Abstract
BACKGROUND Measurement-based care improves depression outcomes, yet implementation remains uncommon. Digital remote patient monitoring (RPM) tools collecting patient-reported outcomes (PROs) with automated alerts are increasingly deployed, but the respective contributions of algorithmic detection and clinician response remain unclear. OBJECTIVE To describe real-world use of a CE-marked digital RPM device for mood disorders and characterize symptom trajectories around algorithmic alerts, distinguishing alerts overall from those followed by clinician response. METHODS Multicenter retrospective cohort study using post-market clinical follow-up data from October 31, 2025, to May 4, 2026, implying thirty-four French psychiatric centers using a Class IIa CE-marked RPM device. We included all consecutive adults (≥18 years) treated for major depressive episode or manic episode, who completed at least one weekly questionnaire and did not opt out of research use of their data. The questionnaires consisted in weekly electronic PROs collection (PHQ-9, suicidal ideation item, adapted GAD-2, medication adherence) with threshold-based alerts routed to clinical teams. Alerts were classified as followed or not by documented clinician response. Outcomes included engagement and operational metrics (completion rate, patient response time, alert processing time, response conversion rate) and exploratory pre/post-alert trajectories for suicidal ideation, anxiety, and depression, analyzed using segmented interrupted time-series regression with Newey-West correction at both the alert and response-bearing alert levels. RESULTS Among 111 patients (median [IQR] age, 41 [31-54] years; 76% female), 84% were treated for depressive disorders, and 16% for bipolar disorder. Overall, 92% of questionnaires were completed, with completion rates exceeding 80% through month 5. Median patient response time was 17 hours, median clinician response time after alerts was 50 hours, and 69% of alerts were followed by a documented response. Response-bearing alerts were associated with significant slope inflections for suicidal ideation (−0.007/day; P=0.01) and anxiety (−0.004 per day; P<0.001), with nonsignificant trend for depression. Across all alerts, no inflection was observed for suicidal ideation, whereas anxiety showed a significant inflection and depression a level shift without slope change. CONCLUSIONS In this real-world cohort, symptom trajectory changes were most evident when alerts were followed by clinician response. These findings support further evaluation of closed-loop RPM in randomized trials. CLINICALTRIAL The study was conducted under the French Reference Methodology MR-004 of the Commission Nationale de l'Informatique et des Libertés (CNIL) (research not involving human subjects, in-house studies on health data).

This publication has 22 references indexed in Scilit: