JMIR Publications

ABSTRACT

Background:

Early detection of clinically relevant changes remains a major challenge in mental health, as deterioration typically unfolds gradually through subtle alterations in daily functioning before becoming clinically manifest. Advances in digital phenotyping and smartphone-based monitoring offer new opportunities to capture these changes in real-world settings.

Objective:

This study aimed to develop and evaluate an algorithm capable of detecting behavioral changes associated with clinically relevant deterioration using passively collected smartphone data, and to examine whether instability in daily behavioral patterns can serve as an early marker of relapse or clinical transition.

Methods:

We conducted a prospective, observational, multicenter study including 92 participants with various mental health conditions. Passive behavioral data were continuously collected using a mobile application over periods ranging from 1 month to 1 year. Behavioral profiles were generated using heterogeneous mixture models, and changes in behavioral stability were detected using a Bayesian online change-point detection approach. Model performance was evaluated by comparing detected change points with clinically recorded relapse events, using receiver operating characteristic (ROC) analysis and sensitivity-specificity metrics.

Results:

A total of 40 relapse events with available passive data were included in the analysis. The best-performing model (ADSVWZ configuration) achieved a mean AUC of 82.99% (SD = 2.49). At a false positive rate (FPR) of 5%, the model reached an average true positive rate (TPR) of 40.12% (CI95%: 36.61–43.63), increasing to 61.10% (CI95%: 56.42–65.79) at 10% FPR. Model performance was strongly influenced by the temporal aggregation parameter, with optimal results obtained using minimal accumulation of past observations. Multimodal combinations integrating sleep, activity, and routine structure achieved the highest overall performance.

Conclusions:

Behavioral instability derived from passively collected smartphone data shows potential as a marker for the early detection of behavioral changes associated with clinically relevant deterioration. This approach enables continuous, individualized monitoring and may support more timely and personalized interventions in mental health care.