The integration of LLMs in healthcare represents the next step in medical diagnostics, treatment planning, and monitoring. Most traditional diagnostic methods can only perform on single data types, making them poorly positioned to address the complexity of the human body. The main limitation of unimodal approaches is overcome by synthesizing diverse data modalities, such as text, images, and physiological signals, using multimodal LLMs for enhanced abnormality detection and context-aware clinical insights. This survey identifies the dire need of advanced AI systems for data integration and decision-making in closing gaps in healthcare domains. In this regard, the study has summarized applications, challenges, and potentials of multimodal LLMs in abnormality detection through a review of several research works. Key takeaways from case studies and emerging trends underpin the importance of explainability, equity, and innovation in leveraging AI for a more inclusive and effective healthcare future. This survey thus provides a better understanding of AI and LLMs via its congregation of information from a long list of research works that may inform and guide readers on the technology and its future.