Purpose/Significance To improve the classification and evaluation mode of medical safety incidents, and to improve work efficiency and timeliness. Method/Process The data of previous medical safety incidents are pre-processed, BERT model is used for training, testing and iterative optimization, and an intelligent classification and prediction model for medical safety incidents is built. Result/Conclusion The model is used to classify 466 medical safety incidents reported by clinical departments from January to November 2022, and F1 value reaches 0.66. The application of BERT model in the classification and evaluation of medical safety incidents can improve work efficiency and timeliness, and help timely intervene in medical safety risks.