Objective A scoping review of unplanned readmission(UR) risk prediction models for cancer patients was conducted to provide a basis for clinical practice and research. Methods The UR risk prediction model of cancer patients was focused,and the Chinese and English databases were searched systematically. The extracted information of the model included applicable population,the incidence of UR,modeling methodology,predictors of the model and their performance. Results 18 studies involving 23 prediction models were included and the population focused on postoperative colorectal cancer patients. The incidence of 30 days UR in cancer patients ranged from 8.2% to 19.0%. The model development methods were various,but the overall prediction performance was poor. Comorbidities,TNM,length of stay,age and postoperative complications were important predictors of UR in cancer patients. Conclusion Clinical staff should pay attention to UR risk factors and choose excellent tools to guide clinical practice. Prediction models with high predictive performance and operability can be developed with artificial intelligence and verified extensively and externally.