ChatGPT and Occupational Therapy: A Study of Generated Program Feasibility

Introduction As the application of large language models (LLMs) in the medical field advances, the potential for creating occupational therapy (OT) programs remains unexplored. This study aimed to clarify the ability of Generative Pre-trained Transformer (GPT; OpenAI, San Francisco, CA, USA) to create OT programs. Methods Based on five case reports of patients with stroke and concomitant psychological symptoms, GPT was instructed to create OT programs. Five occupational therapists (OTRs) evaluated the generated programs and quantified the degree of agreement with the programs created by OTRs. Results The programs generated by GPT showed a low degree of agreement with the programs created by OTRs in all cases, with a rating of two points or less. The content was found to be general, lacking in specificity and individuality, and insufficiently specialized. The scores for all programs generated by GPT were 2/4 or lower. Discussion While GPT has difficulty creating OT programs based on patients' life backgrounds and specialized knowledge, it showed the potential to be used in some processes of OT program creation. This is likely due to a lack of pretraining and limitations in information.