Answers are grounded in this paper and cite their sources.
DOI: 10.6011/apj.2025.02 Received : January 24, 2025 Accepted : June 4, 2025 Advance Published : June 25, 2025 Asian Pacific Journal of Health Economics and Policy, Volume8, No1. 19 [Keywords] diabetes self-management, geriatric diabetes care, healthcare technology, artificial intelligence (ai), chatgpt, gpts, health economics, health policy #### Introduction Diabetes management in geriatric patients presents unique challenges due to the interplay of physiological, cognitive, and psychosocial factors. The International Diabetes Federation (IDF) reports that over 67 million older adults lived with diabetes in 2021, a figure expected to increase by more than 50% within two decades 1. This trend contributes to escalating healthcare costs, as managing diabetes in older adults often involves addressing multiple comorbidities and age-related complications. With an aging population and rising diabetes prevalence, the need for tailored interventions is critical. Artificial intelligence (AI) has emerged as a transformative tool to improve diabetes care 2. The utilization of AI technologies in diabetes self-management support is gaining traction. 3 Advancements in natural language processing (NLP) technologies such as ChatGPT show promise in producing human-like response, simplifying complex guidelines, and delivering personalized, real-time guidance. This capacity is particularly relevant for older patients, where cognitive and physical limitations demand tailored support. However, AI-generated responses substantially vary by model, reflecting differences in design and functionality 4. While some models prioritize detailed, structured information, others focus on brevity or conversational tone. This variability poses challenges in clinical settings, where coherence and accuracy are crucial 5. At the same time, individualized, patient-specific guidance remains essential to address unique needs such as health literacy and cognitive function. Balancing these demands̶standardized accuracy and personalized adaptability̶is a key challenge in deploying AI for geriatric diabetes management 6. ChatGPTʼs GPTs functionality allows users to tailor AI models for specific applications. Leveraging this feature, we developed the Diabetes Self-Management GPTs Support System (DSM-GPTs). This study evaluates the performance of DSM-GPTs in comparison with three prominent AI models: GPT-4 Omni, Gemini 2.0 Flash, and Claude 3.7 Sonnet. Through a detailed analysis of their responses, this research seeks to uncover the capacity of AI in diabetes self-management and to inform the integration of AI into healthcare from a health policy perspective 7. #### Methods Study design This study was a comparative analysis of a customized AI-based system; DSM-GPTs and three representative AI models - GPT-4 Omni, Gemini 2.0 Flash, and Claude 3.7 Sonnet - in delivering insulin injection guidance tailored for normal patients and geriatric patients. The study ensured coherence across AI models by using a controlled environment with Windows 10 Home (version: 22H2) and Google Chrome (version: 131.0.6778.87). Prompt Design The instructions input into the AI model are created by a practicing internist with extensive clinical experience and are designed as structured prompts that reflect medical integrity, accuracy, clarity, and practicality. Temperature Settings of AI Models All AI models used in this study were accessed through web-based interfaces rather than via APIs, and the default temperature settings were used for all models. To minimize the effects of variability in output caused by generative AI, all prompt inputs in this study were conducted using a web-based interfaces on May 12, 2025. Therefore, all output results presented in Supplementary data 1 and Supplementary data 2 were also obtained on May 12, 2025.
No takes yet. Share an insight, caveat, or question.
Motoki et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: