Since the emergence of social media, people have been affected by the misinformation from social media. In some areas of health, the lack of expertise in their respective fields leads to a weak ability to distinguish between genuine and spurious information. Detecting misinformation without professional knowledge continues to be an ongoing obstacle. To solve this problem, we design a Knowledge Guided Inductive Graph Neural Network (KG-IGNN) for detecting health misinformation on social media, which utilizes the professional information from a knowledge graph to enrich the semantic features of social media text. The method optimizes the feature representation by using the neighbor nodes of the article with an inductive graph neural network. We use a real-world dataset to demonstrate the effectiveness of our model. Experimental results show that KG-IGNN achieves significant improvement compared to existing methods.