Assess the Utility of a Speech-based Machine Learning Algorithm to Predict Treatment Response to Psychiatric Interventions

This prospective observational cohort (n=200) will enrol outpatients aged 18–68 receiving TMS or Spravato (esketamine) as part of clinical care; speech recordings are collected before treatment, daily during treatment, immediately after treatment, and at 4‑week follow-up.

The primary aim is to train and validate a speech-based machine learning algorithm to predict treatment response, defined as a ≥2-point improvement on the Clinical Global Impression that is sustained through the 4‑week follow-up.

Participants complete ~12‑minute speech recordings responding to six prompts and standard clinical rating scales; the study will also monitor for any distress caused by the speech assessments.