Retrieval Augmented Generations (RAGs) is an emerging field of computing that has exceptional capabilities inherited from the strengths of both Large Language Models (LLMs) and Information Retrieval (IR) models. By integrating external, domain-specific knowledge into the response-generation process, RAG models overcome critical limitations of standalone LLMs, such as hallucinations, outdated knowledge, and limited domain specificity. RAGs bring novelty to its applications to wide range of domains, including healthcare where the need for accurate, transparent, and up-to-date information is critical. This paper presents a scholarly review of this cutting-edge field, with a focus on its adoption in healthcare. It synthesizes recent literature across various use cases, provides a comparative analysis of existing studies, identifies critical gaps in the current research, outlines the limitations in them, and suggests potential directions for future work. Unlike prior reviews, this work provides a structured comparative framework categorizing RAG healthcare applications by task, data source, model type, and evaluation strategy, thereby offering a roadmap for both researchers and practitioners. © 2025 IEEE.