Evaluating ChatGPT-4 as an Educational Tool for Physical Activity Counseling in Patients with Hyperlipidemia

Abstract

Objective: ChatGPT is widely used as a large language model tool for patient education across various medical fields. This study aimed to evaluate the reliability, quality, accuracy, comprehensibility, and readability of ChatGPT-4 responses to frequently asked questions regarding physical activity counseling in individuals with hyperlipidemia. Materials and Methods: The responses provided by ChatGPT-4 to 30 common questions about physical activity and management in patients with dyslipidemia were categorized into five thematic groups (C1-C5). These categories were associated with physical activity in hyperlipidemia, barriers, facilitators, content of physical activity, impact of physical activity on quality of life and health outcomes. The reliability (modified [m] DISCERN tool), quality (Global Quality Scale, GQS), accuracy (six-point Likert scale), comprehensibility (three-point Likert scale) and readability (Flesch Reading-Ease Scale) of the responses were evaluated. Results: The mean scores for mDISCERN, GQS, accuracy and comprehensibility were 3.30, 4.30, 5.43, 2.47, respectively. The mean readability score was 40.05. The ICC values for the mDISCERN, GQS, accuracy, comprehensibility were 0.634, 0.470, 0.742, and 0.567, respectively. Conclusion: ChatGPT-4 demonstrated a moderate level of reliability; high levels of quality, accuracy and comprehensibility in responses related to physical activity counseling. However, difficult readability and lack of source citations were observed in responses. Although ChatGPT-4 may use as a complementary tool to physical activity education in patients with hyperlipidemia, professional supervision is important to ensure safe and personalized health communication.

Ähnliche Arbeiten