Using large language models as a scalable mental status evaluation technique.

Mental health care faces a significant gap in service availability, with demand for services significantly surpassing available care. As such, building scalable and objective measurement tools for mental health evaluation is of primary concern. Given the usage of spoken language in diagnostics and treatment, it stands out as a potential methodology. With a substantial mismatch between the demand for services and the availability of care, this study focuses on leveraging large language models to bridge this gap. Here, a RoBERTa-based transformer model is fine-tuned for mental health status evaluation using natural language processing. The model analyzes written language without access to prosodic, motor, or visual cues commonly used in clinical mental status exams. Using non-clinical data from online forums and clinical data from a board-reviewed online psychotherapy trial, this study provides preliminary evidence that large language models can support symptom identification in classifying sentences with an accuracy comparable to human experts. The text dataset is expanded through augmentation using backtranslation and the model performance is optimized through hyperparameter tuning. Specifically, a RoBERTa-based model is fine-tuned on psychotherapy session text to predict whether individual sentences are symptomatic of anxiety or depression with prediction accuracy on par with clinical evaluations at 74%.