Forecasting relapse with biomarkers

At Psyrin, preventive care is at the core of our mission. A key area of interest is how we can predict relapse episodes in psychosis. To this end, we’re delighted to announce PRIME - our research initiative focused on using speech and smartphone data to forecast relapse risk in people with psychotic disorders.

Launched in collaboration with the Yale STEP Clinic, our broad aim is to turn easy-to-access digital signals into real-time clinical intelligence.

What Is PRIME?

Relapse is one of the most devastating and costly challenges in long-term psychosis care. Despite major advances in early intervention, up to half of patients relapse after their first episode. And right now, providers have no reliable way to see it coming.

PRIME is designed to change that. Through a simple, 5-minute voice check-in every two weeks, alongside optional passive sensing data from mobile phones, Psyrin’s AI models analyze speech patterns and behavioral trends that may signal increased relapse risk. We’re also incorporating digital self-report tools, clinician-reported outcomes, and cognitive data to build a full-spectrum, explainable model of relapse prediction.

Participants are tracked over 12 months, allowing us to compare data across stable and relapsing periods. Our aim is to test whether early warning signals, such as changes in coherence, tone, sleep, or activity, can offer real, actionable insight for clinicians before a crisis occurs.

Why It Matters

Right now, relapse is only visible once it's already happening. That means missed opportunities to intervene early, support patients, and prevent clinical deterioration. We believe a future where relapse risk is visible, predictable, and preventable is within reach: PRIME is a step toward that.

By validating these models in real-world care, Psyrin is building the evidence base needed to embed AI safely and responsibly into psychiatric practice. This isn’t about replacing clinical care. It’s about giving clinicians better tools, earlier insight, and a chance to act before things fall apart.

What's Next

PRIME aims to recruit 50 participants. A subset of participants will also opt into passive sensing.

We're currently engaging with partner sites and expect to publish early results on feasibility, prediction accuracy, and participant engagement within the study period.

If successful, PRIME will lay the groundwork for a scalable relapse monitoring tool that can be integrated into outpatient care and long-term management of psychosis, making support more proactive, timely, and personalized.

For collaboration or research inquiries, reach out to hello@psyrin.ai