Open-domain question-answering systems aim to provide precise answers by understanding a wide range of topics and questions, which is crucial for enhancing the efficiency and accuracy of information retrieval. However, current systems face challenges when dealing with specific domains, including low data quality, inadequate handling of personalized requirements, and insufficient sensitivity in processing critical medical information. To address these challenges, a new strategy combining retrieval-augmented generation technology and information fusion methods is proposed, aiming to improve the performance of open-domain question-answering systems. By constructing a comprehensive real-case dataset that covers a rich text knowledge base of patients, hospitals, and doctors, deep information fusion is achieved. Utilizing secondary retrieval techniques, the system can accurately extract and utilize domain knowledge directly relevant to the question-answering task. Moreover, employing the BART model enables the use of retrieved information to generate precise and highly relevant answers. Experimental results demonstrate that this information fusion-based retrieval-augmented generation approach significantly enhances the accuracy and relevance of the question-answering system in dealing with specific domain issues. Compared to traditional systems, it shows higher sensitivity and adaptability in handling personalized consultations. The generated answers are not only richer and more accurate in content, but also better meet the specific needs of users.