Longitudinal digital speech assessment in progressive supranuclear palsy

Thirty-two native English-speaking individuals with probable progressive supranuclear palsy-Richardson’s syndrome (PSP-RS) were followed for up to 12 months with clinical and digital speech assessments every 3 months. Speech features from reading a standard passage and sustained phonation were related to clinical outcomes using repeated measures correlation over 12 months and used to develop machine learning (ML)-based predictive models. Longitudinal digital speech measures effectively capture the trajectory of disease progression in PSP-RS and demonstrate sensitivity to changes in disease severity over time, supporting their potential utility as scalable, objective, and remotely deployable digital biomarkers for monitoring disease course and evaluating therapeutic response in clinical and research settings. In addition, ML-based models leveraging speech features accurately predicted both concurrent clinical scores and future clinical scores at three months, demonstrating their potential as a prognostic tool capable of forecasting clinical trajectories in individuals with PSP-RS.