Ali Sara, Aslam Atrubah, Tahir Zarmeen, Ashraf Bashair, Tanweer Afifa
Department of Nutrition & Dietetics, School of Health Sciences, University of Management and Technology, Lahore, Pakistan.
J Pak Med Assoc. 2025 Jan;75(1):78-83. doi: 10.47391/JPMA.11173.
The rapid integration of artificial intelligence into healthcare has introduced transformative possibilities and challenges. The current narrative review was planned to explore diverse applications of Chat Generative Pre-Trained Transformer (ChatGPT) across medical domains, ranging from dietary planning and disease management to medical education and clinical decision support. A comprehensive analysis of ChatGPT's healthcare applications was conducted between July and September 2023, reviewing literature from prominent medical journals and databases, including PubMed, Embase, Cochrane library and the Cumulated Index in Nursing and Allied Health Literature. The studies revealed notable limitations, including inaccuracies, bias and potential safety concerns. Quantitative data highlighted ChatGPT's high accuracy rates in disease detection, nutrient sufficiency in ChatGPTgenerated diet plans and various medical scenarios. The predominantly quantitative evaluations might overlook nuanced qualitative aspects, such as users' perceptions, experiences and ethical concerns. Studies often focus on specific domains, potentially limiting generalisability. Evolving artificial intelligence technology warrants longterm impact assessment, including ChatGPT's contextual appropriateness and accommodation of individual preferences. ChatGPT shows promise in healthcare, but needs specialised training for medical use. Ethical concerns, data quality and interpretability require thorough investigation for responsible implementation.
人工智能迅速融入医疗保健领域，带来了变革性的可能性和挑战。本次叙述性综述旨在探讨聊天生成预训练变换器（ChatGPT）在医疗领域的各种应用，从饮食规划、疾病管理到医学教育和临床决策支持。2023年7月至9月期间，对ChatGPT在医疗保健领域的应用进行了全面分析，回顾了包括PubMed、Embase、Cochrane图书馆以及护理与联合健康文献累积索引在内的著名医学期刊和数据库中的文献。研究发现了显著的局限性，包括不准确、有偏差以及潜在的安全问题。定量数据突出了ChatGPT在疾病检测、ChatGPT生成的饮食计划中的营养充足性以及各种医疗场景中的高准确率。主要的定量评估可能会忽略细微的定性方面，如用户的看法、体验和伦理问题。研究往往集中在特定领域，可能会限制普遍性。不断发展的人工智能技术需要进行长期影响评估，包括ChatGPT的情境适用性和对个人偏好的适应性。ChatGPT在医疗保健领域显示出前景，但需要进行专门的医学使用培训。伦理问题、数据质量和可解释性需要进行全面调查，以确保负责任地实施。
