Published Paper Details:

The integration of large language models (LLMs) into clinical practice has marked a transformative shift in the delivery, management, and personalization of healthcare. This systematic review explores the emerging landscape of LLM-based applications within healthcare systems, focusing on their utility in clinical decision-making, medical documentation, patient communication, and educational support. Recent advances in transformer-based architectures such as GPT-4, Med-PaLM, and BioGPT have demonstrated remarkable capabilities in understanding and generating medical language, enabling more natural, efficient, and informed interactions between clinicians and patients. The review is based on a comprehensive analysis of peer-reviewed literature published between 2020 and 2025, drawn from databases including PubMed, Scopus, IEEE Explore, and arXiv. Findings indicate that LLMs are increasingly deployed in tasks such as differential diagnosis, electronic health record summarization, clinical triage, and mental health support, often with accuracy levels approaching or exceeding human performance in controlled settings. However, significant challenges remain, including issues of model hallucination, data privacy, medical bias, and ethical accountability. Moreover, the lack of domain-specific fine-tuning and explain ability continues to hinder clinical trust and adoption. This review also identifies best practices for safe deployment and outlines regulatory considerations for integrating LLMs into real-world medical environments. By synthesizing the current state of research, this paper provides a critical foundation for future developments in AI-assisted healthcare and underscores the need for rigorous validation, interdisciplinary collaboration, and policy frameworks to ensure that LLMs serve as reliable and equitable tools in clinical practice.