Abstract
To mitigate the rapid spread of health misinformation and its negative impact, this study presents a comprehensive literature review on health misinformation detection. A systematic search is conducted using the Google Scholar database, targeting publications from January 2016 to February 2025. Inclusion criteria require full-text, English-language studies proposing health misinformation detection methods. A total of 100 relevant studies are included. The characteristics of health misinformation are identified through a detailed analysis of its concept, dissemination mechanism, psychological impact, and susceptibility. Datasets and evaluation metrics are reviewed, with issues such as class imbalance and inconsistencies in annotation standards being identified. The strengths and limitations of various detection approaches are examined. Machine learning approaches perform better when using ensemble methods, feature selection techniques, and embedding-based representations. Deep learning algorithms are strong in automatic feature extraction and high-dimensional semantic modeling, though they often face challenges such as high computational cost and low interpretability. Advanced detection methods show clear improvements in accuracy and explainability, while also introducing AI-generated misinformation and associated ethical concerns. This review provides a panoramic view of the current state-of-the-art in health misinformation detection. It further underscores the importance of interdisciplinary collaboration, human-centered design, and ethical considerations for the development of effective and clinically relevant detection systems.