Abstract: Objective: To systematically evaluate and analyze the risk prediction model for 30-day unplanned readmission of stroke patients to provide reference for clinical practice. Methods: A comprehensive search was conducted for articles related to risk prediction for stroke patients with 30-day unplanned readmission risk from the inception of the databases to February 17, 2024. Screened literature and extracted data, and used PROBAST, a bias risk assessment tool for predictive model studies, to analyze the bias risk and suitability of the included studies. Results: A total of 10 studies on the development of risk prediction models for stroke patients with unplanned readmission were included. The area under the receiver operating curve of the model was 0.62-0.955. The 10 studies showed relative high risk of overall bias, good applicability, different performance, namely, each had its own advantages and disadvantages. Conclusion: The research of 30-day unplanned readmission risk prediction model for stroke patients is still on the way, and the prediction performance needs to be improved. Meanwhile, healthcare professionals should focus more on patients with advanced age, long hospital stay, indwelling tubes and multiple chronic comorbidities, and take reasonable preventive measures in a targeted manner.

Key words: stroke; readmission; risk prediction model; systematic review; evidence-based nursing