Emergency medicine is the primary place for acute diagnosis and treatment in hospitals, and undertakes urgent diagnosis and treatment services for patients who come to the hospital for emergency treatment. As most of the patients coming to hospitals are in emergency, critical and serious conditions, the prevention of pressure injury is often neglected, and this group is the high-risk group for in-hospital pressure injury[1]. Studies have reported that the incidence of pressure injuries in emergency patients is as high as 19.1 %[2]. Accurate risk prediction of pressure injuries in emergency patients has positive clinical significance, which has gradually attracted the attention of researchers. Research evidence shows [3], [4], [5] that the risk factors for pressure injury in emergency patients have specialty characteristics, and the existing universal pressure injury risk assessment scales such as the Braden Scale and the Norton Scale do not have satisfactory prediction accuracy for emergency PI, and there are problems of over-assessment, low consistency of assessment, and incomplete assessment, etc., and there is a pressing need to develop pressure injury early warning models with specialty characteristics of emergency medicine. Early warning model. Machine learning algorithms can analyze a large amount of textual information and images contained in pressure injuries, and more accurately and efficiently carry out the assessment and prediction of pressure injuries, however, different machine learning algorithms have different prediction efficacies for different diseases. In this study, based on three machine learning algorithms, we propose to construct a pressure injury risk prediction model for emergency patients and conduct a validation study, in order to provide an assessment tool for the prevention and early intervention of pressure injury in emergency patients.