Digital Phenotypes for Predicting Depression_Validation

This is a single-site, 12-month prospective observational study conducted at Korea University Anam Hospital. A total of 200 participants will be enrolled, including 100 patients with DSM-5 major depressive disorder, bipolar I disorder, or bipolar II disorder and 100 healthy control participants. Participants will be classified into the patient or comparison group according to prespecified eligibility criteria, and both groups will undergo the same 12-month prospective follow-up. Throughout the study, participants will continuously wear a Fitbit Charge 6 and will complete five scheduled clinical assessments at baseline and months 3, 6, 9, and 12. Clinical assessments will include structured diagnostic and mood-episode interviews, clinician-rated measures of depressive and manic symptom severity, and self-report assessments of mood, sleep, circadian rhythms, anxiety, functioning, and related clinical characteristics. Throughout the study period, participants will continuously use the My Mental HERO smartphone application in conjunction with the wearable device. Passive digital phenotypes will include heart rate, heart-rate variability, sleep, physical activity, GPS, ambient light, smartphone use, and communication-related metadata. Participants will also complete a brief daily eMoodChart assessing mood, energy, anxiety, irritability, alcohol use, medication adherence, and menstrual-cycle status where applicable. At each follow-up assessment, a structured mood-episode interview will be conducted to identify mood episodes occurring since the previous visit. For each suspected episode, trained raters will establish onset and offset dates, assess DSM-5 criterion symptoms and functional impairment, determine the predominant mood pole, document relevant specifiers, and assign an episode-level Clinical Global Impression rating. Episode classifications will be adjudicated using available clinical records, eMoodChart data, and other clinical information. The primary scientific objective of the study is to characterize digital phenotypes associated with mood episodes and to evaluate whether passive behavioral and physiological features can identify mood states and predict clinician-adjudicated mood episode onset. The dataset is specifically intended to serve as an independent validation cohort for digital phenotypes and prediction models developed in the preceding MentalHERO prospective observational cohort (ClinicalTrials.gov Identifier: NCT07151846). Because the present study uses a closely aligned design, assessment schedule, clinical evaluation framework, and digital phenotyping protocol, data from the two cohorts may also be pooled for secondary analyses when scientifically appropriate.