Rising amounts of patient records, and unstructured medical information become a problem for healthcare systems looking for insights. NLP, a branch of artificial intelligence, provides innovative solutions through natural language understanding for the extraction, analysis and interpretation of large amount of healthcare textual data. This review aims at discussing how NLP can be useful in tasks including, record extraction from the large volumes of unstructured records, optimizing clinical decisions, and improving predictions. It also describes the practical applications of the method, new developments like pre-trained language models, and multimodal methods, and covers issues of data confidentiality, model scaling, and ethical concerns. However, as the following sections of the paper will show, this paper also identifies current and potential developments for progress in NLP application to improve patient outcomes, optimize processes, and enhance the shift to evidence-based, individualized medicine. Research implications and future directions are presented, highlighting the concept is promising for healthcare.