Background: Health education aims to encourage individuals to voluntarily take action to improve their health. Throughout this process, adult learners can make informed decisions to change their behaviours and enhance their well-being. In medical institutions, employing appropriate instructional strategies to facilitate changes in health behaviours is crucial for enhancing patients' health knowledge, attitudes, and skills. Objective: Health education courses within hospitals are based on the Selection-Instruction-Practicing-Evaluation (SIPE) model, which is grounded in the self-determination theory (SDT). Due to constraints in human resources, the traditional SIPE (T-SIPE) mode is generally carried out with one-way practice without evaluating learners' outcomes or providing personalised feedback. This limitation hampers patients' self-care abilities and self-confidence. By referencing the embodied cognition theory, a robot-based SIPE (R-SIPE) approach was proposed to enhance patients' cognitive engagement. Method: To assess the approach, a true experiment was conducted in a health education course by randomly assigning 90 subjects to the experimental and control groups, adopting the R-SIPE and T-SIPE modes, respectively. Results and Conclusions: The results revealed that the R-SIPE group showed better learning achievement, self-confidence, technology acceptance, and learning satisfaction compared to the T-SIPE group. It was concluded that using robots to provide interactive learning, personalised materials, practice tests, and evaluation had great potential in health education. It is therefore suggested that different health education courses be developed to enhance patients' self-care abilities and increase their confidence in dealing with diseases.