Yvonne Leung 1,2, Jiahui Wong 1,2, Mathew Gancarz 1, David Gratzer 2, Mary Jane Esplen 1,2
1 de Souza Institute
2 Department of Psychiatry, University of Toronto
The proliferation of online patient support groups (OSG) to meet the demand of mental health services warrants research on their safety and efficacy. This project will build on an existing natural language processing system to design an ‘Artificial Intelligence Co-Facilitator (AICF)’ to enhance the safety and quality of therapist-led text-based OSG in cancer populations. AICF will monitor each participant by analyzing their texts posted during OSG sessions to: a) identify those at-risk for significant distress and/or dropout and alert therapists for follow-up, and b) generate discussion summary and emotion profiles (e.g. anger, depression, anxiety) for psychosocial service tailoring and better client-treatment matching.
The current project will build on the ‘Patient-Reported Information Multidimensional Exploration (PRIME)’ framework designed to provide active surveillance for prostate cancer patients. PRIME is an automated ensemble of deep learning and natural language processing (NLP) techniques to impose structure on unstructured texts. Existing data of OSG participants with breast or gynecological cancer will train PRIME to recognize patterns associated with distress and/or dropout. Therapist feedback will guide the improvement of AICF until a high level of performance is reached. Next, 320 women with breast or gynecologic cancer will be recruited for usability testing. AICF performance will be compared to standardized measures on its ability to describe participant emotional states and predict significant distress and dropout.
AICF has the potential to improve the service and safety by tracking patient experiences and outcomes during a session, thereby enhancing online therapists’ capacity to manage psychological responses in a timely manner. Information can also be used to identify patients who benefit most from OSG. Ultimately, AICF may improve health service access by supporting online therapists, and patient adherence and satisfaction.