•  
  •  
 

Document Type

Conference Proceedings

Abstract

Objective: This study examined ChatGPT’s ability to generate culturally relevant, evidence-based lifestyle recommendations for women with gestational diabetes mellitus (GDM), aiming to support clinicians in personalized counseling and shared decision-making. Methods: A descriptive pilot study design was employed using ChatGPT-5.0 to develop tailored lifestyle interventions for nine culturally diverse vignettes. For each case, ChatGPT generated three culturally adapted meal plans, two snack alternatives, and a home-based physical activity plan. The model was further prompted to design a culturally sensitive evaluation rubric based on American Diabetes Association (ADA) guidelines. Four independent reviewers and ChatGPT-5.0 used this rubric to assess the generated plans in eight dietary and cultural domains. Readability was determined using the Flesch-Kincaid Calculator. Independent reviewer scores were averaged and compared to ChatGPT’s self-scores to assess discrepancies and identify areas for improvement. Results: ChatGPT produced individualized plans closely aligned with ADA standards. Overall agreement between ChatGPT’s self-evaluation and human reviewers was strong, though researchers rated lower in the domains of Family Beliefs & Health Views, Cultural Sensitivity, and Diet Plan Completeness. The mean Flesch-Kincaid Grade Level was 8.47 (range 7.4–9.7), and the mean Reading Ease Score was 46.01, corresponding to college-level difficulty. Conclusion: When guided with structured prompts, ChatGPT can create guideline-consistent, culturally tailored lifestyle plans for GDM management. Future efforts should focus on optimizing readability and enhancing cultural nuance while addressing ethical, regulatory, and implementation considerations for clinical integration.

Download

Share

COinS