Rain D. Gako, Marianne Joy F. Hassaram, Maria Kaye A. Hathaway, Jonathan Julius D. Jong, Kate C. Katada, Kit Roger T. Lubguban, Katrice Maureen C. Papas, Ma. Jeannette T. Paquera (Silliman University)

Introduction: The purpose of this research is to explore the potential of ChatGPT in stroke rehabilitation, specifically assessing physical therapists' satisfaction regarding its generated responses to stroke rehabilitation inquiries.

Method: This quantitative descriptive and comparative study involved licensed physical therapists from Silliman University batches 2020–2022 with at least one year of clinical experience in stroke rehabilitation. Using purposive and snowball sampling, 81 respondents completed a validated, 3-part online questionnaire assessing satisfaction with trained and untrained ChatGPT responses for stroke interventions and patient education, as well as intention for continued use. Responses were rated using 4-point Likert scales. Data were analyzed with descriptive statistics, Shapiro–Wilk test for normality, Mann–Whitney U test for comparisons, and alpha set at 0.05.

Results: A total of 41 licensed physical therapists participated, with most graduating in 2021–2022 and having 2–3 years of clinical experience; over half frequently managed stroke cases. For stroke interventions, trained ChatGPT achieved higher satisfaction (M = 3.50 ± 0.31) than untrained (M = 3.32 ± 0.32), with spasticity management scoring lowest for the untrained version. Both models were rated “very satisfied” for patient education (trained: M = 3.73 ± 0.39; untrained: M = 3.75 ± 0.37). Mann–Whitney U test showed a significant difference for interventions (p = 0.015) but not for overall satisfaction (p = 0.107). Continued use intention and peer recommendation both averaged M = 2.70 (“Agree”).

Discussion: The study revealed that licensed physical therapists were generally very satisfied with both trained and untrained ChatGPT responses for stroke interventions and patient education, with trained ChatGPT performing better in intervention-related content. These findings suggest that while baseline ChatGPT capabilities are adequate for delivering patient education, specialized training using clinical guidelines enhances its accuracy and relevance for complex intervention topics. This implies that integrating AI tools like ChatGPT into clinical practice, particularly when optimized with domain-specific resources, may support decision-making and patient management in stroke rehabilitation.

Keywords: physical therapy, ChatGPT, trained ChatGPT, untrained ChatGPT, physical therapists, satisfaction

Funding: Not Applicable

Ethics Approval: Silliman University Research Ethics Committee (2024-409-Hassaram)

Mr. Jonathan Julius Dianala Jong is affiliated with Silliman University and is based in Dumaguete City, Philippines. He is actively involved in academic and professional pursuits, with interests centered on higher education, research, and knowledge development. Through his association with Silliman University, Mr. Jong is committed to continuous learning and contributing to academic excellence and community development.