Advances, reception and potential of ChatGPT as a tool for healthcare delivery and research: a systematic review.

Abstract: ChatGPT gained widespread attention for its capabilities in natural language processing, enabling machines to assess human language inputs and generate complex, yet evolving answers. As large language models (LLMs) continue to develop, clear guidelines are needed to help healthcare providers and educators maximise their benefits while mitigating potential risks. This review assessed the utility and accuracy of applying ChatGPT in healthcare assistance, specifically in understanding clinical knowledge and guiding clinical practice and research. A search on PubMed/MEDLINE for ChatGPT-related articles from 30 November 2022 (ChatGPT's release date) to 14 March 2024 yielded 2690 articles. After screening and reviewing, 2141 articles were deemed relevant to the clinical and research domains. Of the articles, 60.3% were supportive of ChatGPT, highlighting its immense potential for automating routine tasks, enhancing decision-making processes and addressing complex challenges in health care. However, 0.9% were not supportive of ChatGPT's utilisation in its current form, given the unresolved ethical implications and concerns regarding accuracy, bias, privacy and legal. Additionally, 38.8% had an equivocal stance, suggesting for further research to fully understand the rapidly evolving capabilities and potential impacts of ChatGPT in healthcare. This review presents a newly created conceptual framework, the 'ABCD model', to facilitate a systematic approach for researchers and healthcare practitioners to navigate ChatGPT's strengths and limitations. The model aims to align the development and deployment of ChatGPT by providing guiding principles, which ChatGPT and other emerging LLMs should incorporate into further developments to ensure their suitable application in health care.