Criteria Construct Relevance
Journal of Medical Internet Research ·
Background: Depression is heterogeneous and dynamic, and changes in observable behavior may accompany some of its clinical manifestations. Digital devices offer new opportunities to capture this behavior in naturalistic settings continuously. However, the specific digital features captured through device interactions and their relationships to depression-related outcomes have yet to be synthesized, and their potential relevance to functional constructs is unclear. The research domain criteria (RDoC) may provide a conceptual framework for exploring potential links between digital behaviors and functional constructs. Objective: This scoping review aimed (1) to identify and summarize key online behavioral modalities and their specific features for depression monitoring and assessment, (2) to summarize existing evidence on associations between specific behavioral features and depression-related outcomes, and (3) to explore whether the behavioral content captured by digital measures has the potential conceptual links to RDoC constructs. Methods: We conducted a scoping review of 5 databases (PubMed, Scopus, Web of Science, Embase, and PsycINFO) from inception through January 2026, supplemented by citation tracking and gray literature searches. Included studies focused on online behavioral digital phenotyping for depression and reported specific behavioral features. Two reviewers (QR and JC) independently screened studies and extracted data, with disagreements resolved through discussion or consultation with a third reviewer (YH). Findings were synthesized descriptively. Two reviewers (QR and Kaili Tang) independently assessed potential conceptual links between each behavioral modality and RDoC constructs using the official RDoC construct definitions. Results: In total, 74 studies were included. Five primary behavioral modalities were identified: call behavior (n=31), SMS (n=21), screen use (n=41), app use (n=26), and keystroke dynamics (n=14), comprising 36 core features. A total of 31 (41.9%) studies reported findings attributable to individual features, but no single feature showed a consistently directed association with depression-related outcomes across studies. An exploratory RDoC assessment suggested that candidate construct relevance primarily involved social processes, sensorimotor systems, cognitive systems, and arousal and regulatory systems. Conclusions: Online behavioral digital phenotyping captures a range of observable behaviors, but current evidence does not support interpreting individual digital behavioral features as stable markers of depression or as direct proxies for specific RDoC constructs. Future research should improve the transparency of feature definitions and reporting, strengthen the reporting of feature-level findings, and use independent validation and construct-informed study designs to enhance interpretability.