JMIR Publications

ABSTRACT

Background:

Major depressive disorder (MDD) is characterized by persistent depressed mood, loss of interest, recurrent thoughts of death, and significant physical and cognitive symptoms. Despite effective treatment, up to one-third of patients relapse within months after discharge from depression care, revealing a critical gap in post-discharge surveillance, especially in rural areas with limited access to care and follow-up engagement. The Collaborative Care Model (CoCM), a widely adopted framework for managing depression in primary care, has improved detection and treatment outcomes; however, monitoring symptom recurrence after patients leave structured care still remains challenging. Digital technologies, especially smartphones, offer a promising way to close this gap by capturing passive behavioral and physiological data, including psychomotor activity, sleep, movement, social interaction, and light exposure. Building on prior work showing that passively sensed data can capture depression severity, mood fluctuations, and antidepressant use, we propose developing a clinically integrated digital biomarker to predict depression recurrence.

Objective:

(1) Test whether patterns in smartphone sensor features predict next-month recurrence of depressive symptoms in post-discharge CoCM patients with an AUC ≥ 0.8; and (2) to evaluate whether Electronic Health Record (EHR) integration of a passive sensing–based prediction model is associated with reduced MDD recurrence over six months compared to usual care.

Methods:

First, we will recruit up to 120 patients who have been enrolled within the Collaborative Care model to install the HIPAA-compliant MoodTriggers app on their personal smartphones to passively collect multimodal sensor data and complete monthly PHQ-9 assessments for six months. ConvLSTM models will predict next-month PHQ-9 scores, and Shapley values will identify influential features. Acceptability will be assessed through participation metrics and interviews. Second, a randomized trial of up to 200 patients will test effectiveness by evaluating whether EHR-integrated alerts based on the digital biomarker reduce depressive symptom recurrence over six months following CoCM discharge.

Results:

Funded 2025–2029, this project aims to create a scalable, clinically embedded digital biomarker for post-discharge depression care.

Conclusions:

N/A Clinical Trial: The study procedures have been registered on clinicaltrials.gov at NCT07174557.