Social media can facilitate rapid dissemination of misinformation, with consequences for political opinion, public health, and other collective decisions. Existing approaches to identifying misinformation spreaders often rely on writing style, content, profile information, or engagement statistics, but deceptive content is frequently designed to resemble authentic information. This work proposes a deep-learning model that combines content-based features with patterns of social interaction and information-propagation structure. Modeling these multiple facets provides a broader representation of users involved in misinformation diffusion. Experiments on COVID-related social-media data show promising improvements over comparison methods, supporting the use of interaction and propagation information alongside content when identifying likely misinformation spreaders.