Comparative evaluation of ChatGPT-4o and DeepSeek R1 in addressing hyperuricemia-related inquiries

Recent advancements in artificial intelligence (AI) have brought significant changes to healthcare communication, particularly through large language models (LLMs). While AI’s role is often debated, current consensus suggests that AI is designed to complement rather than replace healthcare professionals, highlighting the importance of human–AI collaboration. ¹ Hyperuricemia, a common chronic metabolic condition, requires patient education and lifestyle management, making it an appropriate focus for evaluating the utility of AI models in healthcare communication.

We evaluated the responses of two LLMs—ChatGPT-4o and DeepSeek R1—to 12 typical patient-centered questions related to hyperuricemia. Each question was entered once into both ChatGPT-4o and DeepSeek R1, and the resulting response was used for evaluation. Three clinical experts in endocrinology and rheumatology, each with over 10 years of experience, independently evaluated the responses. They scored the outputs on a 0 to 10 scale based on accuracy, completeness, and clarity. ChatGPT-4o received an average score of 8.44 ± 0.77, and DeepSeek R1 scored 8.58 ± 0.50, with overall high agreement among reviewers. Our findings suggest that both models offer valuable information, with ChatGPT-4o providing simpler and more patient-friendly explanations, whereas DeepSeek R1 offers more technical and detailed responses. This observation aligns with recent studies demonstrating ChatGPT's significant contribution to healthcare information delivery, particularly within informational domains. ²

Notably, while both models generated generally reliable answers, occasional omissions of clinical nuances were evident, underscoring the importance of integrating AI outputs with human expertise. These findings are consistent with the literature emphasizing that the true potential of AI in healthcare lies in enhancing patient education and supporting collaborative decision-making. ³

In conclusion, our evaluation suggests that ChatGPT-4o and DeepSeek R1 can provide reasonably accurate and accessible health information regarding hyperuricemia. However, their outputs should complement professional medical consultation to ensure safety and comprehensiveness. As AI tools continue to evolve, efforts must be directed toward structured integration frameworks that uphold human oversight and patient engagement to optimize AI's utility in healthcare.